*** AMD "Zen" thread (inc AM4/APU discussion) ***

[SiDeards73]

Looking at results from that thread, 4770k are getting like 1min 30, 83xx/9590 are getting 2min 30+

So I would say yes, your Haswell is almost twice as fast as an AMD chip in Blender on that test

 

[jagga1968]

slower than a 6700k slower than 5820k by quite some bit.so no im not worried or bothered.

so if it is faster you will be worried and bothered just like the other intel fanboys here.

i have a great intel machine here and if amd smash my 6700k on price and performance i say "fair play well done"

 

[Martini1991]

Looking at results from that thread, 4770k are getting like 1min 30, 83xx/9590 are getting 2min 30+

So I would say yes, your Haswell is almost twice as fast as an AMD chip in Blender on that test

Which surely would make people question the benchmark. Because it doesn't happen anywhere else, even with programs which do "the same" ; Cinebench etc.

2 weeks ago if I used a blender results to state how fast my 4700K was in comparison to an FX83 I'd get slaughtered, because there's no way in this green Earth that an FX83 owner would blindly accept that their FX83 was half the speed of a 4770K (Because it's not in anything else)

I got 1:18 using latest Blender and their AMD image at 4.4GHZ. People with FX83's are getting double the speed in the same conditions. I can't accept a result which shows a performance disparity that doesn't exist elsewhere.

 

[SiDeards73]

In fact if you read that thread it's obvious that AMD chips don't do well at all in Blender versus Intel chips, if anything the Intel chips are smashing AMD chips all over the place.

Which makes that RyZen result even more intriguing

 

[~>Dg<~]

as pointed out they already said its upto 40 percent better than last gen for amd which isnt even 4770k !

a 6700k/5820k is close to twice as fast as a 3770k

work it out.

so nope not worried.just work it out.stop living in dream land.

 

[jagga1968]

as pointed out they already said its upto 40 percent better than last gen for amd which isnt even 4770k !

a 6700k/5820k is close to twice as fast as a 3770k

work it out.

so nope not worried.just work it out.stop living in dream land.

no you got it wrong they said they have surpassed the 40% gains.

 

[Martini1991]

In fact if you read that thread it's obvious that AMD chips don't do well at all in Blender versus Intel chips, if anything the Intel chips are smashing AMD chips all over the place.

Which makes that RyZen result even more intriguing

No, it should make you question Blender.

Cinebench is meant to do the "same thing". Yet there's no where near that type of performance disparity in supposedly the same workload.

This would be called cherry picking had it been used as a pro i7 slant.

Adding to the fact that I don't think AMD confirmed what version of Blender they were using (An older version completes the AMD benchmark at 1:00 for me)

 

[flopper]

RyZen is about as amazen as it zen to be.
buying asap

 

[SiDeards73]

Which surely would make people question the benchmark. Because it doesn't happen anywhere else, even with programs which do "the same" ; Cinebench etc.

2 weeks ago if I used a blender results to state how fast my 4700K was in comparison to an FX83 I'd get slaughtered, because there's no way in this green Earth that an FX83 owner would blindly accept that their FX83 was half the speed of a 4770K (Because it's not in anything else)

I'd say that benchmark favours the high multithread performance of Intel chips and their better IPC than AMD, those AMD chips just can't make up the ground even with more physical cores and higher clock speeds.

So it's interesting that the RyZen chip runs so well, if anything it shows how different Zen architecture is to Bulldozer etc, and possibly points to their IPC and SMT being decent now?

Must be some other wizardry at play too with Ryzen

 

[jagga1968]

This is a decent thread from AMD Reddit where one guy got people to benchmark blender with that RyZen benchmark

https://www.reddit.com/r/Amd/comments/5i6w0w/submit_your_blender_test_for_comparison/

Makes for interesting reading

great post...thank you!

 

[Mei]

In fact if you read that thread it's obvious that AMD chips don't do well at all in Blender versus Intel chips, if anything the Intel chips are smashing AMD chips all over the place.

Which makes that RyZen result even more intriguing

bcos its a total redesign
its nothing like a fx chip
the last time intel did a redesign this big was core-duo? i duno i was prob playing out on my bike

 

[Martini1991]

I'd say that benchmark favours the high multithread performance of Intel chips and their better IPC than AMD, those AMD chips just can't make up the ground even with more physical cores and

Intel's going to come out top 4700K against FX83 in the type of FPU heavy workload like Blender, but not twice as powerful.
No other program, even with the same workloads shows that disparity.

2 Weeks ago this would be blasted (And rightly so) as cherry picking. Humbug and the like would never believe any benchmark which shows an i7 4C/8T doubling the performance of an FX83 at the same clocks.

 

[jagga1968]

Intel's going to come out top 4700K against FX83 in the type of FPU heavy workload like Blender, but not twice as powerful.
No other program, even with the same workloads shows that disparity.
.

give it a rest your theory is flaky and your like a stuck record.

many men here already told you unless you know for certain what settings were used your results dont mean ****.

 

[Cenuij]

great post...thank you!

Aha! They (AMD) used 100 samples per pixel instead of the default 200...

I get 19 seconds to render the first frame instead of 29 seconds at 200 samples/pixel

 

[jagga1968]

Aha! They (AMD) used 100 samples per pixel instead of the default 200...

I get 19 seconds to render the first frame instead of 29 seconds at 200 samples/pixel

Trouble is the camera footage looks to low to actually make out the numbers correctly without messing.

As i understand it someone took a high res photo of the render in action, blew it up and doctored it to look like AMD were cheating, like they had set the AMD chip to cycle 100 renders and a much higher number (200) on the Intel chip.

Basically trolling, or a paid Intel shill put there to try and seed doubt on its performance and make AMD look like liars.

Aha! no one knows what settings were used.

you didnt have the common sense to match clock speed either.

 

[mattyfez]

So fanboyism aside it seems like theres not much to go on either way for gaming?

 

[Martini1991]

So fanboyism aside it seems like theres not much to go on either way for gaming?

There was nothing to be negative about looking towards the future in regards to buying a Zen and gaming.

 

[SiDeards73]

Aha! They (AMD) used 100 samples per pixel instead of the default 200...

I get 19 seconds to render the first frame instead of 29 seconds at 200 samples/pixel

Actually they used 150 samples and the build version they used was 2.78a x64

https://www.reddit.com/r/Amd/comments/5ie7f0/summoning_uamd_robert_how_can_we_do_the_blender/

Which makes it even more impressive i think?

 

[d_brennen]

Zen is hardly a huge die, $400 is entirely reasonable. The premium we pay Intel is just that, a premium because there's no choice

 

[Cooper]

More interested in the 6c/12t chip TBH.

Hopefully will come with a base clock near 4Ghz and be overclockable to at least 4.5Ghz.