2080 ti 2080 Ti loss of 30-50% performance

[alienwar9]

So my GPU was running fine for about a year, or at least it seems. I am unsure exactly when performance dropped significantly. I benched and OC'ed it when I got it and it ran as expected, but I have been running it and my CPU without OC for most of the time (too many other problems going on). I was playing Division 2 at 1440p max settings in HDR, and only getting around 60fps (have a 144hz monitor). That seemed out of place, so I ran Superposition with 1080p extreme and only got a 4600 score. I believe my original scores were 8000+, and others with the same card have gotten 120fps averages in Division 2 with the same settings. Superposition shows 1600-1700mhz clocks while running.

 

Got a 2080 Ti Fe with a waterblock and custom loop triple rad, also cooling a mobo monoblock, running on a DDC pump. The GPU is vertical mounted with a riser cable, and custom wired with cables I made myself. I've fresh re-installed drivers, nothing physically has changed inside the PC prior to the dropped performance. The temps immediately rise up to and max out at 57C, idle at 28C. I've set MSI Afterburner to reset to default.

 

I've called my PC Sisyphus cause of all the problems I've run into with it, pushing areas I don't have much experience on. I'm at a loss as to what has caused this issue, or where to take the diagnosis process next. I'm planning on taking the whole thing apart, but that's a pain and I hope it's something more easy to fix.

 

Could the riser cable go bad? Could the wiring not be supplying enough power? Could the thermal pads somehow have shifted on the GPU waterblock, or a VRM/something have fried? I don't even know how to check any of these things apart from taking everything apart.

[.Apex.]

17 minutes ago, alienwar9 said:

So my GPU was running fine for about a year, or at least it seems. I am unsure exactly when performance dropped significantly. I benched and OC'ed it when I got it and it ran as expected, but I have been running it and my CPU without OC for most of the time (too many other problems going on). I was playing Division 2 at 1440p max settings in HDR, and only getting around 60fps (have a 144hz monitor). That seemed out of place, so I ran Superposition with 1080p extreme and only got a 4600 score. I believe my original scores were 8000+, and others with the same card have gotten 120fps averages in Division 2 with the same settings. Superposition shows 1600-1700mhz clocks while running.

 

Got a 2080 Ti Fe with a waterblock and custom loop triple rad, also cooling a mobo monoblock, running on a DDC pump. The GPU is vertical mounted with a riser cable, and custom wired with cables I made myself. I've fresh re-installed drivers, nothing physically has changed inside the PC prior to the dropped performance. The temps immediately rise up to and max out at 57C, idle at 28C. I've set MSI Afterburner to reset to default.

 

I've called my PC Sisyphus cause of all the problems I've run into with it, pushing areas I don't have much experience on. I'm at a loss as to what has caused this issue, or where to take the diagnosis process next. I'm planning on taking the whole thing apart, but that's a pain and I hope it's something more easy to fix.

 

Could the riser cable go bad? Could the wiring not be supplying enough power? Could the thermal pads somehow have shifted on the GPU waterblock, or a VRM/something have fried? I don't even know how to check any of these things apart from taking everything apart.

Take a screenshot of Superposition in the middle of the run, also what's your CPU?

[Shimejii]

Max settings is always a horrible idea. By far one of the stupidest things to do no matter the hardware. 

 

Your GPU Die temps are fine, its can be either VRM temps if they are not being cooled properly, or just power. You can see in afterburner if its a power limitation. Put the power slider at max, see if there is any difference. Then see if you were overclocked before, test things like that.

[alienwar9]

9 minutes ago, _Syn_ said:

Take a screenshot of Superposition in the middle of the run, also what's your CPU?

CPU is Ryzen 2700 (I know. I had it half a year or more before I got the 2080 Ti that I wasn't planning on getting in the first place. Planning on upgrading when I redo my setup). Quick OC it to 4.1Ghz, but otherwise run it around 3.4-3.8.

 

 

[alienwar9]

7 minutes ago, Shimejii said:

Max settings is always a horrible idea. By far one of the stupidest things to do no matter the hardware. 

 

Your GPU Die temps are fine, its can be either VRM temps if they are not being cooled properly, or just power. You can see in afterburner if its a power limitation. Put the power slider at max, see if there is any difference. Then see if you were overclocked before, test things like that.

