#off-topic-tech
1 messages · Page 140 of 1
i'm in it rn
I'm just trying to see what the other offerings are
7840HS, 32gb 5600 ddr5, 500gb SN770 nvme, RX 7700S, $2000
with each component being upgradable later, i don't find it too offensive
That seems pretty pricey tbh
you're paying extra for boutique vendor plus forward upgradability
if you want to chuck out a laptop every other year, then yea, other laptops are fine
it's also gonna be more expensive in the long term
I mean the laptop I have will last me for many years and it was $1k
A $2000 laptop should be one you won't have to upgrade for a really long time
at $2000 with framework you'll be able to swap out the cpu in 5 years. lemme know who else can do that
I mean if you're spending $2000 you shouldn't need a cpu upgrade for at least 5 years
i don't think you're understanding that framework has modular construction where you don't have to chuck out the entire laptop
No I do understand, I just think it's somewhat unnecessary
Especially at that price range you might as well build a PC if you want upgradability
A laptop you just use for work doesn't really need to be upgraded unless you're doing CAD or video rendering on the go
Even then if you invest in a good laptop now it'll last a long time
you don't need a hub if you can swap out the ports. I know it's not a laptop for everyone. But for some it's really cool.
and saying "just buy a desktop at that price" is kinda missing the point of a laptop
Heavy workloads are really better suited for a PC
but chucking around a PC for people who travel a lot isn't ideal 😛
yup
Yeah that's why I bought a gaming laptop to bring when I move around
That's great until you need heavy workloads not from a desk
Otherwise I'll just use a PC at home
Yeah my laptop can handle heavy workloads very well
yea, personally i'll never buy an expensive/high end laptop, i'll have a cheapo one as a supplement to the dekstop muscle. but that's just me, and other people's lives have different requirements and restrictions compared to me
My main issue is that you can't swap the CPU without swapping the entire motherboard just because of the way mobile CPUs are
So the main thing you'd want to upgrade requires you to take the whole guts out of the laptop and the only thing you keep would be the frame, screen, and SSD, maybe RAM if you're not going from DDR4 to DDR5
just did a 5700X3D/7800XT desktop build for $1500
7800XT is def overpriced rn at $650
GPUs as a whole are messed up rn
yea but i didn't want to go intel or drop below 16gb vram
so $1500 still gets you in what i like to call the "mini monster" tier
$1500 is enough to get a PC that'll last a while tbh
Even $1000 can get you enough to last for a good amount of time
a couple months ago i priced out a 9800X3D/7900XTX for $2000 🤣
it was definitely in the "you could, but you probably shouldn't" with some of the components
but it's technically possible
I cringe every time I see a PC part picker link with a shady power supply to save like $20
my new PSU has the teeny tiny connectors on the PSU side of things, will be interesting to see how they go
also i wouldn't have gone with this case if i had a choice
it didn't review well with GN, and all of their criticisms were valid
i had to unbolt the psu from the case to add in 1 more sata power plug
That sounds extremely annoying
oh it was extra dumb because that was attempt #3
attempt 1: space between drive cage & PSU was too small to fit fingers plus plugs
attempt 2: unscrewing the drive cage left it hanging by a tab that you couldn't just slide the cage out of the way
attempt #3: screw drive cage back in, unscrew psu
The drive cage blocks the psu?
Plugs on the side seems like an odd choice
Oh it's one of the cases with the PSU up high
yea, non-shift would probably be better, but i didn't really have any say in it
yea, up high & tilted
I could imagine that would cause some problems with the two combined
Well obviously it did lol
lil bit 🤣
also really annoying when trying to do it all at like 2am
the good news is it's worked basically flawlessly since day 1
Oh yes, 2am is also when I think "let's build a system" 🤣
i couldn't sleep and 2 drives needed to be installed at some point
plus i kinda wanted to see for myself if GN's criticisms were accurate
i mildly disagree with the flimsy top filter/panel thing, but also i have it as exhaust so i don't have it on
it's definitely flimsy tho
so you don't disagree?
i disagree that it's a problem. no panel, no problem
i do find the $75 to swap out the finished paneling to be an insane price
they'd probably actually sell them if they were like $40
i do appreciate that the bottom filter pulls out from the front
i got it for free, so i don't feel comfortable complaining too much 🤣
it's actually easier to work in than my P400A i had before
i'm kinda tempted to burn $35 to switch it to the air front panel
but that'd bump it up to 18 fans, which really just feels ridiculous
Spot the dead pixel
I seeeeee itt
nice phone choice, if the watermark didnt get in the way.
Where Is it
Bottom right corner
Or do you have 3 dead pixels or 2 pieces of dust and 1 dead pixel
oh my windows compression sucks
but its around there
Uh
Y'all are smoking something
I'm fairly confident that it's directly above the Asus logo on hte left monitor?
Looks like 5cm up or so?
It's the phone
It's dust that you see
The phone. Is a pixel. And it's dead
Please tell me someone at least thought that
No, my door is <- that way...
Intel's 28a node is getting 20-30 percent yields, let's goooooo
That doesn't tell anything.
The actual measurement is defect density.
As that yield is product of that defect density and die size.
And did you mean 18A?
In September Intel said the defect density on 18A was 0.4 defects per cm^2 which compares with the 0.33 def/cm^2 defect density of TSMC's N7 and N5 processes three quarters before mass production.
how is that compared to previous gen yields?
So that last publicly announced 18A defect density was little higher than TSMCs N7 and N5 published defect densities at about same timeframe before mass production starting.
That was the time when headlines yelled about 10% yields on 18A.
But point is, yield changes massively with die size, on same defect density.
So it is completely useless metric without that accompanying die size.
Doing lot of small dies vs. doing max reticle size dies will cause huge differences in yield, on same defect density.
Interesting
From that last stock price manipulating scheme starting that "10% yields" news cycle:
https://www.techpowerup.com/329613/intel-18a-yields-are-actually-okay-and-the-math-checks-out
It was actually traced to be that I mean.
So I'm being manipulated
When using the older data of the Intel 18A node, that d0=0.4, we must check a few different designs and compare their yields before drawing any conclusions. We are today measuring yields using the SemiAnalysis Die Yield Calculator tool, integrating all known yield models, Murphy, Exponential, Seeds, and Poisson. At the EUV reticle size limit of 858 mm² and with an applied 0.4 defect rate of 18A node, the most pessimistic estimate is the Poisson model, which gives a yield of 3.23%. However, the most optimistic (Seeds) model yields 22.56%. The default Murphy's model is yielding only 7.95% of usable chips. That is five good dies on a 300 mm wafer with 59 leftover dies.
That is for that september public defect density, and max size chips/dies that could in theory be produced.
(reticle limit)
A while back, we covered a leak of "Panther Lake," Intel's next-generation Core Ultra 300 series CPUs. The leaked package thankfully included information about die sizes, which we can input into the calculator to find out the yield, assuming Intel is manufacturing Panther Lake compute tiles on the 18A node, using the aforementioned d0 of 0.4. The die number four with a CPU and NPU on it, measuring 8.004x14.288 mm, a 114.304 mm² silicon die, gives a yield of 64.4% on the default Murphy's model. Moore's model is the most pessimistic, with only about 50% of usable dies. The die number five of Panther Lake, housing the Xe3 GPU, is even smaller and measures only 53.6 mm², yielding an impressive 81% of usable dies. Even with the most pessimistic assumptions, the yield curve drops to 60%.
