GPU Clock decreases as temperature goes up

[ww_ea]

Hello fine ladies and gentlemen. This is my first post in this forum

So straight to the problem... About a month ago I bought a MSI Lighting Z 1080 Ti. It was a great card but its RGB LEDs was malfunctioning so I had to sent it back. Then I decided to buy a Gigabyte Aorus GTX 1080 Ti Xtreme Editon since Lighting Z wasn't available anymore.
But the problem I have with my Gigabyte Aorus is that GPU Clock decreases constantly as temperature goes up. At first (when running a game), Afterburner shows GPU Clock as 1987Mhz, then temperature reaches 58c and GPU Clock decreases to 1974Mhz. Then temperature goes up till 63c and again GPU Clock goes down to 1949Mhz. When temperature reaches 71c, GPU Clock decreases to 1932Mhz.

I know this is probably nvidia's GPU Boost which is working but why I didn't had this problem with my Lighting Z 1080 Ti? On my previous Lighting Z 1080 Ti, regardless of what temperature was, GPU Clock was constant 1962Mhz (as stated in this page). Sometimes the temperature was as high as 66c but still 1962Mhz. But my Aorus GTX 1080 Ti keeps decreasing GPU Clock as temperature goes up. So am I missing something? Is there a way to prevent thins using Afterburne or somethingr?

My second question...Why I can't reach these clock profiles with my Gigabyte Aorus GTX 1080 Ti Xtreme? I never had 2025Mhz or 1993Mhz of GPU Clock.

Thanks in advance.

 

[Ferrum Master]

It should start to decrease from 45C, and it is like that. Read up first...

 

[W1zzard]

GPU clock decreasing as temperature increases is a feature of GPU Boost 3.0. As far as I know there is no way to change its behavior or disable it.

Lightning Z

 

[RejZoR]

@W1zzard
Yes there is. Cool it better. Either by improving case ventilation/cooling or by ramping up the fans on graphic card.

EDIT:
I have the regular AORUS GTX 1080Ti and there is another way. Manually bump up the clocks. That also changes how boost behaves. It'll keep higher clocks unless the temperatures really go that high that it'll decrease the clocks regardless...

 

[ww_ea]

GPU clock decreasing as temperature increases is a feature of GPU Boost 3.0. As far as I know there is no way to change its behavior or disable it.

Lightning Z

Yes. I know it's GPU Boost 3.0 but I think in my Aorus GTX 1080 Ti Xtreme it reduces GPU Clock more aggressively compared to my previous Lightning Z. For example Lightning Z had GPU Clock of 1962Mhz while working under ~65c or 66c degrees but on my Aorus GTX 1080 Ti, GPU Boost decreases GPU Clock to 1949Mhz at the same temperature. Why is that?

Also, do you have the same graph for Aorus GTX 1080 Ti Xtreme Edition? that could be more clarifying on the subject.

@W1zzard
Yes there is. Cool it better. Either by improving case ventilation/cooling or by ramping up the fans on graphic card.

EDIT:
I have the regular AORUS GTX 1080Ti and there is another way. Manually bump up the clocks. That also changes how boost behaves. It'll keep higher clocks unless the temperatures really go that high that it'll decrease the clocks regardless...

I tried to change fan curve in Afterburner in order to prevent my AORUS GTX 1080Ti hit upper than 66c but no luck. Even at 100 percent fan spinnig, GPU temperature reaches 71c in FurMark tests. Room temperature was ~20c.

 

[sneekypeet]

@ww_ea please don't double post. Look here, it will help you in the future. https://www.techpowerup.com/forums/...osts-and-use-the-multi-quote-features.234427/

 

[Vya Domus]

Yes. I know it's GPU Boost 3.0 but I think in my Aorus GTX 1080 Ti Xtreme it reduces GPU Clock more aggressively compared to my previous Lightning Z. For example Lightning Z had GPU Clock of 1962Mhz while working under ~65c or 66c degrees but on my Aorus GTX 1080 Ti, GPU Boost decreases GPU Clock to 1949Mhz at the same temperature. Why is that?

Turbo Boost is very fine grained (13mhz per step), so much that differences in silicon quality can make 2 cards of the same model reach slightly different clocks. Your card is fine , if you want more OC.

 

[ww_ea]

Turbo Boost is very fine grained (13mhz per step), so much that differences in silicon quality can make 2 cards of the same model reach slightly different clocks. Your card is fine , if you want more OC.

