|
How the gently caress does the DirectX team not have a presentation room where they can record proper audio? Jesus christ.
|
#
?
Sep 29, 2019 14:37
|
|
|
|
| # ? Jun 4, 2024 16:34 |
|
|
Hello GPU thread, this question is a bit too big for the short questions thread and I barely know what I'm doing here so I would appreciate some advice. I have a ZOTAC AMP RTX 2080 (the bigass one you could dome someone with) and my performance is pretty far below what one would reasonably expect, even with a better-than-average stable overclock (+100/+750). Without the overclock the performance is even worse. Could one of you more experienced folks let me know if something jumps out at you that could be the problem? Here's a copy/paste from a reddit thread I made about the issue:quote:Hi, I have a ZOTAC AMP RTX 2080 that I've modified for watercooling and I've got some questions about performance that have been bugging me for a while. Specifically, my gaming performance doesn't seem to match up to benchmarks performed by sites like Digital Foundry, Techpowerup and similar. I'm not sure if there's something I'm doing wrong, or if I'm reading things incorrectly, or if I'm just being paranoid and it's actually fine. Granted, this is still pretty high performance and not something to complain about, I'm just curious as to why it's so far under what every source I've found says it should be.
|
|
#
?
Sep 29, 2019 22:40
|
|
|
CJacobs posted:Hello GPU thread, this question is a bit too big for the short questions thread and I barely know what I'm doing here so I would appreciate some advice. I have a ZOTAC AMP RTX 2080 (the bigass one you could dome someone with) and my performance is pretty far below what one would reasonably expect, even with a better-than-average stable overclock (+100/+750). Without the overclock the performance is even worse. Could one of you more experienced folks let me know if something jumps out at you that could be the problem? Here's a copy/paste from a reddit thread I made about the issue: I don't know a lot about AMD products but is it possible the 2700x is bottlenecking the system?
|
|
#
?
Sep 29, 2019 23:42
|
|
|
VelociBacon posted:I don't know a lot about AMD products but is it possible the 2700x is bottlenecking the system? One of the comments on the reddit thread suggested this too, and yeah after testing some more games I'm starting to think that's the answer. For Control specifically, my fps doesn't actually get much worse if I remove my overclock and set it to stock speeds (maybe 5 fps difference). I chose that game because it's very GPU-reliant, so based on the fact that my overclock doesn't change anything I think the GPU may not be the problem after all.
|
|
#
?
Sep 29, 2019 23:48
|
|
|
CJacobs posted:One of the comments on the reddit thread suggested this too, and yeah after testing some more games I'm starting to think that's the answer. For Control specifically, my fps doesn't actually get much worse if I remove my overclock and set it to stock speeds (maybe 5 fps difference). I chose that game because it's very GPU-reliant, so based on the fact that my overclock doesn't change anything I think the GPU may not be the problem after all. Bench it using Firestrike or a Unigine run and compare that to a reviewer's results maybe?
|
|
#
?
Sep 29, 2019 23:55
|
|
|
Also, for sanity what are your temps running while your benchmarking? I see "modified for watercooling" so I would be a little interested to make sure it wasn't running hotter than you expect and throttling.
|
|
#
?
Sep 29, 2019 23:59
|
|
|
It idles at 25-30c and peaks at about 45-50c when it's at the highest utilization. I guess I should clarify that I'm using an AIO cooler on it, I just made my own brackets because the NZXT Kraken doesn't fit on the bigass zotac card by default.Endymion FRS MK1 posted:Bench it using Firestrike or a Unigine run and compare that to a reviewer's results maybe? That's a good idea, I'll try that. GPU benchmarks would surely be better than just trying a bunch of games repeatedly. edit: Ran UserBenchmark to compare against others with a similar setup, I'm in the 95th percentile for the RTX 2080. Meanwhile I'm only in the 5th percentile for the Ryzen 2700x. So yes, I think I can conclude that the GPU isn't causing my problem. I'll look into CPU options further, thanks for the advice folks. CJacobs fucked around with this message at 00:50 on Sep 30, 2019 |
|
#
?