Smaller die size on same defect density from september.
Very interesting
And there is reason why almost no actual dies are anywhere near those reticle limits.
Outside of largest Nvidia enterprise product dies.
Some IBM chips too before.
But very high priced enterprise products, as the yields will be low even on mature process because of the huge size per die.
were you around for the intel 10nm promises
++
they said that 10nm was in mass production as promised, satisfying a pledge to shareholders
it was actually so fucking shit they couldn't even get the onboard GPU working, it was dramatically less efficient than 14nm, and only one or two laptops in China ever used it
but it technically got sold in a product!
so job done
do not believe a single thing intel tells you about yields or when something will be available for purchase
read this entire thing. it's worth it
Well, this seems to have originated with Reuters, not Intel.
These latest "Intel 18A yields are crap" headlines.
Previous set was started as part of real stock manipulation scheme to short Intel stock.
i think intel's own announcements are enough to tank their own stock lmao
I ran a 5820k up until recently.
I am very, very new to semiconductor fabrication processes and terms.
I just know that I'll never buy Intel again after the 13/14th gen "news" broke about the defects. I had just purchased a 13600k 6 months or so before Wendel started reporting on it.
I saw the writing on the wall and dumped my Intel stock before the mainstream outlets got ahold of it.
Basically, it's more or less proven that they KNEW about the issue and continued to sell processors.
I do not.
bigger bullshitters than nvidia
True. Also, Pat Gelsinger was a fall guy. Whether or not Intel could have been a successful fab, idk. But he was screwed the second he accepted the job.
Also, if I were an Intel employee during their stock crash and saw his tweets about praying, I would lose my mind.
I'm glad to see AMD make a resurgence. I hope they continue to advance hehe
he was naively optimistic, but they needed an optimist
they needed a true believer to make the plan work
without him, company's gonna get sold for parts with every failure
NVIDIA GeForce RTX Black Screen Fix already available for some, others will get it this week It may be a BIOS update or a simple driver fix. This is one of the most frequently asked questions we receive: when and how will NVIDIA fix the black screen issues that some gamers are experiencing? It turns […]
don't wanna hear about nvidia drivers being the good ones ever again
So I've read, there is sucha thing as ROP and ROP partitions?
And that one ROP partition has 8 rops?
Nvidia said only 1 ROP was missing and they probably meant partition.
One ROP, that can do 8 operations per clock.
Ah, article I am reading states total different.
Which Nvidia Marketing has then turned into 8 ROPs to make the number bigger.
In every GPC, you'll find two banks or partitions of eight ROPs, for 16 in total.
So, I am not wrong.
But yeah, can be named multiple ways.
Except for those unlucky few who have bought a shiny new Blackwell GPU that has one ROP partition disabled
Well, it's important detail if Nvidia say 1 ROP while rest say 8 ROP.
While Nividia means 1 ROP partition. That's why I ask. It doesn't seem trivial.
Maybe they even said it on purpose 
Weird question, how can a CPU with way cache (L1/L2/L3) and more cores (but less threads) perform worse than one with less cache and cores (but more threads)?
I was comparing framework systems yesterday, and the Ryzen™ AI 9 HX 370 looks really powerfull.
Those Intel comparison chips have just few performance cores and then two different types of efficiency cores.
And cache amounts cannot directly be comprated between such different chips.
So comparing raw specs of Intel vs. AMD doesn't work.
Better architecture?
I know the passmark is kinda bullshit, but 25%. Samples size seems to be the same.
and the Intel core ha more cache? Is the cache actually bottlenecking? Or unused in the passmark tests? Which is pi etc.
nvidia say only 1 ROP was disabled, but in their marketing, they say a GPU has 96 ROPs. they wanna have it both ways
Intel L3 is split between multiple islands that isn't shared.
These are some of the tests:
Yup.
Not sure if these tests hit the cache or not.
That kind of "combined cache numbers for all cores in CPU" don't mean much.
Oh I get it. Bad news: 1 ROP (partition) but for marketing, they don't count partitions, but ROPs 
How do you know L3 cache is split? Techradar?
because it is
That's not the question.
yeah i know, just agreeing that it's the one to buy
Eh, have you seen the price? 🙂
alas yes
That was for intel core.
€2,429
DIY version can be cheaper.
And it's already gone.
as for the intel, that's meteor lake right? deeply flawed processor
if your only example of chiplets was meteor lake, you'd conclude it was a terrible idea
higher memory latency than prior intel chips without any effort to compensate
Yes Meteor Lake
and still less efficient than strix point (AI 9 HX 370)
so it won't be running at those advertised clock speeds for very long
not when power constrained
Cache L1: 112 KB (per core)
Cache L2: 2 MB (per core)
Cache L3: 24 MB (shared)
E-Core L1: 96 KB (per core)
E-Core L2: 2 MB (per module
I found it Baldur. Techpowerup is nice.
AI 9:
Cache L1: 80 KB (per core)
Cache L2: 1 MB (per core)
Cache L3: 16 MB
E-Core L3: 8 MB
Funny, L3 is just 16MB.
calling the ryzen ones ecores is a bit misleading
Not "per core" or "shared" or "per module"
yeah, the intel L3 is partitioned per core but all partitions are connected via the ring bus, so it acts as one shared lump
I think it's just their template.
Two L3s.
One for the Zen5 cores at 16MB.
And second for Zen5c cores at 8MB.
Two CCXs.
But the AI 9 doesn't have shared L3. And still performs pretty well it seems.
Those Per module on Intel mean per 4 E-cores.
it's got two core complexes, two separate ring buses
Shared is for all.
Per core is for each core alone.
Per module is per set of cores.
No Baldur. Wait, you misunderstand my question.
The information is missing. What does that mean?
.
But that should have been Zen5 and Zen5c?
Intel 7:
see the left; four big cores get 16MB, eight small cores get 8MB
So per core. Got it.
No, two CCXs:
One with 4 Zen 5 cores and 16MB of L3.
One with 8 Zen 5c cores and 8MB of L3.
Where that L3 is shared between the cores in CCX.
two separate L3 caches
Like current CCDs have one CCX with 8 cores and 32MB of L3 on the desktop CPUs.
Monolithic die with two CCXs.
Zen 5 Strix (if trustworthy)
that's strix point
TechPowerUp should inlcude such images. I find them interesting and helpful.
i find it easier to remember codenames rather than Ryzen AI Max Double Plus 369 ++
Strix reminds me of Asus though.
But is it me, or does the Strix also have some sort of shared memory?
Only not between Zen 5 Cluster and Zen 5C cluster.
yeah I don't like how they picked that name
there's no shared cache between the 5 and 5c CCX
Yes.
No?
There is that Zen 5 CCX, then Zen 5c CCX, then IGP with cache.
And then memory controller for the RAM.
Then what is the definition of shared?
To me, all the Zen 5 cores use a "shared" cache. And not dedicated cache per Zen 5 core for example.
There isn't.
Just that it is common to some number of cores.
Shared inside the CCX.
Got it. This naming is confusing.