I was just seeing what it might run at. I had a few settings turned down when normally playing, including AA. It was just easier to compare with others on max/ultra, since I didn't have to checklist each and every setting.

 

When putting the power slider to max, I got 4400 on Superposition. But Sp has varied quite a bit, going from 4300 on the low end to 4900 on the high end. Doing OC scanner or manual OC seems to actually get lower results, but I'm also not being terribly stringent on background tasks (iCue might have been running during a test, for example). If I have more time I'll try to do more controlled testing, try and isolate some variables.

 

What gets me is that it has been running fine for a year, so something like VRMs randomly failing to be cooled feels out of nowhere. But I'm not 100% confident on my waterblock installation skills, so sure....maybe a thermal pad slowly squeezed its way out?

[.Apex.]

3 minutes ago, alienwar9 said:

CPU is Ryzen 2700 (I know. I had it half a year or more before I got the 2080 Ti that I wasn't planning on getting in the first place. Planning on upgrading when I redo my setup). Quick OC it to 4.1Ghz, but otherwise run it around 3.4-3.8.

The 2700 is alright as long as your expectations of it are reasonable, but this isn't a CPU bound scenario, performance is terrible but everything else looks fine.. it's weird

Try downloading GPU-Z and click on the "?" icon next to the "Bus Interface", that will test your PCIe slot to check what speed it's running at, it should say "@ x16 3.0" or at least "@ x8 3.0" when you start the render test
https://www.techpowerup.com/download/techpowerup-gpu-z/

 

You can also go to the Sensors tab in GPU-Z and it will list all the sensors on the card, if it's possible to run Superposition in windowed mode do that and take a screenshot of GPU-Z while it's running, and expand it so it shows all the sensors at once

[alienwar9]

13 hours ago, _Syn_ said:

The 2700 is alright as long as your expectations of it are reasonable, but this isn't a CPU bound scenario, performance is terrible but everything else looks fine.. it's weird

Try downloading GPU-Z and click on the "?" icon next to the "Bus Interface", that will test your PCIe slot to check what speed it's running at, it should say "@ x16 3.0" or at least "@ x8 3.0" when you start the render test
https://www.techpowerup.com/download/techpowerup-gpu-z/

 

You can also go to the Sensors tab in GPU-Z and it will list all the sensors on the card, if it's possible to run Superposition in windowed mode do that and take a screenshot of GPU-Z while it's running, and expand it so it shows all the sensors at once

So the Bus interface seems to be fine (@ x16 3.0). The sensor results, not so much. Looks like a power limit, running at 60% TDP, and only 158W.

 

[.Apex.]

4 minutes ago, alienwar9 said:

So the Bus interface seems to be fine (@ x16 3.0). The sensor results, not so much. Looks like a power limit, running at 60% TDP.

Do you have GPU overclocking software on your PC? I suggest you remove all of them and use DDU to reinstall drivers, OC software can mess with the registries, it doesn't seem like there's anything wrong with your GPU, but you can double check GPU VRM temperatures using HWiNFO, just click on "sensors-only" when you first open it

 

Should look like this

 

DDU:
https://www.guru3d.com/files-details/display-driver-uninstaller-download.html

 

HWiNFO:
https://www.hwinfo.com/download       
 

[alienwar9]

32 minutes ago, _Syn_ said:

Do you have GPU overclocking software on your PC? I suggest you remove all of them and use DDU to reinstall drivers, OC software can mess with the registries, it doesn't seem like there's anything wrong with your GPU, but you can double check GPU VRM temperatures using HWiNFO, just click on "sensors-only" when you first open it

 

Should look like this

 

DDU:
https://www.guru3d.com/files-details/display-driver-uninstaller-download.html

 

HWiNFO:
https://www.hwinfo.com/download       
 

Uninstalled all OC software, ran DDU, still the same results. I can't seem to find those sensors in HWiNFO. Is there some setting to change?

 

(Also, I think HWiNFO is conflicting with iCUE, as when it runs all my fans spin up full blast. Honestly, iCUE has been a neverending shitfest.)

[.Apex.]

2 minutes ago, alienwar9 said:

(Also, I think HWiNFO is conflicting with iCUE, as when it runs all my fans spin up full blast. Honestly, iCUE has been a neverending shitfest.)

You mean case fans? I don't know if iCUE has any control over GPU Power/Clocks

3 minutes ago, alienwar9 said:

Uninstalled all OC software, ran DDU, still the same results. I can't seem to find those sensors in HWiNFO. Is there some setting to change?

