Got a used RTX 2060 Super, just want to make sure these stress tests check out

Joined
Jan 3, 2009
Messages
644
I got a HP OEM RTX 2060 Super. Since this is going in an older Dell system anyway I was preferring to try to get a smaller card like this to make sure it will fit (And as a bonus the 8-pin connector is on the side like the current card in the Dell is instead of the top which I wasn't sure I would have the clearance for) and OC capability was not a big concern since I will likely be CPU-limited by the Xeon 2667v2 in it anyway.

(To clarify, the card is currently in my main system, a 11700K system in a Cooler Master HAFX case and 64GB of RAM while I am testing it.)

Specifically the GPU looks like this:


GPU-Z seems to verify all the specs appear correctly:


Although looking the card up on TechPowerup just gives me something generic... and the wrong number of output ports:


Anyway, so first thing I did was run Furmark for 30 minutes to see if there are any issues. So far so good it seems, but I am not certain on my core clocks and temps:



I heard that the card should be hitting 100% load, but the load is at 96% and stays there. Also the memory is running at max clock speed but the GPU core is going between 1440 and 1475, mostly staying at 1455. That is the base speed... but not the boost speed. I also noticed that the temperature pretty quickly rose to 80C and stays there. I tried looking up what normal temperatures are and I saw someone mention max is 88C but I am not certain if that was accurate.

Is this normal behavior for a Furmark test? Or is this card thermal throttling? Wouldn't be too big a surprise considering the compact one-fan design, but the TechPowerup page says it should be able to go a bit faster on the GPU core. I would replace the pads and paste, but I can't find any specifics on this exact model card to know how to correctly take it apart and what pads I would need, I don't have any on hand. Plus since it's a 2019 card would it even be time yet to replace the pads/paste? I recently re-pasted/padded my old EVGA GTX1070 (Mostly due to that VRM overheating defect) but even the paste on that old thing was not dried out.

Also tried MSI Kombustor's stress test (Not even sure which option to choose as there are a lot), I didn't see it mention my GPU's clock speeds but with the default test it also had the temps hover around 81C and load was at 98%. TDP was varying between 96% and 106%, sometimes dips to lower or a spike to 108%.

So far it SEEMS to be operating like it should although it doesn't appear to be boosting (No idea if that's an OEM lockout or thermal throttling) but I wanted to make sure. Also, all of these seem mostly designed to stress the GPU core, is there any good software to stress test the VRAM to make sure it's operating fine? I tried looking for a list of such applications but they suggested software like Furmark or Kombustor which did not seem close to using all of my VRAM (Kombustor specifically mentioned only using 2GB) and the others all seemed to be random old applications from a decade or more ago and even had broken links, and when I tried them they only tested half of my VRAM.

Two things that do worry me so far though. 3dMark's Stress Tests all failed (and by failed I don't mean they crashed or artifacted, but gave me a "Not Passed" score) at around 95%, 3DMark's Stress Tests generally expect a 97% or more for passing. I know these tests are for framerate stability, so I don't know if this is because OEM cards are more likely to have instable frames from their lowered cooling/TDPs or if this is not normal, or even something to be worried about. And another is that the Port Royal benchmark had parts where the shadows would stutter/jitter. This did worryingly look like something wrong, but what's weird is that it was not random, they always happened in the same place in the same way.

I posted a video of it here:


The timestamps where it jitters are in the description.
 
Last edited:
Your card is fine. OEM coolers are not designed for performance. They are designed to be good enough and cheap. Usually you get a basic aluminum cooler and a couple of heat pipes.

You have GPUZ so open the sensors and they will show you what's holding the card back - thermals or power, etc.

You ran Furmark and it didn't crash or artifact. That's going to push the GPU harder than you likely will using it regularly.

