|
Come THE gently caress ON, AMD.
|
#
?
Jul 12, 2022 01:09
|
|
|
|
| # ? May 29, 2024 23:56 |
|
|
SwissArmyDruid posted:Come THE gently caress ON, AMD. this has been happening on my old intel system, so. Come THE gently caress ON, MICROSOFT.
|
|
#
?
Jul 12, 2022 03:40
|
|
|
but what if the usb device really did malfunction?
|
|
#
?
Jul 12, 2022 03:56
|
|
|
Rinkles posted:These have to be state sponsored attacks, right? Dear Leader needs a new gaming rig.
|
|
#
?
Jul 12, 2022 04:13
|
|
|
BrainDance posted:but what if the usb device really did malfunction? then about five minutes after the first time it happens on my new computer, I am ordering a new DAC and changing all my reviews.
|
|
#
?
Jul 12, 2022 08:55
|
|
|
Many people here are very interested/concerned with the vcache consumer chips. Here's Phoronix with a look at the current state of the Epyc 7773X (64 cores, 768MB L3, 280W). There've been some kernel and BIOS improvements for the platform since release, which is what this review focuses on. The results are, as you might expect, very workload-dependent. In the ones where it shines, there's some good old-fashioned Intel shaming, with one EPYC literally doing the same work as two Xeon Platinum 8380s (aggregate of 80 cores and 540W). https://www.phoronix.com/scan.php?page=article&item=amd-epyc7773x-redux&num=1
|
|
#
?
Jul 13, 2022 02:34
|
|
|
wargames posted:this has been happening on my old intel system, so. This happened to me last week on an old Intel system at work, too. I figured it was from having too much bullshit in the USB ports, and got things working after plugging some stuff from the faceplate into the rear ports, but it's weird to see an error I don't remember seeing before now happen three times in this thread.
|
|
#
?
Jul 13, 2022 04:24
|
|
|
SwissArmyDruid posted:Come THE gently caress ON, AMD. wargames posted:this has been happening on my old intel system, so. yeah i kept getting this when moving to ryzen with a MIDI keyboard but it works perfectly on linux lol just vendors being absolute loving shits i bet, microsoft included
|
|
#
?
Jul 13, 2022 12:59
|
|
|
Unbelievable in TYOOL 2022 that Linux has better USB performance than Windows. Could this finally be the year???
|
|
#
?
Jul 14, 2022 01:37
|
|
|
NewFatMike posted:Unbelievable in TYOOL 2022 that Linux has better USB performance than Windows. Could this finally be the year??? USB was a mistake, next gen chipsets / PCHs need 10s of 16550As and parallel ports. Or make SuperIOs great again.
|
|
#
?
Jul 14, 2022 05:09
|
|
|
movax posted:USB was a mistake, next gen chipsets / PCHs need 10s of 16550As and parallel ports. Or make SuperIOs great again. Throw FPGAs on everything and have their high speed serial transceivers negotiate something. Just think of the tech support questions like “Why does my video card come up on 100G Ethernet?”
|
|
#
?
Jul 14, 2022 05:17
|
|
|
The only thing that fixed my USB issues is disabling c-states in the bios which is probably bad for some reason I don't know about but hey at least my usb devices stopped disconnecting and I no longer live in fear of hearing a usb toggle sound in windows. Whole thing just made me want to go back to Intel though.
|
|
#
?
Jul 14, 2022 05:21
|
|
|
You can run into the same issues with Intel too unfortunately. I have a older Ivy Bridge era Intel based PC that has similar issues with USB and power save mode. Its perfectly stable and runs for months between reboots but power save and sleep modes flat out don't work at all no matter what I do and it has random USB disconnects/reconnects all the time. Always has no matter what I do with fresh OS installs, drivers, updates, or BIOS options. My current AMD system doesn't have these troubles but it seems to be totally luck of the draw on whether it happens or not no matter what you buy.
|
|
#
?
