AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[sirmo]

This is just marketing from AMD, look how old this video is. Another case of "AMD likes to promote things in slides that they don't support in real wold"....only that it's a video

And VCE is poor compared to Quicksync and NVENC.

It is not marketing, it's real, here is the actual PR: https://github.com/HandBrake/HandBrake/pull/88/files

Also source on VCE being poor compared to quicksync and NVENC?

I use VCE in OBS all the time by the way and it works without issues. It actually has much better throughput than my 4770K QuickSync.

 

[AtenRa]

http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-eventDetails&EventId=5243376

Lisa Su talks about ZEN, VEGA, ARM, Semi-custom and more. There are some very interesting things said here.

Seeking alpha transcript
http://seekingalpha.com/article/402...ts-credit-suisse-technology-media-and-telecom

 

[Dresdenboy]

They did?

I think, he meant this:

 

[Glo.]

Excavator desktop CPU with 65W is Athlon X4 845 If I am not mistaken.

 

[KTE]

Excavator desktop CPU with 65W is Athlon X4 845 If I am not mistaken.

Yup. Nice little chip for std use.

Sent from HTC 10
(Opinions are own)

 

[Phynaz]

Also source on VCE being poor compared to quicksync and NVENC?

https://forums.anandtech.com/threads/awful-amd-vce-performance.2489847/

 

[sirmo]

https://forums.anandtech.com/threads/awful-amd-vce-performance.2489847/

Not sure what I am supposed to see there.. just bunch of people having issues installing the plugin. I ran the benchmark just yesterday, for 1080p streaming I get 119.0 fps on the rx480.. on speed setting. balanced is like 96 FPS. Which seems like more than enough for streaming.

 

[mikk]

It is not marketing, it's real, here is the actual PR: https://github.com/HandBrake/HandBrake/pull/88/files

Also source on VCE being poor compared to quicksync and NVENC?

I use VCE in OBS all the time by the way and it works without issues. It actually has much better throughput than my 4770K QuickSync.

It is. Even a Handbrake developer said it's just a marketing promo from AMD when someone asked in their forums. This video is almost a year old. What do you get from it? Nothing. It's a promo and marketing, nothing more. Write back when it is available for the public, probably never. And even if it is in a few years, it won't have a chance against Quicksync and NVENC.

 

[AtenRa]

It is. Even a Handbrake developer said it's just a marketing promo from AMD when someone asked in their forums. This video is almost a year old. What do you get from it? Nothing. It's a promo and marketing, nothing more. Write back when it is available for the public, probably never. And even if it is in a few years, it won't have a chance against Quicksync and NVENC.

All i see is your word, at least he posted a link. You know we could all benefit more if you would provide a link for what you saying.

 

[mikk]

All i see is your word, at least he posted a link. You know we could all benefit more if you would provide a link for what you saying.

https://handbrake.fr/
https://forum.handbrake.fr/


Here is the source. Nothing is available, it is marketing according to Handbrake. Wake people up when it is available, otherwise it is another AMD buzz.

 

[Fanatical Meat]

For a CPU novice like me, why is everyone making a big deal about a handbrake benchmark?

 

[AtenRa]

https://handbrake.fr/
https://forum.handbrake.fr/


Here is the source. Nothing is available, it is marketing according to Handbrake. Wake people up when it is available, otherwise it is another AMD buzz.

Where is the developer reply link and where does it say Handbrake doesnt support AMDs VCE ???

ps. Im not saying it does, but you have to provide the link. I dont believe you are expecting me to search for it

 

[AtenRa]

OK i did some searching and OBS does support VCE

Link here

Also i have found the link the Developer talks about OBS and VCE support from back in 2014

Link here

 

[TheELF]

Where is the developer reply link and where does it say Handbrake doesnt support AMDs VCE ???

ps. Im not saying it does, but you have to provide the link. I dont believe you are expecting me to search for it

Best proof available...just download the damn prog and search for VCE.
https://handbrake.fr/nightly.php

 

[mikk]

OK i did some searching and OBS does support VCE

Link here

Also i have found the link the Developer talks about OBS and VCE support from back in 2014

Link here

You are such a......

We are talking about a Handbrake version with VCE based on a 1 year old promo video. The truth is such a version does not exist, expect for AMDs marketing. Nobody is denying the existence of VCE in general. Some people are just strange.

 

[bjt2]

They did?

Any reference for that, including this bit:



please?