Sep 30, 2019 00:20
|
|
|
CJacobs posted:It idles at 25-30c and peaks at about 45-50c when it's at the highest utilization. I guess I should clarify that I'm using an AIO cooler on it, I just made my own brackets because the NZXT Kraken doesn't fit on the bigass zotac card by default. Have you tweaked anything on your 2700X at all? Wonder if that or maybe an outdated BIOS could be causing some of the differences. Might be worth seeing if you can kick up some clock speed or looking at UserBenchmark to see if the other Ryzen scores are at stock, or are overclocked & that's why your scores are so different. Maybe even check to see if your chip is throttling from temps, I doubt that's the case but you never know.
|
|
#
?
Sep 30, 2019 14:19
|
|
|
Gonna guess crappy ram or ram not set to the proper speed.
|
|
#
?
Sep 30, 2019 15:13
|
|
|
K8.0 posted:Gonna guess crappy ram or ram not set to the proper speed. This is good advice in general: Always be sure to enable the XMP profile, otherwise you're not getting any value out of buying faster RAM.
|
|
#
?
Sep 30, 2019 15:21
|
|
|
Don't worry, I am aware that RAM timings have a huge effect on PC performance, I'm running the proper XMP profile with 32GB of Corsair Vengeance LPX @ 3200Mhz (the model of which I also made sure is cleared for my motherboard). This is what I mean, by all accounts my PC should be able to do pretty much whatever I want with it, but it just doesn't reach the peaks it "should". Honestly I'm starting to think it boils down to "every computer is a little different even with the same parts inside". Either way, I nabbed a 3700x because I wanted a Ryzen 3 CPU anyway and pushed it to 4.3Ghz and... nothin'. There was essentially no improvement in gaming, only in benchmarks. But that's not really in this thread's wheelhouse so I'll take it elsewhere unless anyone has any other suggestions. CJacobs fucked around with this message at 15:28 on Sep 30, 2019 |
|
#
?
Sep 30, 2019 15:25
|
|
|
CJacobs posted:Don't worry, I am aware that RAM timings have a huge effect on PC performance, I'm running the proper XMP profile with 32GB of Corsair Vengeance LPX @ 3200Mhz (the model of which I also made sure is cleared for my motherboard). This is what I mean, by all accounts my PC should be able to do pretty much whatever I want with it, but it just doesn't reach the peaks it "should". Honestly I'm starting to think it boils down to "every computer is a little different even with the same parts inside". You have something else going on here. It's weird that it's only hurting things a little bit but if you're clocking out at the lowest 5% on the CPU tests something is going wrong. I agree on the RAM, you might have bought the right stuff but you may have a bad chip. If you have 16 x 2 try taking one out testing and then swapping them. Otherwise you might be one of the extremely rare people who have legit MB problems. Those rarely show up as performance issues but it's not totally unheard of. If you were seeing 5fps drops I'd say "There's always some variance" but this is significantly worse than that. It's almost like you're running a really old driver or something, but you're doing all the right stuff. Lockback fucked around with this message at 15:32 on Sep 30, 2019 |
|
#
?
Sep 30, 2019 15:29
|
|
|
Yeah, it very well may be my motherboard itself, especially with the edit I just did re: the new CPU. Man my computer is hosed up!!
|
|
#
?
Sep 30, 2019 15:30
|
|
|
Make sure your power plan is set to Ryzen Balanced (for 3XXX) and Balanced for 2XXX. Try setting your minimum CPU usage to 90% just to see if it changes performance. Could also be some fans going up and down if you're able to run something on a second monitor to chart temps with. Control is a very weirdly optimized game that runs horribly, but I'm assuming you've tried with other games.
|
|
#
?