Jul 14, 2022 07:16
|
|
|
Speaking of, Where *is* USB 4? I imagine that if Apple were still on Intel silicon, they'd have USB 4/TB4 macbooks and the like, but USB 4 has been.... weirdly quiet? I think the only thing I've heard was something earlier this year about how USB 4 latency is absolute garbage and less suited to eGPU than TB3 was.cheesetriangles posted:The only thing that fixed my USB issues is disabling c-states in the bios which is probably bad for some reason I don't know about but hey at least my usb devices stopped disconnecting and I no longer live in fear of hearing a usb toggle sound in windows. Whole thing just made me want to go back to Intel though. ....that's exactly the problem, though. I don't hear the USB toggle sound... BECAUSE IT'S THE DAC THAT CRAPS OUT.
|
|
#
?
Jul 14, 2022 08:23
|
|
|
Just lol if you don't have all Windows errors/warnings trigger the motherboard beep. I don't know if motherboards even can beep anymore
|
|
#
?
Jul 14, 2022 12:34
|
|
|
Kibner posted:Just lol if you don't have all Windows errors/warnings trigger the motherboard beep. most motherboards I've encountered still have a pin-header for plugging in those tiny internal speakers so they can give you a POST code-beep, but they don't come with the speakers anymore, so you have to source those separately (sometimes they come with the case accessories instead)
|
|
#
?
Jul 14, 2022 12:42
|
|
|
which imo is kind of a funny given the increasing ubiquity of little mobo displays for post errors or whatever. if we hadn't gotten rid of the little beepy guy we wouldn't need it!
|
|
#
?
Jul 14, 2022 12:52
|
|
|
CoolCab posted:which imo is kind of a funny given the increasing ubiquity of little mobo displays for post errors or whatever. if we hadn't gotten rid of the little beepy guy we wouldn't need it! The Dr debug leds are more convenient than trying to look up beep codes though. Although it feels like even though motherboards are getting more expensive some of those useful features are really going into just the high end boards. Like the x99 workstation board I bought years ago has the dr debug led, power and reset buttons and a bunch of other stuff. The x570 strix board I bought which is one step down from the Crosshair boards only has a dr debug. I think even my old Z97 middle of the range matx board had more features than the x570 one.
|
|
#
?
Jul 14, 2022 13:09
|
|
|
cheesetriangles posted:The only thing that fixed my USB issues is disabling c-states in the bios which is probably bad for some reason I don't know about Means your CPU is using more power while idle. C-states are the power save feature where the CPU progressively turns off various bits of silicon. (Running at higher or lower frequency is P-states.) There's a buildzoid video where he fixes some severe USB hitching with c-state. But the thing is, you can look at what's happening with his poo poo and see that it's a different USB problem. Like, that's not random occasional disconnects, it's progressive complete failure. He's saying it's happening on the USB2 ports, other people say they have problems specifically with 3. That buildzoid vid is showing something super defective, RMA the CPU/mobo type issues. PC LOAD LETTER posted:You can run into the same issues with Intel too unfortunately. It's such a hosed situation, I wonder how many people have become conscious of things that may have happened with other PCs. It's like microstutter in games: that was happening for ages with all SLI systems and nobody was complaining about it. Then TR and PCper write articles about the phenomena and suddenly everyone sees microstutter (or thinks they do). SwissArmyDruid posted:....that's exactly the problem, though. I don't hear the USB toggle sound... BECAUSE IT'S THE DAC THAT CRAPS OUT. If you're trying to troubleshoot by tweaking various BIOS stuff: you could try running USBlogview, which is a really light app. That'll catch events when windows would show a full disconnect/reconnect message. OTOH if that doesn't log anything when you experience the problem, you'd have to step up to event trace logs which are way more complicated.
|
|
#
?
Jul 14, 2022 14:27
|
|
|
It's me, I'm the one who bought a pack of PC speakers and installed one in my desktop because desktops are supposed to go "boop" when you power them on. I'm pretty sure it's the law.
|
|
#
?