Where any core can access any part of that shared cache, inside the same CCX.
if a core in the zen5 cluster is running a program, then the other zen5 cores also get to share that cache; you only need one copy of the data. if the zen5c cores need to help out, they need their own copy within their own L3 cache
And need to do cache coherency snoops etc.
yeah there's extra communication involved to keep everything synced up
ye
But we don't call it shared cache.
we call it L3
I am typically speaking of information I find like this:
Cache L1: 112 KB (per core)
Cache L2: 2 MB (per core)
Cache L3: 24 MB** (shared)** ❗
E-Core L1: 96 KB (per core)
E-Core L2: 2 MB (per module)
yeah that's some bs
And that doesn't tell if that 24MB is shared only among the P-cores, or with the E-core modules too.
Probably only among the P-core cluster and E-cores don't have any L3 access.
Would need to check.
you wanna know what's the worst bs though? when AMD adds the L2 to those numbers and calls it Game Cache. so a 32MB L3 turns into 40MB Game Cache
marketing nonsense
This kind of cache implementation stuff isn't possible to distill into simple single words between multiple architectures.
the topology of cores is really complicated
Everyone is just using same words most of the time that got implemented during early days of multi-core.
Fuck this.
I am just back to watching benchmarks and look at a % to see what CPU is best
Higher % better right. Rest doesn't matter.
When you only had per core, or shared among all cores, caches.
this is best
understanding topology can tell you why something is faster, but it's futile trying to look at specs and predict speed
I am a bit black and white here, but it's so confusing these specs. Especially if there are no definitions and rules and brands can just make shit up on the spot.
you get it now
"We have 10 apples, but the consumer wants 20 apples, I have 10 bananas here, from now, these bananas are apples, voila now I have 20 apples" kinda of logic.
AMD's approach is to have multiple, separate caches so each core gets a fast cache; Intel want one big L3 for everything
both have trade offs
the AMD approach means their 16 core desktop chips, with two 8 core dies, have an extra performance penalty from it, but they're able to make them for much cheaper; Intel's approach is running into problems where the ring bus within the L3 is getting stretched really thin being connected to everything
What problems does L3 cache give for Intel? The "stretched really thin".
so a ring bus is like a ring road motorway/freeway, just a big ring you put cores and cache on. the more stuff you attach to it, the bigger the ring, and the higher the latency
arrow lake's ring bus is suspected to be shitass
And lot of the instability with 13th and 14th gen seems to happen in the ring bus too.
comparing it to a freeway will help you understand the problem they're trying to solve; traffic management
it's literally a ring
For example pics of single-ring, dual-ring and mesh as used in various Intel Xeon CPUs.
they tried selling mesh CPUs as gaming chips but meshes have much worse latency, and they sucked
but they're great for having a shitload of cores, good for servers
I wanted to say; "does L3 have issues with throughput" but I think throughput isn't the correct word. So I said nothing.
the ring bus is the bottleneck there
Too many cars on too small of a road 👍
Zen 2 used all-to-all direct connectivity inside CCX.
Zen 3 switched to ring inside CCX.
yeah, when you have 4 cores, just directly connecting them all to each other is easy. but with 8 you can't do it, so you compromise; fortunately ring buses are pretty great
then amd went further and attached the X3D cache to that ring bus
just one great big lump of L3 shared for all 8 cores
no topology nonsense
that's why it goes so fucking hard
but yeah, there's no one perfect topology, it's impossible to summarise in a spec sheet, and everyone is thinking up new clever ways to connect stuff even now
What happened lol
CPU talk. off-topic-tech can go hard on tech. And anything else.
a brief history of cpu
only saps use sap
I have no choice :[
Wtf is going on here. A beer glass for coffee?
depends on the society you find yourself in
😄
@willow pike you were a barista right?
nyet, I only make it at home
If you want to buy 2nd hand delonghi, what should I take a look at?
Oh. Nvm.
i had cheap delonghi machines when i first started making espresso, they were trash and broke within months
Well, this one, I checked multiple reviews, and got very good score for it's price/quality.
now i have a big stronk breville/sage machine and it's been going for ten years
new between 300 - 400 (depend on recent model).
Right? And with decent quality. Just a few missing features. Not full automatic cleaning etc.
i wouldn't say it's good, the grinder is gonna be dire, but you don't like raw espresso so you'll probably get decent enough milk drinks out of it
starbucks quality at home
filter coffee atm 😉
maybe better if you dial the grinder in right
a week ago I was drinking nespresso cups.
I cannot with nespresso
but I miss good tasting cofee 😦
Nespresso is so much plastic garbage. We can have like 10 cups at the end of each day.
yeah, they're in every hotel and they never clean them so they taste like ash
I just want regular black coffee or cappuccino. No small cups. No weird flavors. No espreso. Nespresso. Not the flimsy pads either.
yeah it's definitely worth it for 125 euro
find yourself a good specialty coffee subscription to go with it
the most important thing for coffee beans: buy ones that have the roast date printed on the bag. if it doesn't tell you when it was roasted, it's been sitting in a supermarket warehouse for 6 months
I get a new bag of coffee beans delivered every two weeks automatically
roasted just before they're shipped
good value too
like 40-50p of beans per double espresso
Yeah, just thinking of pro's and cons. More maintenance. Better tasting coffee. Less plastic (garbage).
Filter coffee, the problem is that it's hard to make exact portions, you just fill a pot for the whole day.
And put it into a can.
Also, if I want "normal" coffee, which for some reason is called Americano 🤦♂️, they recommend making a espresso and just add warm water to it.
You ever done that?
yeah that's a real common drink
it approaches filter coffee but you get to make it at espresso speed
the story goes in WW2 American troops ordered a coffee in Europe, got served espresso and it was too much, so they asked to dilute it to match what they were used to, hence americano, or cup of joe
Yeah, I had that experience in Paris.
Ordered a coffee, got espresso for 5 euro. One shot awnd it was gone.
I like to enjoy my beverages. It's not a practical thing (it has to be quick and give me caffeine boost).
average espresso user
its time to drop the coffee and introduce max monitor brightness in that case 😂
When I could drink that's how I was with beer. I drank one or two with a meal from time to time. I wanted it to enjoy the flavor, usually something ranging from pale ale to stout.
Not trying to slam them down and get wasted or something
Or a folder made of usb sticks
Or write a script that will create folders on a usb stick until you reach the filesystem limits...
yee but im too dumb for that so i have to do it manually
Or... play some modded minecraft
wait that's my idea for me 😄
Or Final Fantasy VIII...
yikes so a bunch of copies of the new Assassins Creed leaked? oops
mmm new kernel, time for the ole shutdown -r now
key authentication makes ssh so much nicer...
don't care much about AC anymore.
i tried playing the OG one on PC and it was laggy weirdo control garbage
how can a game that old stutter like it was...
There was so much negativity around the new AC title before it got released.
I can crank Origin up at 1440p, but a game from 2007... nahhh enjoy the random hitches
That I already lost interest. And it's the same crap every game. With slightly different models and graphics.
Do this 100x times.
I loved the OG when it came out
Yeah, me too.
I will probably play Odyssey.. that'll be plenty of AC for me
I haven't played anything past 3
But it got stale. Not innovating game mechanics etc. Just copy paste and reusing already existing game mechanics from other titles.
Well I'll admit the first one was repetative once you get into the game a bit. The second was much better though.
