Sajin
I said the cpu doesn't affect 4k graphic tests in post #78. I also said the cpu does affect the final total score in post #84. 

 
I know.  I think there is just some confusion.  After being pissy last night, I was just trying to help clear it up a little.

Now what we'd need to test is does HT/extra cores give any extra FPS.
 
Sajin: Go to bios and disable cores to 4, and disable HT. Then your CPU is equal to my 7600k (apart from core clock).
Then post Timespy Extreme Graphics Test 1 fps @ 1950/12000 clocks.
 
It is possible, although unlikely that the ~10% performance difference vs kingpin comes from my 4 threads vs your 16 threads.

Edome
Now what we'd need to test is does HT/extra cores give any extra FPS.
 
Sajin: Go to bios and disable cores to 4, and disable HT. Then your CPU is equal to my 7600k (apart from core clock).
Then post Timespy Extreme Graphics Test 1 fps @ 1950/12000 clocks.
 
It is plausible that the 5-10% performance difference vs kingpin comes from 4 threads vs your 16 threads.

Whether or not more cores makes a difference versus higher core speed is very benchmark dependent. 3DMark does a pretty good job of isolating GPU performance from different processors which is shown that, even with a wide disparity in CPU performance similiar GPU will have similar performance.

Sajin

HeavyHemi
Oh...barely... the second test is 1.8%. That's the difference between how many places when you're benching? Lets remember the context here. We're 'boasting' about fractions of a percent to gain a few points and now you deem that irrelevant. Okie dokie, there's no point in trying for that extra fraction of an FPS. Over and out.

You know as well as I do that that difference wasn't because of your cpu clock speed. 3dmark results vary as much as 3% from test to test. Over and out.  

OMG...seriously. I can run that test 100 times and see the same percentage drop. Come to think of it  your 'you know as well as I do...' is essentially accusing me of being dishonest. Over, and out.

Sajin

HeavyHemi
 
Come to think of it  your 'you know as well as I do...' is essentially accusing me of being dishonest. Over, and out.

So you're saying you didn't know that it can vary by 3%?

You're just making yourself look bad. You know perfectly well I know how the bench works.   Why are you doing this?

HeavyHemi

Sajin

HeavyHemi
 
Come to think of it  your 'you know as well as I do...' is essentially accusing me of being dishonest. Over, and out.

So you're saying you didn't know that it can vary by 3%?

You're just making yourself look bad. You know perfectly well I know how the bench works.   Why are you doing this?

I'm making myself look bad? lol. I don't think so. The differences in your graphics scores had nothing to do with your cpu frequency as the results are within margin of error. Have fun believing that your cpu frequency made a difference to your graphics scores in time spy extreme.

Well the CPU does play a huge role in graphics score. At least what type of CPU you use not so much the clocks.
 
For example:
I got a 2.5% higher graphics score across the board (which like heavy said is a huge deal in benchmark language) on my 4790k @4.6G compared to my new 7920X @4.6G
One could argue that I didn't optimize the 12 core and it's mesh and memory very well. But when comparing all other TimeSpy (Extreme) results using the same CPU in their system, I'm actually #1. So clearly on my measly overclock on an AiO I already squeezed the best out of memory and cache.
 
Then on the other hand on an older 5960X or 6950X you'll see much higher graphic scores. So this is just how the cookie crumbles.
Maybe future updates will make better use of SkylakeX, I wish...

Sajin

HeavyHemi
If only the 'margin of error' wasn't always ~-1.5% with about a 10% reduction in clock speed. Repeatable every time.

Guess your system must be busted because mine shows 0% difference. Below are three back to back runs at 3.5GHz & 4.4GHz...
 https://www.3dmark.com/compare/spy/2836117/spy/2836129/spy/2836135/spy/2836162/spy/2836175/spy/2836184

Thank you, this confirms cpu clock does not really affect Graphics Score.
 
A more interesting test: could you kindly disable HT & cores in the bios and see if the fps changes. 
 
It looks possible that Kingpin isn't 5-10% faster, but the CPU plays a part in the graphics score too:
https://www.3dmark.com/compare/spy/2817576/spy/2816492/spy/2813401/spy/2623396/spy/2623419/spy/2801731/spy/2524858/spy/2554002/spy/2815046/spy/2794334
I picked similar gpu core & memory clocks for this comparison.
 
Although most likely is that the kingpin is like 6-8% faster (memory timings or VRM), and the CPU/chipset gives a maximum of 2% boost.
 
Anyone has a way to contact Kingpin and ask him if there's any reason the 1080ti kingpin would be faster clock-to-clock than other 1080 Ti cards?

4 cores @ 3.5GHz with ht disabled... https://www.3dmark.com/3dm/23656139
 
No difference.

Edome
Thank you, then it's all about Kingpin being faster.
 
33.7 with 2012/mem stock?
 
I'm getting 30,6, so that's again 10% difference :)
 
Do you get the same result with K-Boost off?

All the runs above were with k-boost off.

Although 10% difference is pretty big... it could be also that the BIOS secretly unlocks some of the titan XP features such as extra shader units/TMUs/ROPs.
Edit: 384 bit Titan XP memory bus width is not possible with only 11 memory modules of 1080ti. 
 
Sajin: Check with GPU-Z it still shows 88/24 ROPs/TMUs for the extreme bios.
 
 

Sajin
 
It doesn't unlock anything.

Aight, then it just has other optimizations (although GPU-Z wouldn't necessarily detect the extra features for a 1080ti bios).

Edome
 
For confirmation's sake, please do a final check and see if the Graphics Test 1 fps is indeed lower with one of the other 3 BIOS.
 
The listed changes for the watercooling extreme bios are only faster fans & higher power limit, so it does help with 8K superposition but shouldn't with 3Dmark Graphics Test 1.

No difference on the stock oc vbios.

Edome

Sajin
 
No difference on the stock oc vbios.

Thanks, I think it's time to make the final conclusion:
 
Kingpin PCB is somehow ~10% faster than the rest of 1080 Ti clock-to-clock. This improves the price-performance of the product :)

That's in synthetic benchmarks. If gaming is your thing and not benchmarking, you probably won't notice any difference while gaming.