Jul 14, 2022 14:36
|
|
|
My computers are supposed to play thunder.wav when they load Windows. It was installed by some Creative soundcard as a startup sound ages ago (in Windows 98?) and since then I've been dragging the audio file along for bootup/login sounds.
|
|
#
?
Jul 15, 2022 09:38
|
|
|
https://www.youtube.com/watch?v=Nd9-OtzzFxs so this piece does benchmarking into running games with background tasks to evaluate the relative performance of adding additional cores on ryzen 5000 with the 5600 and 5700x as to extrapolate over the lineup to express the idea: more cores for gaming is still not overwhelmingly rational. this plays to my personal bias on the issue ie that's what i would like to find evidence of, so i am instinctively distrustful of it and want to pick apart the methodology a bit before i use it when talking to people in eg the building thread. the results are not what i expected - in a handful of scenarios as you would expect the 5600 loses some 1% lows, in the worst case as much as 10%, but never imo to a particularly significant level. and in some cases the gains seem to actually revert - the 5600 and 5700x equalize when doing some heavy task is busy which seems bonkers and backwards. first i know our paul has some issues with HUB and steve in particular using super highly tuned DDR4, i can't think that would be super relevant here but maybe? maybe their test systems are too clean and they're not getting representative just random background tasks deciding to all spike at once and intensify that 1% low behavior? is a 1080p 60 youtube video not a realistic backround scenario, should he have been making discord cause a bunch of notifications maybe? i'm scratchin my head.
|
|
#
?
Jul 15, 2022 13:03
|
|
|
my extremely kneejerky reaction to that post is that I've seen my Task Manager hit over 16 gigs of RAM usage, which meant I would have gone over had I not upgraded to 32, and that seems to be a more "real" problem with having lots of background tasks than running out of cores
|
|
#
?
Jul 15, 2022 13:16
|
|
|
crosspost:CoolCab posted:oh, followon to our chatting about ddr4/5 the ryzen 5800X3D, best AM4 socket gaming chip by a bit and a contender for best available gaming CPU although that one would be with a pretty big asterisk, is ddr4 only and doesn't really require insanely tuned ram to put up those numbers. it's a very strange chip in a lot of ways (plz amd make a 5600X3D i'm so ready) and it's VERY, VERY title dependent, but 3600 CL16 stuff which sidebar i would still call high end but goes on promo dirt cheap performs almost as good as extortionately expensive 12th gen DDR5 monsters like the 12900KS, better in some titles. coward 
Paul MaudDib fucked around with this message at 14:40 on Jul 15, 2022 |
|
#
?
Jul 15, 2022 14:38
|
|
|
Paul MaudDib posted:crosspost: :O you must be getting some bottlenecking surely
|
|
#
?
Jul 15, 2022 15:37
|
|
|
SwissArmyDruid posted:Speaking of, Where *is* USB 4? I imagine that if Apple were still on Intel silicon, they'd have USB 4/TB4 macbooks and the like, but USB 4 has been.... weirdly quiet? I think the only thing I've heard was something earlier this year about how USB 4 latency is absolute garbage and less suited to eGPU than TB3 was. Add a PCIe USB card, Renasas. Done.
|
|
#
?
Jul 15, 2022 15:40
|
|
|
CoolCab posted::O 5WB means that’s B-die and Samsung only sold DIMMs at JEDEC speeds (eventually up to 3200). Third party DIMM manufacturers are the ones that bin into “overclocked”/XMP speeds. So, it might be good over DDR4-4000, cl14 and 1.55V or it might be a “dud” and only do the rated DDR4-2666 at 1.2V. edit: 1747 is the date code right? November 2017? So zen 1? Maybe you did run it at DDR4-2666  Technically that was 3 years into the production of DDR4, we’re less than 2 years into the production of DDR5.
hobbesmaster fucked around with this message at 15:50 on Jul 15, 2022 |
|
#
?
