#off-topic-tech
1 messages ยท Page 9 of 1
oh, well, thats a completely different topic: MS defaults EVERY DAMNED THING ON BY DEFAULT
๐คฆ
no, it's not a different topic. Backwards compatibility and Microsoft hauling 30% of old code over from version to version is idiotic.
disable powershell too while you are at it.
i wonder how much MS thinks they are 'stuck' doing that, vs wanting to do that ... (if they actually even WANT too)
imagine what would happen if ms pushed like win12 that basically killed off 90% of all those old apps people still use and old games people still fire up or buy and play for the first time ...
it would be armageddon
there is always a choice
Pretty sure if you pull the plug of old and insecure stuff, people will understand if you communicate it properly.
nope ๐
it's like car manufacturers recalling cars. People find it annoying, but people also like it better than having no brakes while driving.
we've been through it with apple doing it... and people STILL dont understand when told why
So what do you do as a car manufacturer? Take the lesser of two evils.
Same for old ass features/code imo
so many old favorites people cant play cause of it ... they are so angry. we STILL get people today demanding refunds on shit they bought 15 years ago because it wont run on 'new' macos
(maybe because mac users are clueless to begin with ? LOL)
๐ฆ
๐คฆโโ๏ธ
Can you play a PS1 game on a PS5?
and do people moan about that too?
during the change of macos... we saw a massive dip in general mac gaming overall. both in new games coming out, old games not being updated, and people just stop showing any interest in games
had someone on the PCPP server recently who was having a lot of weird issues - he was on Windows 7 and refused to go away from it
im not even going to discuss consoles, thats an entirely different realm
That's just sad ๐ฆ
I mean, I know it sucks. I had to do it. Massive refactors in code. Spending days replacing imports and other stuff. Just because some big dependency made a major version. But you don't have a choice, that's part of the job.
and if you have to justify it to anyone, like a manager for example, just throw a few scary terms in their and they will follow your logic and let you work in peace.
i used to just say 'its for security' to justify anything
but now that im the boss ... i just do whatever i want
exactly ๐
Tbf, I do make a difference for home use (depending what you do with machine) vs professional
you can also open task manager with Ctrl + Shift + Esc as a small funfact
omg ๐คฃ So that's, ctrl + alt + delete / windows key + task / right click menubar / ctrl + shift + esc / right click the menu icon ... how many ways are there to task manager?
Win + R -> Write taskmgr. CMD -> Write taskmgr, launching the exe from system32 via explorer 
make a shortcut on the desktop ๐
Search Task manager in searchbar 
Like Microsoft wanted to be sure we can finds task manager ๐
"ugh, we spend all this time and effort on SeCuRiTy, everything's going much slower, yet we don't ever get hacked"
the infosec responsible where I work has to continuously fight against those people wanting for stuff to happen
not like we'd store the health data of thousands of government officials
You don't install a fire alarm after you have a fire and your house burned down
The same goes for security. It's prevention.
"We never had a fire, so why have a fire alarm?"
Takes down fire alarm Why do we even need that?
Two days later FIRE!! Loses his precious pan because he didn't had fire alarm
The generalized format: Why should we pay for IT team when everything always works?
renaming is shown with icon on top of the new menu
The IT paradox:
Everything works, why do we even have the IT?
Nothing works, why do we even have the IT?
Yup.
"look the under paid IT staff is doing shit. Ahah I knew it was not worth it ! let's cut it down even more"
then pr talk that it "was more worth it that way" and that "they were forced by the economic context" to do so
.. even though they actually increased cost by having poor IT capabilities
What version is satis on?
maybe they'll do it for Christmas
didn't snutt say something about it, ๐
I mean, they can't keep chasing versions else they will never release the game ๐
I still don't like it that foliage doesn't grow back in satis ๐ฆ
but the nanite stuff has been in unreal for a while no?
yeah, since 5.0 - nothing new
how so?
still not sure I can follow you
but bio fuel man
anyway, tech related: unreal 5.3 is only as good as the developers who implement it. So a lot of potential to waste.
I dislike how epic is being so aggresive and gives 0 fucks about gamers.
They see us as a big product to steal from steam, that's all they want.
They don't have any passion for games or our culture at the top.
Atleast steam was founded by someone with a passion for games.
Steam isn't what it used to be. But aslong as Gaben can keep steam from going full bonkers like the rest, I am happy with it.
If Tim Sweeny was a gamer he wouldn't steal games from steam and make em exclusive.
Man, how you defend anyone who makes exlusive contracts to fuck over the competition and then say it's for us gamers ...
Yeah, so don't tell me it's with our interest in hearth
The free games? Just to steal players from steam.
They play it so dirty they lost the battle before it began for me
just enjoy whatever you like
and I do the same ๐
been using steam since I was 12 years old, so yeah ...
can't teach an old dog new tricks I guess
though one argument is that it is probably not hard to go from 5.2 to 5.3
I think that it is mostly same assets as any feature in 5.2 is in 5.3 too
Depends on implementation. Haven't read patch notes.
For example, they added Lumen to satisfactory, but didn't optimize it.
and why would that be any problem for going 5.3 ?
?
It's pretty well optimized from what I've seen, just not officially supported
It's good enough to play with it full time
changing engine version only mean RE implementing anything if the engine change in that aspect, if it is only new features then mostly nothing "has" to change
For many it performs better than u7
what do you mean
it would make it longer because of currenlty having to support multiple branch
because they were in a rush to fix update 8 crashs
not longer longterm
going from a bike to a moto is hard,
going from a moto, to a moto with a reverse gear is not hard
going from a moto with broken gear box, to a moto with broken gearbox with a reverse gear is stupid
Wrong channel
what kind of rip off of statisfactory that game is ?

Where should I ask my question?
ty
๐ค
And devs said they couldn't do that anyways because on how the foliage collection system works.
Upgrading between UE subversions is hard if you have done any modification to the engine.
And you almost always have to do that.
And Satisfactory has a lot of those.
Even when 5.1 to 5.2 was much easier than UE4 to UE5.
yeah, i can imagine they are almost having to recode everything theyve done
Manual merging for the parts that CSS has edited and were changed for new engine version by Epic.
And then lot of bug fixing
i dont get all this whining and complaining about the introductory mac M3 having 8gb ram .... if someone wants 16gb or more, pay for it. its no different than building a pc and having to pay more for 16gb or 32gb of ram over 8gb. seriously wtf? so apple offers a dirt-base model for the lowest cost they can offer ...
its like these apple nitwits seem entitled to getting 16gb for the price of 8gb
(and yes i know apple pricing is always bloated, but its not really the point)
The difference is that apple massively overcharges for the added ram or storage
On the order of 5-6x consumer price
if apple ONLY released the 16gb model for the price of the 16gb model and didnt even offer an 8gb model ... no one would complain
but they decided to offer the 8gb, for less ... oh the world is on FIRE now
๐
granted i wish apple would get more reasonable with their ram pricing .... but it is what it is
was nice back in the day where you could actually swap out ram with any brand aftermarket ram ...... always buy the base model, then go buy some g.skill
same with the hdd ... go buy a 2.5" ssd with more space
not anymore! the whole board is one solid piece of hardware ๐ฆ
GN relaunched their text site.
