Question [3080ti] Low FPS on certain games, rebooting fixes - SEVERAL formats + hardware change

[Jasper Burnell]

I got a NZXT build with the following specs back in March:
Nvidia GeForce RTX™ 3080 Ti ( founders edition)
ASUS ROG Strix Z690-E Gaming Wi-Fi DDR5
Intel Core i5-12600K 10-Core 3.7GHz
G.SKILL Trident Z5 Black DDR5 6000MHz CL40 6000 MHz (Maximum Speed) 32GB (2X16GB)
XPG Core Reactor 850W Gold
Western Digital SN570 1TB

After a while I noticed that after reboots in certain games I would get lower FPS. A specific test I do now to see if this is a problem is load up limgrave in elden ring where the first real enemy camp is. If I am in the 40s-50s I know I am screwed for that boot cycle. Another game this really stands out in is fall guys, absolute dogsh** FPS when other people are on the screen.

After a couple months of this my GPU started acting weird, eventually dying. After several formats, installing drivers when it tried to switch displays would show a corrupted and flashing screen. After this, I could never get back into windows. I tried older drivers and formatted a few times and it would do this every time when installing nvidia drivers. I have also tried different HDMI cables(I am hooked into a sony x900h) and probably most other things you see people suggest.

I RMA'd the card with NZXT and they sent a new one (ASUS rog strix 3080ti) and at first everything seemed fine. I rebooted one time though, and this problem came back.

When I was waiting on the new card, I used my 2070 from another system and never experienced this. It could have been luck, or the fact that it doesn't need nearly as much power.

I am down to two ideas now:

1. PSU - It is a PSU issue and the issue slowly killed my other card. I have tried different rails.
2. Unknown motherboard issue.

I want to just replace the PSU but I'm positive that will void my warranty.

The behavior to me is very strange, because while this was going on previously... if I was on a "good boot" i would like never crash and could play games for 8 hours or so. I absolutely HAD to reboot to get rid of the low fps issue.

Has anyone seen anything like this ever before? Trying to diagnose this has been exhausting especially when my hands are tied. I have a Corsair PSU sitting by. If I install it, that's it for my warranty. https://www.amazon.com/CORSAIR-RM850x-Certified-Modular-Supply/dp/B079H5WNXN

Any help would be appreciated.

 

[Phaaze88]

3080 and up Founder's Editions have a problem with thermally throttling memory. Probably why it kicked the bucket.
The new gpu should be fine on that front.

Remove the gpu. Are you able to boot up and log in on the cpu's iGPU without issue?
Take a look through the following, it may be affecting the current system: https://www.techpowerup.com/forums/...intel-cpu-performance-we-have-the-fix.295877/

 

[Jasper Burnell]

3080 and up Founder's Editions have a problem with thermally throttling memory. Probably why it kicked the bucket.
The new gpu should be fine on that front.

Remove the gpu. Are you able to boot up and log in on the cpu's iGPU without issue?
Take a look through the following, it may be affecting the current system: https://www.techpowerup.com/forums/...intel-cpu-performance-we-have-the-fix.295877/

I have a 2070 from another comp that does not ever do this in the system. I also have a new 3080ti from NZXT but it still does the thing where it will have low FPS on some boots. Rebooting one, or however many times, seems to fix it. Not sure if you read my whole post...

 

[Phaaze88]

Sorry. I missed the 2070 part.

The XPG Core Reactor is supposed to be a great unit... unless you think you got a lemon?

I also have a new 3080ti from NZXT but it still does the thing where it will have low FPS on some boots.

Have you confirmed it's the gpu? How far have you gone to check this?
Gpu core, gpu hot spot, memory junction temperatures are fine?
Gpu is boosting normally under load? Or is it stuck at a very low frequency?
Low utility - waiting on work from the cpu+ram?
You can use Gpu-Z(sensors tab, stretch out the window to see all parameters at once) or HWINFO(gotta double click on parameters to make graph appear) to monitor those. There's even a Reason for Performance Cap on them.

 

[Jasper Burnell]

Sorry. I missed the 2070 part.

The XPG Core Reactor is supposed to be a great unit... unless you think you got a lemon?



Have you confirmed it's the gpu? How far have you gone to check this?
Gpu core, gpu hot spot, memory junction temperatures are fine?
Gpu is boosting normally under load? Or is it stuck at a very low frequency?
Low utility - waiting on work from the cpu+ram?
You can use Gpu-Z(sensors tab, stretch out the window to see all parameters at once) or HWINFO(gotta double click on parameters to make graph appear) to monitor those. There's even a Reason for Performance Cap on them.

I think the key takeaway here is that when I’m on a good boot, I never crash really and get good performance and no overheating. I never got overheated regardless, even on the hotspot people look for in hwinfo, memory bridge or something. On a good boot it lasts as long as I am booted that way, until I have to reboot and end up on a "bad boot". I then have to reboot until I get a good one.

On a bad boot I get the fps issue. Please note this is also two diff 3080ti as well.

 

[Phaaze88]

Good boot/bad boot speaks to me of a software or driver issue, and not hardware, as those are more consistent in reproduction:
Does it do this every 3rd reboot, or when it feels like it?
It works just fine if you shut it off and power on normally, regardless of how many times you do it?


