Tom's Hardware Superposition Benchmark Thread

[QwerkyPengwen]

First off I'd like to say that this test is crap. The score it provides is in no way truly reflective of the actual performance of my GPU. My reasoning for this is that this benchmark doens't really benchmark your GPU the way it should because it clearly isn't advanced enough to push the your card in a way that causes your clock speeds to be maxed out. So in the end it basically causes my voltage to not hit it's max limit which in turn results in heavy clock speed fluctuations that are most of the time on the lower end and that in turn hurts my FPS badly which in turn hurts my score.

I made a thread about this crap with nvidia's latest "curve" settings in GPU Boost 3.0 that you can read at the following link if you like.
http://www.tomshardware.com/forum/id-3648245/2000mhz-series-gpu.html

But here's my score and it's the best I could make this card do on this dumb benchmark:

 

[TechyInAZ]

^^^ I think you completely misunderstand how GPU Boost 3.0 works. The GPU only overclocks itself if the card has enough voltage, power limit, and temp headroom.

The benchmark is perfectly fine, it's designed to push your card to the max, it does on my GTX 1080 just fine. You aren't going to hit peak clock speeds because it is impossible unless your idle temp is your load temp also.

 

[jankerson]

Exactly, this is likely the best GPU benchmark around right now and it does max them out for everything they can do.

 

[aquielisunari]

^^^ I think you completely misunderstand how GPU Boost 3.0 works. The GPU only overclocks itself if the card has enough voltage, power limit, and temp headroom.

The benchmark is perfectly fine, it's designed to push your card to the max, it does on my GTX 1080 just fine. You aren't going to hit peak clock speeds because it is impossible unless your idle temp is your load temp also.

Can you explain the idle and load temp. and its correlation to peak clock speeds? GPU Boost 3.0 doesn't start ease back the throttle in 13MHZ increments until about 56c. Before that I don't see how the temperature negatively impacts the peak clock speed. The voltage or power limits would have to be looked at. MSI AB tells me I occasionally hit my PWR limit. My core clock will boost oto and will stay at 2000MHZ but it may slip to 1987MHZ. Still tweaking my fan curve so it won't hit 55c. I just purchased EVGA's 1080 Ti FTW3 Hybrid last month. I haven't looked that closely at my voltages yet. https://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/15

 

[QwerkyPengwen]

^^^ I think you completely misunderstand how GPU Boost 3.0 works. The GPU only overclocks itself if the card has enough voltage, power limit, and temp headroom.

The benchmark is perfectly fine, it's designed to push your card to the max, it does on my GTX 1080 just fine. You aren't going to hit peak clock speeds because it is impossible unless your idle temp is your load temp also.

Then explain to me why my voltage fluctuates heavily along with my clock speeds when in this benchmark but when in a game, for example Ghost Recon Wildlands, I my clock speeds sit perfectly still near 2Ghz and at a constant 1.093v? whereas with this benchmark my voltage and clock speeds (which are connected and correlate with each other BTW) are fluctuating heavily? I'm pretty sure if my clock clock speeds and voltage were to be pinned then I would get a better score in this benchmark. My power limit is set to 112 with temp limit at 92 and voltage unlocked and pushed to 100 on the slider and fans at 100% in open air to maximize my overclocking potential and to maintain a higher clock speed when under load so I don't see how your remark about power and voltage limits is relevent? I wasn't talking about hitting the maximum most ultimate clock speeds but that my typical speeds and voltage when in a heavily demanding title are constant and in this benchmark they drop down really low and then go back up and back down low again constantly fluctuating so that tells me that it's not pushing the card at all and not pinning it down pushing it to it's limit because if it was my voltage and clock speeds would reflect that.

 

[RememberThe5th]

^^^ I think you completely misunderstand how GPU Boost 3.0 works. The GPU only overclocks itself if the card has enough voltage, power limit, and temp headroom.

The benchmark is perfectly fine, it's designed to push your card to the max, it does on my GTX 1080 just fine. You aren't going to hit peak clock speeds because it is impossible unless your idle temp is your load temp also.

