[SOLVED] Is there something wrong with my 3080????

[AnonymousEnthusiast1]

I am getting very inconsistent scores when benchmarking my EVGA FTW3 Ultra 3080. I know userbenchmark isn't a great tool to replicate this, but I am getting no change in effect when overclocking or underclocking the card vs stock. It's overall underperforming. It's the bottom 4-17 percent with my setup on timespy extreme. These benchmarks from userbenchmark show that my overclock is performing about the same or even worse than not having an overclock. Do you think I am running out of room on my power supply? I am pulling max 240 watts with my overclock on my 10850k and I have heard the GPU alone can spike to 460 watts. I am running a 750 watt 80+ gold power supply with a ton of RGB fans and some RGB strips. Any help would be appreciated.

https://www.userbenchmark.com/UserRun/35636562 (1000 memory OC with 100 core frequency OC)
https://www.userbenchmark.com/UserRun/35636977 (0 memory OC with 0 core frequency OC)

System:
CPU: i9-10850k 5.1ghz all 10 cores (pulls max 240watts)
CPU Cooler: Cooler Master Masterliquid ml360r argb (push/pull configuration)
GPU: EVGA FTW3 Ultra 3080
Motherboard: Asus Rog Strix z490-E gaming
Power Supply: EVGA 750W SUPRNVA G1+ 80+Gold FM
Case: Cooler Master h500p mesh (came with rgb 200mm fans and swapped them to the top with the 200mm argb fans)

 

[Zerk2012]

If you look at the DX 9 and DX 10 FPS they both went up with the overclock.
And yes that is a rather poor site.

 

[rgd1101]

cpu/gpu temp?
take out all the rgb light?

 

[AnonymousEnthusiast1]

If you look at the DX 9 and DX 10 FPS they both went up with the overclock.
And yes that is a rather poor site.

Yeah, I understand, but at the same time I mentioned timespy being rather low scores. I should have mentioned that with the same CPU and GPU they had higher scores despite having lower core clocks on the CPU and GPU as well as memory. Firestrike also shows my GPU on the lower end of the spectrum despite not crashing anywhere between 0-1000 on the memory and 0-120 on the core. It does crash after 130 on the core because it boosts past 2130mhz.

 

[AnonymousEnthusiast1]

cpu/gpu temp?
take out all the rgb light?

The CPU never gets above 85 degrees at full synthetic load and is not ramping the fans up when benchmarking so I am assuming not above 75 degrees. The max the GPU got was 77 degrees with the 1000 mem and 120 on the core. I used 100 on the core just in case it wasn't stable since it crashed at 130.

 

[AnonymousEnthusiast1]

I did some power testing. I am pulling 711 watts during timespy with 1000 on mem and 100 on core. The CPU is pulling 189 watts and the GPU is pulling 393 watts. The other system components are accounting for the higher power draw. I think I'm going to upgrade my PSU for extra headroom.

 

[RTX 2080]

I did some power testing. I am pulling 711 watts during timespy with 1000 on mem and 100 on core. The CPU is pulling 189 watts and the GPU is pulling 393 watts. The other system components are accounting for the higher power draw. I think I'm going to upgrade my PSU for extra headroom.

Seems like a good idea. I personally have had great results with my Corsair AX1000. The extra headroom would mean that your higher up on the efficiency curve of your PSU, and the Titanium rating also means that the curve is higher overall. The extra cushion of available wattage doesn't hurt either.

 

[AnonymousEnthusiast1]

I have figured out the issue guys. It was a software issue and not hardware. It took a while to research what was happening.

*Final score results were with +750 on mem and +100 on core clocks.

1. The first issue with the frame jumping had to do the EVGA's Precision X1 software. I didn't know you can overclock the card and then close the program and it will keep the overclock until the next restart (I also verified that it will keep the overclock after waking the PC up from sleep mode). Games and benchmarks run normally with that program closed. Without the stuttering in the benchmarks, my GPU scores in Time Spy increased from 15700 to around 18800.

2. Windows 10 power management was holding back my CPU scores significantly. There is an Ultimate Performance mode that's hidden in Windows 10 for some or most people. This increased my CPU scores SIGNIFICANTLY. My 10850k was getting 9500 scores in Time Spy when it should be getting at least 13000. I switched to the ultimate plan mode and my CPU score increased to 13500 with all 10 cores running at 5.1ghz. This is inline with the 10900k at 5.1ghz at around 13200 scores. I am not sure how people are getting 14800 or above with lower clocks than mine in Time Spy, but I don't think there is an issue with my CPU and windows 10 anymore.

This is the command you will have to run to show the ultimate power plan if it's not showing. I ran mine in the windows 10 search bar and it worked for me.

powercfg -duplicatescheme e9a42b02-d5df-448d-aa00-03f14749eb61

There are two ways to get to the power options in Windows 10. You can go directly to the control panel > System and Security > Power Options or you can search "power & sleep settings" in the windows 10 search bar and then scroll down a little bit and click "Additional power settings" link under Related settings.

3. Conclusion. My overall scores before doing these two steps were 15290 for Time Spy and around 29500 for Fire Strike. After those two changes, I reached an overall score of 17774 in Time Spy and 31842 in Fire Strike. These aren't the top of the line results, but they are the expected results that I wasn't getting before and thinking that my GPU was faulty when it wasn't the issue. My results compared to similar CPU and GPU in Time Spy before these changes was the bottom 4-14 percent. They are now Top 12 percent checked on 11/21/2020 for Time Spy and Top 22 percent in Fire Strike. Link below to the two scores I was referring to.

https://www.3dmark.com/3dm/53336261? (Time Spy run)
https://www.3dmark.com/3dm/53336612? (Fire Strike run)