GPU VRM's should be in a different section, the ITE IT89xxxx.. you see in the previous image is the VRM controller, I don't think you have to change anything?

 

[.Apex.]

10 minutes ago, alienwar9 said:

Uninstalled all OC software, ran DDU, still the same results. I can't seem to find those sensors in HWiNFO. Is there some setting to change?

It seems like not all RTX 2080 Ti GPU's have VRM temp sensors, I thought all of them had it

 

But anyway it's unlikely for the VRM's to be overheating, the whole system would turn off, I'm confused on what the issue is

[xg32]

1 hour ago, alienwar9 said:

So the Bus interface seems to be fine (@ x16 3.0). The sensor results, not so much. Looks like a power limit, running at 60% TDP, and only 158W.

 

1680 core clock is very low

[jones177]

Your 2080 ti does not seem to be getting the power.

 

With My FTW3 Ultra the power target(limit) can get stuck at 66% and this can survive a reboot. Fortunately it does not survive switching the computer off then on again. 

When this first happened I thought it was the CPU causing it since I could not overclock it and the GPU at the same time. The problem I thought went away with a new CPU but came back 10 months later. I did not think it was the PSU because it ran two 1080 tis with no issues before the 2080 ti.

 

Both times this occurred I was doing benches with multiple monitoring/overclocking software. 

I also tested the card in another computer and it had no issues.

 

To test if you are having this type of issue use MSI Afterburner or Precision X1 to see if your power target/limit is stuck. 

 

6 minutes ago, _Syn_ said:

It seems like not all RTX 2080 Ti GPU's have VRM temp sensors, I thought all of them had it

 

But anyway it's unlikely for the VRM's to be overheating, the whole system would turn off, I'm confused on what the issue is

With test that I have done with my EVGA FTW3 Ultra the components that get the hottest are the vram between the VRM and the GPU. they can get 20c hotter than the GPU and 30c hotter than the VRM. 

 

Fortunately my overclock of + 800 on the ram does not add to this heat. 

 

 

 

 

 

[.Apex.]

17 minutes ago, xg32 said:

1680 core clock is very low

Clock is only 8-10% lower than a normal stock 2080Ti but he's getting 50-60% performance loss comparing the AVG FPS to this, the score difference is even higher

 

[alienwar9]

17 minutes ago, jones177 said:

Your 2080 ti does not seem to be getting the power.

 

With My FTW3 Ultra the power target(limit) can get stuck at 66% and this can survive a reboot. Fortunately it does not survive switching the computer off then on again. 

When this first happened I thought it was the CPU causing it since I could not overclock it and the GPU at the same time. The problem I thought went away with a new CPU but came back 10 months later. I did not think it was the PSU because it ran two 1080 tis with no issues before the 2080 ti.

 

Both times this occurred I was doing benches with multiple monitoring/overclocking software. 

I also tested the card in another computer and it had no issues.

 

To test if you are having this type of issue use MSI Afterburner or Precision X1 to see if your power target/limit is stuck. 

 

With test that I have done with my EVGA FTW3 Ultra the components that get the hottest are the vram between the VRM and the GPU. they can get 20c hotter than the GPU and 30c hotter than the VRM. 

 

Fortunately my overclock of + 800 on the ram does not add to this heat. 

 

 

 

 

 

When I run MSI Afterburner, turning power limit all the way up to 123, and temp limit to 88C, I end up getting a lower score (4300) than when running default (4900). TDP is still at 60%, 150W power.

 

...

 

Also, it seems 57C is 10-20C hotter than it should be running without OC and without CPU OC, with a CPU concurrently at 25C on the same exact water loop running right before the GPU (and that idles at 28C, so a 4C delta). I have a triple rad setup, 6 120mm fans on 2 rads plus a 240x60mm beefy rad (HWLabs), all fans on full blast, ambient temps in the room around 20C (Corsair LLs and HDs).

 

So it seems like the issue is not just power, but also temp.

 

The only other discussion I found with similar results was here: https://www.techpowerup.com/forums/threads/weird-throttling-issue-rtx-2080-ti.252138/

 

But the OP ended up just getting a new card. I guess I am still within the 3 year warranty period...I really don't want to go through that.

 

...