But yeah it did seem like it got stale but that's from the outside looking in really
I can't say "AC such and such sucks" because I haven't played it personally
All these little icons is the same thing over and over again.
Origins I played just a tiny bit of and it was... different.
To each their own.
I actually kinda like the crafting aspect
I have played up to ody. Didn't play egyptian one. Neither the viking one. And not playing samurai one.
I wonder what they come up with next. Cyberpunk AC or something.
That's the only thing they do. Putting game in new "theme".
AAA gaming right now is so... meh
I installed the FULL eXoDOS retro game package... I am in retro pc gaming heaven now
It's almost 1TB of games, manuals, box art, retro pc commercials and other videos...
Almost like a digital functional museum really..
yeah, I stick to games I know now.
last game I bought was enshrouded, indie game from germany. They build their own voxel engine.
Really cool stuff.
Oh yeah and a metric shitton of retro magazines
the old age of empires 2 is still on my PC.
I have the original Civilization on floppy around here somewhere 😮
Oh, I still have my floppy disk games somewhere around
And the Falcon 3.0 manual. Which was a GIANT spiral bound book...
That game would have been big box even if it weren't the big box era just for the manual
get a external floppy 😄
believe it or not the older disks are actually better than the later era ones
Yeah. l want so much stuff.
the quality went down as we got deeper intot he 90's
Just have to be a bit careful with spreading out costs.
lol the size of that book, mine is spiral bound
See... you felt like you actually bought something then
Now it's... click a link, download/install :/
pros and cons.
yeah super convenient
too bad we can't just have both... a big box of stuff and a d/l lol
Some games still do that. Special editions n stuff.
And like you didn't have to do some pre-order or pay way more to get all the goodies :/
LOL
Ultima had a cloth map, spell reference/command reference card. Lore book, basic manual
spell book... a little medallion of some kind
I got skyrim with a map and some other games, I forgot
they are all stored in moving boxes.
We would have our map hanging on the wall behind our monitor
I didn't unpack it. Need to move in a few months.
Something to look forward to later
and I wont probably unpack it then.
Since I will eventually need to move again 🤷♂️
lol
I am a tech nomad at this point.
Just pack it back up lol
roll it up, put it in a tube
won't even crease it
I just wish... I still had all the OG Star Wars toys I had when I was a kid. They're worth money now :/
Wish I had all the consoles I owned too. 2600, Colecovision, NES, Genesis, SNES... N64
The Colecovision was one ugly ass console I'll tell you that... even for that era that thing was butt ugly
With some pretty terrible controllers lol
What's interesting is Atari could have possible been a huge winner over even the NES. The 7800 was released a bit earlier, was almost as powerful graphically. They decided to use the old sound chips so it looked NES but sounded 2600 lol
..wonder what the reasoning was behind that, I feel like I know and can't remember
Anyone here thinking they're going to nab the framework desktop?
I'd go for that over a mac mini, would honestly make a super neat little office machine
For a non gaming desktop system I'd still get a mac mini personally, but I'm really not interested in either
It would do most of the stuff I'd wanna do but even the low end model is way more than I actually need for that anyhow. Looks like nice hardware
Yeah, that's kinda my problem
The most intensive thing I do rn is compiling mobile apps, and my current build is plenty for that 🤷♂️
I like the tiles a lot for the case
No longer doing heavy 3d modeling a lot anymore, so that's not a need
Pretty neat setup on that one
Yeah it's pretty creative
Mostly what I'd do is basic desktop stuff like web/email and audio stuff, maybe a bit of video editing.
warning for everyone using linux 6.13.4 udiskctl power-off -b seems to trigger kernel panic init process problem
and for me a 500 dollar mac mini is just as good
and in some regards with the audio software slightly better, also has a good audio stack
My thing is, though, I can do just as well with a macbook and it's portable. So eventually I'll get one.
Guitar, amp and laptop is all I'd need...
Well and the interface but it's USB powered and quite small
And also hooked onto my strap anyhow so it's basically just part of that guitar at this point heh
I know someone who bought one.
i feel like an overlord over here "compile that into a jar for me"
Ugh, I hate coding. For two hours I am telling myself to eat something.
But then I keep just coding 🤣
also vram=funny
yea apple’s auto focus is pretty bad
We're excited to introduce all-new 10 Gigabit UniFi Cloud Gateways:
-
Dream Router 7: Desktop 10G Cloud Gateway with integrated WiFi 7, PoE switch, microSD storage, and full UniFi application support.
-
Cloud Gateway Fiber: Desktop 10G Cloud Gateway with integrated PoE switch, selectable NVR storage, and full UniFi application support.
-
E...
Finally a cloud gateway with sfp port so I can plug in fiber directly to the gateway!
And get rid of ONT
I need to check which sfp modules are compatible though
in 2023:
European average fiber rollout: 55%
My country: 21% 😢
France: 77%
Netherlands: 68%.
How can large countries, win this race vs my small country?
My guess is it's because of population being scattered which makes investment cost higher
@night girder I have good news though: "The top 5 of the annual growth rates in terms of homes passed is headed by Belgium (+43%), the United Kingdom (+38%), Germany (+37%), Serbia (+30%), and Croatia (+28%). FTTH/B coverage rate in EU39 now amounts to 69,9% (up by 6,5 percentage points vs 2022) and coverage rate in the EU27+UK now reaching 64,5% (vs 55,1% from 2022). These data re-confirm the continuous upward trend observed for several years in a row now."
Guess I should pat myself and my coworkers on the back for bringing fiber to areas with no commercial interest from any ISP 😄
I was also thinking, maybe other countries or cutting corners.
And we are doing it "the proper way". Maybe we go for quality and not quantity.
Or ... we are just slow.
EU basically gave my ISP grants to cover areas that will never be profitable. At least something good from EU lol
The Netherlands is the size of my country.
And they are way a head.
It has the same density and spread of population afaik.
My country just doesn't give 💩
Not sure if ChatGPT is wrong but it said Belgium has decent cable coverage which means ISP's are less motivated to deploy fiber
Yeah, coverage is decent. Speed/price, not really.
The main problem with cable is that you are sharing bandwidth with your neighbors
If a lot of people are using internet at the same time, your speed will decrease
Yeah, I noticed that since my neighbour moved out, I got better speed 
Sadly, this weekend a lady moved in.
Let's hope she doesn't download much.
You also share bandwidth on fiber but fiber has much higher bandwidth compared to cable btw
True
But eventually, they have to kill the regular cable right?
Maintenance etc. Sounds harder to keep up two.
At some point they will. Cable (docsis) also consumes a lot of electricity compared to fiber
another good reason for fiber
Fiber is also future-proof once you have laid cables
You just need to upgrade equipment at the customer and ISP central-office as the time goes
Or to be technically correct: demarcation point
Can't you fix fiber for me?
That's how we call central office here (the building from where fiber goes to customers)
Just send someone on vacation to my country. Who knows how to roll out some fiber 
I haven't even fix my home yet lol 🤣
Annoying my coworkers right now and asking them when are they going to do micro-trenching in my street
My house is complicated because I live in a townhouse
Or atleast I think that's the term when multiple houses are physically connected together
Basically, I have a telephone pole for my house while other houses don't so the ISP has to do micro-trenching on the road to run fiber to each house
An example:
That seems fun.