Should be getting the same content as YouTube, one week delayed or so?
GN? hm... im not familiar
Gamers Nexus
oh ah
I know it'd be super useless
But what if we could engineer insiad of dual GPU cards have dual die GPUs. Where it's still the same gpu or two GPUs acting as a single one.
Those were a thing already.
Ah, sorry, read wrong.
But multi-GPU-die cards are already thing.
Both with Apple and with latest AMD professional number crunching card.
AMD Radeon Instinct MI300X:
4 GPU dies
https://www.techpowerup.com/gpu-specs/radeon-instinct-mi300x.c4179
And higher end Apple M# configs use multiple dies as one.
For both GPU and CPU.
But the difficulty is still price, because you need active silicon interposer to connect the dies.
The AMD for multi-die CPUs use almost normal substrate.
And the 7000-series Radeons use cheaper and simpler interposer.
๐งด
But the tech is finally here.
But it is expensive and very production limited at this time.
TSMC just doesn't have the capacity in their assembly lines.
Intel version is starting to come online with their next gen mobile chips.
Oh so that's a thing
But that seems to be at same level as the Radeon 7000 series one possibly?
Just I'm guessing too expensive for the gaming market
Higher end Radeon 8000 might have been planned to use it, but might have been cancelled because of the TSMC production line limitations.
Hmm
Can a Ryzen 7 7600X bottleneck a 7900XTX?
Yes but not likely
Look at CPU review, then look at GPU review.
The performance will be general near the worst of those two, at the select resolution used for the GPU review.
For CPU review the resolution doesn't matter.
Im running at 3440 by 1440
Not going to go look.
Just instructing HOW to look.
ultra widescreen 2k? neat
Yup. I love it
https://www.techpowerup.com/gpu-specs/titan-ada.c3985 WHY WAS THIS NEVER REEASED
Because most of the stuff in there is speculation based on possible leaks?
fair ig
Itd be funny
And all the good enough dies probably go into these:
https://www.techpowerup.com/gpu-specs/rtx-6000-ada-generation.c3933
And the various Nvidia L40 variants.
I would expect that fully working dies are almost unobtainium anyways, which is why all those are not such.
And that finding ones that could do high clocks no matter how insane voltage you used would be even harder.
@twin dew Should i try to get 3733Mhz Cl14 going for the funny on my ram.
I know thats basically the hardwall for Vermeer IMC
am i reading this right 1.5Gbps to read this video???
@jagged snow so in my experience and the biggest issue (Due to me working on modding Minecraft and creating modpacks for it) was OpenGL support Radeons had, expect issues my friends have reported that they couldn't resolve for a long time.
Oh interesting, I've never had problems with opengl support
I wonder if they were using bugged drivers/had improper configurations
What are you doing wrong with your encoding settings to get those insanely high bitrates besides encoding at 144fps which is absolutely ridiculous?
Well... few thousand people having the same crash because they have Radeon can't be a fluke :P
Modders had to often do workarounds if they detected Radeon due to those weird crashes
Oh interesting
hehe this one was on pupose as a test lol
To be fair, there are also similar issues on nvidia though
my HDD can in fact not read that
They just have so much more industry adoption that the workarounds are standard practice
I know, just in my specific use case of computer, I prefer to have Nvidia. I need to say tho that the control panel AMD has is great, and would much prefer that than the... thing nvidia provides lmao
Geforce Experience isn't great, and control panel is hella outdated, it does the job tho... most of the time
But I also don't say Radeons are terrible, it's just my experience. If you have good Radeon Card, Nice, it will definitely serv ya well. In my case, It would just cause a lot of headaches probably

Well CPUs are pretty good. I have no idea above 5th gen as I completely ignored latest releases lmao, but I have r9 5900x and I need to find yet an app that I use that has would require better CPU
just because every amd ive ever bought and used in my life has sucked balls... doesnt mean that they do
Yea pretty much tho for you all of them do at this point
same here haev nerver had a single good expirence whit amd
Says the person with a dead nvidia card
i dont go out of my way to hate on them on purpose but i woudlent buy there products
What amd products have you owned?
Specificity please
i still keep an eye, cause when i go to build, i weigh both sides each time (since i build years apart, and stuff changes)
well i guess i am singlehandedly keeping like 7 acer crapboxes alive curently (all on AMD platforms) but i dont own any of them im just fixing them cuz i have nothing better to do
I have a laptop with Radeon 6700 HD or 8700 HD, Can't remember, but the card is sadly ded for past 3 or 4 years :P
they all suck balls anyway and bluescreen every 30 minutes so even i cant make it worse
So you're saying that you've never owned an amd product with equivalent recency and condition to the nvidia products you're comparing it to
not curently but i use to have an amd card and i got to try a amd cpu like 2 years ago
both are mid but allright i guess
What do you mean by mid?
the amd chip lost whit more speed and the same core count as my 10700F at the time
So?
My ford taurus has a v6 and four tires just like an r35 gtr and that doesn't mean anything
@jagged snow after trying to fix the hotspot issue once again my thermals got much worse
I applied slightly more thermal paste in the same star shaped pattern on my GPU cold plate
Immidatly hit 120c hotspot under load and shut down
I'll try again
You guys know computer stuff, right? My m.2 SSD drive reached 100% usage and my pc shut down, now my bios doesn't detect it. Is my drive completely broken?
It's also my system drive so my computer won't boot
how old is the drive ?
M.2 nvme goodram 1tb, idk the exact model
Nope, 100% usage in task manager
Shouldn't be an issue, but I dont know anything about goodram as a brand
Have you tried a full system power cycle?
Not yet, is it something I can do from bios level?
Yes, just turn off psu switch, hold power button for 10-15s, then turn psu switch back on and try to boot.
I'll try that, thanks, and if it doesn't work? Are there any more things I could do?
Yes
I thought about taking it out and cleaning the socket, but idk if that's a good idea
I'd just reccomend trying a different m.2 slot
Goodram ssd...
I know it's not great, alright?
That sucks. If you can't see the device in the bios then that's not good
It shows up as N/A
@rocky lily did you try a full power cycle?
Yeah, with no luck
Still doesn't work
You tried another slot?
Idk if I have more slots
What board is it?
(looks around to make sure Letterdief isnt around) ... Wow this new System Preferences layout is so much nicer to deal with than the old shit
Gigabyte H410M H V3
It doesn't look like it has more slots
Not unless there's one on the back which is unlikely
Nah, I checked the specs, only one slot
Are there more things I could try or should I just get it serviced
Unless you have another system to put the drive in, not much
Do you think it's a problem with the drive, or the board?
Hard to say for sure
Drive seems more likely since those four lanes are probably direct to cpu instead of chipset but ๐คทโโ๏ธ
Eh, I'll just get it serviced then, thanks for help
If it comes to buying a new drive, what good m.2 SSD would y'all recommend? I need at least 1TB, read and write speeds don't matter that much to me
Prices on 2TB and 4TB even tend to be pretty darned good these days
The Crucial P3+ 4TB can drop below $150 during sales.
