Tech Lab forum - General Discussions!

[mossy gull]

@jade scaffold what's your opinion on CAMM2?

[jade scaffold]

I think it's actually pretty smart

In theory it has more uses out side of just laptops because it's designed for LP DDR it's super super good at signal integrity which means you might actually be able to more stabily do full speed DDR on it if you wanted to

[mossy gull]

I personally think it brilliant, I welcome it, even on desktops, especially smaller desktops.

[jade scaffold]

It's also easier to mount a heatsink to

So super high power overclocking memory is a possibility if you wanted it

[mossy gull]

Do I hear "Faster and better cooled memory"?

Not that I care, I am still on 5200MT/s😅

[sonic meadow]

And I give to you.. the reason this pc was noisy

[jade scaffold]

Aaaaaaa

[sacred seal]

At least there is a fan. I once opened up a pc to find it missing

And that was on a i7 4790

[sacred seal]

Bitlocker is now off, every single time my computer restarts its asking for that stupidly long key

Pretty sure it's not supposed to do that, but I just moved to Japan and I don't feel like trying to fix it.

[surreal moss]

Is your TPM fried?

[sacred seal]

No clue, this is a basically brand new laptop

Got it like a month ago

[surreal moss]

I'd exercise that warranty

[sacred seal]

Unfortunately that is not really an option for me

Having just moved to Japan this is my only PC for the next few weeks

And I don't think I can claim a canadian warrenty in Japan

Other then that though, it seems to be working fine.

[sonic meadow]

you probably can claim the warranty, the issue being that stuff needs to be (or end up) in canada

[sacred seal]

Yeah, that is what I mean. Having to ship the laptop back to Canada only to have them ship it back to Japan would be super expensive

[cinder lagoon]

Ohh dang nice good luck setting up life there ^^

[balmy flicker]

not used Linux can some one explain what the instructions mean?

To run the Linux version, unzip the file, cd starsector, and then run ./starsector.sh.

[sonic meadow]

so, its expecting you to open a shell and do everything from there by the smell of it
hopefully... you should just be able to double click on the extracted starsector.sh (its a shell script file, akin to a .bat file on windows)
if it doesnt you will need to open a shell (doesnt matter which) and navigate to the directory (using cd) you extracted it to, and type ./starsector.sh in to the shell and hit enter

[cinder lagoon]

Most likely it will want some kind of options like -f -r flags or whatever (it can be found out by running the file with --help)

[sonic meadow]

Well they just said "run", not "run with --whatever-the-fuck"

[cinder lagoon]

Tbh if you are not able to run a shell (bash) file then you're probably better off not bothering with linux.

[cinder lagoon]

While that is true, the options you often need will be found out by this

[cinder lagoon]

Non-AI-Google + chatgpt are probably your friends if you do want to try anyways

Found a thread with detailed explainations tho:
https://fractalsoftworks.com/forum/index.php?topic=27345.0

[balmy flicker]

main issue is new to linux and the related device

[surreal moss]

Everyone has to start somewhere

[balmy flicker]

well i ran the executable but the game gives a resolution error which is annoying

[timber barn]

use a bigger screen

[balmy flicker]

the steam deck has a resolution of 800x1280

i have a plenty large screen

it is just in the wrong orientation

[balmy flicker]

found a solution i need to edit a json file

[surreal moss]

watching someone else learn linux like

https://tenor.com/view/star-wars-journey-to-dark-side-evil-luke-skywalker-emperor-gif-19918388

[balmy flicker]

well with what ms is up to atm i might have to make the switch

[balmy flicker]

It works

[surreal moss]

Hell yeah!

[cinder lagoon]

Nice one, didn't really believe in it ^^'

[jade scaffold]

All sorts of weird RTX 5090 rumours flying around

-1st: one and the most weird IMO, one leaker has reported the 5090 will use a Dual slot
How IDK but other leaks suggested the massive titan ada 4 slot, however all leaks seem to agree the FE 5090 will have 3 pcbs...

-What could be happening here is that NVidia are simply testing things for the purpose of maybe outlining laptop cooler specs or perhaps for a quadro type card that might benefit from lower power and a dual slot design to make NVLink more convenient. so its actually possible for both leaks to be true here.

-2nd the 5090 might have a 448 Bit Memory bus cut down from the dies potential max of 512 Bit leaving room for a 5090TI or what ever if this turns out to be true this could result in the 5090 having 28GB of Memory rather than 32GB unless the card uses custom built GDDR to get an unusual amount.

NVidia are Due to announce in the coming week a New ARM SOC from a partnership with Mediatek and rumours earlier last month suggested the 5080/5090 would be announced though not released

[surreal moss]

I’ll just plan on a 40 series

[cinder lagoon]

I've seen pictures of an apparent vertical mount RTX 5090 where there are multiple PCBs in different planes of the card

[jade scaffold]

that wasnt the 5090

that was the Titan Ada cooler

was built for a 900W GPU

[cinder lagoon]

?? It had a gicantic cooler

[jade scaffold]

but its going to be off the same design

[cinder lagoon]

Ahhh

[jade scaffold]

hold up ill get a pic of it

This absolute monstrocity

[cinder lagoon]

Yeah somewhat like that

[jade scaffold]

i believe thats the cold plate on the bottom

[cinder lagoon]

I saw images of pictures without the brass frame, just the cooler itself

[jade scaffold]

has 3 fans

this cooler is actually off its rocker

[cinder lagoon]

It is

[jade scaffold]

allegedly it can do 900w though i doubt blackwell will be anywhere close to 900w
but a 550w gpu on a cooler that can do 900w should run really damn cool

jeesus christ look at the power cable

thats a THICK chunk of copper

[cinder lagoon]

I wasn't sure what that even is - does it use this chunk of copper to transmit the power from the connector the the actual PCB?

[jade scaffold]

i believe so

that side where thoes 2 copper contacts are is the underside of the card the side that faces the Mobo

the 12+4 pin power is on the opposite side of the card to that

Here it is compared to the 3090 and 4090
monster i really hope we get this cooler makes having to get a water block so much more optional

[cinder lagoon]

Just insanity

It's literally the 5090 memes come true

[jade scaffold]

yep xD

i bet that thing is gonna break some PCIE Slots

will be interested to See what ever prototype sparked the Dual fan 2 slot card rumour

[cinder lagoon]

Possibly even while not being moved

[jade scaffold]

nah it should be fine when not being moved

NVidia screw their IO plate into the Cooler properly on the FE cards which when not done is what actually causes card bend

and most manufacturers dont do it for what ever stupid reason

[sacred seal]

Well I just unpacked my computer I had shipped in a suitcase. Found this at the bottom

Fortunately I don't have anything with a green PCB in my computer so I was trying to figure out what the heck it was.

Turns out I forgot to unplug my USB Bluetooth receiver and it just sheered off.

Have to figure out how to remove the bit that is still stuck in the port, but the port seems undamaged

[cinder lagoon]

Ooof lol

In terms of replacement: i recommend going the extra mile and getting an M.2 WiFi+Bluetooth card from intel

[sacred seal]

Unfortunately M.2 is a premium on this board. Only one slot, and I already have an expansion card to add a second.

Fortunately I don't really use bluetooth, I origionally got that adapter for the old VR headsets

[cinder lagoon]

Tbf the bluetooth adapter is internally hooked up via USB anyway

[cinder lagoon]

If you'd possibly want more:
https://www.waveshare.com/pcie-packet-switch-4p.htm
25 bucks a piece, actually worth it. Bought two. Pcie switches are basically driverless, the OS/firmware just needs to support enumeration (which basically everything does, except the boot code of the rPi)

[sacred seal]

I take it that that particular fact was learned through experince?

[cinder lagoon]

..more or less ._.

And the bootcode being rPi's only closed-source component... unfixable :/

[thorn osprey]

Hey, so, I just got new internet and I connected my computer to it. It seems to be doing really well except for the fact that in Discord, it takes around 30 seconds to connect to a VC, then it says I have 5000ms of ping, and doesn't let me talk. My Steam is also acting weird. I came out of offline mode and switched my friends status from offline to online, but no news for games comes up, the friends list doesn't come up, no games say they need updates even though I know they need to be updated, it doesn't let my download new games because I have "no connecting", and when I signed out of my account to log back in and see if that fixes it, it acts like I'm not online still. It doesn't even give me the QR code to use the mobile sign in.

Anyone know what to do?

[sacred seal]

Speedtest?

Might be an issue with dns, though that would not explain the high ping

[mossy gull]

You could try changing your DNS to 1.1.1.1 and 8.8.8.8.

[surreal moss]

Is it hardwired or wifi?

[mossy gull]

Even WiFi shouldn't be that bad

[surreal moss]

Still, start troubleshooting from the end rather than the middle

[mossy gull]

Sometimes WiFi can be faster than wired, that's why I never take WiFi as an excuse these days😅

But still, it's odd the Internet for them is that slow.

[thorn osprey]

So uhh, some other folks ended up helping me, we tried evertyhing that ya'll had suggested here, none worked, and we gave up. Came back a few hours later and it just... worked

[cinder lagoon]

Just as I thought - probably the router was not yet correctly logged into the ISP.

[mossy gull]

So they replaced the XGMobile connector with USB-C 4 with external GPU support, why didn't they just go with Oculink like the Chinese manufacturers.
USB4 is too slow for current gen GPUs, then they better could've kept the XGMobile connector and make it more widely available and cheaper like Oculink.
https://youtu.be/iAJ-J1R7juY

[cinder lagoon]

Probably because hot plug ability.
Honestly I think I would choose USB4 any day of the week over oculink. The couple extra lanes in cost of adaptability aren't worth it imo

[mossy gull]

Not only extra lanes, USB4 has PCIe 3 speeds, while Oculink currently has PCIe 4 speeds, Oculink scales with the PCIe speeds of the CPU.

[cinder lagoon]

Well, it's basically a miniature PCIe connector in a cable.

