Question EVGA RTX 3080 Ti Underperforming

[Aazure]

Hi all, coming here again for some more help regarding my GPU. I was curious about my PC's performance after my friend with the same GPU and similar build and I were playing Battlefield 6 and his system was performing much better than mine at the same preset settings, so I ran a quick benchmark to take a look.

I know there's a lot of controversy regarding UserBenchmark but I ran their tool to see if I could get a general idea of where my system was underperforming and it turns out my GPU was scoring very low relative to other 3080 Ti's. I know that UserBenchmark doesn't take into account overclocking of any kind so it's not the best way to compare, but my GPU was in the bottom 5th percentile of their 3080 Ti scores.

https://imgur.com/a/vLgJrIC

I used DDU to remove GPU drivers and directly installed the latest from NVIDIA (580.88) and I then ran the benchmark again with HWiNFO open to see if there were any issues with thermals. No thermal throttling was detected and hotspot temps were all normal (still received same score), so I'm confused what could be causing the low performance. If anyone has any suggestions about what could be causing this issue it would be greatly appreciated!

PC Specs:
CPU: Ryzen 9 7900x
CPU cooler: NZXT Kraken Elite (2023)
Motherboard: ROG Strix B650E-F Gaming Wifi (BIOS ver. 3265)
Ram: Corsair Vengeance RGB 32gb 6000Mhz DDR5
SSD/HDD: WD Black SN850x 2TB (boot), Samsung 980 Pro 2TB (games)
GPU: EVGA RTX 3080 Ti
PSU: Corsair RM850x (purchased and installed November 2023)
Chassis: Hyte Y70 Touch
OS: Windows 11
Monitor: LG 27GP850-B (main), Acer XF270H (side), Hyte Y70 Touchscreen

 

[Third-Eye]

Don't use userbenchmark to compare your performance, because that place is a cesspit of unreliable data points now. Run actual game benchmarks at minimum and compare that to other review benchmarks that used the same benchmarks with similar hardware configurations to your own.

 

[Eximo]

Userbenchmark is bad for comparisons between various brands/models/etc. Like for like it is good enough to find big issues. Here a bottom of the expected range 3080 Ti is abnormal. With a bell curve you would expect most people to land in the middle.

So to follow up, run another benchmark, like 3D Mark TimeSpy or one of the newer ones. And check there, but while doing so, monitor temperatures. The card may be fine, could be your case airflow, or you need to re-do the thermal paste on the GPU. Pay particular attention to your memory temperature.

HardwareMonitor64 and GPU-Z will be valuable to you for keeping an eye on things. GPU-Z will show the perfcap reason as well. So if you are bumping up against power limits, that will show too.

It is a newer game, might not be super optimized for the older 30 series cards either.

 

[Aazure]

Userbenchmark is bad for comparisons between various brands/models/etc. Like for like it is good enough to find big issues. Here a bottom of the expected range 3080 Ti is abnormal. With a bell curve you would expect most people to land in the middle.

So to follow up, run another benchmark, like 3D Mark TimeSpy or one of the newer ones. And check there, but while doing so, monitor temperatures. The card may be fine, could be your case airflow, or you need to re-do the thermal paste on the GPU. Pay particular attention to your memory temperature.

HardwareMonitor64 and GPU-Z will be valuable to you for keeping an eye on things. GPU-Z will show the perfcap reason as well. So if you are bumping up against power limits, that will show too.

It is a newer game, might not be super optimized for the older 30 series cards either.

Yea, that was the only reason I brought up UserBenchmark -- it just seemed way too low of a score. I went ahead and ran 3D Mark TimeSpy via the demo while logging with GPU-Z. I'm not exactly sure how to interpret the information in the log but my score on TimeSpy is actually slightly above average for similar builds.

https://imgur.com/a/hizLK3q

I'm unsure if I'm allowed/able to post the file of my GPU-Z log but I do have that available if it's helpful.

As for thermals, I have 3x bottom intake fans feeding directly into my GPU along with 3x side intake fans. I have 1x back outtake fan and 3x outtake top mounted with my radiator. I have been thinking about repasting the card but my temps didn't look out of the ordinary so I thought I was fine. I may just repaste to be safe now.

Regarding my friend and I's performance in BF6: the only reason I brought it up was because while he was averaging 130 FPS on the low preset settings, I was averaging about 100-110 on the same settings (both DLSS off). We both game at 1440p and have the same 3080 Ti model as well. The main differences in our rigs is he runs a 13700k with 64gb of RAM. For fun, I had him run a UserBenchmark test as well where he scored in the 86th percentile for 3080 Tis.

https://imgur.com/a/b5AkgJG

I don't want to make him spend too much of his time on this but I may ask him to run a TimeSpy as well to see if we differ wildly there too.

Edit: Got my friend to do a TimeSpy run and weirdly enough, his graphics score is lower than mine. I'm not sure what's causing these discrepancies because our GPUs seem to be operating around the same parameters when using this benchmark.

Embed isn't working, here's link:

https://imgur.com/a/timespy

View: https://imgur.com/a/timespy-results-hizLK3q

 

[Unolocogringo]

Userbenchmark favors Intel based systems and their conclusions are baffling.
What temps are you seeing? GPU memory and hot spot?
What boost clocks do you get?
From the graphs it looks to be 1750 +/- ?
As a suggestion a little more cooling can go a long way.
A custom fan curve for the GPU can drop temps a lot without creating excess noise.
Most 3000 series cards would clock in around 2000mhz boost clocks when cooled well.

 

[boju]

Does BF6 require latest drivers? Not sure the current state of Geforce drivers for 3rd/4th gen cards as there were performance and stability issues throughout numerous driver additions surrounding RTX 50 series. Find out what version your friend is using. If BF6 doesn't require you to use stipulated drivers then try 566.36, this driver is considered the gold standard for our cards, which im still using for my 3080ti.

 

[Aazure]

Does BF6 require latest drivers? Not sure the current state of Geforce drivers for 3rd/4th gen cards as there were performance and stability issues throughout numerous driver additions surrounding RTX 50 series. Find out what version your friend is using. If BF6 doesn't require you to use stipulated drivers then try 566.36, this driver is considered the gold standard for our cards, which im still using for my 3080ti.

He was also on 580.88 during BF6 as well as the benchmarks he ran to compare with me. It's in the screenshot of his UserBenchmark results but it's a bit blurry.

 

[Aazure]

Userbenchmark favors Intel based systems and their conclusions are baffling.
What temps are you seeing? GPU memory and hot spot?
What boost clocks do you get?
From the graphs it looks to be 1750 +/- ?
As a suggestion a little more cooling can go a long way.
A custom fan curve for the GPU can drop temps a lot without creating excess noise.
Most 3000 series cards would clock in around 2000mhz boost clocks when cooled well.

In Time Spy during Graphics Test #1/2, these were my peak temps according to GPU-Z:

• GPU Temperature - 70.3 C
• Hot Spot - 79.1 C
• Memory Temperature - 80.1 C

I normally run an undervolt but was experiencing instability with BF6 specifically so I reset to stock settings for my card. My friend also runs his card at stock because he doesn't like to mess with this type of stuff, which is why I kept my card at stock as well during the benchmarks. And yes the default boost clock on this card is 1800.

I'm mainly concerned about the discrepancies between two seemingly identical cards running at similar temperatures, but I will keep in mind the cooling.