I got my Acer GM7000 4TB for $280, I know it's gone down to $200 for Prime Day and maybe Newgg sales too.
I snagged my xs70 2tb for 80
youre always welcome
I've just seen a decent 512GB drive for 22โฌ
I paid 200โฌ for my 4TB NM790
As cheap as drives must be for OEMs, it's probably only for upsell purposes that 256GB drives even exist for laptops.
Happy with my Acer GM7000, one of the big things for me when looking up reviews was that it performs well even if it starts to get full.
There were also reviews specifically with the 4TB version.
I forget, but when it comes to SSDs, most of them all use the same few underlying controllers.

SATA or m.2
source ?
any website that aggregates shops/prices?
I'm on mironet now
Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.
budget and usecase?
-_-
for 4TB, 990 Pro
Crucial P3 also viable since it's 5666 instead of 7728
and for 2TB, SX8200 Pro or Kingston Fury Renegade
fury renegade is kingston
sx8200 pro is adata
ranges from shit to great
nope
These brands produce their own SSD tech instead of using standard components:
Kioxia, SK Hynix (Solidigm), WD, Samsung
idk about Crucial
so this killed access to my satisfactory server locally...
but I can still access it through a vpn.
so rule enabled.. access is blocked, even using external ip.. if I launch nord vpn that gets around it xd
@jagged snow I know it's semi late but I'm have odd operating system issues.
Trying to boot to normal windows yeilds a BCD error
Wile booting to safe mode works
Wile booting to winRE does not
0xc0000098
Run bootrec
In startup shell type bootrec /rebuildbcd
Kk thanks man
@jagged snow
It shouldn't say total idenified windows instillations 0
it fixed it tho
@jagged snow how the actual fuck
Is my hotspot still 110c
Is there somehow just not enough mounting pressure
Sounds like it, or you have a thermal pad that's torn and holding the Coldplay off the die, which seems more likely
That sounds like a bad mount of some sort
Don't overtighten
I'll take it apart again and see
What happened and respresd the paste
As I'm out of paste
Shouldn't cause issues as long as you have good coverage still
You want the paste to squeeze out
It's also odd that I had my BCD die wile doing this
How does one get their BCD damaged on a ssd
๐คทโโ๏ธ could be any number of things
Also
Found some MX4
Cleaned the die and added a prolly too much but that not gonna be an issue
@jagged snow
New bsod and I'm inna bootlop
How the fuk did I get here
Why the fuck do I have a CPU light on
Something is either failing or is gonna fail
@jagged snow BAD_SYSTEM CONFIG INFO
Sure sounds like it
That's not a good error to get
Fuck the moment I spend money elsewhere not on my PC I get massive hardware issues
Something is def fucked
I'm Guess ng somehow the mobo decided right now of all fucking times
Is the right time to die
Board shat itself
Oh? How do you figure?
CPU light repeatedly coming on and with multiple attempts to start yielding the same issue
Something has to not be going through and I ran a drive test
System drive is om
Can you get to bios?
Once in a blue moon atm
Try a cpu reseat
That doesn't explain the bad system config issue.
That's a drive based issue when the Drive isn't corrupted
It could
Bad data being written to disk, same thing can happen with memory
i kinda want to see ryzen with p/e cores ๐คฃ
It's a thing actually
Similar idea but not exactly big.little
big.little was a thing before intel did it
i don't think that'll stop them from trying
look at all the shit apple's sued over
will be fun to see what reviewers have to say about it
True
RX 7800 XT vs 4070 for satisfactory
Thinking 7600
Intel wouldโve sued Apple a long time ago if Intel had any chance of winning with big.little
If you have the budget x3d performs extremely well in this game specifically
also Iโm pretty sure Apple/ Qualcomm had e and p core on the market way better intel
Yeah waiting to see what happens on Black Friday sales
But hoping to get one
does anyone know of a chart that shows peak amp draw for different CPUs
iirc a review of my mobo basically said "don't throw anything too powerful in it, the caps can't handle it"
No but if you're really worried you can just set current limit in bios
[pumps 2.2V into 5950X]
"only" about 190MB/s.
Because the traffic is still coming from your local LAN even if you use external IP for the server.
With VPN it is coming from internet.
That rule is blocking all traffic from Default LAN to Game-Server, always.
I have no idea WHY you would set such rule?
Anyways, I really thing I now have FCLK, UCLK, MCLK and RAM timings down...
Still need to test if I can tighten couple of IMC internal timings that got exposed in AGESA 1.0.0.7 with the NITRO RAM training settings.
Damn...
Still need to run full testing between tRRD_S 6 and 4...
Having it on 4 causes max bandwidth regression, but I hadn't ran full set of actual benchmarks between those two at any point.
@twin dew sorry to ping you
but I'm fiddling with my 5900X setting
and I suspect my motherboard to do weird things with frequencies
so I'd like to set fckl manually though I'm not sure how it needs to be depending on ram setting
I target 3200 ram so I thought to set it to 1833 (I don't want to go above 1866 as I don't have oc board and I suspect I didn't win at lottery)
but I wanted to check how would fckl interact with ram frequency with your knowledge on the platform
as my 'safe' target for ram is 3000
I felt like 18xx may be odd
for "3:2" on 3200 ram it would mean 2066 fckl which is way too high
I want that network isolated from my local network.
wdym by isolated here ?
No traffic to or from.. if that machine gets compromised it can't be used to access my local network
then the whole machine should not be in your local network
Well, with that rule it's not
what rule ?
The one that's preventing me from accessing my own server.
keep in mind if the machine gets compromised no rule will put off unless it's port based
It's on its own vlan, and the rule blocks all traffic to and from all other networks
I mean physical ports
It's port based
then what do you want ?
To still be able to play on my own server.. I'll probably just use the VPN ig
Maybe I'm just being paranoid but
With AM4, normal would be 1:1 (so 1:2 when looking at the "effective" RAM).
And anything else causes latency problems.
So 1600MHz FCLK for 3200MT/s RAM.
Reason for the suggestion of 3600MT/s RAM as the max "should always work" FCLK is 1800MHz.
I'm stupid
The 2:3 is for AM5.
I wish they didn't do that
๐
if you use public address and you can't join your server
it means no one can
and the server can't access internet
because it can't communicate with your router
He just blocked traffic from his internal LAN to the other.
So internet to the server works fine.
But not from his own computer to the server.
And using public ip doesn't change that.
Because the router sees the traffic coming from the internal LAN, not from internet, even if the machine on the internal LAN tries to connect to the public IP.
it should change that
hairpining should make it
pc -> public -> server
pc <- public <- server
not
pc -> public (server)
pc <- server
No it doesn't.
It never enters the public address space when hairpinning.
I thought it would work.. but apparently the unifi is like nope
Hairpinning is just being able to use the public IP at all.
it enter the router sections
it should leave l2
When the connection is coming from the internal LAN, back to internal LAN.