You mean GPU boost right? Cause Turbo Boost is Intel's technology in CPUs.
But anyway...yeah I think compared to my previous card which was Lightning Z, somehow silicon lottery plays a role in here too. I guess right now the only way to maintain high clock rates is to provide better cooling for the card. Though as I said, I've already tried %100 fan carve at ~65c-66c but again, the temperature goes up until 71 and thus, GPU clock decreases even more (more than Lightning Z at the same temperature). Shame really

 

[Vayra86]

You can try to give the core clock a small bump and see how stable that is. In PrecisionX, you can also set clocks per boost bin, make your own curve.

Don't expect miracles, GPU Boost 3.0 is already fine tuned as it is, and pretty locked down.

 

[thebluebumblebee]

GPU temperature reaches 71c in FurMark tests

Are you only seeing this problem with Furmark? Is the card able to stay cool when running something like Heaven? You really should stay away from Furmark.

 

[eidairaman1]

Hello fine ladies and gentlemen. This is my first post in this forum

So straight to the problem... About a month ago I bought a MSI Lighting Z 1080 Ti. It was a great card but its RGB LEDs was malfunctioning so I had to sent it back. Then I decided to buy a Gigabyte Aorus GTX 1080 Ti Xtreme Editon since Lighting Z wasn't available anymore.
But the problem I have with my Gigabyte Aorus is that GPU Clock decreases constantly as temperature goes up. At first (when running a game), Afterburner shows GPU Clock as 1987Mhz, then temperature reaches 58c and GPU Clock decreases to 1974Mhz. Then temperature goes up till 63c and again GPU Clock goes down to 1949Mhz. When temperature reaches 71c, GPU Clock decreases to 1932Mhz.

I know this is probably nvidia's GPU Boost which is working but why I didn't had this problem with my Lighting Z 1080 Ti? On my previous Lighting Z 1080 Ti, regardless of what temperature was, GPU Clock was constant 1962Mhz (as stated in this page). Sometimes the temperature was as high as 66c but still 1962Mhz. But my Aorus GTX 1080 Ti keeps decreasing GPU Clock as temperature goes up. So am I missing something? Is there a way to prevent thins using Afterburne or somethingr?

My second question...Why I can't reach these clock profiles with my Gigabyte Aorus GTX 1080 Ti Xtreme? I never had 2025Mhz or 1993Mhz of GPU Clock.

Thanks in advance.

Different cards, different coolers. Gpus depend on unobstructed ample airflow and ambient temperature which is room air temperature. The cooler the room air temperature the better gpus run.

 

[newtekie1]

Different cards, different coolers. Gpus depend on unobstructed ample airflow and ambient temperature which is room air temperature. The cooler the room air temperature the better gpus run.

That definitely plays a part. It sounds like the Gigabyte card's cooling isn't performing as well as the MSI card. Which isn't surprising considering Gigibyte's crappy quality.

But at the same time, Jayztwocents just did a video about how he had two identical cards, same manufacturer and model, and they boosted to different clocks. So my theory is that asic quality plays some role in it as well.

 

[eidairaman1]

That definitely plays a part. It sounds like the Gigabyte card's cooling isn't performing as well as the MSI card. Which isn't surprising considering Gigibyte's crappy quality.

But at the same time, Jayztwocents just did a video about how he had two identical cards, same manufacturer and model, and they boosted to different clocks. So my theory is that asic quality plays some role in it as well.

That's another great factor too.

It's interesting on a 7770 thread i stated not all cards are created equal.

 

[Vayra86]

Realistically we're talking about less than 2-3% deviation in clocks between top-end cards from different manufacturers. Worst case scenario, you'll see 4-5%. Honestly even if you count in silicon lottery, this is pretty neat. Real life performance of all these cards is remarkably close, at typical res of 1440p or 4K that's what, 3-5 FPS.

 

[Toothless]

Haaaaaa.. my card doesn't downclock at higher temps unless I allow it to pass 65c. Stays at the 2012mhz I have it hit.

Honestly as others have said, keep it cool. Gigabyte might have their profiles more key on cooling than MSI does.

 

[HD64G]

Stop using Furmark to check clocks or thermals. Use a bechmark as 3D Mark or Heaven. AMD and nVidia have included years ago security checks into their drivers to not allow the bashing Furmark does to their GPUs.

 

[EarthDog]

While true, the OP never mentioned furmark??? He said "game".

 

[EntropyZ]

Point is, GPU boosting technologies still suck. All they do is prevent proper overclocking out of the box.

It's good when it's there, but makes your day horrible when it's not needed and gets in the way.

Btw, Gigabyte's cards are pretty weak sauce on average when it comes to overclocking/thermals because getting a good GPU die from them is pretty rare IMO. They can slap a fancy cooling solution on it, it doesn't make the processor itself better quality.

I haven't noticed much difference between OC, G1 and AORUS cards. While EVGA segments tend do to something.