Sep 30, 2019 15:44
|
|
|
Alright, I've updated the Ryzen chipset drivers for the new CPU, so I have the power plan set properly now. And I have tested it with other games, yeah, and the performance is even weirder the higher it goes. Control was just an example of a game that's GPU dependant (which I now realize is not the issue lol). GTA 5 is a common whole-PC-benchmark for example and I should be pushing 120+ at 1080p, but instead I get like... 80. Which is still really great don't get me wrong, but it's a far cry from the expected result. Well, with all the tweaks that've been suggested and the overclocking I just did, it is a little better but still not what it's supposed to be. I think Lockback is probably right and the motherboard is causing problems, it's relatively new but it may be defective in some way so I may as well replace it. I'll save up for an X570 motherboard to replace this one because ~wave of the future~ and all that, and hopefully that'll make a difference. Thanks for all the advice and suggestions, I appreciate it.
|
|
#
?
Sep 30, 2019 16:06
|
|
|
Did you check your pcie link speed...? I don’t know if it could have THAT big an effect but it’s the last thing I could think of.
|
|
#
?
Sep 30, 2019 16:14
|
|
|
 CPUZ says this, but I don't actually know what it means. I've got it in the first/top PCIE slot. edit: My board is an X470 Taichi for reference.
|
|
#
?
Sep 30, 2019 16:22
|
|
|
That's strange - should be x16 for both. You're only getting half the bandwidth.
|
|
#
?
Sep 30, 2019 16:28
|
|
|
Uhh, weird. How do I... fix that? I ran the render test in GPU-Z that is intended to push it to the highest state you have it set to, and it stayed at x8, so it's definitely not set right or working right. edit: vv haha yeah that didn't happen unfortunately, stayed at x8 during benchmarking, so it appears to be hosed.
|
|
#
?
Sep 30, 2019 16:30
|
|
|
It won’t hit the max rate unless you’re doing something with it, the fact that it says 16x max means it’s probably fine (if you want to be sure you can run hwinfo and see what it’s max value is after running a game or benchmark) Edit: oh. Uh, make sure it’s in a 16x slot? Make sure there’s not something weird in your bios? Sometimes if you have more than one pcie device plugged in, another device steals lanes and turns slot speeds down.
|
|
#
?
Sep 30, 2019 16:30
|
|
|
Dogen posted:It won’t hit the max rate unless you’re doing something with it, the fact that it says 16x max means it’s probably fine (if you want to be sure you can run hwinfo and see what it’s max value is after running a game or benchmark) My PC is idle, but still at x16.
|
|
#
?
Sep 30, 2019 16:31
|
|
|
Mine always negotiates downward when not in use, like right now it’s reading pcie 1.1 but it goes up to 3 under load.
|
|
#
?
Sep 30, 2019 16:37
|
|
|
Lambert posted:That's strange - should be x16 for both. You're only getting half the bandwidth. I don't think that would impact performance all that much, a GPU doesn't come particularly close to saturating.
|
|
#
?
Sep 30, 2019 16:42
|
|
|
Lockback posted:I don't think that would impact performance all that much, a GPU doesn't come particularly close to saturating. Yeah, the performance difference between a 16x and 8x PCIe 3.0 slot for a modern GPU is <2%. Whatever you've got going on wouldn't be because of that.
|
|
#
?
Sep 30, 2019 16:53
|
|
|
I think it is indicative of the problem being with the motherboard regardless though. Reseated it, no change, and it wouldn't even acknowledge the card in the second main pcie slot. Think it's safe to say this thing is defective in some way. edit:  Uh... scratch that I guess? Reseated it again and now it's at x16 regardless of performance increase/decrease. Yeah I think this mobo is busted. CJacobs fucked around with this message at 17:03 on Sep 30, 2019 |
|
#
?
Sep 30, 2019 16:59
|
|
|
One of the things that frustrates me the most about going Ryzen is that it's extremely difficult to find a basic benchmark to get a sense of whether stuff is running OK or not, especially wrt frametimes. Even Overwatch won't bother with a benchmark.
|
|
#
?
Sep 30, 2019 17:24
|
|
|
Userbenchmark is a pretty quick and dirty way to check against the average.
|
|
#
?
