cancel
Showing results for 
Search instead for 
Did you mean: 

Bad 1080 ti overclocking results?

wonderin
Level 8
So recently I've decided to overclock my Rog Strix 1080 Ti OC a bit with gputweak.

78424

Every time I push the core voltage slider over the 10% state, it doesn't pass the 3D Mark Fire Attack stress test – it just stops.
Without changing the core voltage offset I can push the core clock only up to 2012 mhz (+70) without failing the 3dmark test. I guess i can get a bit more from my 1080ti oc with the right adjustments. The temperatures are fine (<65C with a custom fan curve).

What am I doing wrong? Or is it just the bad luck in the lottery?
5,384 Views
18 REPLIES 18

Silent_Scone
Super Moderator
Hi,

More voltage is not always beneficial depending on the sample.

Increasing the voltage slider unlocks further voltage points, but this is not necessarily helpful, as some GPU will be stable at higher clocks whilst at a lower voltage point. In order to maximize potential one needs to experiment with the VR (voltage/frequency) curve which can be adjusted by selecting User Define under the clock offset.
13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

Silent Scone@ASUS wrote:
Hi,

More voltage is not always beneficial depending on the sample.

Increasing the voltage slider unlocks further voltage points, but this is not necessarily helpful, as some GPU will be stable at higher clocks whilst at a lower voltage point. In order to maximize potential one needs to experiment with the VR (voltage/frequency) curve which can be adjusted by selecting User Define under the clock offset.


I was trying to play with that curve but the results were far from good. It's a hard job for noob like me. Haven't found any specific tutorials*

JustinThyme
Level 13
First things first, back off your memory to stock and push the core first.
Are you on air or under water? What are your temps doing? Are you hitting the power target? You can leave the measurements running to see where you are.
Temps will get you every time, add voltage, temp goes up. Your not far off the mark from what I got with a pair of them in SLI. I think I topped out at like 1766 with the volatge slider maxed but I had them under water with a serious loop and when running benches like firestrike ultra 16 fans would ramp to max on two 480 rads. If you are on air open the case up and put a fan blowing in there to get a better idea. Looking at your specs if thats current and your case is closed up and you are on air thats not a bad clock. All the stupid fast clocks you see are on extreme cooling with chillers in the least. look at your compare scores with similar rigs both CPU and GPU and make sure you have it checked for single card. if you see leaders like way out in front open up their results and look at the -20C temps LOL



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:
First things first, back off your memory to stock and push the core first.
Are you on air or under water? What are your temps doing? Are you hitting the power target? You can leave the measurements running to see where you are.
Temps will get you every time, add voltage, temp goes up. Your not far off the mark from what I got with a pair of them in SLI. I think I topped out at like 1766 with the volatge slider maxed but I had them under water with a serious loop and when running benches like firestrike ultra 16 fans would ramp to max on two 480 rads. If you are on air open the case up and put a fan blowing in there to get a better idea. Looking at your specs if thats current and your case is closed up and you are on air thats not a bad clock. All the stupid fast clocks you see are on extreme cooling with chillers in the least. look at your compare scores with similar rigs both CPU and GPU and make sure you have it checked for single card. if you see leaders like way out in front open up their results and look at the -20C temps LOL


There are a few tricks of the trade that can seperate the playing field without the use of cold water, too (when it comes to perceived scores on FM). Plus, Havok physics engine used in the CPU tests are quite sensitive to memory latency, surprisingly so infact.

https://www.3dmark.com/compare/spy/4584307/spy/5862356
13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

Silent Scone@ASUS wrote:
There are a few tricks of the trade that can seperate the playing field without the use of cold water, too (when it comes to perceived scores on FM). Plus, Havok physics engine used in the CPU tests are quite sensitive to memory latency, surprisingly so infact.

https://www.3dmark.com/compare/spy/4584307/spy/5862356


Yes it does, not surprising at all, Ill take 3200 C14 over the stupid fast 4000+ with C19. Thats just the CPU score though and it can separate by 2000 points or so. That's why its important to see what those scores are doing so you are not chasing your tail. When you see a CPU score 10 12000 and you are pumping half that then that a different critter to chase that cant be done with GPU tweak. One part of the system at a time and one thing with that one part. Im still tweaking myself, CPU is about as dialed in as its going to get, need water for GPUs but.......no blocks to be had that are worth having. Ill take what I have, went from 18K on timespy with 7900X and a pair of 1080Tis under water to topping 25K with 9940X and the 2080Tis. CPU score up by 4K.



“Two things are infinite: the universe and human stupidity, I'm not sure about the former” ~ Albert Einstein

JustinThyme wrote:
First things first, back off your memory to stock and push the core first.
Are you on air or under water? What are your temps doing? Are you hitting the power target? You can leave the measurements running to see where you are.
Temps will get you every time, add voltage, temp goes up. Your not far off the mark from what I got with a pair of them in SLI. I think I topped out at like 1766 with the volatge slider maxed but I had them under water with a serious loop and when running benches like firestrike ultra 16 fans would ramp to max on two 480 rads. If you are on air open the case up and put a fan blowing in there to get a better idea. Looking at your specs if thats current and your case is closed up and you are on air thats not a bad clock. All the stupid fast clocks you see are on extreme cooling with chillers in the least. look at your compare scores with similar rigs both CPU and GPU and make sure you have it checked for single card. if you see leaders like way out in front open up their results and look at the -20C temps LOL


i have kind of a good airflow inside my case, custom fan curve keeps it below 65c under full load.
i was overclocking core first, then jumped to memory clock. moving voltage slider brings in instability.
so I guess I should leave it ad it is or go for custom voltage curve :)*

JustinThyme wrote:
First things first, back off your memory to stock and push the core first.
Are you on air or under water? What are your temps doing? Are you hitting the power target? You can leave the measurements running to see where you are.
Temps will get you every time, add voltage, temp goes up. Your not far off the mark from what I got with a pair of them in SLI. I think I topped out at like 1766 with the volatge slider maxed but I had them under water with a serious loop and when running benches like firestrike ultra 16 fans would ramp to max on two 480 rads. If you are on air open the case up and put a fan blowing in there to get a better idea. Looking at your specs if thats current and your case is closed up and you are on air thats not a bad clock. All the stupid fast clocks you see are on extreme cooling with chillers in the least. look at your compare scores with similar rigs both CPU and GPU and make sure you have it checked for single card. if you see leaders like way out in front open up their results and look at the -20C temps LOL


so i couldn't relax all the time 🙂
today i've tested +60 mhz, +100% voltage and set the fans to const 55% and voila! - it gave me 2035 mhz boost, FSEXTREME stable!
setting the fan curve to the same values as at the guide doesn't pass FSE regardless it's actually a good curve. I guess at higher clocks PASCALs begin to be temperature-dependent in terms of stability

https://www.3dmark.com/fsst/1033670

Silent_Scone
Super Moderator
wonderin wrote:
I was trying to play with that curve but the results were far from good. It's a hard job for noob like me. Haven't found any specific tutorials*


Every GPU is different, so walkthroughs will only get you so far when tuning the VF curve manually. The OC Scanner now supports Pascal GPU, I'd recommend trying this (button can be found on the opening page). The process can take anything up to 30 minutes.

78443
13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

Silent Scone@ASUS wrote:
Every GPU is different, so walkthroughs will only get you so far when tuning the VF curve manually. The OC Scanner now supports Pascal GPU, I'd recommend trying this (button can be found on the opening page). The process can take anything up to 30 minutes.

78443


thank you, will definitely give it a try today if yoy say that it's a reliable tool.