[mossy gull]

Exactly, so I don't understand why they take the slower port over the faster 1😐

[sonic meadow]

The upside of usb4 is that.. it's still a usb port

[cinder lagoon]

Because!! The oculink port can only do PCIe. And the ginourmoud data rates of gen3x4 are really more than enough imo.
If you have a low-power mobile system that you wanna equip with extra punch while stationary I think that you don't need the highest possible performance, just some extra power budget for a little bigger gpu

[sonic meadow]

Which means it's usable for usb things. Alt-mode video, charging, usb. Assuming of course it is configured for all them

[mossy gull]

Then give the device more USB ports while also keeping Oculink 😐

[sonic meadow]

Cost

[cinder lagoon]

Size

Complexity

[sonic meadow]

Routing

[cinder lagoon]

Ergonomics

[cinder lagoon]

i like how we both had the same way of reacting lol

[sonic meadow]

It could also be as simple as.. that's what the engineer who was working on it knew how to implement. Or was comfortable doing so

[mossy gull]

I just feel like big companies don't want to take risks with handhelds

[sonic meadow]

Handhelds are a risk

Err, I should have emphasized "handhelds" not "are"

[mossy gull]

But people want them, they're pretty popular

[cinder lagoon]

Honestly- even USB- Dual-Role- PD is difficult af.
Can't imagine how difficult USB4 would be

[mossy gull]

USB4 with GPU support is more complex than Oculink

[cinder lagoon]

Given that oculink is a physical standard, it's probably way easier.

[sonic meadow]

Pc handhelds are a relatively new market, and despite their apparent popularity are still fairly niche.

[cinder lagoon]

I'm fairly sure that companies will once they have a big enough foot in the door

[mossy gull]

Like AyaNeo and GPD?

[sonic meadow]

The bigger companies getting in to it are still testing the water. Seeing what they can make, seeing what customers want, seeing how they can get extra sales out of a single initial product

[sonic meadow]

Who and who?
(And yes I know who they are)

[mossy gull]

Okay, OneXPlayer😃

[cinder lagoon]

Those are ally startups/indiegogo retailers lol
They sell a very limited amount of their products, mostly built to order so to say.

I mean the asus-es and alienware-s. Both of which seem to be too profit-oriented to release actually good products

[sonic meadow]

need framework to build a handheld, or just use this guys plans 😆
https://www.youtube.com/watch?v=0MAnyzWsupc

I still need to watch part 2 actually

[cinder lagoon]

They gotta employ this guy so he can make a proper one lol

[mossy gull]

I seen them a lot in the wild though, especially because I travel a lot with public transport.

[mossy gull]

I was thinking on doing this, but the motherboard of even the FW13 is pretty large.

[cinder lagoon]

They are the biggest ones in the category but their category was pretty niche up until some months sgo

[sonic meadow]

yeah, thats something he talks about a bit. big motherboard, pain to use 😆

its also something you see often when people post buildlogs of laptop->handheld conversions (well donor motherboards from laptops)

[mossy gull]

I am planning to build my Asus ROG G513 into a tablet.
It's powerful enough to do a lot of stuff, but I have to replace the cooling for it.

[cinder lagoon]

I want Framework to sell me a device that lets you take a usb-port and make it Dual-Role- able for power delivery
I.E. a laptop usb-c port upgrade for embedded desktops

[mossy gull]

From what I understand, all USB-C ports on the Frameworks support power delivery

[cinder lagoon]

Exactly

I want a breakout board for diy projects tho

Like you give the thing a 3.2 USB A port, and two terminals for power delivery (for USB- devices) and power reception (for charging)

[cinder lagoon]

The USB-C modules are just extensions for the "integrated" c-ports all modules connect to. The charging and whatever logic is in the laptop without the modules

[mossy gull]

I know

[mossy gull]

Been looking at the specs:

• The Asus ROG Ally X has the AMD Z1 Ultra and 2 USB-C ports, HALL effect joysticks and 24GB RAM

• The AyaNeo Flip has the AMD Ryzen 8840U, 2 USB-C ports(1 USB-C 4 with eGPU support and 1 USB-C 3.2) and an Oculink port, HALL sticks and a finger print reader, and max 64GB RAM.

Both start around 900.- in the Netherlands

[jade scaffold]

On the topic of ovulink there's no reason you can't hotswap oculink
The issue is because it's oculink no alternative peripherals exist for it only pcie devices

Basically you'd need a dongle for everything

It's also not a particularly long lasting port

It's also on the loosing side of server connectors

[cinder lagoon]

As I said - the Asus-es and alienware-s are too greedy to release great products
(The Z1 supports eGPU via a proprietary connector for which they make a senselessly overpriced stupid GPU module

[cinder lagoon]

?
Isn't it not just a physical PCIe adapter? No software involved?
Bc theoretically PCIe supports hotswap, it just needs to be implemented correctly

[jade scaffold]

Yes but if you want to turn your oculink port into anything else

You'd need a dongle just for that

And a expensive one too

[mossy gull]

Same if you want to connect a PCIe device to USB4/Thunderbolt 🙃

[jade scaffold]

Yes

But I wouldn't need a special dongle to charge run video or usb over usb 4

Just a wire built to the correct spec

Oculink also can't handle higher speed pcie

There's a reason it practically doesn't exist in the server space

[mossy gull]

Neither can Thunderbolt

[jade scaffold]

Actually it can

Intel have drawn up the specs for it it should work

Oculink has no support anymore because it's just inferior to MCIO

[mossy gull]

Thunderbolt 4 supports up to 40Gbps, Oculink currently supports up to 62Gbps over PCIe 4.0.

[jade scaffold]

I've seen more oculink on consumer devices than on server boards

[mossy gull]

Mostly on tiny PCs from the Chinese market.
AyaNeo, GPD, OneXPlayer, I think even Minis-Forum has it

[jade scaffold]

Oculink is just a trash standard

Usb4 is both more convenient and more reliable

[mossy gull]

Oculink is an amazing port I think, better than the proprietary ports from Fujitsu, Alienware and Asus.

[mossy gull]

USB4 latency makes it a pain to use with GPUs though

[jade scaffold]

Oculink has literally been ditched

[mossy gull]

Not by Chinese handhelds 😅

[jade scaffold]

It's not good for long term it has no plans to scale in to the future

[mossy gull]

We will see what the Chinese will do with it

[cinder lagoon]

Ahh yes now I get it - an oculink to PCIe adapter with and integrated USB controller or whatever

[mossy gull]

For me it's just the latency of USB-C and Thunderbolt that makes it for me not interesting for eGPUs

[cinder lagoon]

It can't be that bad.. I suspect it's something like an extra frame in latency, maybe two..

It probably also makes a difference whether you're using the integrated display or one connected to the gpu

[mossy gull]

If you take the 4090, it loses almost 50% performance on USB-C 4.0, only 30% on Oculink

[sonic meadow]

ah yes, a device that can actually utilize a substantial amount of the performance of a full pcie slot loses performance on a compromise external connector

[mossy gull]

That 20% difference is a lot

[sonic meadow]

which sounds like a bandwidth rather than latency issue

[jade scaffold]

You antnt plugging a 4090 into a CPU that can only just about keep up with a 4080

[mossy gull]

Latency on USB-C and Thunderbolt is also bad enough that the 1% highs and lows are more aggressive and last longer.

[sonic meadow]

latency

[mossy gull]

Also latency is around double, ±15ms on Oculink, 38ms on Thunderbolt and USB-C

[mossy gull]

The 1% highs and lows has to do with latency

Even Alienware Graphics Amplifier port feels more smooth than USB-C

The unstable performance over USB-C and Thunderbolt is probably why people stay away from external GPU

[jade scaffold]

Oculink is still trash

Use a connector rated for more than half a use

One that actively has alternate peripherals for it would be great too

[sonic meadow]

bring back pcmcia

I need to use my 28.8k modem

[jade scaffold]

That or intel should just release the thunder bolt 5 spec as open source

[sonic meadow]

open source or unlicensed?

[cinder lagoon]

The reason why people stay away is because eGPU docks are vastly overpriced lol

[jade scaffold]

What ever they did last time

Probably unlicenced

The thunderbolt spec is genuinely kinda amazing

[mossy gull]

For Oculink it's lower than 100.- for the converter😅

[cinder lagoon]

Still really expensive for a physical only adapter

Tbf most egpu docks have a PSU built-in so ther's that

[mossy gull]

Thunderbolt docks for GPUs easily cost easily 300+ these days😅

The adapter only will cost me a whopping 150.-

[jade scaffold]

Funky BOI

[cinder lagoon]

Actually wayy more stoked about the 9000 series desktop CPUs

But DAMN thats alotta cores!!

[mossy gull]

288GB HBM3E🟡

[jade scaffold]

yea MI350 is kinda mid af

2025 availability for a product that at its release is gonna be 1 bordering on 2 generations behind

not looking great for AMDs AI Accelerators

[mossy gull]

Sometimes the working ram is more important than just raw horsepower

[jade scaffold]

yea thats cool

Except

AMD dont have a good scale out option

They have to go over Ethernet

which kicks latancy in the balls which really hurts the efficient useage of the RAM they have

[mossy gull]

At my job we always upgrade to the cards with the most VRAM, not the most horsepower.