Sep 30, 2019 18:09
|
|
|
I reinstalled Windows and that lost my power limit slider in Afterburner. What to do?
|
|
#
?
Sep 30, 2019 20:54
|
|
|
Sininu posted:I reinstalled Windows and that lost my power limit slider in Afterburner. What to do? Do you have "Enable hardware control land monitoring" checked in General Settings?
|
|
#
?
Sep 30, 2019 21:03
|
|
|
Stickman posted:Do you have "Enable hardware control land monitoring" checked in General Settings? Yep.
|
|
#
?
Sep 30, 2019 21:08
|
|
|
CJacobs posted:Alright, I've updated the Ryzen chipset drivers for the new CPU, so I have the power plan set properly now. And I have tested it with other games, yeah, and the performance is even weirder the higher it goes. Control was just an example of a game that's GPU dependant (which I now realize is not the issue lol). GTA 5 is a common whole-PC-benchmark for example and I should be pushing 120+ at 1080p, but instead I get like... 80. Which is still really great don't get me wrong, but it's a far cry from the expected result. You definitely have an issue somewhere, but I think 90 sounds about right for a 2700x on GTA 5. It's a pretty heavily CPU-limited game. Have you double-checked that you're running your RAM in dual-channel mode rather than single-channel?
|
|
#
?
Sep 30, 2019 21:09
|
|
|
Ps x570 shouldn't make any difference. Unless the board is buggered.
|
|
#
?
Sep 30, 2019 21:11
|
|
|
Sininu posted:Yep.  That's all I got, except maybe try some of the other compatibility settings, or do a clean reinstall of the gpu drivers and a restart.E: Quick search suggests it's been happening occasionally over the past 5 years, but no one seems to have a guaranteed fix.
|
|
#
?
Sep 30, 2019 21:18
|
|
|
Stickman posted:You definitely have an issue somewhere, but I think 90 sounds about right for a 2700x on GTA 5. It's a pretty heavily CPU-limited game. This is with the 3700x, should have specified. And yeah it's in dual-channel mode, the RAM is running at the advertised speeds.  GRINDCORE MEGGIDO posted:Ps x570 shouldn't make any difference. Unless the board is buggered. Yeah I figured as much performance wise, mostly wanted to get one for future proofing since an x570 would last me a looooooooooooooong time whereas the x470 will only last me a loooong time.
|
|
#
?
Sep 30, 2019 21:18
|
|
|
I'd hit up the AMD thread personally.
|
|
#
?
Sep 30, 2019 22:26
|
|
|
Yeah I think I'm about question-ed out. Thanks for letting me make this The CJacobs' Computer Thread for a bit, grateful for all the help folks.
|
|
#
?
Sep 30, 2019 22:31
|
|
|
346.48 Performance updates for Tom Clancy's Ghost Recon Breakpoint and Asgard's Wrath. No major changes. https://www.geforce.com/drivers/results/152007
|
|
#
?
Oct 1, 2019 16:55
|
|
|
FWIW 346.48 also fixed a problem I was having with Tetris Effect, where it capped at 100fps and ran with weird jittery frame pacing for some reason. Now it's back at smooth unlocked FPS. edit: scratch that, as soon as i connected my bluetooth controller to actually play the framelock and bad frame pacing came back. the presence of the controller is somehow breaking the renderer what the gently caress lol repiv fucked around with this message at 18:41 on Oct 1, 2019 |
|
#
?
Oct 1, 2019 17:13
|
|
|
Interesting. There is a price comparison site in Norway called Prisguiden.no, and if you mistype rx 5700 you get an entry for the AMD Radeon RX 5500: There are no entries for rx 5300/5400/5600/5800 or 5900, just the 5500 (and of course the 5700). Wonder if it means anything.
|
|
#
?
Oct 1, 2019 19:38
|
|
|
|
| # ? Jun 4, 2024 16:34 |
|
|
RX5500 has already leaked a bunch https://twitter.com/VideoCardz/status/1179004381653557252
|
|
#
?
Oct 1, 2019 19:40
|
|