 

I guess I can do a full takedown and try the stock cooler with stock cables. See if that changes anything. Maybe plop the card in my other PC.

[jones177]

59 minutes ago, alienwar9 said:

When I run MSI Afterburner, turning power limit all the way up to 123, and temp limit to 88C, I end up getting a lower score (4300) than when running default (4900). TDP is still at 60%, 150W power.

 

...

 

Also, it seems 57C is 10-20C hotter than it should be running without OC and without CPU OC, with a CPU concurrently at 25C on the same exact water loop running right before the GPU (and that idles at 28C, so a 4C delta). I have a triple rad setup, 6 120mm fans on 2 rads plus a 240x60mm beefy rad (HWLabs), all fans on full blast, ambient temps in the room around 20C (Corsair LLs and HDs).

 

So it seems like the issue is not just power, but also temp.

 

The only other discussion I found with similar results was here: https://www.techpowerup.com/forums/threads/weird-throttling-issue-rtx-2080-ti.252138/

 

But the OP ended up just getting a new card. I guess I am still within the 3 year warranty period...I really don't want to go through that.

 

...

 

I guess I can do a full takedown and try the stock cooler with stock cables. See if that changes anything. Maybe plop the card in my other PC.

Thanks for the link. 

 

It is not the power used(watts)and it is not the power limit.  It is the power% used. It has its own graph.

As you can see the power % graph jumps around a lot during a bench but when I get stuck at 66% it is a flat line like it is when the GPU is at idle.

Here is the same GPU with an overclock and the power limit turned up.

 

I have had lots of strange issues with my 2080 tis that did not happen my 1080tis or 980tis. 

 

Before doing an RMA test the card in another system. My FTW3 Ultra would have been RMAed but I put it in my computer that had the XC 2080 ti first and it ran perfectly.

My hardware likes to make a fool out of me any chance it can get.

 

 

[Dean.P.]

my 2080ti aircooled and overclocked runs at 59C yours seems a bit hot. Also as mentioned the frequency is way down you should be like 2000MHz+ if OC'd. Maybe look into your PSU perhaps that could be indicating that the GPU is not getting enough power.

[alienwar9]

3 hours ago, jones177 said:

Thanks for the link. 

 

It is not the power used(watts)and it is not the power limit.  It is the power% used. It has its own graph.

As you can see the power % graph jumps around a lot during a bench but when I get stuck at 66% it is a flat line like it is when the GPU is at idle.

Here is the same GPU with an overclock and the power limit turned up.

 

I have had lots of strange issues with my 2080 tis that did not happen my 1080tis or 980tis. 

 

Before doing an RMA test the card in another system. My FTW3 Ultra would have been RMAed but I put it in my computer that had the XC 2080 ti first and it ran perfectly.

My hardware likes to make a fool out of me any chance it can get.

 

 

I'm waiting for a friend to test out in his rig, as my alt PC is so thoroughly stripped that I might have to melt some case parts to get it back together. 

2 hours ago, Dean.P. said:

my 2080ti aircooled and overclocked runs at 59C yours seems a bit hot. Also as mentioned the frequency is way down you should be like 2000MHz+ if OC'd. Maybe look into your PSU perhaps that could be indicating that the GPU is not getting enough power.

Tried to pull the PSU from my old PC, but it has 2x6 pin connectors instead of 8, and I only have 1 adapter cable. I'll have to wait. That's my first test. If that doesn't work, I'll have to drain the loop and swap the air cooler back on. 

 

This PC is a nightmare to drain, though, and that's with a drain port at the lowest point. I have to cartwheel the PC to get it all out. 

[alienwar9]

An update on testing. Managed to pull the card out and test it on my older PC with its original air cooler. Worked perfectly.

 

Then tested it with the air cooler on my current PC. Worked perfectly. Tested it with the PCI-e riser cable. Worked perfectly. Then plopped a different waterblock I had for it; did a quick loop. Worked perfectly. 

 

So I'm not entirely sure what was even wrong with it. Could have been the waterblock got unseated somehow, but that's unlikely given the number of screws tightly clamping it together, and the fact that it didn't go past 57C. 

 

Could have been the riser cable somehow unseated partially, though that doesn't explain the momentary 100% power usage (same as putting to question the possibility of the custom power cables being at fault). 

 

I honestly have no idea what went wrong. But I'm glad it is somehow fixed.