AMD Ryzen Blender Challenge

[Massman]

Alright, put this one up. Looking forward to see the comparison results

[Splave]

intel score poorly on this?

 

 

Perhaps intel should have kaby running xtu on one side and amd with a black screen on the other smh

[NoMS]

intel score poorly on this?

 

 

Perhaps intel should have kaby running xtu on one side and amd with a black screen on the other smh

 

Intel is fine on Blender... Throttling issues maybe?

 

Anyway, for this challenge should we use default 200 samples or 100 like AMD used in Ryzen demo?

Edited December 15, 2016 by NoMS

[speed.fastest]

Hi, this is what make all score so slow...

[skulstation]

ware do we find the link for it? like to try it on some intel setups

 

edit : i found the files on https://hardforum.com/threads/amd-ryzen-blender-benchmark-scores-thread.1919896/

 

the skulltrail is slow comparing to the rest

Edited December 15, 2016 by skulstation

[skulstation]

plz chek ranking. slowest System is on top

Edited December 15, 2016 by skulstation

[Splave]

Hi, this is what make all score so slow...

 

Okay that makes much more sense

[moparman]

The scoring is wrong for the comp. The Fastest is in last and slowest in first place.

[FatBoyNotSoSlim]

Oooh, this looks like fun.

[I.nfraR.ed]

FX-8320E@5117MHz running Linux Mint 18 x64. On Windows 7 SP1 x64 it was more than 2 times slower!

 

[BeepBeep2]

@I\.nfraR\.ed

 

Really interesting - can you (and others) try on Windows with the build that The Stilt posted here:

https://forums.anandtech.com/threads/first-summit-ridge-zen-benchmarks.2482739/page-114#post-38629328

 

He and others had huge gains, because it seemed that the 2.78a build for Windows was not taking full advantage of AVX.

 

 

For the sake of the Blender Challenge I think those should be limited to the official binary for now?

[Leeghoofd]

what an unreliable benchmark this is... fluctuates here 0.5-1 sec between runs on air clocks ... lol

Edited December 17, 2016 by Leeghoofd

[I.nfraR.ed]

@I\.nfraR\.ed

 

Really interesting - can you (and others) try on Windows with the build that The Stilt posted here:

https://forums.anandtech.com/threads/first-summit-ridge-zen-benchmarks.2482739/page-114#post-38629328

 

He and others had huge gains, because it seemed that the 2.78a build for Windows was not taking full advantage of AVX.

 

 

For the sake of the Blender Challenge I think those should be limited to the official binary for now?

 

Shaved 1 minute off. Still ~10 sec slower than Linux, but I'm on a very old Win 7 installation. It might be better on e.g. 10.

So I guess the official Linux build is newer than official Win build.

 

PS: Checked both timestamps and they match: both are compiled on 24.10.2016.

Wonder if there will be such an improvement if someone compiles the linux version (update: building it right now).

 

v2.78.4 for Linux (UNOFFICIAL, compiled from source)

blender-2.78.4-unofficial-linux_x86_64.tar.bz2

Edited December 18, 2016 by I.nfraR.ed

[davestarrr]

done