Then explain to me why my voltage fluctuates heavily along with my clock speeds when in this benchmark but when in a game, for example Ghost Recon Wildlands, I my clock speeds sit perfectly still near 2Ghz and at a constant 1.093v? whereas with this benchmark my voltage and clock speeds (which are connected and correlate with each other BTW) are fluctuating heavily? I'm pretty sure if my clock clock speeds and voltage were to be pinned then I would get a better score in this benchmark. My power limit is set to 112 with temp limit at 92 and voltage unlocked and pushed to 100 on the slider and fans at 100% in open air to maximize my overclocking potential and to maintain a higher clock speed when under load so I don't see how your remark about power and voltage limits is relevent? I wasn't talking about hitting the maximum most ultimate clock speeds but that my typical speeds and voltage when in a heavily demanding title are constant and in this benchmark they drop down really low and then go back up and back down low again constantly fluctuating so that tells me that it's not pushing the card at all and not pinning it down pushing it to it's limit because if it was my voltage and clock speeds would reflect that.

I've had bug with msi afterburner, if i max out memory and max out core clocks, after run the clock drops, after i wanna re fix it (soo it stays 1973 but drops to 1961 soo i increase to 1983 soo it can drop to 1973 the curve randomly increases and decreases in weird way assuming that increasing +1 won't increase it but after +13 will do it, after setting memory to 0 then back to max it would increase also core clock from 1973 to 2GHz lol).

As for voltage, mine was hitting before 1.2V on older version of MSI afterburner, but on latest it won't (i think its a bug).
I tried also Precise from EVGA but same result,
but now as I've remembered i was using 38X? driver and i could hit 1.2V with older version of MSI AB (even 1.25V), i think Nvidia restricted in drivers for overvoltage and overclock something.

 

[TechyInAZ]

^^^ I think you completely misunderstand how GPU Boost 3.0 works. The GPU only overclocks itself if the card has enough voltage, power limit, and temp headroom.

The benchmark is perfectly fine, it's designed to push your card to the max, it does on my GTX 1080 just fine. You aren't going to hit peak clock speeds because it is impossible unless your idle temp is your load temp also.

Then explain to me why my voltage fluctuates heavily along with my clock speeds when in this benchmark but when in a game, for example Ghost Recon Wildlands, I my clock speeds sit perfectly still near 2Ghz and at a constant 1.093v? whereas with this benchmark my voltage and clock speeds (which are connected and correlate with each other BTW) are fluctuating heavily? I'm pretty sure if my clock clock speeds and voltage were to be pinned then I would get a better score in this benchmark. My power limit is set to 112 with temp limit at 92 and voltage unlocked and pushed to 100 on the slider and fans at 100% in open air to maximize my overclocking potential and to maintain a higher clock speed when under load so I don't see how your remark about power and voltage limits is relevent? I wasn't talking about hitting the maximum most ultimate clock speeds but that my typical speeds and voltage when in a heavily demanding title are constant and in this benchmark they drop down really low and then go back up and back down low again constantly fluctuating so that tells me that it's not pushing the card at all and not pinning it down pushing it to it's limit because if it was my voltage and clock speeds would reflect that.

Sorry for misunderstanding you. That's really odd. I've NEVER had that problem on any of my pascal cards.

My Non overvolted GTX 1080 always runs at the exact same speed and voltages as in games.

And my GTX 1060 G1 Gaming when clocked at 2152mhz 1.093v stays there for about half to 3/4ths of the benchmark, only then do i see it go down to 1.081v. But games eventually do bring the vcore down to 1.081v aswell so no strange behavior on my cards.

 

[Rogue Leader]

I also do not see any issues like that on my Vega.

Others not seeing this with Nvidia cards, as well as my experience, tells me the problem probably lies with your system or Afterburner than the Benchmark.

 

[QwerkyPengwen]

Sorry for misunderstanding you. That's really odd. I've NEVER had that problem on any of my pascal cards.

My Non overvolted GTX 1080 always runs at the exact same speed and voltages as in games.

