Watercooled 980ti vs. 1080 vs. 1080ti

[Cheddle]

One thing that reviwers often fail to do is proper, solid, long term custom loop overclock benchtesting... They also have a bad habbit of listing a bit table of comparison frame rates without clarifying the clock rates, cooling, brand or any other detail that is critical to comparing two GPUs - so I thought I would show the differences Nvidias Flagships have made over the years while water cooled. I do have results from a 980 but its only air cooled (if anyone is interested I can add them) ive been collecting some data for some time - ill add more as I go through more benches and find some older 980ti/1080 stuff (no longer have the cards) but see below for some 3dmark comparison.

 

all of these cards are basically balls to the wall using full cover EK water blocks

 

the 980ti (1,505mhz) had a 1.23v custom BIOS

the 1080 (2,114mhz) had a resistor/shunt mod to remove the power limit

the 1080ti (2,114mhz)  is totally stock and the TDP wall is being hit quite often...

 

the results I use to draw my conclusions are below: 
 

Time spy:

http://www.3dmark.com/compare/spy/1384479/spy/1061457/spy/72810#

Graphics score:
980ti: 6077
1080: 8250 (35% on 980ti)
1080ti: 10931 (32% on 1080 / 79% on 980ti)

FPS GT2:
980ti: 34.39fps
1080: 47.97fps
1080ti: 63.10fps

Firestrike:

http://www.3dmark.com/compare/fs/12012710/fs/9549548/fs/7381702

Graphics score:
980ti: 21483
1080: 24382 (13% on 980ti)
1080ti: 31641 (29% on 1080 / 47% on 980ti)

FPS GT1:
980ti: 103.88fps
1080: 117.84fps
1080ti: 153.31fps

Conclusion:
Firestrike is tested at such high framerates that its hard to say if its a very good test these days to use to compare GPU to GPU and draw a conclusion on how this difference will be realised in an actual gaming scenario - if you are looking to replace your 980ti with a 1080ti you might look at the firestrike results and decide is not a very good idea however if you look at the timespy results you might just rush out and buy one... Timespy however is known to be a sort of 'quasi' DX12 test with instructions 'spoon fed' to the GPU rather than, as they would in a game, 'force fed' at the GPU (read more about that here) so again it cant be used reliabily to decide what to do.

----
 

1) tests are showing that the 1080ti is in fact a reliable 30% faster than the 1080 (see here) seeing as these cards have a similar stock to overclocked ratio it is very fair to apply this as a consistant rule to all variations of 1080/1080ti's
2) tests have shown that AIB cards do not acheive higher clock rates than FE cards (once cooling has been eliminated) - (see here and here) - this further supports the '30% faster being a consistant rule' theory

3) tests have shown that the 1080ti does in fact just about dobule the performance of the 980ti (here) HOWEVER! the 980ti overclocks HUGELY compared to the 1080/1080ti

Below I explain why, if you have a 980ti and are looking to upgrade, you should think twice! (you might see above how my 1080 only got 13% firestrike performance over my 980ti... turns out once you add overclocking into the mix, you get some interesting results....

individual stock vs. overclocked - and why you can trust the reviewers to say "980ti is 50% the speed of the 1080ti"
stock 1080ti vs overclocked 1080ti (stock air blower vs. overclocked EK-FC WB):

http://www.3dmark.com/compare/fs/12012710/fs/12010758
Graphics Score:
1080ti stock: 27,743
1080ti OC: 31,641

1080 ti overclock percentage: 14.1%

stock 980ti vs overclocked 980ti:
Graphics Score:
980ti stock: 16,923
980ti overclocked: 21,483

980 ti overclock percentage: 26.9%

note: these percentages will be MUCH smaller on an AIB card as they are generally overclocked out of the box by anything from 0-20% over the reference clocks... In my experience AIB cards do not acheive consistantly higher overclocks than reference cards they are just faster out of the box - the silicon lottery is still very much the only thing you can bet on - I've owned three 980ti's, two AIB and one reference - one AIB reached 1,490mzh stable, the other 1,560mhz... 

to summarise - based purley on Firestrike GPU score:

stock 980ti vs stock 1080ti: 63% gap
OC 980ti vs OC 1080ti: 47% gap

 

conclusion: Always take a reviewers comparison tables with a grain of salt and look to the community for their own findings. Overclocking should always be a thought when choosing a video card. All of the benchmarks listed here are repetable within margin of error and are not outliners/suicide runs with fans set to 100% or anything crazy. 

NOTE: I didnt double check anything ive wirtten above, have made countless speeling mistakes and used calc.exe to work everything out.

[Masada02]

Thank you for posting this. I've been interested in seeing true overclocked results between the 3 because none of the reviewers seem to have results for all 3 to do real comparisons on.