[jade scaffold]

neat

but that doesnt change the fact NVidias B200
A will Smoke AMDs MI350 before it even releases

B that NVidia will drop a new generation next year again

AMD are Gambling on the fact their cards are cheap for their RAM

but there comes a point where being able to do somthing in 1 5th the time as somthing helps so much with power efficiency that its just cheaper to buy more GPUs up front

[mossy gull]

We had an issue while back where our stuff took 4 times as long to load in because Nvidia accelerators ran out of memory

[jade scaffold]

no you had an issue of not having large enough infrastructure

the idea is B200 can do anything MI 350X can do in 1/5th the time and thus at like a quarter the power

the upfront cost of these accelerators doesn't matter what matters is how much money they will cost you long run

[mossy gull]

Power consumption isn't an issue as we use the heat to heat up small parts of the building 😅

[mossy gull]

We upgrade once in 2 years😅
I am actually allowed to buy the old ones, but I don't have 20K

[jade scaffold]

AMD are running over ethernet this means junk data gets stuck waiting arround in RAM till it can be verified to have been sent to where it needs to NVidia are using NVLink which is a few orders of magnitude faster thus less waiting

AMDs cards will still end up with more useable VRAM at a time but not as much as you think

and certainly not enough to justify them costing 4x more to run

or for justifying just the time savings alone

[mossy gull]

It depends on how you do stuff😅
Here in the Netherlands it becomes normal that buildings start to use the heat from servers

[jade scaffold]

when you can just have a higher upfront cost of B200/R200 and end up saving millions on time and on power

[jade scaffold]

yea except no building is actually big enough to be really heated

in a world where servers produce as much heat as a small nuclear reactor

if your server isnt producing that kind of heat your a small company and the expensive NVidia products arent targeting you thats what Intel and AMDs are trying to target

but its also why AMD have minimal market share despite actually selling lots of MI300s

Intel also have the capability to easilly out produce AMD with superior products while amd are stuck wating for TSMC to be able to produce them

[mossy gull]

We also can heat multiple buildings, we also use the heat from factories and power plants to heat up parts from a city

[jade scaffold]

AMD are rapidly loosing

their accelerators sell well pretty much exclusivly because the NVidia ones are so bought up by other big companies

[mossy gull]

Except Microsoft, a company with 1 of the biggest servers farms, have already said they are going to use AMD😅

[jade scaffold]

they arent tho

Microsoft put in one of NVidias biggest orders

Microsoft are buying what ever they can get their hands on

another issue is NVidia can just release a GPU with the same ammount of VRAM it costs them nothing pretty much to do and puts AMD in an even tighter spot

though the biggest and most pressing issue AMD have
they cant get their GPUs made

they literally cannot produce the supply to meet the demand
Intel have their own fabs intel could just build their own GPUs them selfs
NVidia is elbo deep in TSMC
and while samsung are an option Samsungs node is uh crap atm

i dont see AMD being atall competitive or winning any market share not when their next gen accelerator is already going to be last gen when it releases

[mossy gull]

People said years ago that AMD CPUs will never be better than Intel

[jade scaffold]

Yes and amd proved them wrong

Amd are currently proving people right

They haven't had a real awnser to H200 till today. Their awnser to B200 isn't till later next year

Which means they have no awnser to R200 when it releases next year

Then no awnser to what ever comes out the year after that and so on.

Their 1 advantage is they usually provide more ram for their price which is good for start ups and smaller companies

[mossy gull]

Nvidia is now on top, but even they eventually walk against the limits

[jade scaffold]

Not unless AMD grow in size by 10x

AMD Simply do not have the money to Out RND NVidia

[mossy gull]

We still have ARM as well

[jade scaffold]

which NVidia also make ARM CPUs specifically for Data centre

and Intel too
Though AMD are clawing alot of datacentre CPU back from Intel

I really want that 500 core Intel CPU that seems really interesting to mess with

this goofy aaaaaa socket

Need to see if i can scrounge a cheap one when they eventually drop in price

will be good for hosting lots and lots of little servers

[jade scaffold]

also yea thoes are REALLY interesting

Look at the cache ammount

me thinks they can actually fit more cores into a CCD

40MB per chiplet as apposed to 32

also also

LOOK AT THAT EFFICIENCY

65wTDP on that one thing i dont like is USB4 on all Mobos

this means USB4 is either gonna have to share PCIE with a internal slot or a internal slot is gonna be ditched

mind you USB 4 is only Gen 3x4 so you only need a Gen 5x1 lane to fuffill that bandwidth

big memory controler improvements yay allways good to see on AMD Chips

[cinder lagoon]

Pretty sure USB4 pcie doesn't use more lanes than the regular USB controller it uses, the PCIe stories are somewhat spftware-defined, ain't it?

[mossy gull]

Nope, USB itself does use PCIe lanes

[jade scaffold]

Intel have officially announced what this is

It's actually really nifty

40% overall power decrease

Huuuuge IPC improvement on the E cores specifically between 38% and 68%, only 16% on p cores same as Ryzens getting meaning how intel do performance wise I tinks gonna depend on how much they can get their clock speeds up

If they can get their clock speeds up by like 10% they might end up ahead of AMDs none x3D chips

50% increase in graphics too

I really like this ram on chip approach I really hope we get to see some of that on desktop

[jade scaffold]

Intel have ditched hyper threading on this new chip using the area and power saved to cram in 2x as many E cores with the new faster E cores in theory multi threaded workloads should be waaaay faster

This means intel believe a E core can do well in excess of 30% the performance a P core can do

[cinder lagoon]

Duh, I mean it's not separate PCIe lanes that are then routed through the controller. I mean the USB controller and PCIe tunnelling share lanes

[mossy gull]

That's what I meant as well😅

[mossy gull]

In practice the E-cores often felt slower than the hyper threaded cores, or else they would be kicking AMD's ass in the consumer market.

[jade scaffold]

That's cos the old ones were just not amazing

They were really efficient and performance was decent but they weren't a substitute for the 30% performance increase hyper threading gives

However these new E cores are actually dangerously close close to 12th gen P cores in performance and with now 2 cores replacing the hyper threading ontop of the e cores the chips would have already had otherwise that's alota performance

Way more than what a thread can do

[mossy gull]

And funny enough, in laptops the Intel E-cores lost from AMD p-cores in efficiency and performance 😅

[jade scaffold]

They didn't tho not in efficiency

You make the most random stuff up sometimes

The old intel E cores perform like intel 8000 at worst a 10th the power consumption

[mossy gull]

How is it that Intel laptops lost in efficiency and in performance against AMD then?

Whenever I look at overall performance and efficiency on laptops, Intel always had worse battery life and performance was never better last few years.

[jade scaffold]

Because of the behaviour between E cores and P cores

This is literally like the 5th time I've had to explain this to you

[mossy gull]

But you can disable the P-cores and AMD still is more efficient in laptops.

[jade scaffold]

What used to happen was the P cores always had to be active as data was always passed to them first. As previously discussed hyper threading uses power especially so when idle and with intels P Core architecture being way less efficient than AMDs this means they kinda drink power

[jade scaffold]

Yea but they aren't

AMD dont have a 6w chip that can boddy the 7700k/3300x

AMD are also generationally ahead of intel

Intel do this weird thing where they innovate make a bunch of new stuff then basically do refreshes for 4 years before repeating the cycle

So a Ryzen 7000 chip is literally like 2 generations ahead of current intel chips which is funny

[mossy gull]

Except Intel CPUs always perform above their spec, I seen an 8650u pull 72w when the thermals were right, that's a 15w efficiency CPU by the way, I think I saw it's power limit on 91w in XTU, but I never saw it reaching it.

[jade scaffold]

Might even be 3 gens

With the upcoming generation Intel's chiplet tech has actually surpassed AMDs it's waaaay faster

[jade scaffold]

So do amd

And just missreport their power

65w on Ryzen 7000 is actually 88w

[mossy gull]

I never seen the 5900HX use more than double its power.
45w chip never use more than 80w.

[jade scaffold]

Also this is again intel just trying to stay relevant with a last gen product

I've seen the 6w chip in usage the CPU portion won't go over 6w with GPU maxed out with the CPU it'll do like 9w

Anyway next gen is a complete architecture overhaul

[mossy gull]

I heard that often from Intel and it's constantly just a 14nm++++++ effect

[jade scaffold]

No you haven't you've had them get stuck rereleasung the same thing over and over because they couldn't make their node work on their machines

Intel next gen are directly targeting Qualcomm X elite

They've completely redesigned their GPU architecture the NPU in their new chip is faster than AMD and apple's offerings combined

And the core architecture is unlike anything we've ever seen

[mossy gull]

That's what they say now, I want to see real performance benchmarks.
Not from Intel's cherry picking list.

[jade scaffold]

The chiplets are built on top of a proper silicone interposer making it wildly fast. Intel have an opportunity here to actually compete

At power consumption 40% lower than last gen

120 TOPs of ai performance is insane

[mossy gull]

Last time I heard Intel being efficient their CPU started using 300w in games and then they started blaming motherboard manufacturers.

I wonder how AMD will treat the AI shinanigans, we haven't seen much from them.

[jade scaffold]

We literally have tho

They just announced new Ryzen AI chips that have their new 890m GPUs in them

Nifty little things

80TOPs of ai performance on the top end model

GPU is a node shrunk rdna3 one that has 2 more CUs should be pretty substantially faster

[mossy gull]

I mean, Benchmarks can say anything, where Intel CPUs 90% of the time dominate AMD, and in the real world are either barely faster or the other way around, slower.
They both are amazing at cherry picking benchmarks so I'll wait till the real world performances are out.

Since Intel's 28-core 5GHz overclocked and chilled Xeon pretending to be an i9 getting beaten at 5.1GHz 32-Core Threadripper on air cooler I tend to not really listening to Intel.

I had tons of problems with Intel in the past, and usually had to do with compatibility, so Intel has to seriously step up their game in trust before I will consider them in any system.

[jade scaffold]

Intel tends to be faster in the 1st gen of their weird little cycle

[mossy gull]

And 1st gen also has the most annoying bugs that exist.
I remember programs being pushed to the e-cores on 12th gen while the programs needed the P-cores.

[jade scaffold]

But judging from the markets Intel's new SOC AMDs new APUs are aimed it I'd say Intel's should be rather substantially faster at around the same efficiency probably higher max power draw of 80w much like with the X Elite

[jade scaffold]

Windows scheduling moment

Though scheduling issues for both sides shouldn't be a thing next gen

The distinction between different cores is better made both in low level firmware and in window's scheduler

[mossy gull]

I want real world performances, not once again the 5GHz Chilled moment

[mossy gull]

Windows's scheduler sucks no matter what.
And it'll make things run worse than they should.

[jade scaffold]

It doesn't tho

W11 scheduler is way better than pretty much any other OSs one

[mossy gull]

I doubt that, I always had way worse performance on W11 than on latest Ubuntu, not to mention I don't know if it's my systems, but everything just takes longer to load in general.