And my GTX 1060 G1 Gaming when clocked at 2152mhz 1.093v stays there for about half to 3/4ths of the benchmark, only then do i see it go down to 1.081v. But games eventually do bring the vcore down to 1.081v aswell so no strange behavior on my cards.

That's ok. When I run the medium version it stays a little more stable with occasional fluctuations but it still ends up pegging at like 1.05v instead of 1.093v. I don't know why it's fluctuating so much when I go to the extreme settings. In fact, when I run 8K optimized it pretty much gives me the same score lol. So something is up with the Extreme setting and my card not cooperating together. I've run the program as administrator in compatibility mode for Windows 7 as well and no dice.

 

[EquineHero]

Sorry for misunderstanding you. That's really odd. I've NEVER had that problem on any of my pascal cards.

My Non overvolted GTX 1080 always runs at the exact same speed and voltages as in games.

And my GTX 1060 G1 Gaming when clocked at 2152mhz 1.093v stays there for about half to 3/4ths of the benchmark, only then do i see it go down to 1.081v. But games eventually do bring the vcore down to 1.081v aswell so no strange behavior on my cards.

That's ok. When I run the medium version it stays a little more stable with occasional fluctuations but it still ends up pegging at like 1.05v instead of 1.093v. I don't know why it's fluctuating so much when I go to the extreme settings. In fact, when I run 8K optimized it pretty much gives me the same score lol. So something is up with the Extreme setting and my card not cooperating together. I've run the program as administrator in compatibility mode for Windows 7 as well and no dice.

Yeah, I even got my 1070 over 2GHz which is nice for a change, tired of 1850MHz.

Unigine Superposition is designed to kill graphics cards - it set my GTX 980 on fire, literally.

 

[aquielisunari]

Sorry for misunderstanding you. That's really odd. I've NEVER had that problem on any of my pascal cards.

My Non overvolted GTX 1080 always runs at the exact same speed and voltages as in games.

And my GTX 1060 G1 Gaming when clocked at 2152mhz 1.093v stays there for about half to 3/4ths of the benchmark, only then do i see it go down to 1.081v. But games eventually do bring the vcore down to 1.081v aswell so no strange behavior on my cards.

That's ok. When I run the medium version it stays a little more stable with occasional fluctuations but it still ends up pegging at like 1.05v instead of 1.093v. I don't know why it's fluctuating so much when I go to the extreme settings. In fact, when I run 8K optimized it pretty much gives me the same score lol. So something is up with the Extreme setting and my card not cooperating together. I've run the program as administrator in compatibility mode for Windows 7 as well and no dice.

Yeah, I even got my 1070 over 2GHz which is nice for a change, tired of 1850MHz.

Unigine Superposition is designed to kill graphics cards - it set my GTX 980 on fire, literally.

So you shared it with the class. Admirable

To those that don't know, VRAM.

 

[EquineHero]

So you shared it with the class. Admirable

To those that don't know, VRAM.

You mean VRM? The VRMs on my 980 were fine - it was the power limit controller that got fried.

 

[aquielisunari]

So you shared it with the class. Admirable

To those that don't know, VRAM.

You mean VRM? The VRMs on my 980 were fine - it was the power limit controller that got fried.

Yes I did. The 970 had that issue.

 

[RememberThe5th]

"Cough".... R7 250 "Cough"... The loading screen took around 15-20m min to load because pagefile+slow HDD (hahahaha)
This was with modded bios 1200Core and 1150Mem, with 1.25V i think i could get higher result with my PC (main) but, why not (vram limit was 1.8GB) but ran it as 2.5GB.

 

[EquineHero]

"Cough".... R7 250 "Cough"... The loading screen took around 15-20m min to load because pagefile+slow HDD (hahahaha)
This was with modded bios 1200Core and 1150Mem, with 1.25V i think i could get higher result with my PC (main) but, why not (vram limit was 1.8GB) but ran it as 2.5GB.

Since memory is such an issue, run it at 1080p Medium.