It doesn't make public to internal rules apply, the internal to internal rules are what are used.
do you know what shape an actual hairpin have ?
it is implemented in the routing
But what is the whole point of that blocking blanket rule?
Just add exception for the ports that are needed?
not in the bridging
there should be no local ip to local ip direct transfer
i.e the server should not see it at coming from local
it should look at coming from router
the hairpining principle is applying nat rules on the wan interface loopback
I tried to add an allow rule before the block rule that would allow the local ip of my machine with ports 7777, 15000, and 15777 to the vlan.. it blocked it
or at least behaving like it is
The server might, depending on the implementation.
But the router doing the routing, NATting and firewalling probably doesn't, again can be implementation specific.
If the server and client are on separate VLANs with separate IP areas, then they will be routed anyways.
But I really don't understand what you are trying to do?
I've added it because that machine keeps getting probed from not nice places if I have certain servers up. Ib don't have any issues with satisfactory.. but mine craft and ats/ets keep getting probed
But then the rule is completely wrong?
Because you aren't blocking traffic from internet, you are blocking traffic between the two internal networks.
No, the server has to be internet facing
So it is the opposite of what you want.
Internet to game server is wide open.
Same for Internet to LAN.
just LAN <-> Game server was blocked in the screenshot.
This is what I have open..
Those are port forwardings, I was talking about the earlier firewall rule pic.
Firewall and Port Forwarding are separate.
You just need both hole and forwarding for the forwarding to do anything.
Was talking about this pic.
Right, so with that rule in place, I need to add a hole for my local machine to be able to play on it
Yes, because that is blanked rule to block all traffic between the Default Local Network and the Game Server network(?).
When I tried, it didn't work.. but it did just occur to me that after I added the rule, I tried to use the external ip. Maybe it didn't work because I needed to use the internal ip
The allow rule would need to be higher priority than that rule.
And there would be need a rule for the traffic in the opposite direction too, if the firewall isn't stateful.
Don't know enough about the specific implementation on whatever device that is for.
I'm not quite sure how to do priorities.. I'm still learning.
Usually the rules are just checked in order.
And there might be generic "Allow, then deny" or "deny, then allow".
If the deny blanked rules are supposed to be first or last.
You mean my network setup, and what I'm trying to accomplish?
But I'm not in a state today to read the manual(s) for how to configure whatever router you are using.
I'll probably have to use my pc.
Not the setup, just how the device logic works on the various things.
My router is a UDM-pro
Yes, but I haven't ever looked into the specific UI for configuring Ubiquiti routers.
And cannot do that today.
Ah, well I appreciate the input. It gives me a starting point.
If it trusted the people who play on my server, I'd probably just have them use wireguard and vpn into my network. I don't.
But if there is way to reorder the firewall rules you make, then you probably need to put the allow ones from the client to server in there above the blanked deny you added.
And by default they would end up below, when created later.
There's a giant firewall list rule page if i open it with a normal webbrowser, that's probably it.
The most common logic is just that traffic is checked against the rules from top to bottom, and first rule that applies specific Allow or Deny action to it is found that is it, rest of the rules aren't checked.
Usually with couple of rules turning the firewall into stateful one at the top.
NAT punchthrough etc. will not work at all with stateless firewall.
That's oddly specific behavior
Normally it will give the boot erro 0xc000001 first then 0xc0000078 next
AM5 has current bug that you cannot have Memory Context Restore enabled without RAM Power Down Enabled.
It will fail on second boot if both aren't set.
And the Windows Fast Boot can also cause issues with later boots.
Any other attempt results in the CPU light turning on and the machine not posting
I went though the steps to repair any possible damage to the operating system.
This is am4
Ok, because didn't get answer to the What machine part.
I still get the same BAD SYSTEM CONFIG INFO bsod. Then 0xc0000001 and shortly after trying to debug again 0xc0000078
Would try disabling that Fast Boot in Windows Power settings, and then Fast Boot in BIOS.
Already disabled.
Ok.
Still no avail.
So far now it's attempting to boot twice and then just throwing on the CPU light.
Need to think.
And look into old messages about this.
Brain Not Work Today.
Np take your time man
I won't have a PC for a few months if the board actually died anyway
That moment when you explain powerline adapters to people and they think you are some kind of a genius who invented an alien technology lol
Powerline adapters exist since early 2000's if I am correct
And have you tried without the CPU OC/UV or similar?
And RAM at auto?
Disabled all OC
Ram kit is at 2666mhz
CPU is at stock.
CPU light still comes on after two attempted boots
They've been around for a long time
Can't find the exact year they were first introduced in the market but I guess they were available for a long time.
Reset BIOS to complete defaults?
And what kind of drive interface in use for the boot drive?
SATA, M.2 SATA, M.2 NVMe?
Yes
Meanwhile, my sister's boyfriend was like: "Why has nobody ever told me about powerline adapters?" 
M.2 NVME
Also attempted using SATA on a separate drive to boot off of. No dice
Are Samsung m.2 SSDs good?
Always depends on the models. But from my experience (and this is only from experience) yea.
They are good, they just aren't price competetive most of the time.
980 is shit, 990 Pro is great.
Prices are almost always too high
980 competes with the cheap SSDs, badly.
980 Pro is top tier, but pricy too.
970 Evo is sometimes viable in EU.
2TB 980 Pro and 990 Pro dropped to 100 USD in the recent past, making them viable picks too
980 is just misleading, it's even worse than 970 Evo
(i gave up on samsung products awhile ago)
they are just riding their name now, and other brands have comparable quality for lower prices
Every Samsung product I've used has been terrible
Their flash and flash controllers are nice, most of the time.
ive got their washer and dryer ... still going ok... and their fridge... had to get them to replace it cause it crapped itself ....
I know one line of their phones was prone to random bricking because of a firmware bug in the flash controller
And no-one is making anything even coming close to this line:
semiconductor.samsung.com/consumer-storage/memory-card/micro-sd-pro-endurance/
The Kingston and SanDisk "equivalents" are about tenth the stated endurance.
ive got a samsung galaxy tablet ... its pure crap ... and then an SSD of theirs, which is overshadowed by all my other brand SSDs which were much less
their QLED TVs are nice too
QD-OLED too obviously
In sequential, or random low queue?
AFAIK Samsung is still the king of random read and write.
