GTX 1080 TI strix hitting power limit on benchmarks

[SilverwingX0]

I haven't overclocked a GPU in a long while, and I'm completely new to Pascal, so please correct me if I'm wrong but I was watching a simple overclocking guide to overclock pascal based GPU's and one of the points made was that the card should not be hitting the power limit at stock. Unfortunately that's exactly whats happening to me.

I ran both heaven and firestrike and both benchmarks caused my 1080ti strix to hit it's power limit. I followed up by unlocking the power target slider, unlinked it to GPU temp and set it to 120%. This appeared to ever so slightly reduce the number of times I hit my power limit but it still does quite often in the same benchmarks.

Now my question is, is this normal? I haven't overclocked the card at all, from my understanding, this shouldn't be happening unless I overclocked the card in some way?

Temps never hit above 65C and my top core speed was 1911 Mhz (for only a few seconds). Thanks in advance.

 

[tanckattb]

This isn't very normal, but all cards of the same model are exactly the same. Sometimes they differ a little in performance or heat dissipation or powe usage. The strix 1080ti has a very large factory overclock if not wrong. Asus intentionally does it so that they gain bigger market share I Guess as the higher clock speed attract more consumers.
So what might be happening is your strix might be hitting its stock power limit of 100%, so since you unlocked he voltage, it gives the strix more power to operate on its factory overclock.

To solve this, you could reduce your core clock speed little by little so that after some adjusting, your card won't hit it's power limit anymore.

: D

 

[Lucky_SLS]

1911 for air cooled surely requires quite an oc. furthermore, benchmarks like firestrike and timespy r similar to running a stress test. u wont come across those power draws in gaming scenarios. running a i7 7700k with 1080ti for benchmarks require a 650w PSU to be comfy. a quality 550W would also do but thats cutting it close. is ur psu 550w?

 

[SilverwingX0]

No, it's a EVGA 1000W G2. AFAIK that should be plenty of power. I did notice something peculiar while running benchmarks yesterday. While my power limit might be reached doing 1911 Mhz on a benchmark, that isn't necessarily the highest frequency it can run at. For example, I found a moment in the beachmark that I know for a fact it hits the power limit at 1911 core speed while just using GPU boost. So I went back into afterburner and set a off set of 109+ Mhz to the core and kept the powerlimit at 120% and reran the benchmark. Now I was running at 1989-2000 Mhz stable with essentially the same amount of power (still hitting my power limit but now I'm running a higher clock speed). And that wasn't the highest OC I could achieve, just the easiest and quickest.

TL : DR
It seems that while my card is hitting it's power limit fairly often, it isn't necessarily using that power efficiently. So while my card appears to be hitting a power limit wall, in reality I can achieve much higher clocks with the same amount of power by doing it manually.

 

[Lucky_SLS]

what abt custom fan curves? if u have a bit better cooling solution like optimised fan flow in ur case and custom fan curve, u can reach even higher oc with the same power draw.

 

[SilverwingX0]

My GPU core never hit higher than 60C, I set a slightly more aggressive fan curve since the fans are so quiet. So I guess in the end, hitting your power limit isn't nearly a big a deal as it was made out to me as. I can hit the power limit and still clock higher as there's still wiggle room within the power/voltage already being fed to the GPU, even at "MAX" power consumption.