Jul 15, 2022 15:44
|
|
|
the fun factor is that all of the box clocks are specced at 1.2v memory voltage, it should go a decent bit faster with 1.35v or so. Edit: also yes didn’t b-die production end like mid 2018? Edit: maybe not? https://m.youtube.com/watch?v=orOSDYk6aao Paul MaudDib fucked around with this message at 15:56 on Jul 15, 2022 |
|
#
?
Jul 15, 2022 15:52
|
|
|
A 7600x was apparently spotted in the wild and its benching about 10% harder than a 5950x.
|
|
#
?
Jul 15, 2022 16:14
|
|
|
Samsung completely stopped selling DIMMs to consumers iirc? The ICs still show as “mass production” but that doesn’t mean they’re running as many wafers as say C die. Cao Ni Ma posted:A 7600x was apparently spotted in the wild and its benching about 10% harder than a 5950x. That unfortunately doesn’t mean much without more context. That could either be very impressive or somewhat disappointing.
|
|
#
?
Jul 15, 2022 16:14
|
|
|
Paul MaudDib posted:crosspost: Is this an ECC memory stick, too? I see that it has five chips on one side.
|
|
#
?
Jul 15, 2022 16:18
|
|
|
Are there 5800x3D benchmarks with ddr4 2400? I wanna know how it would perform with my poo poo-rear end Kingsoec memory. Would it be slower, or would the cache make it not matter?
|
|
#
?
Jul 15, 2022 16:22
|
|
|
CoolCab posted:so this piece does benchmarking into running games with background tasks to evaluate the relative performance of adding additional cores on ryzen 5000 with the 5600 and 5700x as to extrapolate over the lineup to express the idea: more cores for gaming is still not overwhelmingly rational. this plays to my personal bias on the issue ie that's what i would like to find evidence of, so i am instinctively distrustful of it and want to pick apart the methodology a bit before i use it when talking to people in eg the building thread. the results are not what i expected - in a handful of scenarios as you would expect the 5600 loses some 1% lows, in the worst case as much as 10%, but never imo to a particularly significant level. I think you just need to realize that a) most games are still not intensely multi-threaded. They have a one to 3 heavy threads that will take up a CPU core, and then a bunch of others that are pretty light and the main benefit is getting them off the cores with heavy threads. b) for the average gamer, nothing else they do on their PC is particularly CPU heavy. Discord calls are audio encoding, a trivial task. Youtube video is likely h/w accelerated decode. They showed the CPU graphs in task manager at the start -- but note that the CPU clockspeed is bouncing around and frequently low! And that's in task manager which reads high. So these tasks hardly use any CPU. CoolCab posted:maybe their test systems are too clean and they're not getting representative just random background tasks deciding to all spike at once and intensify that 1% low behavior? So 1% lows are still a whole lot of frames to average out. If a bunch of background tacks line up every other minute, that's not gonna show up on the 1% lows. It'd be the 0.1% lows -- and 0.1% lows are that much more difficult to get consistent results. You have 1/10th the data so your error bars are much higher. (HUB doesn't show error bars in the first place which sucks. You absolutely get round-to-round deviation just running a plain game benchmark on a clean system.) CoolCab posted:and in some cases the gains seem to actually revert - the 5600 and 5700x equalize when doing some heavy task is busy which seems bonkers and backwards. I'm not really seeing it? They only do 2 tests with a 4k video and they do it with CS:GO and RSS, which are not exactly CPU heavy. The 1440p results are essentially unchanged from the previous 1080p + voice call ones. So the only real change with the "heavier" task is to 1080p, which drops down with both CPUs to about the same results as 1440p HUB's take is it means some "architectural limitation" of Ryzen, which seems like a huge leap to me. I'd guess system memory myself. This is the problem with using games that get 500 FPS -- memory bandwidth becomes a bottleneck more than CPU.
|
|
#
?