Yeah. I will have to wait few months though because a lot of residential buildings near my house are being built and ISP has to increase capacity by running additional cables to the street cabinet
lel
now i'm thinking of that scene from 3rd Rock from the Sun where John Lithgow is cooked on Valium and making bass without the stove on, just grabbing the fish and flopping it over in the pan 🤣
what happens when they resurface the road? because you know, america is always fixing the roads.
When resurfacing, you remove tiny bit from top, and lay new layer on top of the old one.
You don't remove it all.
But contractors for Google Fiber did fuck that up and do it too shallow.
In parts of the network that actually got built.
that's not how we do it here. they rip the whole damn thing up.
all the way down to the roadbed.
Ok, weird way to do it...
Yeah, and by the time they finish it's time to rip it up and do it over again because they part they finished is already ruined.. good job security I guess.
I wonder what the fan info is.. these cards have no fans.
I wish I could've build this system with 4 of them instead of 2 but.. budget constraints.
"You can overclock a 5080"
To make it melts even faster?
Microducts are laid in 40 cm deep so nothing will happen when road asphalt is being replaced
I repeat, if they price the 9070xt in the sub 700 range it's a huge W
if they were smart they'd shoot for $500 & make nvidia shit themselves
go back 10 years to when $700 got you a flagship, and since the 9070XT is supposed to compete with well below that, $500ish is the segment
temporary losses to gain market share, ride the ryzen profit margin for a bit
yea seeing amd blow intel out of the water in datacenter orders last year is fun
think it was 78% boost YoY
and Steve's right, whatever rivalry or refusal to talk to each other on the amd side with radeon is only hurting them
they need to do stuff that meshes ryzen with radeon
whatever it takes to scare nvidia
Time to switch secondary computer from Core 4670K to Ryzen 4650G.
dumb question: is there anything useful that stuffing a 1660 & a 1660S in the same system would do
Not really if you don't need one GPU for main OS and another for passthrough to VM.
Or do something machine learning or like that can run on multiple GPUs.
along those lines, does machine learning need the same brand of GPUs on the system, or can you mix & match
That would most likely to be just running two instances, but no idea.
stuff my old 6800XT right under my 4090 to see what happens 🤣
Most notable change will be Nvidia GPU drivers disabling PhysX acceleration on the Nvidia card, if they haven't changed that again since I last read about that side.
Long long time ago too many people were buying cheapest possible Nvidia cards to just get working PhysX so Nvidia added extra detection into their drivers to disable PhysX if AMD card was detected in same system.
There were then modded drivers that removed that.
Imagine having an AMD IGP
But seems that disable functionality was removed about 7 years ago.
So I was way out of date.
But I was one person contemplating doing that when that disable was added, and dropped the plans as I didn't want to bother with hacks.
So didn't buy any Nvidia GPU at all because of it.
And need to reroute case front panel motherboard power etc. cables, as the ones in Define 6 aren't long enough to be able to reach micro-ATX MB when routed via bottom...
ok
I'm back @twin dew missing agesa updates finally actually bite me
usb device is able to crash chipset soon crashing the whole machine
I patched device and it now seems fine
but this is not feeling good security and integrity wise
nothing to do as there is no agesa update
The new computer is on and didn't give up the smoke yet.
UEFI update now in progress.
ECC works, with errors being reported on using software error injection \o/
So Ryzen Pro 4650G, 16GB of ECC RAM (2x 8GB 1Rx8 3200MT/s DDR4), IGP with HDMI and DP outputs, 256GB Intel OEM TLC NVMe M.2 drive.
Hynix 8Gbit DDR4 D-die
So that computer is properly up and running I might see how fast I can make the memory and IF with lowist voltages.
The RAM could at least do 3600MT/s @ 1.35V, don't remember how high I did try it.
IF on that monolithic die could go at least to 2200MHz IIRC in quick test without problems, never tested to limits.
Might have tested it up to 2400MHz IF.
yeah i considered it too, but physx was only ever used for fancy looking smoke and completely non-interactive effects, so i never bothered
(by this I mean, you can effect physx, but physx cannot effect you)
yeah, mirror's edge glass, arkham asylum trash in the wind etc
But still just eye candy.
i wasn't willing to pay for a whole separate GPU for that
besides I had a full danger den mad scientist watercooling loop at the time and did Not want to add another thing to the loop
Would have been about 70e IIRC for the cheapest new GT-something-30
and no way was I gonna leave a stock air cooler in my pc
So half-height, passive, slot powered crap card.
But would have given same PhysX performance as the best ones.
hm, I was not aware of that
Didn't really matter what Nvidia card, as long as it was one of the supported ones.
imagine paying €70 for a gpu capable of anything these days
Could even have been one of the little more expensive and very rare variants with just physical PCIe 1x slot on the card.
i wonder if people are working on hacking physx to run on newer devices or cpus
we got cuda running on radeons these days so why not
Newer versions work fine on CPU.
yeah i mean the older, abandoned ones
which only run single threaded not even SSE enabled fallback modes
Seems to be possible with just DLL replacement at least in some titles.
(almost like they did this on purpose to sell PhysX cards)
Need to rename some of them to match the older ones.
And some files need to be left at default.
GeForce.com presents an in-motion look at Batman: Arkham City on the PC, which will be released November 15th in North America and November 18th in Europe. Primarily focusing on Hardware-Accelerated PhysX effects, the video compares several scenes from the game with PhysX effects enabled and disabled. Complete details, and the world's first tess...
changed my mind. this looks trash
just cluttering the frame up with shit like it's the star wars prequels
physx can stay dead
Using Physx 5 in my engine as it's open source, but Im only using it for collision detection & trigger regions.
Have done a 2d physics engine myself before and that was a pain, and cant be bothered to make one for 3d (:
Might use it for some basic cloth sim too, but thats for later.
@twin dew what's your thought about about usb device triggered mm rss chipset linked pcie device timeouts bugs in last kernel with ECC memory all ok
having random software sigsegv and other kind of shit like that
No use asking, I have no idea.
never had any problem for 6 month, I plug in gen 3.2 2x2 device on 5G usb port and it full break down everytime
fuck asrock not updating bios
I kind of hate this board ngl
There is the 1.78 Beta with AGESA 1.2.0.C.
Newest otherwise would be 1.2.0.C Patch C (1.2.0.Cc)
the only update I didnt read the fucking changlogs
what wait it does not show up with "show all" then ?
I just checked yesterday sorry maybe ?
No, in the Beta Zone entry only
btw I'm talking through it, if I'm not responding maybe it just crahsed
1.78 from 4/11/2024, with that AGESA with LogoFAIL fixes in it.
When selecting that Beta Zone entry from Downloads.
I mean 1.2.0.C is much much better than 1.2.0.7 right ?
Much newer at least.
I mean it surely is better for ram compatiblity and early microcode patch and chipset controller/driver stuff
Asus put out UEFI version for B550 board with 1.2.0.7 in 2022/05/23.
And then UEFI with 1.2.0.C for same MB in 2024/04/25
some people say thing implying I may have screwed because of core undervolting, which drop under usb burst (usb 3.2 2x2 device certainly is able to burst the usb path, even if connected in 5G mode)
1.2.0.c have mitigations for various security issues, its the main reason I'm interested
And the later versions after that LogoFAIL 1.2.0.C have been just more security issue mitigations.