When discounting Optane and some other very high end stuff.
probably just overall performance tier
just in general ... its not like worlds apart better (or worse) but it definitely has no justified its higher cost
I have a Samsung monitor... it's one of very few I've ever owned that doesn't have a VESA mount. It takes DVI only. It's one solid piece of plastic with a stand, so impossible to mount or repair. The bezel is concave and reflective. The buttons are one of those stupid touch panels. Just crap all around.
by the time the things (any brand) wear out ... its time to replace them anyways because they are just outdated and not as useful anymore
got the C27FG73 back when 1440p was expensive and impossible to run, still one of my prouder investments - tho I don't recommend it any longer
Had a Samsung flip phone back in the day and it could do nothing right. Sometimes the modem would crash and it would just say you had no signal, but if you restarted it, magically you had the signal again. It would also crash about half the time you tried to take a photo
ive still got a super old double decades old viewsonic lcd panel that is still solid..... even IT has vesa mount holes lol
Also their phones today are full of ads
apparently my droid phone is a samsung ....
i never really cared or whatever... its a phone
I used a Galaxy Note 9 as a daily driver for 4 years. Switched over a couple weeks ago
From what I've seen their smartphones are wildly different from one model to the next
it had burn-in, battery was struggling and it had more and more hickups in software
my phone was also 'free' from the phone company, so i never really found any reason to complain about it lol
Still an absolutely awesome phone that served me well
yeah, sometimes you get something that just seems to carry on despite all signs say it should not be
bought mine used for 330โฌ - guy got it from his carrier for free, tested it against the Note 8 he was using, stuck to Note 8 and sold the Note 9
so pretty much brand new for me
nice
my father keeps sending me his extra iphones ... so, free again i cant complain about them
i just use them as tool devices ... no phone service
glorified iPod touches
Note 9 is still in service as 2FA device
incedently i still have and use my first gen ipod touch .... its hooked up in the bedroom and just plays nature sounds/music at night lol
I have a first gen iPad I need to find a use for. Surprisingly it doesn't seem like you can just install Linux?
its battery is shot, but it handles a power outage at least... it just stays plugged into a usb power adapter 24/7
i tried to use my galaxy tab4 for my DJI Drone ... but the damned thing wont install the Fly App ๐ฆ
by all accounts its powerful enough to do so ... but it just wont
I might just give up and sell my desktop at this point
what why
I don't have the money to keep dumping parts into this thing
so .... youd get rid of it then not have anything to use? at least you can still use it as is, right?
Completely unable to boot
oh
I don't understand how you and magic have so many problems with your hardware
I normally don't. Like ever seince I swapped ram it seemed fine.
For like a good wile.
Tho to be fair the mobo was used.
when i have hardware problems, it seems to always be just one part was total crap and needed replacing ......
And quite old
But you also had memory problems... that's really non normal
Oh that makes sense
So I don't think the memory problems are unusual for a used and old B450M
when i had memory problems, it wasnt the memory but the leads on the motherboard were scuffed/faulty .... lol ... took a bit to realize it, so replaced the mb, and the ram has been a solid champ ever since
My only memory issues came from dust in the slots
That and my GPU still is overheating for no reason. Screws are good and properly in. But hotspot still jumped up too high
turns out the leads ran 'under' the heatsink pegs for the CPU ... bad board design really ... the pegs had cut into them when screwed in
Could it be what you said with the thermal pads
Yes it could
That's why I said it
Once I get my machine booting again.
If I get my machine booting again.
I'll test that
you might end up with so many spare parts you can just make a second machine ๐
I sold my spare parts off lmfao
oh oop!
Thinking my machine seems to be running smoothly with the new ram kit
I may sell it back to my friend for parts and buy a laptop.
if i sold off all my spare parts, i could prolly build a top of the line rig with the most expensive series 14 and 4090 shit in it =p
Don't buy a laptop
hear hear
Not gonna use the laptop for much more then say satisfactory
Even worse.
I 100%ed cyberpunk so.
ouch, on a laptop?
No more super heavy game
No on my desktop
But I've seen some reasonable gaming laptops out there
I know the laptop world is so much different from desktop
people running cyberpunk on an M2 laptop emulated and saying 'this is fine' lol
But. I may give my desktop one last go around
See if yet another $160
Fixes the issue.
my laptop plays factorio, and oxygen not included ok ....
Hell I might buy one of those old latitude laptops
Plays zero games but basically will never die
Gaming laptop will cost twice a desktops price for same theoretical performance, and thrice for actual usually.
Because the "cheap" ones have bad cooling and more limited power levels in actual usage.
@twin dew I've been able to pull 3133 1566 Cpu:-5mV Soc:+5mV thank you
after >40minutes (splitted) of stress testing I have no error
and with almost 10mV less Vcpu I draw almost 15W more and have noticeably better performances
(going toward more than 1h of stress testing)
(also compiling chromium project)
Legion 5 for $900 with a 4060 out there recently. I don't think anyone is confused about the shortcomings of laptops, just don't pay too much for one.
But seems that it is better for me to use TrrdS 6 and not 4.
Even when the 4 works and is what the Buildzoid Hynix timings tell to use.
Which is very weird...
Damn
I tryed cancelling the order on my phone
Apparently it's a bit too late
So def Oh well.
You already bought a laptop?
But with laptops you cannot just look at the CPU and GPU.
You need to look at them in the exact config.
So the chassis including exact revision/cooling, exact CPU model, exact GPU model.
And the tester needs to try to run both CPU and GPU at same time.
Most only test them one at a time.
Used motherboard.. worked fine until you replaced the memory, assuming you don't have the old memory to test? If you did and it's still no joy, then my next guess would be the socket was damaged when you installed it, could check by removing all but 1 stick and moving it around.
Could do the same with new memory. If it doesn't work, maybe the new memory is bad? Have you tried to see if it'll boot memtest?
Board just throws CPU light whenever I try to boot @maiden coyote
Remounted CPU twice
Ram is not at fault
Replaced the two old kits with a good Kingston kit
I haven't bought a laptop
And my desktop will have to sit dormant now
And the errors aren't memory related
All the current issues relate to my drive.
Ran memtest86 on my kit.
The ram passes with flying colors
This was before this
So it's not the Kingston ram kit
So you haven't retested since the problems started?
I Can't retest when the machine is unbootable
So you cannot boot memtest now?
Board just gives a CPU light
Or Windows install media?
Or anything else?
And after reinstalling the CPU.
Still no dice
Reseated ram as well
Still no dice
Swapped to the old kit of 16gb
Ok, I had understood you got BSODs in some cases while trying to load Windows.
Still no dice.
And not that it is now degraded to not being able to start UEFI either.
It progressively got worse. Within a 4hr time frame
All the issues started and worsened within a 4hr timeframe
Often times it won't even show the splash screen
It just turns on the CPU light and halts
On the odd chance it does half attempt to boot it won't accept any input from my keyboard
So I physically cannot get into uefi
Remove everything but the minimum?
Including GPU if there is IGP to be used.
All drives, all USB, all other things.
Just PSU, MB, CPU, RAM and GPU if there isn't IGP.
If the M.2 is still in, it isn't at minimum.
If keyboard and mouse are still in, it isn't at minimum.
I pulled the m.2 out
Mouse and keyboard are still in so I'll do that in a hot sec
Keyboard and mouse are out @twin dew
Do I attempt to start it?
Debug lights all flashed once in the normal order and then CPU light hung
So CPU or MB.
Most probably CPU.
Should I remount again.
I still have that degraded 5500
Do we wanna give that a shot
What would it get?
Remove and check the pins maybe?
And try with that other CPU then, it would close the MB out mostly.
I did check pins last time I remounted all of it looks okay
I might take my machine to my buddies place and use his machine to debug and test my parts
Pins look okay on the 5600. Mounted the 5500. I'll try to start again.
