Performance instability after adding 3070

[keltys]

Hey all, I posted last month with an issue. My computer was randomly crashing during gaming. Figured it was my VRAM and ended up getting lucky and finding a 3070.

 

Those crashes stopped, however a few other issues have popped up:

 

1. Games were crashing during high-loads. It had nothing to do with heat. This seems to have stopped after updating drivers, but I have only played a game once since the update. 

2. Screen flickering. This mostly seems to be while using zoom, but it has never occurred before the new GPU. It only happens on my primary screen. It also seems to alternate between flickering my windows and my desktop if that makes sense. It rarely effects the entire screen at once. 

3. Weird lagging and performance issues. Examples would be:

- typing in chrome, every once in awhile there will be a 1-2 second delay before the letters appear. 

- Dragging windows. This is particularly bad. Its not even just a rough drag, the window literally stutters across the screen. This also seems to bring my GPU to nearly 100% usage at times which is odd.  

- I feel like my computer is just less snappy overall. Load times and input lag just seem longer...

 

Any idea what is happening? I'm starting to get a feeling that my GPU wasn't the problem originally.  

 

I have updated all drivers. 

It's not heat. GPU maxes at 75 and CPU at mid 90s on max load. 

I have not done a stress test yet, but that is my next step. 

CPU hovers around 20-60%, doesn't really spike. 

RAM hovers around 50%, spikes at 80% when using lots of chrome windows. 

 

In case it matters, my previous GPU that appeared to fail was a 1080 TI. I had it on the same loop as my CPU and slightly underclocked (due to previous issues with unstable VRAM years ago). 

 

System specs:

Custom loop cooled overclocked i5 8600k (I think have it boosted to about 4.5 ghz)

ASUS z370-f

16 gb ram 3200

EVGA 3070

Corsair RM 850x

 

 

[GhostRoadieBL]

- is the card in the loop or can you swap PCIE slots? it's unlikely but EMI can cause instability with signal traces

- from the PSU are you using individual power cables or the single cable with the daisy chain? shouldn't be this but covers a loose pin or spiking current draw across a thin cable

- have you fully reinstalled the driver with DDU before installing the new GPU? possible windows is still polling for the 1080ti hardware ID through the driver with the new card installed

- In your bios you should be able to save the overclock profile and reset settings to bone stock to see if the instability is part of the overall motherboard power draw vs the previous setup.

 

[keltys]

1 minute ago, GhostRoadieBL said:

- is the card in the loop or can you swap PCIE slots? it's unlikely but EMI can cause instability with signal traces

- from the PSU are you using individual power cables or the single cable with the daisy chain? shouldn't be this but covers a loose pin or spiking current draw across a thin cable

- have you fully reinstalled the driver with DDU before installing the new GPU? possible windows is still polling for the 1080ti hardware ID through the driver with the new card installed

- In your bios you should be able to save the overclock profile and reset settings to bone stock to see if the instability is part of the overall motherboard power draw vs the previous setup.

 

- The new card is not in the loop. I modified it and bypassed the GPU when I got the 3070 since I plan on giving it to my wife and upgrading when the 3080ti comes out. I can definitely try swapping slots. 

- According to Geforce experience, it is correctly identifying the 3070. But I didn't do a clean wipe. I can try that as well. 

- I will try this too. Are you suggesting that the MOBO just can't handle the power draw of the overclock + the new GPU? 

 

[GhostRoadieBL]

14 minutes ago, keltys said:

- I will try this too. Are you suggesting that the MOBO just can't handle the power draw of the overclock + the new GPU?

with the EPS and PCI power both being 12V it's possible that spikes in load on the GPU which draws power from both the slot and the 8pins it can cause a slight droop in voltage to the VRMs which causes instability. This is very very unlikely but any dip in power can cause the 1-2 sec lag symptom. This is typically only seen on overclocked systems at the very edge of stability. 

[keltys]

16 minutes ago, GhostRoadieBL said:

with the EPS and PCI power both being 12V it's possible that spikes in load on the GPU which draws power from both the slot and the 8pins it can cause a slight droop in voltage to the VRMs which causes instability. This is very very unlikely but any dip in power can cause the 1-2 sec lag symptom. This is typically only seen on overclocked systems at the very edge of stability. 

That would make sense. The MOBO isn't exactly new or high end. The GPU switch upped max power draw by about 150wts. I'll dial back my OC and see if that helps. 

[keltys]

On 1/11/2021 at 4:40 PM, GhostRoadieBL said:

with the EPS and PCI power both being 12V it's possible that spikes in load on the GPU which draws power from both the slot and the 8pins it can cause a slight droop in voltage to the VRMs which causes instability. This is very very unlikely but any dip in power can cause the 1-2 sec lag symptom. This is typically only seen on overclocked systems at the very edge of stability. 

It seems undoing my OC has fixed it. I will still be upgrading my MOBO and CPU to something more modern soon, but this bought me a little time. Thanks again!

[GhostRoadieBL]

7 minutes ago, keltys said:

It seems undoing my OC has fixed it. I will still be upgrading my MOBO and CPU to something more modern soon, but this bought me a little time. Thanks again!

you can always walk up the overclock if you need those last few fps to get over 60 but most of the time it's going to be good enough without the overclock. 

[keltys]

On 1/13/2021 at 2:33 PM, GhostRoadieBL said:

you can always walk up the overclock if you need those last few fps to get over 60 but most of the time it's going to be good enough without the overclock. 

It seems I spoke too soon. I've been swamped this week, so I haven't had a chance to reply. The overclock walkback "seemed" to have fixed it, but probably just due to my usage being a little lower for a few days after. Since then the problem has gotten worse if anything. 

 

I've also discovered a few things.

 

So, my main screen is on display cable (1440p), my second is on HDMI 1080p. I only experience the graphical issues on my main screen. If I run a single screen and its my 1440 screen, the problems still occur, if I run just my 1080 they do not. 

 

I've had the following errors:

 

- both my screens turn black for about 10 seconds. My audio disconnects, then re-connects. This is the only thing that effects both screens. 

- My 1440 screen goes black from the middle down, and the upper half becomes pixelated and squished. 

- My 1440 screen frequently flickers black. 

- My 1440 screen lags and the refresh rate takes a huge dive. 

 

I have confirmed that it isn't the cable, the screen, or the individual ports on the GPU. I've also done a clean driver install. 

My last thing to try is to move the GPU to another PCI slot. Im doing that today. 

 

If that doesn't fix it, I think I'm forced to just start buying new parts... It seems incredibly doubtful that I would replace a broken GPU with another broken GPU, and seems more likely to me that there is something else causing these issues. The GPU is new and under warranty, as is the PSU. My gut says its some sort of a MB issue. It's the oldest part in the build, and not a great one to begin with.

 

[GhostRoadieBL]

Trying a second slot would be a good start. After that I'd be looking at rma the graphics card for a vram issue. The half screen issue typically indicates errors in the memory or slowly dieing vram chips. It's probably in the higher chips when you have more of the vram in use than when running 1080p.