[jade scaffold]

The issue with w11 not putting things on the right cores was an issue with AMD for their cpus on intel it was a windows issue I believe

[jade scaffold]

Doubt it all you want it's factually true

W11 is just more full of random crap you don't need so ends up slower overall

Take that way and W11 has easily the best scheduler out of any mainstream os

[mossy gull]

I'll wait for W12, hopefully it'll feel faster and get some customisation back.

W11 is substantially worse if you use unconventional stuff like stacked monitor layouts, or having 1 touch displays instead of all 3.

W11 is like another W8 moment, just a lot worse.

[jade scaffold]

W11 has alot of really nice features but it lacks so much from previous windows

[mossy gull]

Is it me or did W11 go away from the touch UI?
On my SB2 in W10 when I use the touch display the bars on top with the close and minimise button always grew larger so you could hit it easier, in W11 that doesn't happen at all.

On 1 side I am really sad the Tablet mode was removed, it was really nice to have when using the touch display

[jade scaffold]

But yea I'm really excited for the new intel SOC from specs alone and from just the ipc from some of the more pretty objective tests it's gonna be real good can't wait to get some more details on the GPUs configuration and how replacing the threads with 2 new gen E cores works out they seem to be gunning for the snapdragon X Elite and apple M4 which is great and honestly if they did their desktops CPUs like this I don't think id complain

[mossy gull]

The Snapdragon X Elite crushes even Apple in battery efficiency apparently, and considering Intel even gets crushed in battery performance by AMD, I am really curious how they deal with that.

Even a laptop with a 14700H loses from my 5900HX in battery life, which is kinda sad.

[jade scaffold]

Yea that's cos the 14700H is the same gen as 5900HX

[mossy gull]

For me, efficient cores only really matter on mobile devices as I usually disable the Turbo anyways to lower the heat output.

[jade scaffold]

It's literally just a slight intteration on the 12th gen

Which was a 1st gen product with some issues

One thing we know is guaranteed is the new SOC will idle vastly less power than pretty much any x86 chip on the market just because it doesn't have hyper threading

Theoretically the E cores are well over 30% of a P core from 14000 so theoretically a E core should outperform a thread rather substantially

[mossy gull]

1 thing I want from Intel is for a change to not decrease software compatibility over time, I seen it happen on 3 laptops, a 6700HQ, 7700HQ and 8650u, all use to work with XTU, the latter 2 had 2 power profiles, yet compatibility for that was removed after 3 years of use, and using older drivers doesn't work as they're firmware locked after 3 years.

[jade scaffold]

Intel have substantially redesigned the E core and P core. And their architecture generally

Putting the RAM on the same silicone interposer as the chiplet should make the latency and clock speed really really good too which probably helps with how high the GPU performance appears to be along with the total complete and utter redesign of the GPU architecture

[mossy gull]

I personally don't think Hyperthreading isn't the issue if it comes down to efficiency, I think it's how said CPU is used and what said person uses it for.

There's programs that do perform significantly better with hyper threading, especially older games that can deal with Hyperthreading but not more cores.

[jade scaffold]

Hyper threading literally increases power consumption by 20%

It just does that

[mossy gull]

I know, but a program can also load significantly faster.

I tested this once with Fallout 4, where without Hyperthreading the game was slow and rarely hitting 30fps, while with Hyperthreading it reached even 40fps.

[jade scaffold]

Not how that works

All hyper threading does is allow 2 ques on a core

If one que has to wait for data the other can keep going

The performance increase best case is 30%

If a program can use extra threads it can use extra cores

If for the same power you can get more performance out of 2 E cores that's an efficiency gain with literal 0 downsides

[mossy gull]

Yet, with Fallout 4 it supports 2-cores 4 threads, disabling Hyperthreading keeps the game running on 2-cores 2-threads, not more.

Yes I know the specsheet asks for 4 cores, but I never seen the game use 4 cores'

[jade scaffold]

That's a u issue

[mossy gull]

Not really, it's a Bethesda game issue, my brother has the same issue and he mods the game to oblivion

[jade scaffold]

It's also fall out 4 it's about half as well programmed as java MC is

Regardless of that

[mossy gull]

Oh I know, that's why it's a fun benchmark 😅

[jade scaffold]

If intel do what they said they are going to do

The E cores will just take the place of the threads

[mossy gull]

I still have to see how it'll perform

[jade scaffold]

Yea it'll be interesting

[mossy gull]

And how Intel is gonna deal with software compatibility over time

[jade scaffold]

For sure a huge improvement over the last few gens

AMDs Ryzen 9000s are really promising looking especially the 9700x

[mossy gull]

Intel's attitude is why I don't buy their stuff.

[jade scaffold]

Not really a fan of the new APUs they are better but they aren't really interesting

[mossy gull]

The APUs are only the 8000 series no?

[jade scaffold]

NVidia haven't even announced the new SOC so their keynote was booring

[jade scaffold]

You can't read can you

I said earlier they are releasing new APUs

They are Ryzen 9000 gen but they have intel esque shitty branding now

[mossy gull]

Ah, then I am curious how those will perform

[jade scaffold]

Node shrunk rdna3 gpu

With 2 more GPU cores

[mossy gull]

I mean, if they perform amazingly, that's what I look for.

[jade scaffold]

Probably like 20% faster

New memory controller

Otherwise not atall interesting

From the stated AI performance alone the GPU In them either is smaller than Intel's or Intel's is just waaaay better at AI

[mossy gull]

I am curious with that though, what if they take the 7700M, push it into their APUs, and actually make it work🤨

[mossy gull]

I mean, their ARC GPUs were suppose to compete, but they hardly ever did😅

[jade scaffold]

They kinda did tho

In the super low end intel dominated

Infact amd actually lost market share to intel and Nvidia lat gen which is impressive

Been listening to one of Intel's engineers talking about how they redesigned Thier GPU architecture it's WILDLY different

[mossy gull]

Hopefully they don't release it too late like they A780 I think it was.
While I want Intel to be competitive, they need to up their game with these things.

They competed way too late and software compatibility still isn't where it should be.
I once was thinking on getting an Intel GPU for fun.
But they go for 400.-, I can get a 7600XT for that, probably even the 7700XT.

[mossy gull]

In the super low end I see the Z1 Ultra😅
Where Intel lost even after that APU existed for a year.

[jade scaffold]

This new SOC they claim as 50% faster GPU wise

[mossy gull]

Now we need Qualcomm to make GPUs and see how they will perform 😅

[jade scaffold]

Which should put it at pretty ludicrous speeds

[mossy gull]

I hope so, time for AMD to make the Z2 Ultra or 9840U or whatever what they're gonna call it

[jade scaffold]

Tbf iris XE should be capable of matching 780m 50% faster than that actually tracks with the ai performance number being that amount higher than AMDs new HX370

That's what the HX370 is from amd

It's the new 8840u equivalent

It actually has 12 cores on the GPU as apposed to 8

16 GPU cores

4x zen 5 + 8x zen 5C

[mossy gull]

They use an Arc GPU on their latest mobile handheld from MSI no?

[jade scaffold]

It's iris XE but it had some odd power configuration

It was really weird

[mossy gull]

Ah okay, no wonder it performed worse

[mossy gull]

That's a nice step up, it now has 12😅

[jade scaffold]

Yee

It's nifty

[mossy gull]

Lets hope they keep AVX-512 😅

Unlike Intel who ditched it😐

Why the in gods tarnation does the 8840U need 20 PCIe 4.0 lanes for, it'll never use that

At tops 4 for the GPU, 4 for the SSD, 4 for an external GPU, that's just shy of 12, not 20

On the other hand, you'll never run out of PCI lanes😅

[mossy gull]

What if AMD response with an 880M and beats its performance by 50%😅
Then they equal in power

[jade scaffold]

That sounds is the 890m

The 870m os on a lower end chip

Which has 10 cores 6C 4 normal
And 12 GPU cores

Oh no sorry that is the 880m

No 870m as of yet

[mossy gull]

To my knowledge the 770m doesn't exist and it drops down to the 760m, they could do the same with the 800m series.

[jade scaffold]

Yea no apu GPU below 880m listed yet

[mossy gull]

Maybe we'll get an 870M in the SteamDeck 2🙃

To be honest, currently numbers don't mean anything to me, I have to see real world performances.

[jade scaffold]

AHH I see how they are maintaining so much core performance without the thread

The prediction engine is turbo built up

Thoes e cores should also be vile in speed on single thread

Each 4 E cores has a Shared L2 cache

4mb of L2 could be taken by 1 core entirely that's actually insane

The NPU is actually smaller than the one in the Ryzen AI chips ever so slightly 48 TOPs Vs 50 TOPs but overall AI performance is 50% higher because of the GPU at 120 TOPs Vs 80 TOPs

[mossy gull]

We still have to see how it'll perform, and not be another 12th gen case where it works like it isn't suppose to

They can say this, and the other thing happens.
Currently everything are numbers, no actual real world performances just yet

[jade scaffold]

If we assume iris XE can hit arround 1060 3gb performance the new GPU should actually hit 1070 performance pretty much exactly

Which for a tiny lil chip with 4P cores 4E cores is alot of GPU

I actually doubt this chips tdp will exceed 45w

It's also built on tsmc 3NB which is a super new node

So that's nice

Most of the architecture changes seem to revolve around obliterating bottlenecks

The chip has 8500mhz ddr5 at super low latency cos ya know it's on the chip

Minimum of 16gb up to 32gb of ram which is nice glad there's no 8gb version

[mossy gull]

The AMD 780m already does hit 7% slower than the 1060 3GB in DX12.
Which is impressive for something that fits between your hands.

And Intel's variant is slower than the 780m

[jade scaffold]

The entire chip has a Shared cache a super engineered prediction engine. It has a all new thread handler too brand new GPU architecture

[mossy gull]

I think even in an handheld, having more than 16GB RAM is a nice bonus, especially if you want to play newer games at lower resolutions.

From what I seen even in 720p, Dragon's Dogma 2 still uses 6GB VRAM, which also means you won't have enough RAM left over as it can pull 12GB on my laptop.

