[SOLVED] GPU sudden 40% frame loss ?

[Nlucz]

Hello, I'm having an issue that has left me stumped. A little backstory to the problem, I installed a new SSD which ran fine. A week later I upgraded ram from 2x8 ddr4 3000mhz to 2x16 ddr4 3600mhz. All was great I noticed load time improvements as well as slight fps increases. A couple weeks went by and the power in the house flickered once. The next day I game for a good 8-10 hrs. playing the new halo at 144fps, at the end of my session while watching YouTube my computer shut off and reset, I figured the power just went out again as it happens quite often and went to sleep.

The next day the computer ran until I launched any games or Heaven benchmarks causing the GPU to take load which would immediately would shut down the PC, all while desktop apps, YouTube, and even CINEBENCH ran stable. I troubleshot thinking it was a memory incapability. Started with swapping in and out my old ram, no luck there. I ran SFC / SCANNOW, CHCKDSK /R, also memorytest86. The memory was ok according to all that. I updated Nvidia drivers, uninstalled them thru windows and reinstalled. I disabled fast boot in windows, all were no help. At this time, I was also getting Kernel Processor Power event 37 in the event viewer, The speed of processor 0 in group 0 is being limited by system firmware. I then updated my BIOS which wasn't done since 2015, and also reset BIOS settings to default.

The BIOS update resolved the event 37, but I was still shutting down so I figured it must be the PSU. I replaced the PSU and all the cables with an equivalent size just a different brand that my local store had in stock. The first PSU arrived dead on arrival it would not boot or pass a paper clip test. The second PSU however booted the PC just fine and also solved the issue of shutting down after taking GPU draw. I let Heaven benchmark run for about 45 mins before rewiring everything back in. After sitting down and loading into Halo I noticed my in- game settings did not change but my Fps went from 144fps to about 65-75 with a lot more frame drops than usual. I tested other games that I play frequently and they all struggled with what seems to be a 40% percent average frame loss or more.

MSI afterburner shows according to my memory, normal stats for usage, clock speeds, memory usage, also voltage seems normal according to other users. I don’t monitor my computer constantly being I'm usually pretty good keeping it clean and updated, I have never had problems with heat but I do keep an eye on the basics especially FPS . So, the next step I took was making sure all wiring was seated correctly, I exchanged the PCIE cable going to the GPU thinking make I was unlucky and the second PSU was faulty as well. I went thru and made sure 99 percent of all my drivers were updated. I doubled check bios settings made sure nothing odd was disabled in PCIE settings, still no luck.

Do you guys think the GPU could have taken damage from the power outage or the forced shut downs while troubleshooting, I would say it happened around 10 times before I solved that issue? Also, my Userbenchmark score is fairly bad. https://www.userbenchmark.com/UserRun/48893919 I called Evga they told me to take a 3-4 hr benchmark and send them the results, I don't know how to run Heaven or 3dmark that long and have results recorded. Any help would be greatly appreciated. My specs are as follows:

CPU: I7600k 4.00ghz
GPU: RTX 2070 Super XC Ultra
Board: MSI Z170A Pro Gaming Carbon
Ram: Corsair Vengeance DDR4 3600
PSU: Corsair RM850X ( Replacing old EVGA G1 850)
Drives: 1x Samsung SSD 850 EVO M.2
1x WD 1TB SSD
1xHDD 2TB
Cooling: Corsair H115i also 5 case fans.

 

[Nlucz]

Also I forgot to mention, my temps all are normal and i've been thru nvidia control panel multiple times checking vsync, dsr, triple buffering, all are off. Power management set to opitmal power . Discs are defragged . Im running my games at 1080p with display port to a Asus VG248qe

 

[DRagor]

Usernenchmark shows that your RAM runs at 2133 Mhz. Check in BIOS that you have XMP enabled. You know all this components swapping could turn it off.

 

[Nlucz]

I had my games running fine with the new ram installed and speeds were still 2133 before all these issues, all while not using XMP. I was happy enough tohave the additional memory space. I did try enabled xmp but got blue screened. I didnt want to start tweaking voltages to get xmp to work when halo was running like butter without it.

 

[alceryes]

I had my games running fine with the new ram installed and speeds were still 2133 before all these issues, all while not using XMP. I was happy enough tohave the additional memory space. I did try enabled xmp but got blue screened. I didnt want to start tweaking voltages to get xmp to work when halo was running like butter without it.

Are you absolutely sure it was running at 2133MHz previously? I bet the original swap out of RAM caused it to drop down to the base/default speed. When you reinstalled your GPU drivers some settings may have changed. You may also want to completely uninstall Halo (including deleting left over folders) and reinstall from scratch.

