[1080 Ti] max. Voltage draw seems kinda low to me

[fuechschen_twitch]

Hey,

 

I recently bought a used 1080Ti and installed it. After playing around in Afterburner I found myself pretty good settings regarding OC'ing. But looking in the stats of the GPU via GPU-Z I saw that the card only draws a maximum voltage of 0.85V under full load (running a FurMark 1080p-Test), which seems kinda odd to me. After searching I found some results claiming to have a voltage of 1.xV which (with my "not good" knowledge) seems kinda better.

 

What further raises some eyebrows for me is that with two Afterburner settings I have, the card runs stable in a FurMark test but crashes as soon as I start up a game (in my case GTA V), which further implies to me that the voltage isn't right. The seller claimed to have undervolted the card, so I guess that could be the "issue" here. Also when running FurMark, GPU-Z will show that there is a PerfCap for Pwr, which seems also odd, having a 650W PSU and the GPU only drawing a maximum of 300W. 

 

My question overall is: Is there any way to reset the Undervolting (IF there is any) and will that run the Overclock more stable? I didn't found any results helping me with resetting the settings the seller made, so I am asking you guys.

 

For further information any settings/hardware is listed below.

 

Thanks in advance

 

Hardware:

GTX 1080 TI Palit Super Jetstream (Specifications found here)

Corsair CX650F Power Supply (Productpage here)

any more hardware found in my Profile (CPU etc.)

 

OC Settings in Afterburner (which are unstable when starting games):

Core Voltage: +0 / +100
Power Limit: +120 / +120

Temp. Limit: 90 / 90

Core Clock: +150 / +190

Memory Clock: +500 / +565

Fan Speed: Auto

 

OC Settings in Afterburner (stable with any workload):

Core Voltage: +0
Power Limit: +120

Temp. Limit: 90

Core Clock: +130

Memory Clock: +450

Fan Speed: Auto

[Godlygamer23]

If the undervolting was only done in software, like Afterburner or Precision, it shouldn't affect you at all(settings are only loaded once the program opens if it's set that way. In other words, the program is required every time you want to apply the settings). If they modified the card physically or the BIOS, then that's a different story.

[fuechschen_twitch]

1 minute ago, Godlygamer23 said:

If the undervolting was only done in software, like Afterburner or Precision, it shouldn't affect you at all(settings are only loaded once the program opens if it's set that way. In other words, the program is required every time you want to apply the settings). If they modified the card physically or the BIOS, then that's a different story.

Is there any way to check for tempered BIOS settings? And the other burning question is: On what voltage should the 1080Ti usually run, because I didn't found anything regarding that.

[Godlygamer23]

2 minutes ago, fuechschen_twitch said:

Is there any way to check for tempered BIOS settings? And the other burning question is: On what voltage should the 1080Ti usually run, because I didn't found anything regarding that.

I'm not really sure if you can check for it, but I did run Superposition on my 1070 SC, and it spiked at 1.065V. Now, I do have it overclocked and overvolted(at least the slider is set that way) currently. You might be able to save the BIOS through GPU-Z and compare it to a known good BIOS of the same card if you can, or even upload it here to have someone look at it.

[fuechschen_twitch]

Alrighty, so in case anyone can/wants to check if the BIOS of my GPU was tempered with, I have the file attached to this post. I don't know if it's a needed info or not, but UEFI is checked in GPU-Z. 

 

Also for maybe more help regarding that I have ran a Sensor Log in GPU-Z while having a FurMark-Benchmark active. The results for that can be found in this Google Spreadsheet.

 

Thank you so much in advance

GP102.rom

[Godlygamer23]

12 minutes ago, fuechschen_twitch said:

Alrighty, so in case anyone can/wants to check if the BIOS of my GPU was tempered with, I have the file attached to this post. I don't know if it's a needed info or not, but UEFI is checked in GPU-Z. 

 

Also for maybe more help regarding that I have ran a Sensor Log in GPU-Z while having a FurMark-Benchmark active. The results for that can be found in this Google Spreadsheet.

 

Thank you so much in advance

GP102.rom 256.5 kB · 0 downloads

Also, one other thing...When running Furmark, drivers are automatically set to downclock GPUs, which will also have an affect on voltage. What games have you tried to run altogether? A free benchmark you can use that would tax the GPU in a more realistic manner would be Unigine's Superposition benchmark.

[fuechschen_twitch]

1 minute ago, Godlygamer23 said:

Also, one other thing...When running Furmark, drivers are automatically set to downclock GPUs, which will also have an affect on voltage. What games have you tried to run altogether? A free benchmark you can use that would tax the GPU in a more realistic manner would be Unigine's Superposition benchmark.