[jade scaffold]

I need to check what they are refering to gpu side when they say 50% faster

Also 780m is faster because it has 2x TDP

[mossy gull]

The entire package of the 8840U has a TDP of 28w, same as the 155H

[jade scaffold]

8840u doesn't hit thoes 1060 performance numbers

It doesn't even get close

[mossy gull]

The 155H does have a turbo of max 115w

[jade scaffold]

8700g or the much higher power 54w variants do

[mossy gull]

I said 7% shy of the 1060 3GB, but that's through Passmark

[jade scaffold]

Iris XE GPUs cap their power at 28w they don't seem to financially share it with the CPU

[jade scaffold]

And that will be at 54w

Because the 8840u can do 54w

[mossy gull]

Max power?
That like half of the 155H max power😅

Which does 115w

[jade scaffold]

Yea

Which the GPU is only allowed 28w of

AMDs 780m will pull any available power it can

If the CPUs not being hammered then the GPU can actually run faster

[mossy gull]

115w in a handheld, that's toasty😅
Even a 25w in a aluminium tablet is already toasty

[mossy gull]

I forgot how it's called, but it's fun to see in the AMD driver where most of the power goes to😅

[jade scaffold]

This is an interesting graph

Can really see where the new architecture makes improvements in rendering

[mossy gull]

Yayy, more stuff majority never heard off 😅
Not even me, and I have a weird API to make Maya into simulation software 😅

[jade scaffold]

Thats the render pipeline

[mossy gull]

Fun fact: the 8840U can encode on 1080p864.

AMD, why the fuck do we need that?

[jade scaffold]

1.6x faster at raytracing
1.2x at tesselation
4.1 mesh shader

[mossy gull]

Cyberpunk likes a word😅

I'll have to see how good RT will be, so far nobody beaten Nvidia yet, okay not under the companies we know of.

Nvidia, release your damn CPU for the consumers

[jade scaffold]

Yea they aren't gonna be matching Nvidia In raytracing

I don't think you understand how tiny this lil chip is

[mossy gull]

I do😅
But Apple chips are also tiny and they do pretty decent with raytracing

[jade scaffold]

Apples are bigger

Like a 3rd of its just the ram chips

They may do a bigger one this is only really to compete with the snapdragon X elite not apples max/Ulta/whateverthefuck

So smol

Also tdp is allegedly 15w or 17w

16% IPC boost on P cores which is inline with AMD depending on clock speed improvements it could be slightly faster or slightly slower in single thread

[mossy gull]

Apples Super Duper Ultra chips?
Yeah those can be pretty big, I was talking about the normal 1, not the Pro or whatever the fuck they have currently😅

[mossy gull]

Bigger than my 5900HX 😃

[jade scaffold]

Remember it's got ram on it

And a npu and a GPU

It's a rather dense little thing

Looks to be only slightly wider than a m.2

[mossy gull]

I meant the CPU part, not the entire chip😅

[jade scaffold]

Apples m series is slightly wider but about the same height

[jade scaffold]

The CPU part does have alot more in it

It's literally like 30% cache the chip part

[mossy gull]

Fair enough, still big though 😅

[jade scaffold]

Well yea but again it's got a huge npu in it a bunch of cache a GPU and a lil octocore CPU

Love how modular it is

[mossy gull]

Modularity should've been a thing all along in my opinion

[jade scaffold]

It's got loads of tiles actually there's a few filler tiles maybe extra room for more cpu if needed or GPU

[jade scaffold]

Yea

Intel wanted their chiplet tech to be perfect I think

Like I said it's way more advanced than AMDs chiplet tech

Well more than what they use on ryz nans epyc

It's pretty much the same as Nvidia and AMD and intel use on their data centre accelerators

But that's a fun little bit of tech to trickle down

[mossy gull]

Been a while I seen a new Xeon Phi, lately just been replaced ARM, Nvidia or AMD

[jade scaffold]

Yea intel have ditched phi

It's just not needed CPUs can now do what phi did

[mossy gull]

My job made the jump to those 128-core Epyc a while back. Intel was still stuck on 28-cores

[jade scaffold]

As can GPUs and accelerators for the most part

Intel haven't been stuck on 28 cores for ages

They've had 56 cores since at a minimum 12th gen

[mossy gull]

It was fun to play with them though, I learned how to run multiple VMs on just the Phi cards

[mossy gull]

Ah, the lesser cores overall I can understand they still jump to Epyc

[jade scaffold]

Oh also because of the new scheduler being split Into 2 different schedulers that's added to the overall die area

[jade scaffold]

Per core thoes 56 are faster than the 128 core epycs

[mossy gull]

Hopefully it'll work better, and understand to play games on the P-cores

[jade scaffold]

Could also do 8(?) sockets in a system on intel intel are still king for raw density and are king for efficiency and pcie

[mossy gull]

They split the cores into equal systems and not make them dynamic.
This is to prevent that 1 simulation will hinder the other.

We make the simulation stuff on a local system, then send it to a VM on the server.

[jade scaffold]

The 128 core still is overall faster

But the idea with intel is if you need slot of cores you put 8 CPUs in 1 system and end up with 436 cores

But if you only need the ability to virtulise and not the density or performance epyc works out better from that point for the efficiency

Also eww epyc Bergamo simulations

Whyyyyyyy

It's terrible at it

Genoa is so much better at Sims

[mossy gull]

We also have the CPUs and GPUs in 1 case, like 8 GPUs and 2 CPUs.

[mossy gull]

All I know is they're Epyc and roughly 1.5 years old by now.

Most of our budget goes to space research, so cost also could've played a role.
We only have an IT department the size of 4 people, with a capacity of max 200 people in 2 buildings.

[jade scaffold]

I'm looking really deep into this architecture

And I found something weird

[mossy gull]

Good or bad weird?

[jade scaffold]

Good

Just unusual

Most CPUs only have 1 integer multiplier unit per core right

The p cores on lunar lake have 3

Which means it can simultaneously on 1 core multiply 3 different numbers per clock

For certain workloads that's actually kinda game changing

[mossy gull]

Hyper thread only with 3 threads effect?

[jade scaffold]

Threads can't multiply

That's not how they work

[mossy gull]

I had to visualise in a way😅

[jade scaffold]

On a normal CPU with 2 threads and 1 core

You would still only be able to multiply 1 integer per clock

This architecture on the P cores can do 3

Vector execution has gone from 3 per clock to 4

[mossy gull]

I have no idea what that means but if it means I can't play my older games, then it'll be a no-go😅

[jade scaffold]

You can really see how much more efficient this new architecture is

It's waaay better at lower power than less gen and actually scales worse with more power than last gen

Still at full power faster than last gen but it doesn't get as much of a boost from power running half the world's power grid through it

[jade scaffold]

No you can

Still fully compatible

Basically cores are a lie

Cores are not 1 processor

They are a bunch of little processors wearing a tench coat

They just added more little processors in the core so it can do things faster

[mossy gull]

That's what they always say, I still remember Spore warning me I had 0MB VRAM left or that my CPU was "too slow" 😅

[jade scaffold]

That's spore being poorly made tbf

[mossy gull]

I know, but when I said old games, I meant old games😅

[jade scaffold]

All compatibility issues you will have are pretty much exclusively gpu driver end or os end

[mossy gull]

Skyrim thought I had an 8800GS, that fucker told me my GPU was "too old"

I have a 7900XTX by the way😅

And a 6800M

[jade scaffold]

I've seen someone's Ryzen 7 2700 report as like a pentium 4 or some shit

[mossy gull]

I can list all GPUs but the rest is useless, not looking at you Vega FE

[jade scaffold]

https://media.discordapp.net/attachments/1238450897466687499/1246892209903505408/image.png?ex=6660049c&is=665eb31c&hm=d2427618c2ecc8fa3f2d39c1733a8086d0e9413e5544eb984558320330ad60af&

This

That's a Ryzen 7 2700

[mossy gull]

My dad had his old Pentium D to be reported as an Intel Core2Quad by GTA 4😅

[sonic meadow]

at its heart its a massively funky and screwed up pentium

[mossy gull]

AMD Intel Ryzen Pentium 2700 0😅

I didn't know that was a thing

Best Intel AMD collab ever

[jade scaffold]

I swear this thing is a different size in every picture of it

from that picture the entire chip RAM and all is smaller than the DIE of my P100

[mossy gull]

Sounds suspicious

[jade scaffold]

its not actually it has to be just perspective

but yea its REALLY small

looks to be about 2.5cm tall

[jade scaffold]

It's officially 2.75cmx2.75cm

[frozen igloo]

How are 8th gen intel mobile cpus, would it be worth me getting the laptop to a functional state again?

[jade scaffold]

They're allright

If it's just a CPU+igpu it's not gonna game or anything but itlld do.youtube and so on fine and battery life should be serviceable

[mossy gull]

I would say avoid them.
Intel dropped software support for them almost 4 years ago.
Intel also removed a lot of functionality from them.

[mossy gull]

Is it me or are they going full in on CAMM2 @jade scaffold

[jade scaffold]

Whom

I like camm2

[mossy gull]

Same, I already seen Lenovo, Asus, AsRock, MSI and Gigabyte doing it.
So they're going full in, I know my next board will have CAMM2

Funny enough, Dell is nowhere in the picture, even though they usually like wacky stuff

[jade scaffold]

I think my next board I might go CAMM2 if I don't go server again

[mossy gull]

Let me have a consumer board that can do PCIe 4.0 16x times 2 and not go down to X8 and I'll be happy

At this moment almost anything is faster than a 3970X

[jade scaffold]

Or just dual 5x8 would be fine

Same bandwidth

[mossy gull]

It's for 2 7900XTX, unless AMD gets a replacement, but till then, I want both again in my system for Maya3D

[cinder lagoon]

Noctua will release coolers with different contact plates, optimized for different CPU lines.
It the version will just be called G2 etc as a suffix

[jade scaffold]

Noice

They are also working on a themosyphon

[mossy gull]

G-Skill CAMM2

[jade scaffold]

someone made 10200mhz CAMM2 modules

which is uh

insane

[mossy gull]

I read, that's double of standard DDR5 speeds😅

I have been reading blueprints, and CAMM2 is amazing, it also lowers the latency between the CPU and RAM

I would love to see some cool coolers for CAMM2 though

[cinder lagoon]

I think the camm2 form factor doesn't make much sense on desktop motherboards. The signal integrity seems to be wild tho

[jade scaffold]

Well it's better signal integrity and allows for better cooling

It's only disadvantage is capacity

Atleast capacity to price ratio past a certain point

[sacred seal]

Well it also lacks in density compared to desktop ram. Board space is still somewhat at a premium on desktop

Just looking at it, it seems like a single module has a similar footprint to 4 dimm slots

[jade scaffold]

Yep

That's what I mean

If you want the same capacity you can get on a desktop you need chips that have more memory which are spensive

The limit for total men on desktops is entirely cpu side ATM

[sacred seal]

I think the modules cap out at 128gs?