A not quite as good alternative to the psu problem is to try the games with a lower power limit applied. Use Msi Afterburner and try an 80% power limit.

 

[Jasper Burnell]

This is over MULTIPLE FORMATS OF OS and clean / different driver version installs fresh from formatting. Ive BEEN down the software road and that is NOT it hence why I wasted about 2 weeks of my life troubleshooting.

 

[boju]

A friend of mine is running 3080ti on a Corsair Rm850x psu fine (w/ 11600k). With it's power related spikes or whatever, just maybe, the Core Reactor isn't up to task but is fine with lesser power hungry card. Have you contacted Nzxt to suss out your options in changing psu without losing warranty? Some certified service centre you can take it to or a pc shop they would be satisfied with.

 

[Jasper Burnell]

Is it possible this is related to running an Alder Lake CPU on Windows 10 and on some boots the ecores are utilized/locked in differently?

 

[alceryes]

Is it possible this is related to running an Alder Lake CPU on Windows 10 and on some boots the ecores are utilized/locked in differently?

Then the 2070 the OP used for a while would've also exhibited the issue.
OP, run HWiNFO64 in the background (sensors only, logging on). Record at least 15 mins of the framerate being great and a SEPARATE file of 15 mins of the framerate being crappy. Upload the files somewhere we can grab them.

 

[Jasper Burnell]

Then the 2070 the OP used for a while would've also exhibited the issue.
OP, run HWiNFO64 in the background (sensors only, logging on). Record at least 15 mins of the framerate being great and a SEPARATE file of 15 mins of the framerate being crappy. Upload the files somewhere we can grab them.

Thank you. it may take several reboots but i will do this. I would also like to limit this to maybe 5-10 mins given it may be damaging the card for the low FPS one. It's instant, and immediate upon getting in the game. will use elden ring.

 

[Jasper Burnell]

https://file.io/rzTUJnAgylZZ - Good session

https://file.io/4KxGDNoZHM0C - Bad session. I didn't have a lot of time to look but it seems the core voltage is a bit higher on the bad run.

edit: total GPU power looks high too

 

[Jasper Burnell]

file.io keeps deleting the CSVs. Why is it so hard to find hosting for a 10k CSV file but I can easily upload a 10meg image somewhere. I guess it is intended for one download.

 

[Jasper Burnell]

BAD CSV - https://filetransfer.io/data-package/cXgY1wf1#link
GOOD CSV - https://filetransfer.io/data-package/WrqoBEnH#link

There is a limit to 50 DLs on these so please only download it if you sincerely intend to comb through this... thanks

 

[alceryes]

BAD CSV - https://filetransfer.io/data-package/cXgY1wf1#link
GOOD CSV - https://filetransfer.io/data-package/WrqoBEnH#link

There is a limit to 50 DLs on these so please only download it if you sincerely intend to comb through this... thanks

Is your computer doing exactly the same thing (running the exact same programs/games) between these GOOD and BAD logs? Cause these look like two completely different CPU/GPU/RAM load sets.

 

[Jasper Burnell]

Is your computer doing exactly the same thing (running the exact same programs/games) between these GOOD and BAD logs? Cause these look like two completely different CPU/GPU/RAM load sets.

Yes. I rebooted until I got to a good/bad boot, then loaded elden ring in the same area immediately and nothing else but hwinfo.

 

[alceryes]

I think Elden Ring may have too many variables for it to be a good litmus test.
Load up Unigine Superposition using the 4k preset (it doesn't matter if you don't have a 4k monitor). Run HWiNFO64 in the background, logging, and repeat the benchmark runs until you get the log of one good one and one bad one to compare. Remember to wait AT LEAST 5 mins at the desktop before launching Superposition.

 

[Jasper Burnell]

I just sent the whole thing back. this is exhausting.

 

[alceryes]

I just sent the whole thing back. this is exhausting.

Sorry it didn't work out.
The GOOD and BAD files you had uploaded looked like your computer was doing two entirely different things. There were so many things going on I didn't know where to begin.

 

[PcDoktor]

I’ve had same experience on a rig I built for someone same card, ended up pulling out my hair it came down to a windows driver that loaded something GeForce experience so I uninstalled bot the driver GeForce experience and stopped the windows driver don’t remember I can check. I just stopped it display something I think.. And ran MSI afterburner and haven’t had a problem since.. Hopes this works for someone..
Edit: I changed fast boot on the pc and selected normal boot and load only windows files I think..

 

[Karadjgne]

I've seen this once before. It was a driver conflict/date code conflict. What would happen is all the chipset drivers are stored in windows/drivers and not touched by windows nor software. Windows would run, but there'd be software that had different driver versions or date codes. It'd run as a background task. On shutdown, everything in ram is saved to cmos, it's part of fastboot process, but that'd include the funky error. On reboot, bios would try and load the cmos entries, but get an error, so skip loading that driver altogether. On successive reboot or reset, that cleared the cmos, bios being forced into loading from scratch, so loading valid drivers from the driver folder. No issue at boot, but that changed as soon as that funky software ran and the cycle would repeat.

Antivirus programs are often the culprit too as they'll quarantine windows files on boot if it thinks they are dangerous file types.