CPU light hung
@twin dew
So probably the MB then.
Yeah
I have heard normally when a mobo goes it goes in a catastrophic fashion.
And this is certainly catastrophic. 4hrs was all the time it took to devolve from booting fine to completely unusuable.
Not always.
Depends on what fails.
That would be the case if some VRM mosfet or like shorted.
But lot of other options too.
So any ideas what failed.
On the mobo.
Just for shits and giggles
Pulled apart my GPU one last time @jagged snow and the paste had good coverage
Odd that it was still hitting 110c hotspot and only 60c edge temp
The die was properly covered
I'm kind of learning towards cpu too. It's possible the memory controller on the cpu was dying, and now it's dead
But the other CPU (but that was damaged) didn't boot either.
The other CPU could still boot. Normally. Just was degraded beyond daily use
Probably bad mounting order, causing the cooler to sit crooked.
I'll fix it when I get the PC repaired
No point in fixing something I cannot test
I go one screw in one corner then opposite corner then the Corner adjacent to that one then the opposite
So bottom right, top left, top right, bottom left
I start them before tightening
And I tighten in that order
Needs to be properly sitting first, and then careful tightening turn or few turns each, in pattern like that, until all are tight.
All holes were aligned and it clicked into place
www.cnn.com/2023/11/05/tech/nvidia-amd-ceos-taiwan-intl-hnk/index.html
It's cnn. But now AMD vs Nvidia Is a family feud
Could be that there was still bad mount... how thin was the layer of paste?
Can you send a pic?
Anyways, this seems to be the end of the line for the RAM tuning.
Unless I get some weird issues in longer testing.
Very nice
Not the best possible, but limited by the lowish voltages I'm willing to use.
65 passes of Memtest86:s Bit fade test last night to test that lowered tRFC.
450 doesn't even boot.
450 would have been 150 ns.
The paste was on decently think when I put it on
I'll send a pic after lunk
Lunch
I guess setting up a proxmox vm with pcie passthrough and usb passthrough isn't super straight forward
Fuuul
I just remembered that I have a OEM key
And the only way to transfer it
Is to go onto that install and get it to release the ley
From that board
Orm keys can't be transferred
They can
You have to do bunch of stuff to get it transferred
That's how I transferred the key to this board
Last I tried they wouldn't do it
Something along the lines of getting the board to report that the key is not in use essentially deregistering the PC.
And then using the key on the new machine
But you can't do that when the board is inoperable
That only works sometimes...
I've done it a lot and it's not reliable
I got it first round when going from a 3770k to a 5500
I also know a dude who knows how to use some Microsoft enterprise panel to basically fake register your pc
Without ever needing a key
So far with the one try I have used it for
It was perfect
I've had about a 75% success rate
what are windows keys obviously if you upgrade your motherboard, or buid a pc, you must give microsoft your money. That is the honest thing to do.
Fuck no
Im sorry Microsoft got enough money from me
They don't need another $160 for a win 11 pro key
Well.. the dishonest thing would be google search massgrave activation. We're all honest folk here though
Proxmox windows 10 with pci and USB pass-through is working... installing gpu drivers is another can of worms
Amd driver said nope not today. Told me it's not supported.. said something about embedded drivers, oem drivers and something else.. gpuz reports card correctly, I probably misconfigured something
@jagged snow
I am safe to assume I put more then enough paste on
And it spread properly on the cold plate so.
Why the actual fuck are my hotspot temps still jumping to 110c
I did get a spare mobo from a buddie of mine
Bruh ๐
How many tubes of paste was that???
a good quater /j

And why does it look like play-dough
its MX4
takeing the termal pads off Changed the thermal runaray from 60C delta to 30C delta under load
Don't take off the thermal pads!!!
Replaced with thermal paste
on the vram
its messy but did that on the 580
Helped alot
this nas is advertising โpowerful hardwareโ
then lists 1gb ram, 1.7ghz cpu, and 114mbs read/write as its selling points
those are the selling points they chose to highlight
why
That will not work
with a 2 bay nas would raid 1 mean iโd essentially have the capacity of one bay, with the other being purely redundant?
iโm reading the description of raid 1 and i think thatโs what it means, but iโm also dumb
Yes
ok good to know
Raid 1 is just a mirror
ill run a vram test
would raid 1 in a 3 bay nas mean you still only have the capacity of one bay, but with two sets of redundancy?
Yes
Vram at 60C
ok thanks
Bad idea, your vram will run hotter because of that.
With the pads on My Hotspot Deltas 60C
The gpu heatsink has at least .5mm between the memory chips and the heatsink
It will run hotter.
Put the pads on
Hotspot is still out of control
Could it be the MX4
Not handling the 230w
Imma get rid of the sspt
See if running fully stock helps
Yeah
That fixed it
The MX4 is much worse then the thermalgrizzly
hotspot still out of control
Ill get my gpu serviced
So your other issue was the motherboard?
running a spare from a buddie of mine
Windows is yelling at me But at this point ill forge the key /hj
i keep putting off switching my OS over to a new drive to make it recognize that i do indeed have the proper hardware to switch to win11
raid 5 in 3 bay would be the 2 bay capacity with 1 of redundancy
(if same capacity across all bays)
raid 2 3 and 4
are just low performance old raid 1 or 5 versons that are deprecated
hence the naming scheme
Usually you can call the MS registration support and say that MB failed and was replaced, to be able to register the new MB.
And have you ever replaced those pads?
Because if you did, you might have gotten too thick ones.
And if you note, the GPU has small components around the GPU die, which you have smothered in paste...
Even if the paste is "non-conductive", it can still cause issues.
And as the GPU coolers usually don't have lot of force when thightened, too much paste causes issues because the extra doesn't get pushed out and there will be too thick paste layer.
I cleaned it off again
Those are the OEM thermal pads
It's still delta 30c under load for hotspot
Unlikely, I used MX4 with a 3090
It is an old tube
Looking at the pic, it seems like it's not quite making tight contact
Screws are tight
Not overly tight
And order matters, and as the full tighten has much less force than CPU coolers, the paste will not spread out enough.
If you want to spend way too much on smth, could buy a PTM7950 pad, but it doesn't seem like it's necessary.
Bottom left,top right, top left, bottom right. is the order I use
But at once, or in turns?
Start them first then tighten in that order
And like I said, with that much paste, it doesn't matter WHAT you do, it will not be good.
And if you go from just contacting to full tighten one screw at a time you are fucking up on that.
PTM7950 is a thermal paste that's not liquid at room temp, highly regarded in the OC community for use with GPU dies.
Yes, I know.
Expensive af in comparison to most thermal solutions
I do three turns for starting then full
Because there isn't official retail availibility.
So completely wrong.
Get them started to same amount, then max one turn each at a time, in that criss-cross pattern.
I'll tear it apart again this morning.
All the PTM7950 you can currently get is someone buying fuckton from the manufacturer and then manually cutting and selling small parts.
Damn
Because the minimum order is lot.
I'd still expect cheaper prices on aliexpress, but ig not
Because there isn't cheap Chinese source they could sell.