[jade scaffold]

Current ones probs do

[sacred seal]

I think you can do 512 on a single dimm

[jade scaffold]

But you could easy do a 256gb one

You could probs do 512 on a cam module

[sacred seal]

And you can fit 4 dimms into the same footprint

[jade scaffold]

Yes

But

Big but

[mossy gull]

To note here, Windows Home Edition also supports max 128GB, thats where the CPU limit comes from

[jade scaffold]

Ryzen only supports like 128gb

Or 256

[sacred seal]

Pretty sure its 256

[cinder lagoon]

"only"

[jade scaffold]

Yea you could do a 512gb cam module

You would need a register tho like on normal DIMMs

Which consumer CPUs don't support

[sacred seal]

Looked it up, Ryzen only supports 128gb per module

[jade scaffold]

So it doesn't really matter anyway

That's not right

I know for a fact that memory controller can't do 512gb

Maybe it can do 128gb per DIMM upto 256gb

But not 4x 128gb

Wait no

[mossy gull]

Home = 128 GB
Pro/Edu = 2048 GB
Pro Worstation/Enterprise = 6144 GB

[jade scaffold]

7700x caps at 128gb

[sacred seal]

Hmm, interesting. Looks like my source was wrong

[jade scaffold]

So there's no overall capacity disadvantage

Buuut

If you wanted 128gb 4x32 is probs cheaper than 1x128

[mossy gull]

128GB is more than the average joe will use, there are Pro CPUs that can use 512GB I believe

[mossy gull]

It is😃

[sacred seal]

This just keeps getting wierder

[jade scaffold]

I can't even find a 128gb ddr5 DIMM

Only RDIMMs

[sacred seal]

https://www.guru3d.com/story/msi-releases-new-bios-updates-for-amd-ryzen-7000-series-cpus-adds-192gb-ddr5-memory-capacity/

[mossy gull]

AM5 isn't Ryzen 7000, it's the socket that supports an AM5 CPU, which can also be the 9000 series

[sacred seal]

Yeah, but who is asking for 192 gb capacity

[jade scaffold]

On Ryzen

Which has a shit memory controller anyway

[sacred seal]

Like that is 3x64

Or like 2x64 and 2x32

[jade scaffold]

4x48gb

[sacred seal]

Do they make 48gb sticks?

[jade scaffold]

48gb is the biggest ddr5 DIMM I can find

Yep

[mossy gull]

I would, I can fill up 256GB

[sacred seal]

Ah, well that answers my question then.

[jade scaffold]

I'm gonna have so much ram I might use it as a boot drive

[mossy gull]

I even can fill 512GB, but I am more a Pro-sumer than a consumer

[jade scaffold]

How fast does windows load when ran on ram we shall see at some point

I'm sure the advantage of windows loading fast will be far surpassed by the disadvantage of the memory controller taking 10min to get itself in order

Fun fact Ryzen 8000s memory controller is waaaaaay better than normal Ryzens

[mossy gull]

Probably incredibly slow on boot

[jade scaffold]

Well windows itself would load real fast

Program and is loading is something ram and optane does way better than any SSD can

[mossy gull]

I mean, they sort of need it😅

[jade scaffold]

Gen 4 optane out here beating a drive with 3x it's read write speeds

[jade scaffold]

They do it's a shame it didn't make it into normal Ryzen

[mossy gull]

Nah, the normal ones are too old😅

[jade scaffold]

eBay pricing really tempting me rn

12 channels of DDR5 6000 anyone

[mossy gull]

I have 256GB with my 3970X, but they're the slow DDR4 ones😅

[jade scaffold]

That's nearly 600GB/S of memory bandwidth

That's more than Vegas HBM or the 1080tis gddr5x

[mossy gull]

Hey, don't talk bad about my old 32GB of HBM2 VRAM 🤣

[jade scaffold]

Still sad no new Blackwell or Jetson news :(

Atleast intel and AMD announced something

NVidia were just like "remember black well we announced 3 months ago? Good. lmao"

[cinder lagoon]

Lmao nVidea literally announced an announcement before show start just to be before AMD's show opener

[jade scaffold]

Nvidia announced an announcement about a previous announcement

[cinder lagoon]

Yea kinda

[jade scaffold]

Wanted to know specs of gpus

Even B200 would have been nice to gush about for a bit

[mossy gull]

I am sad there won't be an 8900XTX this year🤣
Or 9900XTX next year

[jade scaffold]

And weren't that interesting either tbh

More efficiency and IPC on the desktop chips

The AI chips aren't really exciting because they didn't talk about the actual exciting one

The best thing so far has been Intel's announcement

We got GPU and CPU architecture talk which is atleast interesting

Frore announced an M.2 cooler that's pretty cool

And cooled an iPad with air jet minis

[mossy gull]

For me AMD is still exciting, no matter how bad they are, yesterday I felt the difference between my broken laptop's 6800M and a 2070 desktop, they should be roughly on-par in performance but the 2070 felt way more jittery.

Although I only tested Dragon's Dogma and Star Citizen 😅

[jade scaffold]

Noctua dint announce anything new other than the thermosyphon

AMDs products are fine but the new cpus for desktop were just boring

Good yes

Interesting nah

Got no real technical details just pull less power is faster

Intel did this whole architectural deep dive into how they designed their new chip and how they've totally redone this new mobile architecture and it's CPU and GPU

AMD just did "new cpu! it's a tad faster" bit of an unfortunate theme from them recently atleast the apus are pretty cool

[mossy gull]

I am more curious about their X3D lineup of CPUs.

[jade scaffold]

Same

Hoping they can maybe keep the clocks up with the power utilisation on the new chips being so much lower

[mossy gull]

I want double the X3D memory, then they can show who's boss in gaming

For me completely useless considering Maya doesn't use any of it🤣

[jade scaffold]

Nah double x3D isn't the way to go

[sonic meadow]

just slap some hbm on the cpu

[jade scaffold]

You want to see if you can cram a 3rd cache die separate to the cores maintain clocks upgrade their old ass interconnect allow for shared cache between ccds

[mossy gull]

Fun fact: the 3970X has the same amount of L3 cache as the 7950X3D, yet it's slow as balls in games🙃

[jade scaffold]

How does a 3970x have that much cache

There's no way it's at 8 full CCDs

Wait no it's only need 4 CCDs for that

Mine has 8 and has double 7950X3D cache

[mossy gull]

The 7950X3D has 128MB L3, but so does the 3970X 🤣

[jade scaffold]

Yeee

[mossy gull]

Lol, multi-threaded performance is roughly the same in Passmark, but in games it's slow as balls, come on devs, make it so it can use all cores.

Single threaded the 3970X is 1/3th slower though.

[jade scaffold]

I feel like it should be slot more than 1/3 slower single thread

[mossy gull]

3970X scores 2666
7950X3D scores 4144

In single threaded Passmark

[jade scaffold]

7950x in single thread should be more like 40-50% faster adjusting for clocks and IPC improvements

Unless your running the 3D cores then maybe closer to 35% in certain tasks

[mossy gull]

Maybe it's the X3D that holds it back

[sonic meadow]

1/3 slower is 50% faster

[mossy gull]

The 7950X scores 4286

[sonic meadow]

4000 - 1/3 = 2666
2666 + 50% = 4000

[jade scaffold]

Huh yea makes sense now I think about it

[sonic meadow]

I was about to do some maths on it when it hit me 😛

[mossy gull]

1/3th is 33.3333333333333something%

[sonic meadow]

and?

[jade scaffold]

Yes but

[mossy gull]

Butt?

[jade scaffold]

If something is 2/3rds as fast as something else then the other thing is 3/3rds as fast as itself

How many % more is 3/3 than 2/3?

It's 1/3 more which is half of 2/3 so 50% more

Making the 7950x and 7950x3D in excess of 50% faster from your tests

[mossy gull]

Probably 😅

Math isn't my strong suit

[jade scaffold]

The 7950x3D is 50% faster than the 3970x single core

The 3970x is 66.6% recurring as fast or 33.3recuring slower than the 7950x3D

[mossy gull]

That's roughly in line what I calculated yes😅

Aka the 3970X is slow as balls

[jade scaffold]

Still pretty reasonable

Especially for slower GPUs

[mossy gull]

It is for multi core yes, Maya can use the whole CPU which is nice

[jade scaffold]

Anything sub 3080 class should be fine on it for the most part

[mossy gull]

Fine, I'll throw a 5090 in it🙃

[jade scaffold]

In 1440p

CPU bottleneck boogaloo

[mossy gull]

I have an LG C1 on the wall

What if, hear me out, what if I make games think my 2 3440x1440 240Hz panels are 1 panel🙃

3440x2880 240Hz🙃

That's like 3:2 aspect ratio 🤣
That would be a pain to program for again

[sonic meadow]

wasnt the MS surface 3:2?

[mossy gull]

It is, I use to game on a Surface Book 2, and I can say, Intel + 3:2 gaming is a true pain

1: Intel scales everything full-screen, which means unless something isn't 3:2, it'll stretch.
2: Intel removed the ability to add custom resolutions 1.5 years after I had the SB2.

When I contacted Intel about it, they said it was never possible to make custom resolutions.

That's where my massive distrust for Intel comes from, reason I don't trust them anymore.