Seems 1.2.0.E is just out with more.
Over that 1.2.0.Cc that I thought was latest.
Wasn't yet available for that B550 board I just put into use.
On another Asus AM4 board .E UEFI was released two days ago.
tbh I'm on the "if you update your shit often enough it's way easier to keep track of changes and react fast" team
hopping from .7 to .c surely can bite asrock devs
2nd most interested feature : not broken ECC ram manual settings
so I can up my ram speed
though it may still be broken
79 euro... crazy. What you gonna do with 32GB SSD 🤣 and something between the 4 and 16 GB RAM.
Not even proper SSD, eMMC on MB,.
For that price.
Basically that thing was useless when it was new.
This comes from "Hardware Kings"... some local refurbish shop (or one man army).
Now I know not to buy anything from them.
When looking further into the shop, it's fishy.
And very probably illegal Windows installed.
very likely yes
I have to be careful with my accusations, but my gut feeling tells me this shop isn't 100% legit.
Just the fact they didn't fill in the "about us" section etc. And just leave a phone number. The shop has 1 rating so far, but has pages full of hardware. And is 4 year active.
Same phone number, so they do have a website.
but their website is like a shop template. (feels like every other generic webshop you visit)
holy fuck look at that ai boom
nah i reckon it'll come down fast, and this year
nvidia won't go out of business, they'll forecast the pop and reduce orders with TSMC
it would make them care about gaming again, if people hadn't shown that a $1000 5070 is just fine
Yeah I also read that. And said it ages ago.
Nvidia focus shifted. Only thing that didn't shift, are the loyal fanbase/gamers.
Who don't realise Nvidia gives 💩 about them.
no company cares about their customers, but most have to pretend to
ahhh good ole lip service
What I hate seeing is people taking out their anger on some counter help or something. They didn't do it... and often nothing they can even do about it. No point in giving the little guy heat
A lot of people complain about customer service work but to me people are pretty easy to deal with in those scenarios.
Just keep smiling and being nice, be helpful, remain that way. It will either A) bring them down a bit to match your level or B) they will get more pissed that you're not getting upset.
Either way you win really heh
I worked on a production line, shit always rolls downhill. blame middle managers and decision makers, never the poor bastard actually doing the work
Yep that's why when I have to call customer service numbers for something I don't lash out on the person who answers
People will call and start bitching them out like they did it personally or something haha
At one time I was who you had to deal with if you went all Karen and wanted to see the manager at Autozone lol
I'd just love the ones that wouldn't dial it back and get more angry. Any time I can do my job as a customer service professional to perfection and STILL piss you off, is a good time haha
Because believe me with the filter off he'd learn what getting pissed really is haha
Is that graph made before or after Deepseek?
Maybe care is the wrong word I used.
It's the focus of a company like; young people, old people, gamers, programmers, etc. Target audience etc.
69 Million out of thin air
Nice.
And without need to pay any taxes on it either.
Damn, I want my tax free $69m too
Haha 🤣
Ah, Nvidia 5000-series dropped 32bit CUDA, and also 32bit OpenCL support.
Which is the reason for the low PassMark scores.
As they were still using 32-bit code for some tests that didn't need to be 64bit.
tbf, we went from 1200€ 3070 Ti in crypto to 1200€ 4080 in AI craze
And now those cards are even more expensive because there is almost no stock since production has been halted
the 1200€ 4080 is current pricing
this will pass when people look out the window because there's no electricity and/or internet
Nah
Ill keep huffing hopium
funniest shit from the tech market in recent past: "we didn't wanna full commit on this cuz we thought Intel's stuff was actually good. we were wrong"
Next gen; everybody buys Nvidia 
Hmm
"intels horrible product"
Some guy from AMD
LOL amd should show off the fps of games with 32 bit physx during the announcement

just do a collective middle finger to nvidia, troll the fuck outta them
spend 10 minutes showing off the PCIe power connection, its history, compatibility, and stability
and don't ever mention nvidia, not even in the charts
Which would match what Nvidia 5000-series gives?
Because both would be down to CPU.
It would be a home goal / self own
people have been showing off the 5080 losing out to the 980Ti in 32 bit physx games
I just assembled this cantena from a piece of solid coper cabel and a can I had on the table They don't call me a MacGyver because I'm pretty
use ZLUDA to run 32bit physx
i mean you can't with ZLUDA right now but it would be very funny if amd cooked up a secret version just for trolling
What do we think? Boss wants a new (small) Hyper-V host to move existing users to. All VMs are accessed via RDP, the A2000 will be partitioned across them.
Not asking for advice just opinions
should be fine for 10-20 (concurrent) office users
Yeah we're looking I think 8 so that checks out
It's also going to be hosting one of our proprietary applications. It's heavily single threaded. That's why I picked that CPU specifically.
At the same time, it's a nice fast CPU and it has twice as many cores as the one that he originally picked out, which means I might just be able to get my IT workstation moved over
Because oh my God, that thing is atrociously slow
I was going to install HyperV on my remote server and RDP in. But I don't know the first thing about doing that with HyperV and didn't wanna be bothered researching it.
Almost installed Proxmox which I have some experience with but just said screw it, installed Ubuntu, and called it a day.
Hyper-V is a sin and I hate it
You dodged a bullet
Well I was going to play with it but I kinda figured I'd hate it. Proxmox is really more suited for me anyhow if I want to run multiple OS's since they'd almost all be linux which it's very efficient with.
That being said it's not entirely a joy to work with either but not too bad, just a but I dunno clunky?
Well VMWare Workstation did change their license, so their V17 Pro is now free for Personal use..
didn't know workstation had a free, i know last i looked they had a player that was free but you needed a pre-rolled install to load in
Dunno if i can send link
i usually use virtualbox if i want to just run something inside the OS... well i've done it on a headless machine too
i can find it im sure
thank though 😄
yeah the y have a whole blog post about it lol
Yeh its somethibg they changed like Q4 2024
VMWare is something I could do a lot with because my IT buddy works with it a lot
can do basically anything i want with his help im sure
Atleast compared to Hyper-V, then VMWare you’re atleast fairly free to do as u want in terms of having shared network folder from Host machine, and dedicate your USB to the VMWare as well
Not sure what your use case it, but thought it give a hint as it have saved my arse a lot of times.. And knew it works fine with Ubuntu and Debian
At some point I'm setting up a rack with UPS's, server, switch etc, I'll either run something VMWare or Proxmox on the server
he might be able to get ahole of a licence for what i need
Not familiar with Proxmox.. But WMWare and backup wise, its nice with their snapshot system, which can be intergrated with Veeam
Proxmox can do VM's but it's real magic is containers. Creating a new linux container creates virtually no extra overhead on it's own
But im no techie at all.. Nor anything remotely a server engineer at any level :D
shared kernel and other resources saves memory and cpu cycles
Me either, well with that stuff anyhow
I basically just want virtualization on the server
The storage server when I get to it will be more rack space filled heh
Proxmox does sound like the pros from Hyper-V with shared resources.. Thats one thing VMWare doesnt have going, as u have to specify the CPU, RAM and HDD commitment to your machine
you can assign cpu counts in proxmox and stuff but yeah once loaded in the linux container doesn't use much of anything at all
it's almost like running it as an app really
Windows I only care about on my main desktop..and laptop.
welp guess I'm... about to become the product lol
Think ill give it a look.. Maybe test it out..