Currently considering building a custom keyboard with high quality parts from aliexpress, funny thing is that the parts are cheaper in the US lmao.
Or maybe similarly priced after tax and shipping
But the form the PTM7950 is sold in general is on reels:
And how you buy is "Contact Us", direct with Honeywell.
And the paste variants also come in cans of maybe liter or so each?
Availability
PCM pad is available in both sheet and roll formats.
Thickness Range: 0.2mm-1.0mm
Stencil printable (SP) material is available in 300cc syringes or 1kg jars.
And the information even on what are the "standard" sheets and rolls aren't public.
You need to contact to get the info.
With the 7950 only available in 0.25mm base thickness.
But anyways, normal way if you wanted some is for your companys buying department to call Honeywells sell department, then talk about what form you need, and how much each year you will need, and then agree on price and constant delivery schedule etc.
Damn... At some point my AGESA CBS changes had been reset to all auto, so ECC might have been on...
IIRC Auto for that was ECC disabled in older AGESA versions, but don't know about 1.0.0.7 variants...
Noticed when went to re-enable it to run same test suite as before to confirm there wasn't performance change.
But anyways, set ECC off, booted to normal Windows and now set it to run Y-cruncher for some time.
And later need to set ECC off, boot into safe mode again and run the test set to get comparative numbers to previous ones, to see there aren't any regressions.
Damn you AMD and your incomplete AGESA MCE support!
I need to install Win11 at some point and see if that is because MS hasn't updated Windows 10 WHEA to read the AMD Zen 4 (Family 19h) MCEs.
No, Zen3 and Zen3+ are in the same family.
So the MCEs are the same.
Zen 5 will be 1Ah (h for hex in both)
i know it's clearly a thing, but it's hilarious to see reels of thermal pads
it's like the first time seeing one of those giant tubs of thermal paste
oh god it's honeywell, there's a gigantic plant near me
But company that has no desire to sell in retail.
oh definitely, it'd just be fun to get a tour & see what industrial volume looks like in person
I'm pissed off, I was searching for some tool to check RAM latency, and all I found is people "having the answer" .. pointing at reading spd
I dont want to read fucking manufacturing informations
I want to check the actual RAM latency, which is not even in spd obviously
looks like it's an art to mis the questions
is there linux versions though ? and last time I checked aida it was all in paid versions (though I guess you bought the full version looking at all the test you do)
Example:
Aida has Trial version, which shows some of the info in that test.
IIRC for RAM the Trial shows the Memory Read and Memory Latency.
But as whole, Linux side has much less OC related software.
This one would probably work.
github.com/ChipsandCheese/MemoryLatencyTest/releases
But the results between the different softwares aren't comparable.
Should work in linux
@wanton orchid Perhaps a bit too sophisticated, but you can try it
Rip
Thermalright seems to have started to buy the stuff and sell it as Thermalright Heilos.
But Thermalright doesn't seem to be that interested in selling it either.
As the official retailers/importers in EU cannot get the stuff at this time.
*ooh* some have now listed those with prices, but unknown delivery dates.
Thermalright listed it in end of july.
And it seemed to be available for sale in some places in Asia back then.
Thermalright really baiting us rn
Thermalpads, Frozen Vision, etc
At least their PSUs are somewhat available by now
Goated company as long as you can get their products
But ProShop has both sizes (30x40 for Intel and 40x40 for AMD) listed at 7.99e.
But no timeframe when they will actually get any.
Computer Orbit has them listed for ยฃ5.00, same with unknown delivery.
Thermalright doesn't seem to care much about the marked outside of limited area in Asia, unfortunately.
They just design and produce shit, everything else is secondary
AFAIK they have pretty good retail presence in parts of Asia.
But like shown here, even the official partners in EU and US have major difficulties actually getting stuff from them.
Jimm's PC Store is the official distributor for Finland.
They have 12 Thermalright products listed at this time.
All are in some way out of stock.
3 are "Available from next step in chain" type result, with week to month delivery time estimate, so probably direct from Thermalright.
6 are "Ordered" with no delivery time estimate, and the orders are basically already sold out.
AM5 Secure Frame has 8 coming, 7 of those already sold.
Peerless Assassin non-RGB has 16 coming, 9 already sold.
RGB one has over 25 coming (largest number the site shows), 12 already sold.
3 are "Not in stock", so not even orderable from Thermalright, and that category includes the LGA1700 Contact Frame.
And they have been trying to get those Heilos pads for months now.
Not listed at this time, because they have no idea if they will EVER get them.
And that is the company Thermalright site lists if you go to the "Where to buy" for Finland, for both Distributor and Reseller sections.
the link of Baldur says it is a fork of the clamchowder project
Ah
I didn't check what Baldur sent, just what I remembered as a free aida alternative for ram testing.
Use that then, maybe itโs a bit more user friendly
I tried but got "i/o error"
maybe it's a gui app ? because I wanted to run it headless
That Golder Soldiers version is GUI over some other Clamchowder stuff.
Non-GUI set:
https://github.com/clamchowder/Microbenchmarks
But for those, you need to compile them yourself.
Which is why I linked the specific fork which had compiled version.
But like always with Linux compiled stuff, if it isn't static linked, it might not work on YOUR linux.
*Oeh Noes!*
Enabling the ECC seems to be causing 0.6ns latency penalty for RAM!
Four runs of AIDA64 in safe mode were all 60.4ns without ECC and three runs have all ended at 61.0 with ECC.
And fourth run was 61.0 too.
And it has been pretty uncommon to get that tight groupings for the RAM latency.
And just being in the normal Windows caused it to have about 2ns variance in it between runs for the Aida64 test.
Hmm... would need to do the same test with much lower FCLK too to confirm.
Might be marginal instability that is being corrected without reporting.
And would only show up as major performance loss once too bad and as reboots if there are uncorrected errors.
I suppose you are benching it without ECC for error spotting reasons ?
so if it's so low overhead you probably want ECC back once you got stable settings ?
missing Meshsim, they probably removed some projects and didnt update makefile
looking fine thank you for sharing the links
Because ECC on would have allowed me to get tighter timings than without, because AM5 doesn't support REPORTING ECC errors for RAM yet.
That is the point of this line.
That AM5 doesn't support reporting corrected RAM or Infinity Fabric ECC errors.
And just reboots on uncorrected ones.
yep but I mean you did successfully disabled ecc, and you can probably safely re-enable it right ?
that's what I was asking/supposing
Yes, I just ran ECC on.
But as that enabled 2 more RAM dies on each rank, that means I enable 8 more RAM dies, that haven't been tested at all before that.
And if the any of those is the worst part, then I will have constant corrected RAM ECC errors with ECC enabled.
And possibility of uncorrected errors too.
Each DDR5 ECC DIMM has two extra RAM dies, each used only in half.
fuck you linux, fuck you github, why do it have to be like that, compiled succesfuuly
DDR4 and earlier were 64bits (8x 8bit die) for normal and 72bit for ECC with 9x 8bit die.
DDR5 is 2x 32bit (4x 8bit die) for normal and 2x 36bit for ECC, with 10x 8bit die.
