Why isn't Ryzen faster for encoding video?

[JeremyCT]

Is this something that needs to be worked out in software or is it just a weakness of the architecture? With twice the cores/threads as a mainstream Intel chip it only seems to keep up rather than surging ahead. I find this odd because video encoding is an area where more cores/threads traditionally helps. I haven't seen any articles addressing this, does anyone here have any particular reason why this is the case?

 

[KTRF]

Sorry, nothing to add but I'm interested in a new 4K video edit PC and I was thinking of Ryzen 1700 or 1700k.
I would have thought that it would/should be faster at video encoding, must be something wrong.

 

[Janus67]

My only guess is maybe the operating system and the applications aren't quite coded to handle the methodology that AMD utilizes their cores/modules?

 

[RJ88]

Still it's definitely an improvement and in rendering seen even in CB that's an easy 20-30%+ better than a clocked 8xxx pd.

 

[KTRF]

What Intel chip are compairing it to for encoding speed?

 

[trents]

Is this something that needs to be worked out in software or is it just a weakness of the architecture? With twice the cores/threads as a mainstream Intel chip it only seems to keep up rather than surging ahead. I find this odd because video encoding is an area where more cores/threads traditionally helps. I haven't seen any articles addressing this, does anyone here have any particular reason why this is the case?

What software are you using for encoding? Do you know if it's optimized for Intel architecture or maybe isn't well multi-threaded?

 

[chrisjames61]

Is this something that needs to be worked out in software or is it just a weakness of the architecture? With twice the cores/threads as a mainstream Intel chip it only seems to keep up rather than surging ahead. I find this odd because video encoding is an area where more cores/threads traditionally helps. I haven't seen any articles addressing this, does anyone here have any particular reason why this is the case?

Do you have any proof or evidence to show us? Without that your post borders on trolling.

 

[ssjwizard]

As far as I know handbrake should scale proportionately to the thread count. If you are not seeing a substantial increase from having more threads your software is probably not coded for as many cores/threads as Ryzen provides.

 

[dejo]

Intel Quick Sync maybe?
For transcoding I would guess that it is near the octo thread Intel chips as well

 

[trents]

Hey, we already have encoding benchmarks from Ryzen' first few days of release and they show it kills Intel mainstream gaming chips. About 10% in per core performance behind Intel in games.

 

[unilx]

Enable high performance mode in windows and manually set your chip to 3.7Ghz in your bios.

Edit: Make sure you're using the latest bios for your mobo.

 

[Alaric]

does anyone here have any particular reason why this is the case?

Everything I've seen shows that's not the case.

 

[JeremyCT]

Wow, ok, you guys must be reading different reviews than I am. For reference here, I'm going to look at Anandtech results. Source: http://www.anandtech.com/show/11170...review-a-deep-dive-on-1800x-1700x-and-1700/20

Hey, we already have encoding benchmarks from Ryzen' first few days of release and they show it kills Intel mainstream gaming chips. About 10% in per core performance behind Intel in games.

Do you not consider the i7 7700k to be a "mainstream gaming" chip? Honest question. If you like, I can use the i5 7600k to compare the 1800x to instead. Both of those cost a good bit less than the 1800x, but since that's what I actually compared the AMD chip to when building a machine for a client, I can go that way.

As far as I know handbrake should scale proportionately to the thread count. If you are not seeing a substantial increase from having more threads your software is probably not coded for as many cores/threads as Ryzen provides.

Handbrake is, AFAIK, able to consume as many CPU resources as one can throw at it. The link to the Anandtech testing is above, but other articles I've seen on Ryzen performance have had comparable results. I'll break down my concern/interest here so you can see what I'm talking about a bit more clearly.

H264LQ
7700k: 1327
7600k: 970
1800x: 1345

Gain over i7: 1%
Gain over i5: 28%

H264HQ
7700k: 32.21
7600k: 25.08
1800x: 36.83

Gain over i7: 14%
Gain over i5: 32%

HEVC
7700k: 47.50
7600k: 41.41
1800x: 46.63

Gain over i7: -2%
Gain over i5: 11%

So, despite a literal doubling of the core/thread count, Ryzen manages only a 14% win over the 7700k in one encoding, matches it in another, and loses in the third. A 7700k costs $329 compared to an 1800x at $499. This does not impress or encourage me, and why I'm asking whether it's a weakness of the architecture or a weakness of the software. Handbrake is generally considered to be pretty well optimized, but I don't know if there are gains to be had here. I'd start to wonder about seriously diminishing returns as core/thread count rises, but the 6900k came in tops in all three tests, so it seems the limiting factor is elsewhere.

Do you have any proof or evidence to show us? Without that your post borders on trolling.

Everything I've seen shows that's not the case.

As I mentioned, these numbers match results I've seen elsewhere. I'm not sure what reviews you guys were reading.

 

[Alaric]

The 7700k is what, 500 MHz faster with a higher IPC than Ryzen?

 

[mackerel]

Looks like there was a copying misalignment for the H264LQ numbers, where the 1800X did come on top of tha bunch, and I've also added the lower Ryzen models for a broader picture.

1800X: 1345
7700K: 1327
1700X: 1296
1700: 1213
7600K: 970

I'm not familiar with video coding software in general. In other testing I've done, Ryzen is weak in AVX, to the tune of about half IPC relative to Intel. This was a design choice by AMD. If AVX instructions were used, it would not be a strong area for AMD.

 

[ShrimpBrime]

So, despite a literal doubling of the core/thread count, Ryzen manages only a 14% win over the 7700k in one encoding test out of three. A 7700k costs $329 compared to an 1800x at $499. This does not impress or encourage me, and why I'm asking whether it's a weakness of the architecture or a weakness of the software. Handbrake is generally considered to be pretty well optimized, but I don't know if there are gains to be had here. I'd start to wonder about seriously diminishing returns as core/thread count rises, but the 6900k came in tops in all three tests, so it seems the limiting factor is elsewhere.

Both software and Hardware limitations. The hardware is brand new yet. So software would need to be optimized.

If software is written on Intel platform and optimized for Intel platform, Like many games, it will inherently run better on that particular platform.

Multi-Tasking is the aim with lots of threads. However a 40% IPC gain from FX to Ryzen is a pretty large gap to fill. And the price tag is right on all things considering.

 

[JeremyCT]

The 7700k is what, 500 MHz faster with a higher IPC than Ryzen?

Yes, but half the core/thread count.

 

[Alaric]

True, but that doesn't make the IPC and speed irrelevant.

 

[JeremyCT]

Looks like there was a copying misalignment for the H264LQ numbers, where the 1800X did come on top of tha bunch, and I've also added the lower Ryzen models for a broader picture.

1800X: 1345
7700K: 1327
1700X: 1296
1700: 1213
7600K: 970

I'm not familiar with video coding software in general. In other testing I've done, Ryzen is weak in AVX, to the tune of about half IPC relative to Intel. This was a design choice by AMD. If AVX instructions were used, it would not be a strong area for AMD.

That's what happens when I copy things over in a hurry apparently, good catch. I fixed the post.

 

[ShrimpBrime]

True, but that doesn't make the IPC and speed irrelevant.

Very relevant. Instruction sets also relevant.

But is HandBrake the tell all of processing power? Just wondering.....