Tom's Hardware Superposition Benchmark Thread

[EquineHero]

So why did you remove my 1080ti score? It remains the single card record for this site at 6225.

The 1080ti easily allowed 1ghz OC on ram. I'd take it higher, but afterburner doesn't allow it.

Also you can use nVidia Profile Inspector to change clock speed, you can manually enter any clock speed for any component.

 

[bjornl]

So why did you remove my 1080ti score? It remains the single card record for this site at 6225.

The 1080ti easily allowed 1ghz OC on ram. I'd take it higher, but afterburner doesn't allow it.

Also you can use nVidia Profile Inspector to change clock speed, you can manually enter any clock speed for any component.

I didn't know that. I'll have to see what it allows for memory speed.

 

[jankerson]

Ran it again yesterday myself, OCed as before, higher memory clock this time.

And here is it at 1080P High Setting just for a laugh, was stock speed for this.

I'll add your extreme score later, I'm out of the house right now.

No hurry.

Was at 6K Memory clock this time (12 GBS Effective).

I can barely scratch 4000 with a single GTX 1070, but two of them destroys. If you want a chance at being on top of everyone else, you can SLI some 1080s.

Also, Superposition says your 7700K is not overclocked. It does the same for my Xeon, even though I have the BCLK at 105.8.

How are you getting your memory so high? My 1070s max out at around 4500MHz if I drop my core overclock (power mgmt issue).

The first 1070 I have on the chart had an 8 and a 6 pin and the shunt mod so I could get +400 on mem along with a good +200 on core, but I haven't hardmodded either of my 1070s yet.

I have Word open now. I can't add your new score, it's not 200 marks or 5% above your last one. Every time I update my score I make sure it is. My previous best was 7340, then I got 7605. Pretty big boost +350 on memory gives.

It's GDDR5X and stock clock is 5508 or 11 GBS effective. I have a EVGA GTX 1080 FTW2, one of the ones that came out of the box with 11 GBS Memory.

If I had a Classified I might be able to push the core higher, not sure about the memory though, however the FTW2 is a monster over clocker.

4767 is however +324 points over my stock clocked original score of 4443 so I posted it in the thread as reference.

I doubt I would get ever 4855 point out of it, or 200+ over 4655 or +412 over the original 4443.

 

[threebills11]

Ran it at 1080 p Etxreme
https://imgur.com/Fhv8jVs

 

[EquineHero]

Ran it again yesterday myself, OCed as before, higher memory clock this time.



And here is it at 1080P High Setting just for a laugh, was stock speed for this.

I'll add your extreme score later, I'm out of the house right now.

No hurry.

Was at 6K Memory clock this time (12 GBS Effective).

I can barely scratch 4000 with a single GTX 1070, but two of them destroys. If you want a chance at being on top of everyone else, you can SLI some 1080s.

Also, Superposition says your 7700K is not overclocked. It does the same for my Xeon, even though I have the BCLK at 105.8.

How are you getting your memory so high? My 1070s max out at around 4500MHz if I drop my core overclock (power mgmt issue).

The first 1070 I have on the chart had an 8 and a 6 pin and the shunt mod so I could get +400 on mem along with a good +200 on core, but I haven't hardmodded either of my 1070s yet.

I have Word open now. I can't add your new score, it's not 200 marks or 5% above your last one. Every time I update my score I make sure it is. My previous best was 7340, then I got 7605. Pretty big boost +350 on memory gives.

It's GDDR5X and stock clock is 5508 or 11 GBS effective. I have a EVGA GTX 1080 FTW2, one of the ones that came out of the box with 11 GBS Memory.

If I had a Classified I might be able to push the core higher, not sure about the memory though, however the FTW2 is a monster over clocker.

4767 is however +324 points over my stock clocked original score of 4443 so I posted it in the thread as reference.

I doubt I would get ever 4855 point out of it, or 200+ over 4655 or +412 over the original 4443.

Unless you have a different AIB GTX 1080, the rules are the rules. New scores stay, old scores go out the window simply because they'd clutter the list. My single 1070 bench is still there because that's without SLI, and that's a CMTX 1070. If I were to individually bench my ECGA and my PNY 1070s, those would get separate scores.

The point of this thread is not only for users to show off and test their overclocking abilities but to have a good comparison for different vendors and to use it as a baseline for comparison.

For example, (*opens Word doc*) Sinxar Knights has a GTX 1070 Gaming OC, and my modded Founder's Edition board are both the same GP104. His got a better score than mine, so anyone looking to buy one based on the chart would go for the Gaming OC (also because CMTX is crap).