Heat and power are your two main issues. I would definitely inspect the cooler mount and change the paste. Using a name brand paste may help drop the temps a few degrees alone. OEM cards often use reference designs. If your VRAM is running hot, you can buy a couple pads based on common thickness and change those at the same time. If VRAM is not an issue you don't need to change the pads. Also, make sure there is adequate cooling in the case or flowing across the card. The heat coming off it needs to go somewhere else.

Run MSI Afterburner. You can see if there is any power left to push it further and us their auto OC feature to create a custom curve for your card based on what it can do with power and thermals. Remember, you are still using an OEM cooler, so don't expect miracles.
 
Using a name brand paste may help drop the temps a few degrees alone.

I have some Noctua NT-H2 left over I could use. The issue is...

If your VRAM is running hot, you can buy a couple pads based on common thickness and change those at the same time. If VRAM is not an issue you don't need to change the pads.

I have no idea how many pads it uses, what sizes, and what thicknesses. I heard that some cards can have pads of multiple thicknesses in various areas.

Also, make sure there is adequate cooling in the case or flowing across the card. The heat coming off it needs to go somewhere else.

Part of the reason I still like to use this old case. Two 200mm fans on the top, one 140mm fan in the rear, one 200mm fan on the side (Why don't we have side-fans anymore? They help a lot) and a 230mm fan on the front.... ok, so it's a little annoying that one is a bit more proprietary but the case has a lot of airflow.

Run MSI Afterburner. You can see if there is any power left to push it further and us their auto OC feature to create a custom curve for your card based on what it can do with power and thermals. Remember, you are still using an OEM cooler, so don't expect miracles.

Not really looking to OC, it's going in an older Dell with a 2013-2014 era Xeon after all, just wanted to make sure it wasn't preforming under what it should when the card was new stock or if the card itself wasn't giving me issues.
 
Not really looking to OC, it's going in an older Dell with a 2013-2014 era Xeon after all, just wanted to make sure it wasn't preforming under what it should when the card was new stock or if the card itself wasn't giving me issues.

You can run MSi afterburner to undervolt the gpu and get a much reduced power consumption and better performance due to higher and more stable clocks.

I know its a tier down, but I can do some test if you need or want from my daughters machine which is using a regular Zotac RTX 2060 which doesn't have a top end cooler or binned chip.

About the undervolting, it gave to that card much more stable performance while reducing power about 15%, so less heat, less noise, lower temps, its a win in any way.
 
If I were to undervolt it would probably be best to do it in the system the card is eventually going to go into than the much more powerful one that I am currently testing the card in, I am going to be testing the card in the current system for a while as I have other upgrades to do on the system it will eventually be in first.
 
So 3DMark got back to me on the Steam forums telling me that the glitchy shadows is a known bug with the Port Royal benchmark, and that they don't know how to fix it without invalidating all past benchmark results of it.

Anyway another thing I realized now is that while these benchmarks and stress tests help test that the card is working properly and how it performs... it's not really giving me a way to compare how it performs to others of the same card.

I wanted to know if this card is performing as expected, and I was told that despite how horrible the site is, that Userbenchmark's test is a good one to determine if my card is performing as it should compared to other 2060 supers. So I ran a test, but I don't really understand the results. It claims it's underperforming, yet it also claims it's performance is at 100%?

Can anyone help me determine what these results mean? https://www.userbenchmark.com/UserRun/59330664#GRAPHICS_CARD

Since its a cheaper OEM card, the kind that is lower profile and has one fan, I am not expecting it to perform like some EVGA Superclocked version of the card, I just want to make sure that it's not drastically underperforming either due to wear, thermal issues, or being a fake.

I was also told that 3DMark apparently can do such comparisns showing how my specific card performs vs others of the same model? But I can't find that option anywhere.
 
Userbenchmark is telling you your card is running in the middle of the pack of 2060's. So, you're fine. As you pointed out, you don't have a top end, water cooled, superclocked gpu. Nor do you have one that's broken or gimped.

3D mark gives you a basic comparison on the left hand side of the results page and will let you compare results online clicking the link.
1676768531688.png
 
Back
Top