I only tried the "unstable" settings in GTA V so far. When starting the screens freeze, then go black and then I would get a error message regarding some GFX_3D3 stuff. I will try the benchmark you recommended and will give the results in this thread. 

[fuechschen_twitch]

Short update to my "problem" - It seems to be a FurMark issue as stated by @Godlygamer23 - I ran a Superposition Benchmark on 1080p High and saw the Voltage going up to exactly 1000mV (according to MSI Afterburner). The benchmark ran stable with the higher "unstable" setting (+190 Core Clock and +565 Memory Clock), but anything higher resulted in the Benchmark crashing after the loading screen and the attached error message.

 

But that brings me to another "problem" - I've seen articles stating that the 1080Ti can run up tp 2000 MHz GPU Clocks, while mine is struggling being stable at anything higher than 1671MHz GPU Clock, 1518MHz Memory and 1772MHz Boost (numbers taken from GPU-Z with the "unstable" setting active, as seen in this screenshot). So is it just a bad GPU not reaching it's maximum power or am I doing something wrong here?

 

[Godlygamer23]

10 minutes ago, fuechschen_twitch said:

Short update to my "problem" - It seems to be a FurMark issue as stated by @Godlygamer23 - I ran a Superposition Benchmark on 1080p High and saw the Voltage going up to exactly 1000mV (according to MSI Afterburner). The benchmark ran stable with the higher "unstable" setting (+190 Core Clock and +565 Memory Clock), but anything higher resulted in the Benchmark crashing after the loading screen and the attached error message.

 

But that brings me to another "problem" - I've seen articles stating that the 1080Ti can run up tp 2000 MHz GPU Clocks, while mine is struggling being stable at anything higher than 1671MHz GPU Clock, 1518MHz Memory and 1772MHz Boost (numbers taken from GPU-Z with the "unstable" setting active, as seen in this screenshot). So is it just a bad GPU not reaching it's maximum power or am I doing something wrong here?

 

 

The numbers in GPU-Z are not accurate on the "Graphics card" tab. Anything related to GPU stats should be pulled from something that's giving live clocks like MSI Afterburner or the GPU-Z sensor tab . For example, my GPU runs at about 2000MHz, but in GPU-Z, it only shows up as 1695MHz for GPU clock, and 1885 for Boost. 

 

Also, what is that image from in the post? Something you experienced?

[fuechschen_twitch]

8 minutes ago, Godlygamer23 said:

Also, what is that image from in the post? Something you experienced?

Yes, the image was the error message the Benchmark gave me when trying to go higher than my highest "most unstable" OC written in the first post (+190 Core Clock and +565 Memory Clock).

 

I tried doing a benchmark again using the Extreme Setting, and I was able to get around 1999MHz with 1.05V (with peaks to 2012MHz and 1.081V) with these settings - however, the same issue was showing (Screens freezing, then the Game/Benchmark crashing) and GPU-Z giving me issues about Pwr and VRel PerfCaps.

 

With the knowledge I have, I guess the best option is to go lower with the Core Clock and Memory Clock settings in Afterburner, is that correct?

[Godlygamer23]

6 minutes ago, fuechschen_twitch said:

Yes, the image was the error message the Benchmark gave me when trying to go higher than my highest "most unstable" OC written in the first post (+190 Core Clock and +565 Memory Clock).

 

I tried doing a benchmark again using the Extreme Setting, and I was able to get around 1999MHz with 1.05V (with peaks to 2012MHz and 1.081V) with these settings - however, the same issue was showing (Screens freezing, then the Game/Benchmark crashing) and GPU-Z giving me issues about Pwr and VRel PerfCaps.

 

With the knowledge I have, I guess the best option is to go lower with the Core Clock and Memory Clock settings in Afterburner, is that correct?

Yeah, try backing off on your clocks a tad and see if that helps. Start with memory clocks first though. That seems a bit high.

[Mark Kaine]

pretty sure gpus throttle when running fumark... 

 

hence, protip: don't use it, it's useless for "benchmarking" (and since gpus throttle while running it, for everything else too) 

 

 

[Godlygamer23]

Just now, Mark Kaine said:

pretty sure gpus throttle when running fumark... 

 

hence, protip: don't use it, it's useless for "benchmarking" (and since gpus throttle while running it, for everything else too) 

Whenever I read or watch reviews on graphics cards, I always skip the power draw section with Furmark or similar, as it's totally pointless because of the throttling.

