SSD jumps to 100% under GPU load and freeze Windows

[NotSoFunnyClown]

I'm despair on a strange problem.

My system runs since almost 2 years without any problems. Since about a few weeks I got the problem that when I'm playing a game like overwatch, after a few minutes, textures and 3D models starting to disappear. After a while the whole game will freeze. When I try to tap out the window the other applications in the background starting to freeze and becoming unresponsive.

First thing that came to my head was the RAM. But after I opened the task manager, I saw that the main SSD (Samsung 960 Pro m.2 on DIMM.2) "active time" jumped to 100% and stays that high.

Then I have tried to reproduce this behaivor. I can easily do it when I run Furmark on both GPUs. After a few seconds, SSD jumps to 100% and applications starting to freeze. The strange thing is, it will not happen when I only use one GPU. But it's not one of both GPUs that produce this problem. I've tried GPU 1 and 2 separately without any problems. when I use both, yea... you know.

It's not a temperature problem. Both GPUs run under water.
The SSD temperature is on a normal level.
The system is not OC

How to reproduce:

1. Run Furmark at highest settings to push both GPUs to 100% load
2. Run AS SSD Benchmark
3. Right at sequential benchmark, "active time" jumps to 100% and system starts freezing

What I've already tried to fix this behavior:

• Installed the newest GPU driver and one of February
• Removed Norton Anti-Virus
• Check the system for malware
• Installed all current drivers from the Asus website
• Installed Samsung NVMe Driver 3.0, 3.1 and the Windows NVMe default driver
• Closed all background applications before testing
• Switched from BIOS1 running UEFI 1704 to BIOS2 running on UEFI 0802
• Reset UEFI settings on 1704 to default
• DIMM.2 over PCH instead CPU
• Disable SuperFetch
• Disable PCIe power saving in Windows
• Updated Windows 10 to 1903
• Samsung Magician says "State: Good" and "Newest firmware installed"
• Clean Windows install
• Switched to a new SSD (970 Pro)
• Memtest86 runs without any errors
• Updated UEFI to 1902

PC-Specs:

• OS: Windows 10 Pro (1903)
• CPU: Intel Core i9-7900X (delidded)
• Motherboard: Asus ROG Rampage VI Extreme (UEFI 1704)
• RAM: G.SKill TridentZ RGB 64GB
• GPU: 2x EVGA GeForce GTX 1080 Ti K|NGP|N Hydro Copper
• SSD 1: Samsung 970 Pro M.2 1TB (on DIMM.2 Socket; newest firmware)
• SSD 2: Samsung 850 PRO 2.5" 2TB
• PSU: EVGA SuperNOVA 1600 T2
• SLI-Bridge: ZOTAC SLI HB Bridge / 2 Slot

Any ideas, what the problem could be? The whole behavior makes no sense for me.

Thank you

[NotSoFunnyClown]

EDIT: It was a misbelief Problem still exists.

[NotSoFunnyClown]

Ok... the party started to quickly. Problem seems not to be fixed and came back

[cekeu]

so the problem is when both gpu runs in sli. is it a problem of sharing PCI Express?

[NotSoFunnyClown]

Not only in SLI. but under load, yes. i've tried disabling SLI and run Furmark seperatly on each GPU. Same issue.
So yes it seems to be something in the PCIe area. But its harder to reproduce on the 970 Pro as on the 960 Pro.
I can now easy reproduce that whole thing by runing Furmark to full load GPUs and then run AS SSD Benchmark. As soon AS SSD starts Seq testing, disk usage jumps to 100% and systems freezing.

But why? PCIe lanes shouldn't be any problems or was not any problem since 2 years.

Updating to 1902 changed nothing.

[Wheity]

Very strange. Do you have this also by using SATA drives instead your SSD's? Maybe a small power problem from your PSU.

[JTZ1032]

I agree with Wheity this kind of troubleshooting is necessary. If you can 100% load the SATA and it doesn't crash you have narrowed down the issue.

Out of curiosity; is there any other SSD you could try e.g. a SATA SSD, a PCI Express SSD or even an SSD from another manufacturer? What results do you get if you check the disks status with HDSentinel? It can show issues that few other utilities will pick up. Have you run a CHKDSK on the SSD? Is it just that game that causes the crashing issue? Do any other games cause it?

I know from having a faulty SATA controller many years ago that issues with hard disks are difficult to diagnose since they cause issues that look like failures from other components. If you stress the SSD with another benchmark program does it still crash (I'm trying to eliminate that benchmark implementation as being the cause). Can you 100% that SSD on another PC; does it still cause issues there?

I hope this helps somewhat. Thank you.