[sonic meadow]

I would say we need a third player in the x86 space, but a: intel and amd are never going to let anyone else in, and b: x86... kinda sucks 😆

[mossy gull]

Correction, Qualcomm has to do better than Intel

[jade scaffold]

I mean x86 is great

It's absurdly fast at certain things compared to arm

[sonic meadow]

thats mostly because its an extremely mature platform, not because its good

and there is a looooot of legacy cruft in it, which causes issues. intel (or was it amd?) proposed ditching a bunch of the old stuff recently. by old I mean the entirety of the 16bit layer

[jade scaffold]

Well no not nessisarilly

There are certain things arm literally needs more clock cycles to do

Arm is extremely mature too

[sonic meadow]

well thats almost everything. the difference between a VLIW ISR and.. whatever arm does (forget the term) basically

[jade scaffold]

Yea

X86 is old yes and it's a bit bloated

[sonic meadow]

its been a while since I did computer organisation 🤔

the two are difficult to compare at a micro level, as they do things rather differently.
its mostly at the high level (ie how fast you are able to get algorithms or applications to run) that you are really able to compare

[jade scaffold]

But it's verity of instructions it has for cutting down on number of clocks needed to accomplish a task is a large part of why it's so much faster

But ofcourse there's so so much more to it

[sonic meadow]

mostly in the low power space, its only very recently started to be used at more than a few watts

[jade scaffold]

Arm is much lower power typically using much denser cores and no hyper threading which lets them save HELLA Power

But Arm alone isn't actually that much more efficient

The reason in recent years apple primarily have made it seem more efficient is because apple and also Qualcomm cram absurd numbers of Accelerators into their chips

Like you can build an x86 CPU in the same way as an arm cpu

Ditch the hyper threads compress the cores sacrificing some clock speed and raw performance for compactness and power efficiency

[mossy gull]

I believe X86 is much better at completing complex tasks than ARM

[sonic meadow]

they are able to be more efficient because they need less silicon (well less transistors) for the same function because they dont have the extensive instruction set. just like E cores are more efficient than P cores because they are missing a bunch of stuff. even "dark" silicon is still on and consuming power

[jade scaffold]

Yea if you look at the new lunar lake soc tho it's built kinda like an arm cpu

P cores E cores no hyper threading also equinox will be excited to hear the new thread scheduler will pass things that used to be passed onto hyper threads onto the E cores

[mossy gull]

As long Intel's E-cores work like ass I won't like them

[sonic meadow]

im a bit disappointed we never saw more than two threads per core 😦
I know some of the weird non-x86 server chips did more

[mossy gull]

I seen a CPU 1 core with 8 threads😃

But it was a 1 of a kind prototype

[jade scaffold]

Apparently in the desktop chips they are equivalent to their last gen P cores

The lunar lake SOCs ones are integrated slightly different for better efficiency so are a bit slower

[mossy gull]

Like I said, they can say everything with numbers, and in the end they're slower.
Like on paper the 13900K is faster than the 7800X3D in games, yet the 7800X3D beat the 13900K

[jade scaffold]

Depends on game

[mossy gull]

On average it did

[jade scaffold]

Cache sensitive games the 7800x3D slams just about everything ATM

13900k tends to do better than the 7700x in games where cache doesn't matter so much

Intel's CPUs generally don't venture fit my Ch from having lots of L3

They love L2 tho probably why the new lunar lake chip is like 40% L2 cache by volume

[mossy gull]

I still want to see how it'll do in the real world, numbers mean nothing to me at the moment.

[jade scaffold]

So far I'm pretty impressed architecture wise

Though I do wonder how and got 38% up in geekbench over the 16% average

AMDs last 2 gens IPC wise have been a lil disappointing

Ryzen 7000 saw a big bump in clock atleast

[mossy gull]

Cherry picking

Or extreme overclocking

With a chiller

And a custom designed Asus board

[jade scaffold]

This was in IPC testing

Which means clock speed is locked down

Wonder what changed in the architecture that saw geekbench take such a huge liking to it

Intel tripped the number of integer multipliers per P core which is gonna see certain applications just enormously benefit as well as number of vector units went up by 1 I think

[mossy gull]

Still, Intel doesn't have the most spotless track record if it comes down to performance stuff

[mossy gull]

@jade scaffold I used the new Nvidia software suite, and I can say, it looks nice, but still doesn't have everything in 1 package.

[jade scaffold]

Yea they're still adding stuff to it

Not fully done yet but it's alot nicer than the old one

[mossy gull]

That's true.
But I still miss the colour profiles that AMD has.

[jade scaffold]

You can get old 16gb optanes for like £4 so you glcan get 16 pcie gen 3 optanes ran in raid zero for 12000mbps on a raid card with optane latency

[mossy gull]

I probably won't notice the difference compared to my MP600 Pro😅

[jade scaffold]

Well that's 12gb/s with optane latency

Which basically means it'll load just about everything faster

Intel should bring back optane

Or atleast let solidigm use the optane tech

[mossy gull]

I still won't notice the difference, or it's too small for me for the hassle

[jade scaffold]

Potato reactions

Also that new 890m igpu in the new APUs is allegedly upto 40% faster than the one in the 8945hx

[mossy gull]

Awesome 😃

[sonic meadow]

for the same soc/igpu power?

[mossy gull]

They should go into handhelds so yes

[sonic meadow]

even the handhelds have wildly different design tdp (not the chip necessarily, but the final product design)

[mossy gull]

Even if it's a downgraded variant, it'll still perform nicely, like the 6800M in reality only performs slightly worse than the 6700XT.
And in theory it is the same chip

[sonic meadow]

im not saying its bad or anything, just trying to get a bit more information 😛

[jade scaffold]

One would assume

[sonic meadow]

it would be even more impressive if they did it while cutting soc draw by 5w or soemething 😆

[jade scaffold]

This is the CPU in geekbench

However take that with a grain of salt Ryzen 9000 loves geekbench

Based on that set of graphs GPU wise it should be about equivalent to Intel's claims for their new SOCs GPU

That's geekbench openCL

No tdp given but it's tdp apparently maxes the same as last gen 54w

CPU side 40% faster doesn't seem like a stretch it literally has 50% more cores

[mossy gull]

16 core APU?
Hell yeah

[jade scaffold]

12 cores

[mossy gull]

Still hell yeah

[jade scaffold]

Was comparing there to 8840u

[mossy gull]

The 8840U only has 8 cores

[jade scaffold]

4 E cores 8 p cores I believe

Not really sure I see the point

It's an apu 8 zen 5 c cores would be more than overkill

[mossy gull]

E-cores for streaming software, so if you game on for example an handheld like an AyaNeo, you can stream much easier.

[jade scaffold]

What

Why would you ever use CPU encoding

For streaming

Ah yes id like wise quality worse performance worse battery life with my streaming

[sonic meadow]

they didnt say encoding, they said software

[jade scaffold]

Streaming software itself isn't very demanding

[sonic meadow]

depending on what you are doing with it, correct. which makes it the perfect thing to run on the higher efficiency but lower performance cores

[jade scaffold]

Which are already overkill for a 1060 class GPU

What I'm saying is why 8 p cores and 4 E cores why not just 8 e cores they are absurdly fast as is

Like 3700x fast at a minimum
Which with a 1060 would be considered in all other circumstances stances overkill I mean if you're streaming or running weird bits of background stuff it's always nice to have more cores 8 cores is a sweat spot imo and they'd sip power so more battery or more GPU performance

[mossy gull]

OBS, no matter if it's CPU or GPU encoding, it's demanding on the CPU

[jade scaffold]

Not enough for it and a game to reasonably overwhelm a zen5c core

Not 8 of them

Also OBS hardly even scratches my CPU

And I'm locked to like 4.4ghz

[sonic meadow]

it all depends on what you do with it

[jade scaffold]

If you are wanting to stream at such a level you are hurt by the limits of a 8 core Zen 5C CPU then maxing the GPU in a game you shouldn't be using a handheld as your primary streaming device

[mossy gull]

Low quality streams are enough though

[cerulean helm]

Well, finally setup the appointment to have fiber optic installed at the house, that will be so nice to have. Just annoyed I am forced to use the ISPs Gateway/Router unit. Can not use the router part of it at least. Though its via IP passthrough not bridge mode.

[mossy gull]

The Gateway is called a modem, it's obvious you have to change it

[surreal moss]

Obvious you have to change it?????

[jade scaffold]

we got a small amount of game play on the new Lunar Lake SOC, and in F1 2024 its uh great 1080p High settings over 60fps not bad for a chip ran at 17w lets hope this isnt turbo cherry picked and that it performs similarly in other games

[mossy gull]

https://youtu.be/vfUcKJwbAFk

[cerulean helm]

Huh?
The ISP that has fiber, requires you to use their gateway. A modem/router combo unit. I cannot use my own. I can however put the gateway into IP passthrough to use my own network hardware.

[mossy gull]

Wait, you guys call that a gateway?
Here we just call it a Combo Unit, a gateway is where the fibre or ADSL comes out of the wall.

[cerulean helm]

ATT calls it a gateway as its an all in one unit for fiber. This is what they use.

This has the ONT built into it, along with being a modem and router.

[mossy gull]

That we call a combo unit😅

I personally use a a Ubiquiti Dream Machine SE, so I neither have a router or modem😅

[cerulean helm]

This is for fiber though, not to be confused with cable internet modem/router combo units.

The Dream Machine lineup is listed as a gateway on the website. And still requires a modem hooked up to it.

[mossy gull]

They don't need a modem, just a fibre OTP module and they can connect to the Internet

[cerulean helm]

You still need an ONT box, which for ATT is only available via their Gateway.

[mossy gull]

It's what we call them, also I live in the Netherlands, the word "gateway" doesn't have a Dutch word, so we call them combo units because that's the closest I guess

[mossy gull]

Here we don't 😅
Just an OTP module, which Ubiquiti sells as well, although mine is from Mikrotik I think it's called.
And we need the ISP's network address.

Dream Machines are also listed as alternative to work on the network

[cerulean helm]

Yeah, ATT owns all the fiber network here, and since they have to run the line to the house I have to use their ONT box. And even with an external ONT only box I would still have to use their gateway unit for service.

As those units (Aris BGW320) have the certs embedded in the firmware for authenticating the network.