We usually run VMWare to have somewhat syncronized workstations at work with our engineering software.. The VMWare works fine, but its very very resource intensive
Not sure how good Proxmox is running Windows 11, as atleast in Europe becomes a requirement if you want to comply with European law about cybersecurity ( NIS2 )
Never ran 11 on it, should run fine though. With virtualization it seemed as fast/good as anything else I've used personally
Pretty sure in the enterprise field you mostly see VMWare stuff
Or at least probably don't often come across Proxmox
I guess some places do HyperV...
Yeah we had a lot of talk, because VMWare was changing their License method, so each user has to have a $100-200/year subscription.. And when running 100+ users that can easily become a fairly expensive procedure..
So we were experimenting with Hyper-V, it got the job done, but with alot of restriction..
So think ill try look into Proxmox during the weekend see if thats any good
my company is old and the cto is older
It was really good for me when I ran it because it runs headless with a web interface out of the box.
If there is a screen attached to that system (unlikely), I can't see it from here 😄
I can tell you what rack I'm in, and slot
Lemme guess the CTO often goes.. Back in my days in the 80’s we ran our system like this and this.. And that worked flawlessly
Well the thing is... back in the 80's
that's how we actually did it...
Laptop? HA! They were called luggables
look it up whippersnapper!
They were big, heavy, tiny screen. That's how it was and that's how we LIKED IT!
When I was 14 my dad brought home one of those luggables from work. He worked for the US General Accounting Office. So I was a nerd and knew things he didn't even know I knew. So I was reading over some documents and it was about how they were disposing waste, the cost.
Yeah imagine loading a game like Satisfactory using Punchcards :D
I'm like mmm not sure if should be reading -continues reading-
My first "real" computer had 360k floppies. By real I mean IBM (or compatible)
And we waited like a half hour to download pirated ms pac man one night and it was like 30kb or something stupid
300 baud baby!
Im a much younger guy, so those 5 1/4” floppy was just taken over by 3 1/2” floppy
By the late 80's there was a quite robust global forum network.
like FidoNet
And with phone lines on some BBS's very limited (one or two lines) there was a message packing formmat, QWK. It would download all the new posts/replies then you can view them and write up replies offline then when you dialed back in you'd upload that packet and it'd go out.
The largest BBS here when that era died had 40 dial in lines and data... so at least two T1's
those things were pricey then too lol
like several thousand i think... a month
The 2nd biggest had 20 phone, 2 data
The cool part about those, and one of the aspects I miss, is they were local. The boards would have regular meetups.
Did you ever own a Zip Drive?
It doesnt ring a bell.. But looks like a regular floppy drive
But guess it was early stages of compressing files :P
Similar looking yeah
But would store (in later versions) 750MB on a disk
I had one that was 100, early adopter lol
I got it with a SCSI card I installed for faster transfers
And yet people would still insert Disk 1 of 852 to install their software :D
Well... those drives just wen't common enough for software devs to make them
The multi floppy games was getting a bit out of hand though with the number of disks. Just to install a game that didn't need disks later you'd be swapping in like 8 disks
Dont recall that.. But was also first beginning “gaming” in 89-90s
In the mid 90's UHaul was running Windows NT and backing up on tape lol
...which would then be physically shipped to corporate
I'd love to see what things are like in 100 years...
No he's soviet ukraine old
God damn..
This video was sponsored by the Light Loop AIO from be quiet! : Here is an amazon.com affiliate link to the New Light Loop AIO: https://amzn.to/3NfucW6
I try and get Stalker 2 to not run like a dumpster fire on my PC. I JUST WANT TO PLAY THIS GAME.
Here is a link to the Hardware Unboxed video: https://www.youtube.com/watch?v=g03qYhzifd4
Get...
shockingly, good hardware and software still needs competence to run it properly
It's one of those things
We're at a point where our hardware is so, so much more refined than the software that it runs
To such a degree it really boggles my mind
Refining software doesn't mean software hacks 🤷♂️
It would be 120 FPS solid if optimized properly
Devs don’t wanna make a Mac port tho 😭
Yeah because like a dozen people play on mac
And it's not so simple to port to arm 🤷♂️
Apple isn't just stock Cortex cores either.
Like 10 
you cant know that
That would have it faster than a 7700XT 
maybe even 7800 XT
i'd love to see someone make a game with the optimization from the 80s
Working on it
Stupid thought process..
Generally smaller motherboards, like Mini-ITX has better performance than say an Extended ATX.. The signals travel path is shorter, thus making less latency, thus increasing the performance on communication between CPU and its respective components..
Technically correct, negligible IRL.
Build quality is far more important in that regard than trace length.
Another important factor are open ends which cause signal reflection.
None of this matters to everyday users, but it's stuff that XOC enthusiasts care about when pursuing top ranks on some benchmark's leaderboard
Yup, good thought process, but signals travel at the speed of light, and the difference in .2 meters or .5 meters at light speed is not a factor.
I think the limit of pcie is around 15- 30 " because of signal integrity requirements to transceive data at those speeds.
PCIe can cover ridiculous distances far exceeding 30". Ever seen the LTT video where they just keep adding PCIe extenders?
Ehhhhhhhhhhhhhhhhhhh
Electrical Signals in wires don't travel at the speed of light, the electrical field builds up at that speed tho.
It's slower, but still very fast
Alphaphoenix tested that in response to the Veritasium vid
LTT should do a "how far do components have to be apart before the speed of light becomes measurable/noticeable"
bro what is it with Mexico and having no outlets
Anyways, when you're talking RAM and CPU performance, you're talking about latencies smaller than a millisecond
And iirc one light nanosecond is about 30cm
if it's anything like the US, it's old construction where the most important electrical object in a room was a lamp
Fair enough I guess
with older houses here you'd be lucky to have 2 outlets total in a bedroom
just searching for outlets isent what I want to be doing 22 hours into my day
it can be a pain in the ass when it comes to room layout
But ay atleast the wifi is absurdly good
Bigger problem is completely different libraries and even more importantly that only graphics API available is Metal that is Apple-only.
With active extenders that have signal amplifier chips in them.
Each chip resetting that distance max because of signal integrity.
With PCIe 5.0, you already need amplifiers to make second PCIe slot on MB with it.
I though the footprint of an electron app would be doable. But the minimum seems to be 100 MB because of the shipped browser.
Should have gone for Tauri 2 maybe 😕 (v2.tauri.app)
Basically all the HTML+JavaScript app thingies are bloat incarnate.
No. Simply not true.
The chrome browser is shipped with the .app or .exe or whatever for Electron.
Even discord (built in Electron), has it's own chrome engine shipped with it.
That's why it "works magically" on all platforms.
Yes? And those are very bloated in both install size and memory footprint etc.
but there are alternatives way more lightweight (and not bloated)
This is what you said.
Tauri can be as little as 600KB ...
Tauri is a framework for building tiny, fast binaries for all major desktop and mobile platforms. Developers can integrate any frontend framework that compiles to HTML, JavaScript, and CSS for building their user experience while leveraging languages such as Rust, Swift, and Kotlin for backend logic when needed.
By using the OS’s native web renderer, the size of a Tauri app can be little as 600KB.