Every time a specific GPU gets a new 200/5% score, the score for that particular GPU is updated. Different models generally have different ASIC quality, one reason I specify the vendor if it's provided in the Unigine screenshot.

Ran it at 1080 p Etxreme

Please put images in BBCode tags so we don't have to click a link. I've already done so.

 

[jankerson]

Ran it again yesterday myself, OCed as before, higher memory clock this time.



And here is it at 1080P High Setting just for a laugh, was stock speed for this.

I'll add your extreme score later, I'm out of the house right now.

No hurry.

Was at 6K Memory clock this time (12 GBS Effective).

I can barely scratch 4000 with a single GTX 1070, but two of them destroys. If you want a chance at being on top of everyone else, you can SLI some 1080s.

Also, Superposition says your 7700K is not overclocked. It does the same for my Xeon, even though I have the BCLK at 105.8.

How are you getting your memory so high? My 1070s max out at around 4500MHz if I drop my core overclock (power mgmt issue).

The first 1070 I have on the chart had an 8 and a 6 pin and the shunt mod so I could get +400 on mem along with a good +200 on core, but I haven't hardmodded either of my 1070s yet.

I have Word open now. I can't add your new score, it's not 200 marks or 5% above your last one. Every time I update my score I make sure it is. My previous best was 7340, then I got 7605. Pretty big boost +350 on memory gives.

It's GDDR5X and stock clock is 5508 or 11 GBS effective. I have a EVGA GTX 1080 FTW2, one of the ones that came out of the box with 11 GBS Memory.

If I had a Classified I might be able to push the core higher, not sure about the memory though, however the FTW2 is a monster over clocker.

4767 is however +324 points over my stock clocked original score of 4443 so I posted it in the thread as reference.

I doubt I would get ever 4855 point out of it, or 200+ over 4655 or +412 over the original 4443.

Unless you have a different AIB GTX 1080, the rules are the rules. New scores stay, old scores go out the window simply because they'd clutter the list. My single 1070 bench is still there because that's without SLI, and that's a CMTX 1070. If I were to individually bench my ECGA and my PNY 1070s, those would get separate scores.

The point of this thread is not only for users to show off and test their overclocking abilities but to have a good comparison for different vendors and to use it as a baseline for comparison.

For example, (*opens Word doc*) Sinxar Knights has a GTX 1070 Gaming OC, and my modded Founder's Edition board are both the same GP104. His got a better score than mine, so anyone looking to buy one based on the chart would go for the Gaming OC (also because CMTX is crap).

Every time a specific GPU gets a new 200/5% score, the score for that particular GPU is updated. Different models generally have different ASIC quality, one reason I specify the vendor if it's provided in the Unigine screenshot.

Ran it at 1080 p Etxreme

Please put images in BBCode tags so we don't have to click a link. I've already done so.

I know what the rules say, the reason why I said I posted it for reference.

 

[jankerson]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

 

[EquineHero]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

Holy crap, those clocks are insane. If I wasn't using SLI I could get 2200MHz out of my EVGA 1070, but the PNY 1070 has lower ASIC quality so I'm limited to 2100Mhz.

According to the current chart though it's still not 200 marks...could be 5%...

5% of 4655= 232

Nope.

If you can play with the curves in MSI Afterburner and go somewhere REALLY cold you can get the extra 30-ish points you need. It took me a couple hours to squeeze the most I could.

I don't know if you'd want to but the shunt mod also helps a lot by maxing out the amount of power the card can use.

I think it was either COLGeek or Bjornl that said that the test is memory sensitive, try lowering the core a bit and upping the memory more. I lowered my core clock by 85MHz and raised my memory by 150 from my previous high score, got 250+ points off that.

 

[jankerson]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

Holy crap, those clocks are insane. If I wasn't using SLI I could get 2200MHz out of my EVGA 1070, but the PNY 1070 has lower ASIC quality so I'm limited to 2100Mhz.

According to the current chart though it's still not 200 marks...could be 5%...

5% of 4655= 232

Nope.

If you can play with the curves in MSI Afterburner and go somewhere REALLY cold you can get the extra 30-ish points you need. It took me a couple hours to squeeze the most I could.

I don't know if you'd want to but the shunt mod also helps a lot by maxing out the amount of power the card can use.

I think it was either COLGeek or Bjornl that said that the test is memory sensitive, try lowering the core a bit and upping the memory more. I lowered my core clock by 85MHz and raised my memory by 150 from my previous high score, got 250+ points off that.

I tried that last night, didn't give me any more, Oh well.

Heat isn't the issue, the card never gets hot, stays in the 60's.