Jul 15, 2022 16:22
|
|
|
Kibner posted:Is this an ECC memory stick, too? I see that it has five chips on one side. yea, that’s why it’s 1.2v, 2667 16gb is one of Samsungs fastest and largest ECC udimm bins. In terms of binning, guess I didn’t realize how variable the b-die bins can be but my z390 taichi ultimate should run four sticks at rated speed regardless of any overclockability. Later AMDailires. 
|
|
#
?
Jul 15, 2022 17:04
|
|
|
CoolCab posted:https://www.youtube.com/watch?v=Nd9-OtzzFxs Nah man, it just doesn't matter. It never has. Windows is generally pretty good at scheduling tasks, and most background tasks are utterly insignificant in their demands, and most games are never going to fully 100% utilize your CPU, even with a six-core CPU. On 4-core CPUs, maybe you could run into some issues, but there just aren't any issues with multitasking on a fast six-core CPU. gradenko_2000 posted:my extremely kneejerky reaction to that post is that I've seen my Task Manager hit over 16 gigs of RAM usage, which meant I would have gone over had I not upgraded to 32, and that seems to be a more "real" problem with having lots of background tasks than running out of cores This is also misleading. Windows overallocates when it detects that you have a lot of ram. Like, by a lot.
|
|
#
?
Jul 15, 2022 18:45
|
|
|
I remember that somebody did an article testing a previous generation of ryzens with games + OBS streaming, with the video encoding happening on the CPU. And that had a more noticeable difference with the 6-core being more impacted than higher core count. Of course that's not really relevant now with hardware video encoding being well supported and broadly available on GPUs. Dr. Video Games 0031 posted:This is also misleading. Windows overallocates when it detects that you have a lot of ram. Like, by a lot. Plus, a shitload of the ram the average person has in use is web browsers. Who gives a poo poo if some of your background tabs get paged out to disk? They'll reload fast enough. OTOH doing VM stuff can easily make 16gb feel a little tight, so maybe gradenko was doing something like that.
|
|
#
?
Jul 15, 2022 20:50
|
|
|
Great, I just found out the my Gigabyte B550I Aorus Pro AX is a revision 1.1 re-release with "AMD" wifi (basically AMD branded Mediatek garbage) instead of the rev 1.0 board's Intel AX200 that's in every review and on shop pages. The driver from Gigabyte installs but then won't load because error 39 so now that I need the wifi, I can't actually use it and 99% of the internet's helpful troubleshooting topics are re: the original board and Intel wifi issues 
|
|
#
?
Jul 19, 2022 22:56
|
|
|
Gigabyte is one of those companies that will change up hardware dramatically between revisions. I stopped buying their stuff when a plugging a camera into a USB port killed an entire motherboard in '06 or something, but it seems like they've got drastically varying quality between models and shady practices when doing revisions of products. I can't straight up suggest for people not to buy their stuff since plenty of it is good, but they're definitely on my personal list to not buy from. edit: you might see if there's wifi drivers available on the mediatek site if you can find the right chipset.
|
|
#
?
Jul 20, 2022 04:33
|
|
|
I generally avoid built-in wifi, but oddly enough the PCIe wifi card I use is a Gigabyte adapter using an Intel laptop wifi card. I have pretty much accidentally ended up with mostly Gigabyte builds the last two desktops I've made, and I think I've been lucky with the hardware because it's all been good - two mobos, three GPUs. I avoid their software like the plague, though.
|
|
#
?
Jul 20, 2022 07:36
|
|
|
|
| # ? May 29, 2024 23:56 |
|
|
for what little it's worth, i spent a week ripping my hair out reinstalling the wifi firmware, reinstalling windows, trying different bios versions, etc. trying to figure out what was wrong when my gigabyte motherboard's wifi didn't work out of the box. well, everything worked immediately when i got a replacement.
|
|
#
?
Jul 20, 2022 07:45
|
|
