Titan XP vs 1080 vs 1070 vs Titan XM - Benchmarks

[philliplakis]

Hey guys,


I was able to get the Titan XP to +283mhz with a score of 182.
The clock was at 2163mhz

That was only to run OctaneBench once so not sure of stability - Im running both my 1080 and Titan XP and +100 with no issues.
My 1080 is a bad overclocker and would crash if i pushed it and doesn't show much of an improvement at all.

So i done a comparison against all the Pascal series cards in OctaneBench v3.
Price to performance the GTX 1070 wins easily...

Below are the clocks for the tests:

XM - 1163mhz Base - 1263mhz - 1363mhz - 1413mhz
XP - 1809mhz Base - 1964mhz - 2075mhz - 2113mhz
80 - 1860mhz Base - 1885mhz - DNF - DNF
70 - 1900mhz Base - 2000mhz - 2085mhz - DNF


If you have a chance to check it out link is below.

[mauroborba]

Nice! I've read somewhere that people are being able to push 1080 to 180 points range. I'm not sure how or if this is accurate though.

[Lewis]

Thanks for test Phillip.

I'm also experiencing similar results that TitanX(M) is very overclockable while 1080 is already high by default stock Boost so not big change (not much overclock headroom) like in Maxwell when you overclock manually.

[philliplakis]

Nice! I've read somewhere that people are being able to push 1080 to 180 points range. I'm not sure how or if this is accurate though.

Maybe on water , But on Air I'm not a believer. With the Boost states they have now there clocked super high, I think on water is the only way.

I'm also experiencing similar results that TitanX(M) is very overclockable while 1080 is already high by default stock Boost so not big change (not much overclock headroom) like in Maxwell when you overclock manually..

I agree, With Boostv3 that nvidia has with the Pascal series i feel its already overclock close to there potential.
The XM achieved a 20 point boost with OC where else the XP only 10 points. Crazy clocks on the Pascals, I remember the days when 850mhz was mind blowing now we are hitting 2100+ lol

[Badger]

1080 is such a ripoff compared to the 1070. The price is double and the performance difference is only about 15-20% tops :\ .

[oguzbir]

1080 is such a ripoff compared to the 1070. The price is double and the performance difference is only about 15-20% tops :\ .

Titan XP hasnt been started to selling in stores here. and 1080 is really double the price here as well. I believe gddr5x is worth it but. Not according to my money..

I had to buy 2x1070. And they are so reasonable, needs less power, works cool and silent on air, OC is not bad at all.
Finding 140 score wih OC is very easy. I'll sell my first gen titan ASAP and buy another 1070.

[glimpse]

one thing I don't understand how do some get nearlly 200 with TitanX..think at this point we still do not have clear results..nor optimised build =) I will wait to make any proper conclusions till we see official release of benchmark =)

[oguzbir]

one thing I don't understand how do some get nearlly 200 with TitanX..think at this point we still do not have clear results..nor optimised build =) I will wait to make any proper conclusions till we see official release of benchmark =)

I agree.
The thing with Cuda 8 was the new unified memory, Faster compiling.
I wonder if these new features will be adapted to the optimizations for Octane..

[salmon]

From Abstrax,
viewtopic.php?f=36&t=56895

"Currently it doesn't look like we can improve things considerably for Pascal compared to what we have right now. Maybe a few percent, but not 25%. I think that's because Octane is quite complex compared to most other CUDA applications and mostly memory bound so it's mostly improvements in the memory area that help Octane and not so much improvements in computing performance. But I have spent only a limited amount of time on this, so I might be wrong here.

Switching from CUDA 8.0 RC to CUDA 8.0 didn't make any difference unfortunately. We still use CUDA 7.5 as part of the compilation pipeline, because otherwise things would be up to 50% slower... We are in contact with NVIDIA, but there is only so much they can do without accessing our kernel code. I think, eventually, I will have to have a proper look and check if there is anything that can be done to fix the problem with CUDA 8."

Cheers

[Timmaigh!]

Nice! I've read somewhere that people are being able to push 1080 to 180 points range. I'm not sure how or if this is accurate though.

Its possible, the proof screenshot is posted in the GTX 1080 thread. I think it was on air-cooled card as well, but not 100 percent sure, but there was some voltmod required. Still, even without it, it can do about 165 points at 2088 MHz, which seems to fairly obtainable OC for majority of Pascal cards. In other words, not bad.

Anyway, very nice topic and video, this. Thanks to the OP, would give karma, if i could on these boards. Only shame is the lack of higher OC on the 1070/1080.