 

[EquineHero]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

Holy crap, those clocks are insane. If I wasn't using SLI I could get 2200MHz out of my EVGA 1070, but the PNY 1070 has lower ASIC quality so I'm limited to 2100Mhz.

According to the current chart though it's still not 200 marks...could be 5%...

5% of 4655= 232

Nope.

If you can play with the curves in MSI Afterburner and go somewhere REALLY cold you can get the extra 30-ish points you need. It took me a couple hours to squeeze the most I could.

I don't know if you'd want to but the shunt mod also helps a lot by maxing out the amount of power the card can use.

I think it was either COLGeek or Bjornl that said that the test is memory sensitive, try lowering the core a bit and upping the memory more. I lowered my core clock by 85MHz and raised my memory by 150 from my previous high score, got 250+ points off that.

I tried that last night, didn't give me any more, Oh well.

Heat isn't the issue, the card never gets hot, stays in the 60's.

Thermal throttling occurs at 53C on all EVGA models.

 

[jankerson]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

Holy crap, those clocks are insane. If I wasn't using SLI I could get 2200MHz out of my EVGA 1070, but the PNY 1070 has lower ASIC quality so I'm limited to 2100Mhz.

According to the current chart though it's still not 200 marks...could be 5%...

5% of 4655= 232

Nope.

If you can play with the curves in MSI Afterburner and go somewhere REALLY cold you can get the extra 30-ish points you need. It took me a couple hours to squeeze the most I could.

I don't know if you'd want to but the shunt mod also helps a lot by maxing out the amount of power the card can use.

I think it was either COLGeek or Bjornl that said that the test is memory sensitive, try lowering the core a bit and upping the memory more. I lowered my core clock by 85MHz and raised my memory by 150 from my previous high score, got 250+ points off that.

I tried that last night, didn't give me any more, Oh well.

Heat isn't the issue, the card never gets hot, stays in the 60's.

Thermal throttling occurs at 53C on all EVGA models.

Depends, if the voltage available is bumped up to 100% it doesn't drop on mine at all on the core or memory (Memory never drops)

Reason why I got those high clocks on the GPU, stayed rock solid at 2126.

But if you don't move the slider then yeah it will drop some, slightly on the GPU.

Ended up be max I can get stable though, any higher I get freezing etc, on both GPU and memory clocks.

 

[EquineHero]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

Holy crap, those clocks are insane. If I wasn't using SLI I could get 2200MHz out of my EVGA 1070, but the PNY 1070 has lower ASIC quality so I'm limited to 2100Mhz.

According to the current chart though it's still not 200 marks...could be 5%...

5% of 4655= 232

Nope.

If you can play with the curves in MSI Afterburner and go somewhere REALLY cold you can get the extra 30-ish points you need. It took me a couple hours to squeeze the most I could.

I don't know if you'd want to but the shunt mod also helps a lot by maxing out the amount of power the card can use.

I think it was either COLGeek or Bjornl that said that the test is memory sensitive, try lowering the core a bit and upping the memory more. I lowered my core clock by 85MHz and raised my memory by 150 from my previous high score, got 250+ points off that.

I tried that last night, didn't give me any more, Oh well.

Heat isn't the issue, the card never gets hot, stays in the 60's.

Thermal throttling occurs at 53C on all EVGA models.

Depends, if the voltage available is bumped up to 100% it doesn't drop on mine at all on the core or memory (Memory never drops)

Reason why I got those high clocks on the GPU, stayed rock solid at 2126.

But if you don't move the slider then yeah it will drop some, slightly on the GPU.

I'm gonna have to try that on my SLI setup. Doubt I'd be able to pass the nitro cooled 1080Ti as that's in the 8000s...

 

[jankerson]

Well this is the max I can get out of the card from what I was able to do. I think the Seasonic Prime Titanium helped me get higher clocks with the cleaner power over the Seasonic X-Series.

2126 GPU and 6039 (12078 effective) memory

Holy crap, those clocks are insane. If I wasn't using SLI I could get 2200MHz out of my EVGA 1070, but the PNY 1070 has lower ASIC quality so I'm limited to 2100Mhz.

According to the current chart though it's still not 200 marks...could be 5%...

5% of 4655= 232

Nope.

If you can play with the curves in MSI Afterburner and go somewhere REALLY cold you can get the extra 30-ish points you need. It took me a couple hours to squeeze the most I could.

I don't know if you'd want to but the shunt mod also helps a lot by maxing out the amount of power the card can use.

I think it was either COLGeek or Bjornl that said that the test is memory sensitive, try lowering the core a bit and upping the memory more. I lowered my core clock by 85MHz and raised my memory by 150 from my previous high score, got 250+ points off that.