 

[RememberThe5th]

"Cough".... R7 250 "Cough"... The loading screen took around 15-20m min to load because pagefile+slow HDD (hahahaha)
This was with modded bios 1200Core and 1150Mem, with 1.25V i think i could get higher result with my PC (main) but, why not (vram limit was 1.8GB) but ran it as 2.5GB.

Spoiler

Since memory is such an issue, run it at 1080p Medium.

Shh thats an actual 1GB version
And thats 1080p medium :O

 

[EquineHero]

Shh thats an actual 1GB version
And thats 1080p medium :O

That IS 1080p Med? Ahahaha. Table will be updated shortly...

 

[RememberThe5th]

"Cough".... R7 250 "Cough"... The loading screen took around 15-20m min to load because pagefile+slow HDD (hahahaha)
This was with modded bios 1200Core and 1150Mem, with 1.25V i think i could get higher result with my PC (main) but, why not (vram limit was 1.8GB) but ran it as 2.5GB.

Since memory is such an issue, run it at 1080p Medium.

Shh thats an actual 1GB version
And thats 1080p medium :O

That IS 1080p Med? Ahahaha. Table will be updated shortly...

If i've got time today which i should, ill take it to my main rig and re test it with extreme (the ram should help???) and hope it will run it xD.

 

[QwerkyPengwen]

[strike]OpenGL or DX?[/strike] never mind found out....

I think i nailed this one

1973?1971? Core and max memory (core was 1981 but drops).

Can someone check 1050Ti score on this page 1080P extreme
https://benchmark.unigine.com/leaderboards/superposition/1.0/1080p-extreme/single-gpu/page-1

Because for me it won't load, somewhat bug.

I think you are doing great. I jumped over to my other system that I upgraded from to test.
It has an MSI Gaming X 1050 Ti
During the benchmark it sat at 1100mv / 2000Mhz (Core) / 4004Mhz (MEM) / 40c
i5-4670K OC @ 4.0Ghz
16GB G.Skill RAM @ 1600Mhz

photo web hosting

However, it says your benchmark run was custom. You need to run the 1080p Extreme settings.

 

[RememberThe5th]

Ive found out that it was riva tuner problem, but i can try to retest but im unsure will i get the same oc again...
Ill try in these days and see will i get the same score again.
But i maxed out my memory to 4504MHz
But how you all get over 2GHz but im stuck to max 1973 without crashing test?
Can you post the MSI afterburner screen?

 

[EquineHero]

However, it says your benchmark run was custom. You need to run the 1080p Extreme settings.

All he did was disable fullscreen, he didn't actually change the difficulty, so to speak. RTSS is a pain in the ass for applications like Superposition that need GPU exclusivity.

 

[RememberThe5th]

Yeah i did again benchmark and still gave me errors ... even without RIVA (sometimes), but still im 4 points over record in superposition with same clock,
In my score i ran like 20+ times? (2h in total)
Ill try that in weekend or when i catch time these days.

Here's my run which i wasn't including cause 6+ points (LOL) and with fullscreen i ran like 5 times and scores were 1300(tried 0 memory)-1514 i guess because the 1987? stuck longer at beginning of the test?

http://www.userbenchmark.com/UserRun/7436461

here i managed to run 2GHz with the test lenght and didn't crash. (others i tried higher and crashed obviously)

 

[TechyInAZ]

Ok so I might be keeping my 1080 after all and not selling it. After reading the reddit post below, it seems like it's the power limit that's the problem and not so much the cooler.

https://www.reddit.com/r/nvidia/comments/4tc9wy/zotac_1080_amp_overheating_fix/?ref_source=embed&ref=share

 

[RememberThe5th]

Ok so I might be keeping my 1080 after all and not selling it. After reading the reddit post below, it seems like it's the power limit that's the problem and not so much the cooler.

https://www.reddit.com/r/nvidia/comments/4tc9wy/zotac_1080_amp_overheating_fix/?ref_source=embed&ref=share

It is always power limit, as i said im hitting perf cap on 1973MHz overclock.

 

[EquineHero]

It is always power limit, as i said im hitting perf cap on 1973MHz overclock.

Do the shunt mod, increase power limit by 300%