Would be SO EASY if the MCEs for RAM and IF worked...
it would be nice that you can't have false positive correction
how would you get uncorrectable error with valid direct channels ?
If the ECC die has dual bit error.
When the ECC dies weren't part of the testing so far.
Because if I had ECC enabled, all single bit errors would be invisible and have been corrected.
Before causing any problems.
(5x 8bits is 40bits)
Yes, half of the ECC dies isn't used.
ok
But need one per channel.
You only still have 72 bit data lane even with DDR5 on ECC mode, split into two independent subchannels.
you mean that instead of just working it will reset system for "memory failure" ?
if the ecc bits get wrong
Yes, currently, because the error reporting to OS isn't working in AGESA.
and you have no means of checking the Ecc bits are consistent with the already checked bits ?
๐
I thought it had 2 more bits in ecc than ddr4
using already wired lanes
ECC off: IMC uses 2x 32bit channels, no correction, IMC cannot notice errors.
ECC on: IMC uses 2x 36bit channels, single-bit correction, multi-bit detection. IMC can notice errors.
Normally if there was single bit error, it would be corrected, and IMC would set some CPU registers, that the OS would read and log that RAM ECC error happened but was corrected.
And on multi-bit error the data would be scrapped and IMC again would notify the OS that multi-bit uncorrected error happened, and usually OS would react, either by killing the specific program the failed data read belonged to, or by BSOD or kernel panic or like.
But currently AM5 doesn't support the reporting, and just corrects single-bit errors and doesn't report them, and automatically reboots the system on uncorrected error, with no warning to OS.
Trying to find DDR5 pinout to check...
micron ddr4 v ddr5 presenttatin talks about 80 bit channels for ddr5 instead of 72 for ddr4
Because multiple places talk about 40 bit per channel, but AFAIK it isn't 80 bits total in reality even with ECC, but 36 bit.
At least Micron DDR5 UDIMM core spec talks about two 36bit ECC subchannels.
And all the ECC DDR5 has 80 bits of chips, but AFAIK the pinout only has 72 bits of data lines.
it could have more as the pin layout of ddr5 is different that ddr4 one, even if they use same physical layout iirc
Yes, but they already had to find pins to split them into two subchannels.
the on-board active talking pmic and related components removed quite a lot of pins use
Ok, looked on more and Micron does have both x72 and x80 ECC DIMMs?
Trying to find the datasheets, those early Micron datasheets for DDR5 only talk about x72.
they removed a freaking lot of potential external crosstalk with the new power and grounding layout
On that Micron DDR5 SDRAM UDIMM Core PDF:
the micron doc I was looking at
But need to find if they have newer version somewhere.
btw I just noticed as vddq is now managed by memory package
isnt it a bit tricky
Ah, yours is about RDIMMs and those seem to have 80 bits?
Remember that DDR5 UDIMMs and RDIMMs use completely different pinouts.
And that shows just x72 at 2x 36 bit like I talked about.
yep
Never looked into RDIMMs, because those weren't relevant to me.
hell expensive pltforms ๐
Yes, the extra bits allows for more complete correction.
on Each channels thats a lot beefier correction that ddr4
oof, my mothers LG TV has all sorts of blue burn-like spots in the corners ... quick google suggests her backlight strips have gone bad ... i wonder how difficult it would be to replace them (and how much those strips are, hmm)
with tv's being as cheap as they are, theres a threshold of time/part money vs 'just buy a new tv' lol
waiting until 8995WX EC8 overclockable watercooled build
the most expensive of all time
but the most destroyer of all usecase wise
Even DDR4 ECC can do multi-bit correction in some cases, if the IMC supports that.
But depends on what bits are corrupted.
Any single-bit can be corrected, some multi-bit combos can be corrected, some cannot.
EC8 means any 2 bits are correctable, or at least levelable
And some IMCs only support correcting single bit errors, and no multi-bit correction ever.
Because the extra data lanes and chips are passive, and they just store and retrieve extra data for the IMC to use however.
you have grey area you can choose to correct or not
and the "uncorrectable" state is harder to get wrong
Yes, probably.
But was talking about the DDR4 ECC with 64 bits of data and 8 bits of ECC.
That even that can do some multi-bit correction already on later CPUs.
bit if they can on that , they can better on ec8
Yes, more extra bits, more complex algo you can use.
I meant that on EC8 you really have a clear insight on the error state, it's not a bet, or very much less a bet
To save the data in rebundant way.
with EC8 I believe you can choose to correct or discard on multiple level (depending on data critical score), and uncorrectable error or correctable state have very much less odds of being false negatives
it becomes near impossible to get a corrected uncorrectable error
Anyways, set the computer to run Y-cruncher with 30GB RAM usage on 8 threads (one per core), and Linpack Extreme 30GB RAM stress test on 8 threads (so they sit in the free HT cores).
And will let it to run that combo until next morning.
With ECC disabled.
I will later run ECC on tests too, but for those you can only wait to see if the computer reboots at some point or not.
Until AMD adds the missing MCE stuff to the AGESA.
CPU cache ECC errors already emit MCEs like they should even on AM5.
What do you guys think I would need to run the game high 60+ fps
Medium to big factory
Depends on save size.
From potato to unobtainable depending on the save size.
If you mean Satisfactory, and not for example Factorio.
Because for Satisfactory related, you are on wrong channel.
This is part of Off Topic channels.
But cannot really say, you can get Satisfactory optimized computer, but as the game performance is so save details dependant, not easy to aim for specific performance.
Fair I will get a 1660 super. If anything will upgrade
The problem is more on the CPU front.
But 8GB would be recommended for GPU VRAM.
4GB is current official minimum, and very near to actual minimum for Low textures.
8GB is currently enough for Ultra textures, if you don't use too many unique signs, or some other new to U8 features too much.
I wouldn't get anything below a 5600XT for gaming in 2023
5700XT is great value used
5700x
5700xt
you can't get it wrong
(joke)
me with dual _800XT
I wouldn't
I wouldn't spend any money for iPhone 15
But my recommendation for you depends heavily on a lot of factors
The only reason I'd ever recommend a flagship phone is if it comes bundled with your carrier's plan
lte will be around for a long time yet
@charred pewter update on the GPU: UPS denied the seller's claim, I'm returning the card for full price. It worked out fine in the end, but took two weeks.
I may end up investing that cash into a new macro setup/replacement camera body for my dslr
so the seller basically lost money on it huh
Unfortunately
and yeah ... UPS ... what a ๐ฉ show
I'm assuming that the card will be saveable with a fairly simple board-level repair but I didn't want to attempt it and risk losing my opportunity for a return
our UPS Store is now charging $1 to simply be a 'drop off' point for amazon returns. they dont pack it, they dont even print a label... you have to do it all before you get there. yet Staples down the road will just take the loose items as is, no packing, no label, free return
ive lost all respect in UPS over the last couple years
Oh weird... my "local" ups store(30 miles) accepts amazon returns for free and they do the packaging and everything
You just bring the item and they scan the QR off your phone