I tried that last night, didn't give me any more, Oh well.

Heat isn't the issue, the card never gets hot, stays in the 60's.

Thermal throttling occurs at 53C on all EVGA models.

Depends, if the voltage available is bumped up to 100% it doesn't drop on mine at all on the core or memory (Memory never drops)

Reason why I got those high clocks on the GPU, stayed rock solid at 2126.

But if you don't move the slider then yeah it will drop some, slightly on the GPU.

I'm gonna have to try that on my SLI setup. Doubt I'd be able to pass the nitro cooled 1080Ti as that's in the 8000s...

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

And I would think the Seasonic Prime Titanium PSU helps also.

 

[EquineHero]

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

My SLI setup beats all of the 1080Tis on the leaderboard except that one by KingPin that's been nitro cooled and at like 2.5Ghz or something crazy

 

[jankerson]

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

My SLI setup beats all of the 1080Tis on the leaderboard except that one by KingPin that's been nitro cooled and at like 2.5Ghz or something crazy

Dunno, might be silicone lottery on the GPU and memory clocks.

I think the FTW3 GTX 1080Ti cards can go really high OCed.

 

[EquineHero]

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

My SLI setup beats all of the 1080Tis on the leaderboard except that one by KingPin that's been nitro cooled and at like 2.5Ghz or something crazy

Dunno, might be silicone lottery on the GPU and memory clocks.

I think the FTW3 GTX 1080Ti cards can go really high OCed.

Nah the leaderboard includes all details about the benchmark, temps was min -40 and max -40

 

[jankerson]

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

My SLI setup beats all of the 1080Tis on the leaderboard except that one by KingPin that's been nitro cooled and at like 2.5Ghz or something crazy

Dunno, might be silicone lottery on the GPU and memory clocks.

I think the FTW3 GTX 1080Ti cards can go really high OCed.

Nah the leaderboard includes all details about the benchmark, temps was min -40 and max -40

That's LN2 cooled.

 

[EquineHero]

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

My SLI setup beats all of the 1080Tis on the leaderboard except that one by KingPin that's been nitro cooled and at like 2.5Ghz or something crazy

Dunno, might be silicone lottery on the GPU and memory clocks.

I think the FTW3 GTX 1080Ti cards can go really high OCed.

Nah the leaderboard includes all details about the benchmark, temps was min -40 and max -40

That's LN2 cooled.

LN2

L

N2

L=Liquid

N2=Nitrogen

Liquid nitrogen=nitro cooled

I don't think I should have had to explain this

 

[jankerson]

The GTX 1080Ti's won't be had.

The thing with the EVGA GTX 1080 FTW2 and Classified is they have more power phases than the standard cards, Classified has the most to it would be the highest OCer I would think.

My SLI setup beats all of the 1080Tis on the leaderboard except that one by KingPin that's been nitro cooled and at like 2.5Ghz or something crazy

Dunno, might be silicone lottery on the GPU and memory clocks.

I think the FTW3 GTX 1080Ti cards can go really high OCed.

Nah the leaderboard includes all details about the benchmark, temps was min -40 and max -40

That's LN2 cooled.

LN2

L

N2

L=Liquid

N2=Nitrogen

Liquid nitrogen=nitro cooled

I don't think I should have had to explain this

Explains why he got 2.5 GHz on the GPU.

Would need LN2 to get anywhere near that.

 

[jankerson]

On the EVGA GTX 1070's only the FTW and FTW2 cards have 10+2 power phases, the rest are 4+1.

On the GTX 1080's the FTW and FTW2 have 10+2, the rest are 5+1.

The GTX 1080 Classified has 14+3.

The GTX 1080Ti FTW3 have 10+2, the Kingpin Hydro has 10+3, the others are 7+2.

 

[EquineHero]

On the EVGA GTX 1070's only the FTW and FTW2 cards have 10+2 power phases, the rest are 4+1.

On the GTX 1080's the FTW and FTW2 have 10+2, the rest are 5+1.

The GTX 1080 Classified has 14+3.

The GTX 1080Ti FTW3 have 10+2, the Kingpin Hydro has 10+3, the others are 7+2.

No, I mean the user on the leaderboard has the handle "Kingpin".

 

[jankerson]

On the EVGA GTX 1070's only the FTW and FTW2 cards have 10+2 power phases, the rest are 4+1.

On the GTX 1080's the FTW and FTW2 have 10+2, the rest are 5+1.

The GTX 1080 Classified has 14+3.

The GTX 1080Ti FTW3 have 10+2, the Kingpin Hydro has 10+3, the others are 7+2.

