Question EVGA RTX 2060S sc Ultra underperforming, time spy score of 5925

[Spenser_R]

I have had this issue where when I play games with friends, my buddies rtx 2060 performs better than my rtx 2060 super, this does not make sense to me, even youtube videos people have uploaded with the same card perform significantly better than mine. I'm not sure why this is happening. I do have one lead:
EVGA RTX 2060 SUPER SC Ultra Specs | TechPowerUp GPU Database
Listed under the "Theoretical Performance" are 5 different values, when I do an AIDA64 Extreme analysis of my computer, these values are noticeably lower.

One other bit of helpful information, FurMark benchmark shows my card averaging 96fps with OC on the 1080p preset, and about 84-85fps without OC on the same preset.

I do have XMP (profile-2) enabled and have been tinkering with a bunch of different settings in the BIOS.

Installing the studio ready drivers for my card seem to increase performance in games compared to the game ready drivers.

Pc Part Picker Build:
https://pcpartpicker.com/list/HC2JpH
Time spy score:
NVIDIA GeForce RTX 2060 SUPER video card benchmark result - Intel Core i9-11900K Processor,ASUSTeK COMPUTER INC. ROG MAXIMUS XIII HERO (3dmark.com)
Linus Tech Tips Thread that I started to help identify the issue, not much changed:
Issue with EVGA RTX 2060 super sc ultra, low performance compared to same card in other builds - Graphics Cards - Linus Tech Tips

A lot of my benchmarks that I have taken pictures of are listed on the Linus Tech Tips thread that I have listed above.

Any help would be very appreciated,
- Spenser_R

 

[Phaaze88]

Whatever the problem is, it's not an obvious one.
-core and hot spot look fine.
-it looks to be boosting normally.
-power draw looks fine.
-PCIe speed... ok.
3D Mark has bar graphs. You should've taken a pic of that too.

You are going to need to monitor the card's behavior in real-time. Something may be happening that can't be caught in the static screenshot.
If you have a multi-monitor setup, run Gpu-Z(sensors, stretch the window out so all parameters are visible at once) or HWINFO(double click on gpu clock, memory clock, gpu voltage, and gpu power - double click makes graphs pop up).
If you aren't running multi-monitor, you'll have to run the games in windowed/borderless window so you can see the graphs while you play.

 

[Spenser_R]

Whatever the problem is, it's not an obvious one.
-core and hot spot look fine.
-it looks to be boosting normally.
-power draw looks fine.
-PCIe speed... ok.
3D Mark has bar graphs. You should've taken a pic of that too.

You are going to need to monitor the card's behavior in real-time. Something may be happening that can't be caught in the static screenshot.
If you have a multi-monitor setup, run Gpu-Z(sensors, stretch the window out so all parameters are visible at once) or HWINFO(double click on gpu clock, memory clock, gpu voltage, and gpu power - double click makes graphs pop up).
If you aren't running multi-monitor, you'll have to run the games in windowed/borderless window so you can see the graphs while you play.

Thank you for the reply, I can get you a picture of the graphs from 3DMark, I have looked at them in the past and have noticed nothing abnormal. I'll quote myself I'm a moment once I restart my computer and run the benchmark again, I will be doing it with no overclock applied.

 

[Spenser_R]

Thank you for the reply, I can get you a picture of the graphs from 3DMark, I have looked at them in the past and have noticed nothing abnormal. I'll quote myself I'm a moment once I restart my computer and run the benchmark again, I will be doing it with no overclock applied.

Here is a picture that I took from the 3DMark benchmark, nothing seems out of the ordinary. Maybe you notice something?
I forgot to take a picture of HWMonitor, but I was watching it closely and again, nothing seemed abnormal, only some fluctuation in GPU power, but that's expected I believe,
unknown.png (1514×927) (discordapp.com)

Edit: Sorry I forgot to link the results
NVIDIA GeForce RTX 2060 SUPER video card benchmark result - Intel Core i9-11900K Processor,ASUSTeK COMPUTER INC. ROG MAXIMUS XIII HERO (3dmark.com)

 

[Phaaze88]

Negative. I see nothing especially unusual from those bar graphs... well, there's a 90% gpu load dip before the 1:40 mark, but none of the other bars appear to have corresponded with it.
I mean, load dropped to 90% in that brief moment, but none of the clock frequencies budged at all, and neither temperature graphs showed any sporadic change.


The card appears to be 'fine'... it's functioning as it should, but 3D Mark's average is suggesting otherwise.
Makaroni from the other thread may be spot on about the card's vbios.
You've already tried a Display Driver Uninstaller(safe mode) to fresh driver install?