 

[EarthDog]

Not really..it just does some of it for you.

 

[thebluebumblebee]

While true, the OP never mentioned furmark??? He said "game".

Even at 100 percent fan spinnig, GPU temperature reaches 71c in FurMark tests.

 

[EarthDog]

Oh geez, i guess i didnt catch that in the second post.

Point still remains is he said it he was seeing it in games first..then that morsel came in on top.

 

[eidairaman1]

Point is, GPU boosting technologies still suck. All they do is prevent proper overclocking out of the box.

It's good when it's there, but makes your day horrible when it's not needed and gets in the way.

Btw, Gigabyte's cards are pretty weak sauce on average when it comes to overclocking/thermals because getting a good GPU die from them is pretty rare IMO. They can slap a fancy cooling solution on it, it doesn't make the processor itself better quality.

I haven't noticed much difference between OC, G1 and AORUS cards. While EVGA segments tend do to something.

I prefer my cards operating at full throttle at all times or specifying on my own what profile to use when a program launches, variable clock rates tend to cause some screwy issues.

 

[yotano211]

If you download and install Afterburner you can undervolt the graphics card to run at lower voltage which equals to lower temps.

 

[RejZoR]

Not really..it just does some of it for you.

Not to mention GPU Boost 3.0 does it so well it almost reaches what cards can achieve with manual overclock. I wonder if NVIDIA will be able to bump it up even further to really squeeze out every MHz from it out of the box, basically making manual overclocking unnecessary. I'd actually quite like that. Because I'm lazy and don't want to be bothered by installing 3rd party stuff to achieve same results.

 

[ww_ea]

OK so I did run some tests to find out how I can prevent GPU Boost's behavior in decreasing GPU Clock at higher temperatures.
First and utmost, as is obvious, was temperature. I Couldn't change the fact that GPU Boost downclock GPU Clock from 1974Mhz to 1949Mhz at 63c degrees, however; I did manage to bump up fan speed to keep the GPU cool and under 63c degrees and thus, playing with 1974Mhz stable. This was achievable through games and benchmarks and not in stress tests (for example in Firestrike Ultra stress test)
Second factor was power target. By monitoring power consumption through Afterburner I did find a sweet spot for power target offset in 4K tests which was %140. Any more than this was pointless since power consumption never went upper than this. Of course this required more Fan Speed to keep the GPU cool and prevent GPU Clock from going under 1974Mhz. Also, I must mention that this power target is not necessary in most games at 1080. At 1080, something around %100 or %110 was enough to maintain 1974Mhz Clock rates. More demanding games like Ghost Recon Wildland was exception of course. I hope gigabyte somehow increase the low default power target by a future Bios updates.
And third was Voltage which I did notice some limitations at stock offset but didn't tamper with it because I didn't know what was the safe offset. Again, this limitation was shown in 4K by Afterburner and not in 1080. Plus, I've already reached 1974Mhz stable as I said so didn't feel the need to tamper with it.

So overall, I can say that given the fact of Aorus's price compared to my previous Lightning Z, I'm pretty happy with my card at the moment. It's true that it's not as cool as Lightning Z (71c degrees at full load in Aorus compared to maximum 66c degrees in Lightning Z) but as @Vayra86 said, difference between these clock rates (for example 1974Mhz,1963Mhz or even 1949Mhz) are just 1 or 2 percent max and not even noticeable in in-game FPS. Maybe at 4K this can lead to maximum 1 or 2 FPS difference.

You can try to give the core clock a small bump and see how stable that is. In PrecisionX, you can also set clocks per boost bin, make your own curve.

Don't expect miracles, GPU Boost 3.0 is already fine tuned as it is, and pretty locked down.

Well, my goal was to maintain high clock rates out of the box and without overclocking. Because as I said and you said it too, these GPU clock rates are not that much of difference. But I confess, seeing clock rates goes up and down in games sometimes can make you sensitive or edgy

Are you only seeing this problem with Furmark? Is the card able to stay cool when running something like Heaven? You really should stay away from Furmark.

Well I should've mentioned that this problem only occurs in Furmark stress test (Firestrike Ultra) and not in games. I haven't tried Heaven since games are OK. Maybe you or others can report their temperatures in Firestrike Ultra stress test? This can be helpful.

That definitely plays a part. It sounds like the Gigabyte card's cooling isn't performing as well as the MSI card. Which isn't surprising considering Gigibyte's crappy quality.

I thinks it's true that Gigibyte's quality is actually crappy compared to let's say MSI for example. But my last gen card was Gigabyte GTX 980Ti Xtreme Edition which I think was one of the finest cards at 1-2 years ago. So manufacturers too kinda have their own bumpy road in the terms of product's quality I think.