No, I mean the user on the leaderboard has the handle "Kingpin".

I was just pointing out the major differences in the cards really.

 

[bjornl]

So why did you remove my 1080ti score? It remains the single card record for this site at 6225.

The 1080ti easily allowed 1ghz OC on ram. I'd take it higher, but afterburner doesn't allow it.

Also you can use nVidia Profile Inspector to change clock speed, you can manually enter any clock speed for any component.

Where exactly is this? I went through the nVidia Profile Inspector and it does not allow adjusting clock speeds. I then went through nVidia Inspector and it does. But it caps memory speed below what I am already getting from MSI After burner. nVidia Inspector max = 6000, afterburner max = 6505.

 

[Vellinious]

Actually, running it at a lower setting would push more frame rate, and put more pressure on the CPU. Unigine benchmarks are notorious for being lightly threaded, which is why Valley is nothing more than a CPU benchmark any more. Bottlenecking the CPU intentionally, by running lower settings, and pushing the frame rates higher will show a bigger difference in the scores. By running a test with higher settings, or a higher resolution, you'd be taking the onus off of the CPU, and putting it more on the GPU. I would think that 1080 Extreme or 4k you'd see extremely similar scores between the two systems.

That said....it's a synthetic. Compare frame rates between a lightly threaded DX11 game, or a heavily threaded DX12 game, and you'll see a HUGE difference between them. Assuming, of course, the 1090T can run DX12.....I'm not really sure about that.

No, running it at a higher setting would push the GPU to its maximum making it the bottleneck and therefore any performance differences between the two would be CPU only (being he is using the same GPU). Your theory would be correct if the CPU was being stressed here but its not. Increased frame rate does not necessarily cause the CPU to have to work harder.

There is nothing preventing a 1090T or any other CPU made in the past 7 or 8 years from running DX12, it has little to do with the CPU.

Increased frame rates on lower resolutions WILL stress the CPU harder...especially when you consider the pretty weak IPC of an old Thuban. It will create a bottleneck on the CPU side, which WILL create a vastly different score between the two systems. Look at any high end system on Valley in the last 3 years, and you'll see exactly what I mean.

If you make the GPU the bottleneck by increasing the resolution, the score should be very close to the same. The only difference will be the level of the tech. PCIe 2.0, iirc? Slow memory. With the GPU the bottleneck, the CPU shouldn't matter all that much.

 

[Rogue Leader]

Actually, running it at a lower setting would push more frame rate, and put more pressure on the CPU. Unigine benchmarks are notorious for being lightly threaded, which is why Valley is nothing more than a CPU benchmark any more. Bottlenecking the CPU intentionally, by running lower settings, and pushing the frame rates higher will show a bigger difference in the scores. By running a test with higher settings, or a higher resolution, you'd be taking the onus off of the CPU, and putting it more on the GPU. I would think that 1080 Extreme or 4k you'd see extremely similar scores between the two systems.

That said....it's a synthetic. Compare frame rates between a lightly threaded DX11 game, or a heavily threaded DX12 game, and you'll see a HUGE difference between them. Assuming, of course, the 1090T can run DX12.....I'm not really sure about that.

No, running it at a higher setting would push the GPU to its maximum making it the bottleneck and therefore any performance differences between the two would be CPU only (being he is using the same GPU). Your theory would be correct if the CPU was being stressed here but its not. Increased frame rate does not necessarily cause the CPU to have to work harder.

There is nothing preventing a 1090T or any other CPU made in the past 7 or 8 years from running DX12, it has little to do with the CPU.

Increased frame rates on lower resolutions WILL stress the CPU harder...especially when you consider the pretty weak IPC of an old Thuban. It will create a bottleneck on the CPU side, which WILL create a vastly different score between the two systems. Look at any high end system on Valley in the last 3 years, and you'll see exactly what I mean.

If you make the GPU the bottleneck by increasing the resolution, the score should be very close to the same. The only difference will be the level of the tech. PCIe 2.0, iirc? Slow memory. With the GPU the bottleneck, the CPU shouldn't matter all that much.

No it will not. The CPU in any GPU biased game is doing things like telling the enemies where to shoot, and laying out the map ahead. Higher frame rates do not make that part of the game run any faster or stress the CPU anymore. This is clearly evidenced in that he ran it again on Extreme and got the same results.

If the CPU is already too slow then yes a GPU that is capable of a higher frame rate will put out less FPS due to the CPU in that it cannot provide the most frames it can possibly. But a faster GPU that can put out more FPS does not stress a CPU additionally. If the CPU can't keep up with the game it can't keep up, you can throw all the GPU at it you want.