Unable to reach same RTX 3080 OC that I could in 2023

[King Mustard]

I have an RTX 3080 Founders Edition.

Back in August 2023, I posted the following 3DMark Time Spy score:

https://www.3dmark.com/spy/40882941
GPU Core: 1,995 MHz (avg 1,914 MHz)
GPU Memory: 1,219 MHz (avg 1,210 MHz)
GPU score: 18,297

I posted a new score yesterday:

https://www.3dmark.com/spy/53043106
GPU Core: 1,590 MHz (avg N/A)
GPU Memory: 1,372 MHz (avg N/A)
GPU score: 17,412

To achieve yesterday's 17,412 score, I used MSI Afterburner to max out the Power Limit (115%), max out the Temp Limit (90C), and set the Core Clock to +151 MHz and Memory Clock to +1475 MHz. I could not push the Core or Memory any further without crashing.

I am trying to understand what I would have done differently back in 2023. I would have only use MSI Afterburner then, too. I am not an expert OCer, I just enjoy the basics.

Both the CPU and GPU were under 70C to get yesterday's 17,412 score, so I don't think I had thermal issues.

The Core seems to have dropped a lot since 2023 - I am not sure why.

Does anyone have any ideas?

 

[EarthDog]

I see your memory is way higher than the first run. Maybe lower the memory speeds and try raising the core?

What card, exactly? What clocks is it supposed to have? Maybe post a picture of gpuz...

 

[King Mustard]

I see your memory is way higher than the first run. Maybe lower the memory speeds and try raising the core?

What card, exactly? What clocks is it supposed to have? Maybe post a picture of gpuz...

Sorry, it is a Founders Edition.

I did try a Core of +155 MHz and a default Memory Clock (+0 MHz) but the system crashed.

It appear anything higher than a Core of +151 MHz crashes, no matter than Memory Clock.

 

[Woomack]

New drivers use more graphics card resources and cause worse OC. At least it happened about a year ago after some updates. There is some performance gain, but worse OC.
It can be something else, but I would check the same driver version if you can still download it.

 

[Iron Bars]

How did you rule out other software issues?

 

[=XAF=AfterShock]

I would run the test again with gpuz and see what your memory and hotspot temps are also what voltage is your gpu running at?

 

[King Mustard]

How did you rule out other software issues?

I would have done what I did in 2023 - close all programs from the taskbar and system tray.

I would run the test again with gpuz and see what your memory and hotspot temps are also what voltage is your gpu running at?

I will check after work.

1,995 MHz (2023) dropping to 1,590 MHz (yesterday) seems quite substantial.

The voltage is whatever the stock voltage is. I would not have altered it in 2023, either.

 

[=XAF=AfterShock]

I'm just taking a wild guess that either the thermal paste or thermal pads dried up and pushing hotspot temps and/or memory temps too high causing the card to throttle. In my case it turned out to be the thermal pads that covered the memory chips were more like plaster, after less than a minute of starting a game or any 3dmark stuff the clocks would start dropping but since I had the card capped at .900mv the loss wasn't as extreme.

 

[EarthDog]

1,995 MHz (2023) dropping to 1,590 MHz (yesterday) seems quite substantial.

It is. Also, running your benchmark with GPUz will show any throttling reasons (on the sensors tab)... please post a screen shot showing the main page and sensors after you run the test (again make sure it's running while you benchmark)

If your temps are 70C though like you said in the first post, I doubt its temps on anything (as your memory is clocking higher without issue).

 

[King Mustard]

It is. Also, running your benchmark with GPUz will show any throttling reasons (on the sensors tab)... please post a screen shot showing the main page and sensors after you run the test (again make sure it's running while you benchmark)

If your temps are 70C though like you said in the first post, I doubt its temps on anything (as your memory is clocking higher without issue).

The first screenshot is stock GPU values but with the GPU fans manually set to 70% in MSI Afterburner.

The second screenshot is with the Power Limit set to +15% in MSI Afterburner.

The third screenshot is as below, which is the maximum overclock I can now apply without crashing (which is not as good as I somehow got in 2023; hence this topic).



I took the screenshots at roughly the same point - about 15 seconds into the second scene.



Please keep in mind the three CPU-Z screenshots show 'Current', rather than 'Highest Reading'.

 

[EarthDog]

So far, I don't see the ~1,500MHz core clocks you're talking about... it's running ~10MHz slower than what you listed in the first post.

 

[King Mustard]

So far, I don't see the ~1,500MHz core clocks you're talking about... it's running ~10MHz slower than what you listed in the first post.

Sorry, as mentioned above the two screenshots, they are without an overclock. One is fully stock (apart from increased fan speed), one is the same but with the Power Limit increased to the maximum (+15%).

I will run an overclock and post another screenshot.

 

[EarthDog]

I will run an overclock and post another screenshot.

Perfect. We know it's working right at stock...

I'm curious to see the Perfcap reason when overclocked...

 

[King Mustard]

Perfect. We know it's working right at stock...

I'm curious to see the Perfcap reason when overclocked...