+40% IPC at equal energy per cycle means same consumption at same clock...

 

[bjt2]

He has some pretty darn crazy, non scientific, theories comparing an ARM research test FPU vs XV, applied to Zen in a sounds-good-to-me biblical zeal fashion.

In short, nothing of Zens power@clocks scaling is known yet.

Nor how it compares to XV.

Nor how the process stacks up vs Intels.

Sent from HTC 10
(Opinions are own)

See above...

My theory was to try to figure out clock and consumption. Then, after the publication of the slide posted above, I used that argument to justify that statement... Anyway if you think that a Zen core will draw more that 5W at 2.4Ghz, i think you will be very disappointed soon... The NEON FPU argument was to estimate a consumption at 2.4GHz... Above that we had no data. With the AMD statement of same consumption as excavator core, we know that we can expect same clock as excavator at same core number, and since in that graph we also know that excavator is better than piledriver and other 32nm cores, we can conclude also that Zen 8c will have at least same clocks than BD 8c, e.g. FX8370@125W or FX8370E@95W...

 

[jpiniero]

See above..., we can conclude also that Zen 8c will have at least same clocks than BD 8c, e.g. FX8370@125W or FX8370E@95W...

I have no idea how you got from that graph to "it will have the same clock speeds as Bulldozer"

 

[swilli89]

For a CPU novice like me, why is everyone making a big deal about a handbrake benchmark?

Because people are trying to derail this thread.

 

[Fanatical Meat]

Because people are trying to derail this thread.

Thanks. I was wondering why a pretty minor app was such a big thing. Do people burn that many DVDs or convert that much video they can't wait 5-10% longer?

 

[swilli89]

Thanks. I was wondering why a pretty minor app was such a big thing. Do people burn that many DVDs or convert that much video they can't wait 5-10% longer?

Yeah by the sound of it I guess everyone here spends there days converting video files. I'm just a gamer so that's all I really care about.

 

[TheELF]

+40% IPC at equal energy per cycle means same consumption at same clock...

No it doesn't, using 40% more energy for a given workload but at 40% less clocks(less cycles) is still same energy per cycle.Zen might need all the available TDP to get 3,2Ghz(40% less cycles then 4,5Ghz) .

 

[Abwx]

No it doesn't, using 40% more energy for a given workload but at 40% less clocks(less cycles) is still same energy per cycle.Zen might need all the available TDP to get 3,2Ghz(40% less cycles then 4,5Ghz) .

It is written same energy per cycle, wich is clock cycle...

40% more instruction/cycle at same energy per cycle.

 

[sirmo]

You are such a......

We are talking about a Handbrake version with VCE based on a 1 year old promo video. The truth is such a version does not exist, expect for AMDs marketing. Nobody is denying the existence of VCE in general. Some people are just strange.

I said they were adding support in (not that it's completed), but there is a whole PR with source in it.. handbrake developer hasn't merged it yet.. but you can compile it yourself if you want. It's not marketing there is real code being submitted. Some forum post from 2014 is irrelevant this is from 2016.

I understand you want to bash AMD every chance you get.. but this is ridiculous. Did AMD kill your dog?

The maintainer of handbrake wrote on the Support for VCE Encoder commit:

Core LibHB patch by Michael Wootton from AMD.

And there is 1,684 lines of code in the patch. I reviewed it and it's legit. https://github.com/HandBrake/HandBrake/pull/88/files

Also if you look at that PR it was accepted in June, so 4 months ago.. and it's actually slated for the 1.1.0 release (whatever that version is, going by their milestones in github). So you can either compile it yourself.. there are instructions on how to do it. Or you can wait for the 1.1.0 release. It's not marketing.

Here since you obviously don't know how to use github: https://github.com/HandBrake/HandBrake/labels/Accepted

1.1.0 milestone: https://github.com/HandBrake/HandBrake/milestone/4 VCE support is the oldest accepted item.. so the hold up seems to be handbrake guys. AMD submitted the patch in February, It was approved in June, It's waiting to be merged and the next version will have it.

 

[deasd]

It is written same energy per cycle, wich is clock cycle...

40% more instruction/cycle at same energy per cycle.

Thanks. It's first time I see 'Energy per Cycle' statement. Looks like Zen has similar pipeline stage and FO4 delay compared to EXV, while have much wider engine and SMT.
This could also imply the IPC gain might be optimistic although still couldn't tell how much.