[mossy gull]

Here we have net neutrality, so even if KPN or Dutch Fibre owns the lines, they still are forced to off the services to Odido or Vodafone

[cerulean helm]

You can still use different ISP, for instance. Earthlink in my area (how the hell they are still in business is beyond me) offers fiber, but its via the ATT fiber network, so you again have to use ATT equipment basically.

ATT locked down their fiber because their units also handle NAT, no way to bypass that. Even the folks at DLS reports haven't firmware hacked the new units.

I know the older Aris BGW210 was cracked and people put different certs on them to unlock it, but the new 320 version they said it wasn't even worth doing.

[mossy gull]

Yeah, here you can use Odido equipment on KPN lines, as long you can set-up your equipment to work with your ISP it'll work.
They are forced to allow this.

For example the unit we get for 10 Gbe is a Zyxel T75 I think, it only has 1 10 Gbe port, which doesn't even support WAN, so in theory you only can connect 1 device onto the 10 Gbe.

My UDM SE has a switch connected to it that has 8 10 Gbe ports, even for the access points.
So even on WiFi I have faster than gigabit, the Zyxel doesn't even support this

[cerulean helm]

Aha,
Further digging shows that there is an programable ONT SFP bypass module specifically for ATT in the works. MEaning once this thing comes out, I can copy the serial number from the ATT gateway to the bypass module and plug it directly into my UDM pro eliminating ATT gateway completely.

[mossy gull]

US style 😅

[cerulean helm]

Well of course. What government won't do, hackers will, lol.

[mossy gull]

You guys need net neutrality like us

[cerulean helm]

I'm sure its somewhere on the extremely long list of things that should be done, currently however lots bigger fires atm. So, until then, hacking it is.

[mossy gull]

It's actually very simple thing, just ISP's rule in the US

[surreal moss]

It’s not that simple

[cerulean helm]

Its on the list of things to do, and some net neutrality laws have been brought back by the FCC. But anything else isn't on a timeline right now as other issues are consuming everything.

Ow... $209? wtf

[mossy gull]

Wow, mine was 1/4th of that😐

[cerulean helm]

This one is programmable, maybe thats the difference?

[mossy gull]

So is mine

[cerulean helm]

Wonder if its the same chipset that the Fiber Store module above is. AT least with my reading so far there are a limited option of modules that will support ATT fiber.

[mossy gull]

ATT are just Aholes.
Simple as that

[cerulean helm]

Well yeah, that goes without saying. But Comcast is what I have now and pay nearly 2x for 1200/45 vs the ATT fiber 1gps symmetrical.

[mossy gull]

All fibre connections here are symmetrical.
My ISP told me it's a much lighter load on their servers, because it's both the same speeds.

[cerulean helm]

Comcast is not fiber, its what they call a "hybrid". If I recall its fiber to all their nodes and then cable to the homes. Currently I pay $125 for my internet, and ATT fiber at 1 gig is $85.

[mossy gull]

We have 2 names for that, which I both forgot 🤣
But I know how they work.

With 1 the fibre reaches your home, the other the fibre ends at a switch box in your neighborhood, and the rest to your home is copper.

The latter is cheaper, but speeds can suffer greatly.

[cerulean helm]

Well, thats enough reading today on fiber wavelengths and whatnot. I won't know what module to order until I actually get my ATT gateway next month and can pull the info from it. I'm just assuming its an XGS-PON ONT and not a GPON.

[mossy gull]

1 of the fibre installation guys told me that the 1st few letters in front of PON just have to do with the manufacturer, and not much else.
He showed me that different PON modules which all started with different letters, all worked on the PON from the ISP.

[cerulean helm]

Depending on what Gateway I get I require either of those modules. And then depending on the wavelength I can get specific with what brands modules. And thats if a module is available. There are wavelength options for the Gateway and the highest wavelength there is no programable module on the market and won't be from what I was reading.

The one I screenshotted earlier just launched a couple days ago.

[mossy gull]

ISPs in the US are unnecessarily complex

[cerulean helm]

Remember these are power users bypassing ISP equipment. Otherwise I can just use the ATT Gateway in IP passthrough and my setup would work just fine. I just don't want ATT easy access to my stuff.

This is the info I need off the gateway once I get it:
D23446-STCA GPON 1310nm
3FE46899AB GPON 1310nm

3FE46900AAAA ALT OPTICS 1550nm
3FE46900AB ALT OPTICS 1550nm

3FE46901AC XGS-PON 1270nm
3FE46901-STCC XGS-PON 1270nm

As long as its not the ALT OPTICS, then I can purchase a bypass module. GPON is the older fiber stuff, and the XGS is all the newer deployment. I assume thats what I will have.

[mossy gull]

It's still unnecessarily complex😅

[cerulean helm]

I mean, I'm going to be bypassing what the ISP doesn't want me to bypass but can't legally stop me from doing so. So yeah, I expect some level of headache.

Hopefully by next month the webgui for the Fiber XGS module will be up and won't require me SSH into the the XGS SFP module to configure it.

Currently in beta testing.

[mossy gull]

So I was playing around with an old Alienware Area-51m R1 with a 8700 and 2070, and I disabled turbo and the CPU is 40°C lower😅

Before anyone asks, yes it's a mobile desktop

[iron rock]

can anyone recommend a pc gaming controller? I want wired not wireless and it needs to be cheap. I would like a few extra remapable buttons if possible. main one I am considering is the logitech F310 because its cheap and everything I have read says its built like a tank but it dosnt have vibration, I dont know how much that matters. I will mostly be using it for driving games and in VR for games that dont like touch controllers.

I dont have a strong prefrence for xbox vs ps layout. I have used ps controllers more then xbox but no used controllers much anyway to have built up any strong prefrence or habits

[surreal moss]

You can get xbox controllers for less than $30

Granted I haven’t tried to use mine wired

[iron rock]

the ones from xbox/microshaft are not less then 30... they are more like 50 and that is way more then I want to spend on something I wont use all that much anyway. the F310 is under 20

I just want to make sure I dont get something with sticks that will start driving a week after I start using it

[surreal moss]

It’s been a few years since I got mine so I guess they went up in price.

Generally though the oem controllers have to be reliable and from what I’ve seen they are. I have used Logitech controllers in the past and never developed any issues with them

[mossy gull]

Look for 8BitDo

Most of their controllers are cheap, but still rock HALL effect joysticks

And are fully reprogrammable

[iron rock]

are they cheap because they only come in pastels?

[mossy gull]

They come in both wired and wireless, and the wireless variants are pretty good still, just like older Xbox controllers they come with a dongle so you have next to no latency.

[iron rock]

its not latency I am worried about. I just hate wireless, dont like dealing with recharging shit because I know I will forget.

[mossy gull]

They're a budget manufacturer that actually has quality

[mossy gull]

You know you still can use wireless controllers wired.
1 massive advantage, if the cable breaks you just replace the cable, you don't have this option with wired controllers

[iron rock]

of course you do

[mossy gull]

I have an 8BitDo controller that came with a charging dock, so once I am done I just put it on the dock😅

[iron rock]

redoing a USB cable isnt hard

[sonic meadow]

it is substantially more work than a new microusb/usbc cable

[iron rock]

sigh all the ones that say they have hal effect sticks are in that 40+ price point 😦

[mossy gull]

My bad, I thought they all did, at least they're all pretty good controllers

[iron rock]

guess if my budget is under 30 I am going to just have to deal

probably dosnt matter what I get at that price then

[surreal moss]

Could you wait a month to save up a bit more?

[iron rock]

no

[mossy gull]

If you want programmable, get an 8BitDo, they're fully reprogrammable.
And by games they're just seen as Xbox controllers and you don't need funky software

https://www.amazon.com/dp/B0B9B2ZFGZ

$30.-

[iron rock]

is there a difference between that and the C version? I cant see one other then the C only seems to come in silly colors but for half the price

I am assuming the C means C connector

that would be a little annoying. would have to use an adaptor, dont know if I have one

[mossy gull]

Probably the other is wireless or has a detachable cable

[mossy gull]

They come with a cable, and it's only USB-C on the controller side

You get a USB-C to USB-A cable

[iron rock]

its not wireless so its not detachable

think this one is just ment for laptops and steam decks and similar things that have C ports instead of A

[mossy gull]

8BitDo have wired controllers with a detachable USB-C to USB-A cable

[iron rock]

are triggers always analog on controllers or is that a feature I have to look for?

as I want it for driving games that is one of the most critical things. analog triggers and sticks so I can stop playing the rappid tap game using a keyboard

the logitech seems like the best deal and reputable brand but it dosnt specifiy that the triggers are analog

[mossy gull]

Most controllers these days have analogue triggers, if you aren't sure, you always can look to the bottom and check for lots of spacing between the triggers and the housing.

[mossy gull]

Cheap Logitech controllers are actually crap, their sticks don't have the needed sensitivity stages for racing games.

[iron rock]

saw something mentioned about the sticks not being accurate enough.

I mean anything is better then steering with keyboard

[mossy gull]

8BitDo, unlike Logitech, constantly use new technology and keep refreshing their line-up, Logitech is kinda known to keep using old tech for their not popular products.

How much I like Logitech, unless it's a popular product, their hardware is actually junk

[iron rock]

I hate shoping for stuff like this

[mossy gull]

You and me both, that's why I stick with companies I know make good products

[iron rock]

like Logitech? 😛

[mossy gull]

Only the more popular stuff from Logitech is decent, the unpopular products are junk, like real junk.

At this moment for cheap stuff I rather go to 8BitDo, and for more expensive stuff I go to Corsair, my last Logitech product was my G600 I think it's called, its battery died in 2 years, so I moved to Corsair, and that mouse still works to this day, bought it like 6 years ago.

Before that I had the X52, the X56, a keypad and more, all died in less than 3 years, which is 1 of the reasons I step away from Logitech.

[iron rock]

another reason I wont deal with wireless

thing that worries me about the 8bit ones is I keep seeing stuff about connection issues with them.

[mossy gull]

I have had like 4 limited edition controllers from them, all with the wireless dongle, and connection always been fine.

People who probably have wireless problems use Bluetooth, and Bluetooth experience on Windows can be very finicky depending on the wireless adapter they have