GTX_1080 benchmark Tweet

[smicha]

This benchmark and this whole thread is a bunch of BS !

It should be banned by admin immediately because it can lead innocent people into believing this benchmark gives accurate results when in fact it is completely broken !

Thanks smicha for pointing out how to make accurate comparisons by reading Msamples/s while using benchmark scenes in standalone.

Benchmark results told me that my 1080 is slower then 980Ti when in fact it was 23% faster.

Don't use this benchmark hack !

That is what I am trying to pass you guys all the time:

1. After Marcus (abstrax) Pascals work in Octane due to 7.5/8 cuda hybrid so Pascals are still not rendering at full speed. According to my statistical model
viewtopic.php?f=11&t=53959
Pascals shall be even faster (I'll refine the model soon)

2. 7x1080 OC to 2050 mhz on core and 5500mhz on memory scored above 1100 based on calculation of real scene rendered in PT mode - it was exactly 2.6x faster than 3 Titan X Maxwells that score 426.

[Tutor]

Smicha,

So, would it be your projection that the average Octane Bench (OB) score for similarly overclocked GTX 1080s would hover @ about 158 OB points (assuming that the current base OB GPU comparator stayed the same)?

[KonstantinosD]

I can say that this is how much they score now, but I bet it will be even more if and when the cuda 8 optimizations come to play..

[jibbles10]

Maybe a little off topic, does a mixed GPU system work well with octane? For example a GTX 1080 and GTX 980 Ti?

(I'm currently using five GTX 980 Ti's and wondering if throwing in an extra 1080 in the mix will cause issues)

[uncia]

Cause the benchmark is fake and not optimized well - get the same scene on Maxwells and compare it on Pascals in real PT rendering.

Cool story , bro

[Matufia]

Hi,
curiously i found out that if you run the "fake" benchmark, while also running some CPU burner like Prime95, you get to won a higher score.

Im Running 3 x 1070 FE, getting 394 points standard and 422 while running prime. That's like 10 points more per card. I assume this is score only based, and it's not actually going to benefit in rendering.

Im running the cards with +161 core // +223 mem

Cheers

[MildMustard]

Don't bother with the hacky benchmark, figures are all over the place.

[uncia]

... the hacky benchmark...

Funny

[MildMustard]

No disrespect

[nickmilitello]

Any thoughts on why Im only getting 762 with 7 1080s? About 108 per card. Seems pretty weak.