[fuechschen_twitch]

Okay, now with nearly everything tested I can safely say that FurMark dissapointed me in a BIG way. 

 

I ran some tests with Superposition and found the sweetspot for my GPU which is reliable and also tickles a bit more performance. Also I learned that I shouldn't take for granted what some people write on forums etc. since I saw NO improvement over setting the Core Voltage to +100%, leaving it at 0% is actually more reliable I guess. (GPU-Z givng me VRel as a Performance Cap Reason (the voltage was between 1050-1075mV with 1050mV being the average).

 

The settings below saw the 1080Ti between 1949-1999MHz in GPU Clock (with the average being 1974MHz), 5775MHz in the Memory Clock and a max. Temperature of 75°C.

 

For anyone searching on the Webs and finding this thread, my Afterburner settings are as followed:

 

Core Voltage: +0

Power Limit: +120

Temp. Limit: 90°

Core Clock: +135

Memory Clock: +265

Fan Speed: Auto

 

With these results Superposition gave me (1080p Extreme Test) a score of 6006 (with 3 Monitors hooked up, dunno if this could make a difference).

Overall we can say: My card wasn't undervolted, FurMark is a pain in the butt to benchmakr since the card is not going full voltage and Superposition Benchmark is better for wathcing over something like this.

 

Thanks @Godlygamer23 for the suggestion and therefore helping me with the solution to this topic. 

[Godlygamer23]

Just now, fuechschen_twitch said:

Overall we can say: My card wasn't undervolted, FurMark is a pain in the butt to benchmakr since the card is not going full voltage and Superposition Benchmark is better for wathcing over something like this.

I'm glad your issue is fixed ^.^

Real world tests are the go-to(demanding games). 3DMark can also be a good solution when testing GPU stability. A few years ago, both AMD and NVIDIA released drivers that deliberately throttled cards because Furmark and similar programs were literally killing graphics cards because they would pull so much power. Even before that, Furmark wasn't ever really that great for stability anyway. Cards would pass Furmark, but fail in actual games. 

[Mark Kaine]

15 minutes ago, fuechschen_twitch said:

I ran some tests with Superposition and found the sweetspot for my GPU which is reliable and also tickles a bit more performance. Also I learned that I shouldn't take for granted what some people write on forums etc. since I saw NO improvement over setting the Core Voltage to +100% other than GPU-Z givng me VRel as a Performance Cap Reason (the voltage was between 1050-1075mV with 1050mV being the average).

yeah, core voltage is pretty useless in my experience - I'm not quite sure but I think it's really only useful for hardcore overclockers, or people with "golden samples". 

 

For example my old 1060 did actually clock slightly higher with +100, and therefore also gave a bit better results in certain benchmarks... however, it wasn't stable in most games, and it was utterly impossible to finish a run in Firestrike... 

 

Which would also be my suggestion if you want to test your OC... most other benchmarks like superposition etc, aren't as realistic, they're more for a quick test if your OC works at all... but no guarantee for an actual stable OC. 

 

Also while finding and testing OCs it's advisable to have the fans at 100% since that nearly eliminates the factor of thermal throttling, which otherwise can become an issue quickly... 

 

If you found a stable OC you can still always dial down the cooling, there might be slightly more throttling *but* you will still run at the highest possible OC with your configuration. 

 

And that brings me to my last point, don't underestimate the importance of cooling, even 5-10C can make a *huge* difference, it's much more important than voltage (which I personally would advise to leave at default, in almost all situations) 

[Mark Kaine]

10 minutes ago, Godlygamer23 said:

3DMark

while I didn't even try that many benchmarks, seriously Firestrike and to an extent Timespy are really the closet to a real life test... Some OCs will be stable in one game but not in another etc, but if it passes Firestrike its almost guaranteed to be stable in like 99% of games in my experience. 

 

 

12 minutes ago, Godlygamer23 said:

Even before that, Furmark wasn't ever really that great for stability anyway. Cards would pass Furmark, but fail in actual games. 

I worked for 6 months in a pc repair shop (which was a lot of fun btw) and they *only* used fumark on new builds, not necessarily to test stability, rather than functionality, like they'd let it run for hours... or even over night... 

 

which kind of makes sense, but I'm not sure they really knew that gpus would throttle, ironically. 

 

It's really an outdated program and I'm not sure why people still use it with its reputation of killing cards (even if that's not necessarily true anymore) and otherwise being rather useless. ¯\_(ツ)_/¯