So?
So that talks about the INSTALL size only.
It's still HTML+ Javascript that isn't bloated.
While my .app that I compile is 107MB for Electron.
I did talk about both install size (if including browser in itself), and memory footprint (even when using external browser).
Anyway, pros and cons for everything. But 107MB is a bit rough for a tiny application. Didn't expect it.
I am looking into reducing it with a webpack maybe.
Using WebView2 (for Windows, or similar for other OSes) helps with install size, does nothing to runtime costs.
using dmg
QT + QML with shared QT install is lowest I know for RAM use.
time to play with this AI in Warp Terminal a bit more
I'm in the gym, is the 9070 XT any good
No idea.
At least the start of the stream is absolute marketing crap.
yeah well who's isn't
The readymade PR video.
gotta suck the shareholders off
But the FSR4 being RDNA4 only is because it uses data type(s) that aren't available on the older RDNAs.
eww json
I know. Need to fix that. If possible.
Prices have voiceovers on the stream...
They had already recorded the whole thing and then changed prices...
Think I might need to migrate my project to Tauri ... sigh
So much work 
Tauri is Rust, and Electron js, but I use typescript and then bundle to .js
But it does support Angular framework which I know. And it does support typescript. Nice.
So 9070XT, possibly cheap enough, will see once third party benchmarks etc. come if it is sufficiently cheap to possibly get people away from Nvidia.
9070 non-XT MSRP is too high compared to the XT model.
Macs are REALLY fast if the apps are fully optimized to take control of Apple Silicon
Apps are really fast if they're designed properly.
Especially as Apple has decided to make security opt-in.
By default the CPUs have ton of speculative execution data leaks, and by setting a register flag you get slower but secure mode.
maybe apple should support real APIs then
metal is not vulkan and mac gamers are much, much too small a market to bother making ports for
let alone make ARM versions for
it is a hostile platform
Not ARM as such, but the full library suite, not just having to use Metal for graphics.
If you have the source for the program, and it doesn't include ASM snippets as optimizations, then if the same libraries are available on the other CPU type, it usually just can be compiled to the other CPU.
you'd never wrangle all the dependencies you need
That is that libraries part.
So not about x86 vs. Arm.
But MacOS library availability.
i think a lot of the big Mac (lmao) games that use metal are still x86
this imposes a single threaded penalty which hurts the bigger chips with bigger GPUs
Lot use MoltenVK to actually use Vulkan, that then gets translated to Metal.
yeah I mean the ones apple like to parade about
moltenVK also isn't perfect
also satisfactory's vulkan is still experimental
Yeah, but difference between full rewrite and rejigging something in the Vulkan to work with MoltenVK.
yeag
Satisfactorys Vulkan is a mess, AFAIK just enabled automatic Unreal Engine Vulkan shader creation.
no real good options
or you just don't get a mac and you avoid all of this shit
framework desktop has seriously dented the mac studio on every level
CPU, GPU, AI
if we start seeing strix halo in laptops, that's a real good macbook pro competitor
For people who do buy macs, it is very large part about the OS.
heard it before and it's dead to me, I'm gonna move over to linux one day, that's gonna need relearning
go to where the best product is. if you stay on a platform out of inertia, you deserve all the shit deals you get.
the fact that macs are still so dominant in some industries is Wild given how modern mac pros are prosumer grade desktops rather than real workstations these days
they're good for video editing and that's pretty much it
they used to be good for audio work but M3 is a severe regression on that front and there's a continuing backlash about it
3D? you cannot use graphics cards on it. dead
Even most of the same video editing software are now available for Windows too.
Except Apples own, which they shot in the foot some time ago.
also we're not talking about the difference between like, adobe premier and davinci resolve here. learning a new OS is much simpler
(also people should unlearn adobe too)
My engine will support mac... if you can build it without me changing anything 
pay me $100 a year for Tim Cook's permission to build on mac, plus the cost of buying a mac

then pay for a therapist to deal with my anger issues from how shit macos is
It will? 
Like I said
if you can build it without me changing anything.
Which is highly unlikely
Just throw it on github, let people fork it. 
yup
I think it's 60% or something. Since c is given in a vacuum. Still very fast indeed
Had some back and forth with boss, made this new estimate. He likes this one much more (more redundancy and scalability)
yeeesh
is that USD?
Basically me 3 years ago when scalping shit was preventing me from buying a new GPU
Shit, my nephew just told me my PS4 died.
He was annoyed with the wifi, and pulled out some cables, "suddenly" PS4 starts in safe mode asking for update 
Best part, he just admits that he was angry 🤣 The honesty of kids. I love it.
Make sure you validate the drive before doing anything else
If something got corrupted due to the sudden shutdown and you try to let it update it will brick and you'll need to remake the drive
Honestly, it seems pretty bad already
Not sure how much options I have in safe modus, what I can and can't do.
Do you mean this by any chance?
yes
Sorry its been a hot minute
It's been like over a year since I last messed with ps5 firmware issues
Thanks for the tip
Firefox is so many miles better than anything else out there I'm still thrilled to use and donate it
is many good must have
when you're young, you break stuff when you're angry. when you're older, you break relationships. i'll take broken shit over losing people any day
Monster Hunter World is enumerating GPUs in system every frame...
My 8 month old nephew accidentally knocked a night lamp with his name and birth date engraved recently so I had to buy a new one lol
Hope you fixed PS4 though
World or wilds?
The new one probably I and I had brain fart again
Yeah, Wilds
Hm, maybe I should try deactivating the iGPU then...
Doesn't affect anything.
K
Just doing stuff each frame that should only be done once per start, and possibly when user enters video settings menu.
And throwing the results away.
When the specific calls are not cheap to be made.
Thats the thought, does it get cheaper with less GPUs to enumerate?
No idea, possibly.
Worth a test.
But not by me, I just wanna play for now
Not sure, will see PS4 on sunday.
good talking points about science overall
Go to https://ground.news/sabine to get 40% Off the Vantage plan and see through sensationalized reporting. Stay fully informed on events around the world with Ground News.
In this video I explain why I think that academia will be mostly defunded in the coming decades. In a nutshell it’s because it’s for all practical purposes a planned economy...
Had a tough divorce with intel. Going through a second divorce with nvidia atm 

Should I sell rtx 3060 ti for €250 or €300 and buy RX 7800 XT for €500-600 or just wait for new series?
When the 9070:s are coming for sale in just few days in that price range?
but you got to agree - they screw you hard
but....
Assuming scalpers don't buy everything
Been in production for couple of months now.
Original release date was in January if I have understood it correctly, and AMD pushed it back for non-technical reasons.
If they can get plenty of th ose on shelves it'll be such a huge win for AMD
I'm thinkinking about getting one myself if I can manage it into the budget
I'll have to buy a new CPU though if I get 9070 since ryzen 5600X is going to be a bottleneck
For some things probably yes
the 5800x3d can probably keep up
i think past that it's pretty much new board too
And RAM
That's why I hated this upgrade, I needed to escape AM4
Why would you need better CPU etc. any more than if you bought 7800XT for same price?
Can't believe you need 32 GB of RAM now
So expensive upgrade...
You don't need.
But 2x8GB of DDR5 is about 10% slower than 2x16GB, because those 8GB DIMMs are 1Rx16, and not 1Rx8.