Added the third screenshot

Please keep in mind the three screenshots show 'Current', rather than 'Highest Reading'.

Do you need the three screenshots taking again with 'Highest Reading' set, instead?

 

[EarthDog]

Still working right. I see a current clock speed of ~1995 MHz.

The key here is to show a screenshot of GPUz when it's throttling and around that 1500 MHz you mentioned in the first post. So far, it's not exhibiting the problem you've described.

 

[King Mustard]

Still working right. I see a current clock speed of ~1995 MHz.

The key here is to show a screenshot of GPUz when it's throttling and around that 1500 MHz you mentioned in the first post. So far, it's not exhibiting the problem you've described.

Anything over +141 MHz Clock and over +1475 MHz Memory, I crash.

Screenshot 3 is using those numbers.

 

[EarthDog]

Anything over +141 MHz Clock and over +1475 MHz Memory, I crash.

Screenshot 3 is using those numbers.

Perhaps I'm just tired...

....throw what values you had 'entered' in the past to reach those clocks away...


.....in the last GPUz screenshot, you hit the core clocks you initially reached, didn't you? And going past the memory???

Back in August 2023, I posted the following 3DMark Time Spy score:

https://www.3dmark.com/spy/40882941
GPU Core: 1,995 MHz (avg 1,914 MHz)
GPU Memory: 1,219 MHz (avg 1,210 MHz)
GPU score: 18,297

Assuming that's true, I don't see an issue with the images you've posted. You could just be at the end of your overclock. It's that simple (so far).

 

[=XAF=AfterShock]

I don't see anything that stands out either, your at your cards max of 115% TDP and temps are reasonable. you could unlock voltage control in msi and then crank that slider up and see if it still crashes over the +141 but like EarthDog mentioned if we can't see what it's doing when it gets knocked down to 1590 it's just a guessing game. You could change the graphs so it shows the highest recorded temps and you could run port royal in stress test mode to really push the card and see if you can make it downclock and then screenshot the gpuz results.

Another unlikely thing that comes to mind is when I 1st starting running folding@home on this card I was arrogant with my overclock and after an hour I thought so far so good it must be solid then the next morning I saw my card wasn't getting the points that it should and my gpu was hardly doing anything so I rebooted my pc and it was back to normal for about 15 min and then I watched them tank again. Spent 30 min wiping the drivers and then reinstalling them only to have it happen right away, wasn't getting errors in the folding program so I was so damn confused. I happened to look in the event viewer and saw an entire page of nvlddmkm errors which means the driver crashed, normaly the folding@home programs will fail and it would log it but it wasn't doing your normal gpu crash symptoms. so long story short - see if you can get your gpu to drop it's clocks like it was and then check the event viewer for any errors relating to the driver. I probably could have just simply said that but I haven't been on my pc all day and now my fingers won't stop typing.

 

[King Mustard]

I do not remember what Core and Memory adjustments I entered into MSI Afterburner back in 2023.

However, the 2023 link shows I got a higher GPU score (18,297) than I did this week (17,412) (4.8% lower), despite the fact I cannot increase my current OC this week beyond +151 MHz and +1475 MHz without crashing.

The 2023 link shows:

• Clock frequency 1,995 MHz (1,440 MHz), Memory clock frequency 1,219 MHz (1,188 MHz)

The 2025 link, when using +151 MHz and +1475 MHz, shows:

• Clock frequency 1,590 MHz (1,440 MHz), Memory clock frequency 1,372 MHz (1,188 MHz)

Clearly I am now beating my 2023 Memory clock but the Core clock is nowhere near. I have tried dropping the memory OC and trying to increase the Core above 151 MHz, but even 152 MHz crashes.

 

[EarthDog]

Clearly I am now beating my 2023 Memory clock but the Core clock is nowhere near

From what you showed us, you are reaching ~1990 MHz (see your 3rd gpuz image).

Have you run 3dmark again?

Also, where are you getting that you're running at ~1500 MHz??????? If it's from 3dmark, thats often wrong....

....rerun 3dmark (let it finish) now (with gpuz) and see what happens...

Until you can show us the clocks dropping, I'm afraid things are running 'normal'. I'll repeat...

The key here is to show a screenshot of GPUz when it's throttling and around that 1500 MHz you mentioned in the first post. So far, it's not exhibiting the problem you've described.

Edit: after clicking into your 3dmark scores, I see you changed your system/cpu as well so you're not comparing like things either.

Edit2: I'd bet money t he ~1500 mhz you see 3dm reporting are the base clocks for your card (post a screenshot of GPUz... the first page with all the specs) but that doesn't mean your actually running the test at that speed... again, run 3dm the whole way through with gpuz and check those clockspeeds/post the image here. If you were running at 1500 mhz, your score would be a lot lower.

Your TOTAL score is actually HIGHER on the second 3dm run (because of the cpu/system change). There isn't a problem here at all outside of your gpu score being slightly lower... but that could be driver/overhead. Fromnwhat you shared, you are, clearly, hitting the same clockspeeds.