Run HWiNFO64 (sensors only, logging on) in the background while the Heaven benchmark loops for a bit and then go through the logs to see if anything is amiss. Also, go through a 3DMark Time Spy run and post a link to the results.

Yes, it is possible something happened to your GPU but we're not there yet. How do you have the aux GPU power plugged in? Two separate lines from the PSU or daisy chained? Whichever one it is, switch it to the other to test performance.

 

[Nlucz]

Good catch, I took a screenshot with my phone before I swapped in the new ram and the speeds were at 2400 . I really can't determine when that changed though, if it changed to 2100 when i swapped the ram I still didn't notice issues for a week . But it could of changed during the bios update or reset setting to defaults?

Ok so, I changed from a single power cable to two PCIE cables. Didn't notice a difference.

The tests I ran, but which application should I use to read the logs with HWiNfO64. Here's the 3dMark run hope it's the right link. https://www.3dmark.com/3dm/69793267?

Here's a link to a Userbenchmark with MSI Game Boost running the ram at 2400, also the voltage went up a bit to like 1.35. https://www.userbenchmark.com/UserRun/48901155

 

[alceryes]

Good catch, I took a screenshot with my phone before I swapped in the new ram and the speeds were at 2400 . I really can't determine when that changed though, if it changed to 2100 when i swapped the ram I still didn't notice issues for a week . But it could of changed during the bios update or reset setting to defaults?

Ok so, I changed from a single power cable to two PCIE cables. Didn't notice a difference.

The tests I ran, but which application should I use to read the logs with HWiNfO64. Here's the 3dMark run hope it's the right link. https://www.3dmark.com/3dm/69793267?

Here's a link to a Userbenchmark with MSI Game Boost running the ram at 2400, also the voltage went up a bit to like 1.35. https://www.userbenchmark.com/UserRun/48901155

HWiNFO64 logs can be read with any CSV reader. Libre Calc is a good free one. You can also use Microsoft Excel.
There seems to be something going on with your GPU. 'What' I'm not yet sure.

 

[Nlucz]

Here’s the news, I heard back from EVGA I told them basically the same thing I posted here. They just told me to start an RMA process just like that no troubleshooting. This is the first PC that I owned personally, and I would still like to learn anything I can, I guess I was hoping there was a simple cure and sending a package across the country during the holidays doesn’t sound simple.

So, here’s what I found so far as last-ditch effort. I took a log after running Heaven for about 25 mins and a log running 3DMark. Only thing I noticed that was odd are columns 279 and 280, stating GPU Performance Limits and Performance Limit Power. Both have yes entries. Also over the last 5 Time Spy runs the graphics scores have went down from 7500’s to about 5500.

Here’s the most recent Time Spy score. https://www.3dmark.com/3dm/69812224? The last I posted the score was like 7500.

And here’s the logs in case anyone would like to look thru. https://drive.google.com/drive/folders/1biKD_0DFjNVoLWg0uvsCS2DfRB3HZxwy?usp=sharing

Also here’s a spec list from HWiNfo64 don’t know if this has any good info in it… https://drive.google.com/drive/folders/1nawJ8JtXIwR4mlISIqkymiRB1F4fXknD?usp=sharing

Edit:I forgot to mention after getting the PC to stop shutting down. When I was installing driver updates and going thru windows updates, I installed a couple optional updates. I hope that didn't cause any issues, the only options I cant think of left are reseating the GPU, swapping in my old card, or reinstalling OS. Thanks for your guy's help so far.

 

[alceryes]

You can safely ignore GPU 'performance limit utilization'.

You're trying to run your core and uncore at a locked 4400MHz - this may be causing all of your issues, including the original hard power down. Reset all overclocks/overvolts/timings adjustments to default.

We really need more than a 1 minute or even 6 minute log file. Log should be started before the first 3DMark starts. Then go through three full tests (when one ends start up another) while logging. Try looping Unigine's Valley instead of Heaven for 15+ mins. Upload the whole log file for us to look at.

 

[Nlucz]

I hope that's not the case, the reason those numbers are up is because of MSI Game boost but i've had that running for about a year prior to the events.

I'll try to get those logs posted tonight. I still have these options left, Reinstall OS, reinstall Graphics driver (this time using DDU), trying out my old GTX 960 , or try a third PSU. Should i try any of those and which should I do first ?

 

[Nlucz]

Here's the logs, Valley and 3dMark, though i ran 3 benchmarks for 3D not stress tests. I can do either.

https://drive.google.com/file/d/1HG1JxFLi-ucXyLnk8iXUmdeq2d6t18j1/view?usp=sharing