One other thing that may fly under the radar like this is an unstable overclock. You may not have tinkered with this - in that case, a card like this will have a factory overclock. They only adjust the core clock though; memory clock is left alone.
Take EVGA's Precision software, apply a -50mhz downclock, and test it out. Then try -100mhz and test again.

 

[Spenser_R]

I did what you suggested, I actually did another wipe with DDU and re installed the drivers, no such luck. However, I did notice something when I underclocked it, although it was very minimal and was such a tiny amount, I did see a small increase in fps in some applications.

I know this picture might be a little weird to read, but I tried to make it as simple to interpret as possible, x,y relates to as follows:
x = core clock speed
y = memory clock speed
This was a simple test to determine how fps is affected by the clock speeds.
unknown.png (1169×923) (discordapp.com)

Left is -100 core clock and right is +100 core clock speed, difference is minimal.
unknown.png (773×376) (discordapp.com)

I'm at a loss on how I should go about this. Tomorrow when I wake up I'll do a more Indepth analysis of my computer like you were saying about the downclocking, where I run 3DMark each time I adjust clock speeds.

 

[Phaaze88]

Please don't use Furmark for this. Keep using 3D Mark.
I don't trust it for anything more than seeing how good the gpu's cooling solution is, and even there, other apps do it better.

 

[Spenser_R]

I'll stop using furmark, I was just using it because it was simple to pump out tests for average fps.

 

[Spenser_R]

Update:
I found this thread and decided to follow the instructions he gave to try and isolate the problem more.
even weirder is that the clock speeds are insanely low
See: WTF.png (802×561) (discordapp.com)
Latest Benchmark shown here:
unknown.png (1519×934) (discordapp.com)
NVIDIA GeForce RTX 2060 SUPER video card benchmark result - Intel Core i9-11900K Processor,ASUSTeK COMPUTER INC. ROG MAXIMUS XIII HERO (3dmark.com)

 

[Phaaze88]

even weirder is that the clock speeds are insanely low
See: WTF.png (802×561) (discordapp.com)

That's actually normal low load behavior.
My 1080Ti just sits at 1488-1518(depending on power plan) core clock.
^That's not normal, but a fault with Nvidia's driver regarding power management with multi-monitor setups, and they never seemed to set it straight.
It's only using around 60w of power, so not that bad, I guess. It used to be a little worse.

The most notable thing with the latest benchmark is that the gpu's core and memory clocks relax when the gpu isn't really doing anything, but that's a normal function.
I am quite stumped with this one.
No obvious signs of thermal nor power throttling.
It's holding boost clock well over the base when under full load.
Memory clock is straight as an arrow.
The card is 'fine', but the 3D Mark average and UBM are saying otherwise...


Do you have Resize bar enabled by mistake?
Is the card still under warranty?

 

[Spenser_R]

I have been toggling resizebar on and off as I have been doing testing, none of the data that I have showed has it enabled, it never affected any results. As for the warranty, this was an RMA card as of may 2022, it has not even been 3 months so I assume it's under warranty, I'm contacting EVGA as we speak, but they take so long to respond that I resorted to this and the other forum post I made.

 

[Spenser_R]

I'm going to be getting the card RMA'd and I'll get back to you if the issue persists, If I still have the issue, I might end up getting a new set of ram, I'll link it below if you think that It could help my situation, Its the last thing I was planning on upgrading apart from my gpu, personally I think this ram kit will be beneficial towards getting the most out of the rest of my components. Thanks again for your help

G.SKILL TridentZ RGB Series 32GB (2 x 16GB) 288-Pin PC RAM DDR4 4600 (PC4 36800) Intel XMP 2.0 Desktop Memory Model F4-4600C20D-32GTZR - Newegg.com

 

[Phaaze88]

Wow, those primary timings are looser than a-

Anyway, I'm not a memory expert. But uhh... from what bits I do recall [I look at some of Buildzoid's videos on occasion, though some of what he says does fly over my head], I gather that Rocket Lake is kinda busted with ram.

View: https://www.youtube.com/watch?v=nNkLRHFT72M

 

[Spenser_R]

After reflecting on the timings of the ram kit, I had this one originally in my watch list, I think this might be the better choice, I also found a blog post on performance differences between the different clock speeds, which gear mode it was in, etc. I think the performance is minimal on 11th gen, I'm mostly interested in heavy desktop use, IE: game development, scripting, etc.
All in all, I don't see an issue with what I chose, but more research is going to be done up until that point, and I'm planning on waiting for a sale anyways.
Intel Rocket Lake RAM Speed Analysis: Best RAM for i9-11900K? (premiumbuilds.com)
G.SKILL TridentZ RGB Series 32GB (2 x 16GB) 288-Pin PC RAM DDR4 4400 (PC4 35200) Intel XMP 2.0 Desktop Memory Model F4-4400C19D-32GTZR - Newegg.com