Current state of AMD APU landscape?

[NostaSeronx]

No AMD product currently has the hardware for VP9 decoding, but a hybrid solution should be enabled in the driver for Bristol Ridge and Polaris dGPUs.

Phase 0 driver support allows for fixed function VP9 4K30 & 2K60 and just those two. The actual fixed function device supports 12-bit @ 4K60 for VP9. VP9 uses a slightly modified HEVC part of UVD 6.x (Bristol Ridge/Stoney Ridge/Polaris 10/11). [http://www.webmproject.org/hardware/vp9/ , clean room design of G2v2 via existing HEVC][[This also means it can encode it.]][[[http://wiki.webmproject.org/ffmpeg/vp9-encoding-guide VP9 is the reason HEVC supports 2-pass as well]]]

 

[Abwx]

This is wrong, Skylake has a fully capable H265 decoding support on board since 1 year, AMD only since a couple of weeks. At least this is what you call full support. Both support 8 bit only in hardware.

SKL has no support other than some hybrid solution that do not work well, heck at some point you even said that it has main10 support while Hardawre.F stated that there s none and that they had no answer from Intel when they asked for some clues.

So with Kabylake AMD is far behind, no 10 bit and VP9 support. This is a big disadvantage for them.

You had to end your post with a blatant lie, i guess that you dont even read in the thread, i posted this today just above your post..:

Actually Intel just caught up with KBL, as for SKL it has no adequate support and is stuttering according to Notebookcheck in their KBL/SKL comparison, so much for the actually disadvantaged one..

AMD developers have today sent out patches for implementing HEVC Main 10 provide video decode support to the open-source RadeonSI UVD driver stack.

http://www.phoronix.com/scan.php?page=news_item&px=HEVC-Main-10-UVD

For Bristol Ridge and Stoney Ridge...

It is dated 8 March 2016, so it looks that you dont even bother about AMD capabilities, you re just here to mislead the public and vaunt non existent features on your beloved brand CPUs while negating what AMD did better..

 

[The Stilt]

This is wrong, Skylake has a fully capable H265 decoding support on board since 1 year, AMD only since a couple of weeks. At least this is what you call full support. Both support 8 bit only in hardware. So with Kabylake AMD is far behind, no 10 bit and VP9 support. This is a big disadvantage for them.

Bristol and Stoney Ridge along with Polaris support HEVC Main10.
Carrizo does not.

 

[mikk]

SKL has no support other than some hybrid solution that do not work well, heck at some point you even said that it has main10 support while Hardawre.F stated that there s none and that they had no answer from Intel when they asked for some clues.

Plain wrong. Skylake has a full decoding and encoding support in hardware for HEVC 8 bit which works incredibly fast. AMD only catches up with Bristol Bridge in this regard but will be far behind against Kabylake. I never said Skylake supports 10 bit in hardware, this is a Hybrid solution.

 

[mikk]

Bristol and Stoney Ridge along with Polaris support HEVC Main10.
Carrizo does not.

Is there a proof?

 

[NostaSeronx]

Is there a proof?

https://awesome.nwgat.ninja/rx480/P...roved/ef863980c6d742d6be8285356a039ac7-02.png
https://www.x.org/wiki/RadeonFeature/#index9h2
VCE3.4 = Stoney(Bristol), Polaris10, Polaris11

The Radeon™ RX Series features exceptional accelerated H.265 encoding and decoding, enabling effortless streaming or recording of 10-bit 4K video at 60 FPS4.

If you check VCE2 it supports YUV444, if it can encode then it can decode it. Which means possible support up to Main 4:4:4 12 support and VP9 Profile 3 support.

Encode capabilities always equals decode capabilities, while decode capabilities don't equal encode capabilities. Decode will always be better than encode.

 

[Abwx]

Plain wrong. Skylake has a full decoding and encoding support in hardware for HEVC 8 bit which works incredibly fast.

That s not the point wich is that you once said that it support main10 as well, wich is not the case.
As for HEVC 8 bit even Kaveri can (software) decode it albeit at 60% CPU utilisation with a A8 7600 but with pre Crimson drivers, dont know if they updated the drivers in this respect..

The Skylake hardware was already able to decode H-265/HEVC videos at a low power consumption, but it did not support the Main10 standard for contents with a 10-bit color depth.

http://www.notebookcheck.net/Kaby-Lake-Core-i7-7500U-Review-Skylake-on-Steroids.172692.0.html

AMD only catches up with Bristol Bridge in this regard but will be far behind against Kabylake. I never said Skylake supports 10 bit in hardware, this is a Hybrid solution.

The Hybrid solution is just cosmetic, in laptops it stutter.

As to AMD catching up with BR you are wrong, it is the other way around, it s KBL that catch up with BR since the latter was released before KBL and still, here what NBC say about this "support" :

The added VP9 support is particularly interesting for 4K videos on YouTube, because the videos use this codec by default in Google's Chrome browser. It unfortunately looks like the hardware acceleration does not really work yet: The average CPU load slightly drops from 66 (i7-6600U) to 54 percent (i7-7500U) and therefore also the power consumption (11.0 instead of 16.6 Watts), but it is still not very efficient. We probably need a driver or browser update to activate full hardware support.

So it doesnt work well even in KBL, not even sure that it s full hardware decdoding given the CPU utilisation of 54%, so who is catching up to whom, or rather, who is actually far behind.???.

 

[Sweepr]

So it doesnt work well even in KBL, not even sure that it s full hardware decdoding given the CPU utilisation of 54%, so who is catching up to whom.???.

Nice selective quoting there Abwx.

The Skylake hardware was already able to decode H-265/HEVC videos at a low power consumption, but it did not support the Main10 standard for contents with a 10-bit color depth. Kaby Lake changes that: Our demanding 4K trailer (HEVC Main10, 50 Mbps, 60 fps) is handled smoothly by the i7-7500U at an average power consumption of just 3.2 Watts (CPU Package Power), while the video stuttered noticeably on the i7-6600U despite a consumption of 16.5 Watts.

Regarding VP9:

We probably need a driver or browser update to activate full hardware support.

I wonder if Raven Ridge catches up or we will have to wait till 2018+ for a similar solution from AMD.

 

[Abwx]

Nice selective quoting there Abwx.

It s you that are doing selective quoting since you quoted the presentation of the claimed features at the start of the article while i quoted the results of the tests..

So the tests proved that the claims were not accurate...

Regarding VP9:

I wonder if Raven Ridge catches up or we will have to wait till 2018+ for a similar solution from AMD.

VP9 you said..?.

Looking at NBC tests you are just in full denial mode and spreading innaccuracies as a poor smokescreen by implying that AMD would be late when it s the other way around, Intel are more than late, as for AMD all they need is to catch up with themselves and keep being innovative.

 

[mikk]

https://awesome.nwgat.ninja/rx480/P...roved/ef863980c6d742d6be8285356a039ac7-02.png
https://www.x.org/wiki/RadeonFeature/#index9h2
VCE3.4 = Stoney(Bristol), Polaris10, Polaris11


If you check VCE2 it supports YUV444, if it can encode then it can decode it. Which means possible support up to Main 4:4:4 12 support and VP9 Profile 3 support.

First slider refers to Polaris, I'm asking for Bristol Bridge. We need a real proof. There is no mention of 10 bit HEVC in any of Bristol Bridge slides I have seen.

That s not the point wich is that you once said that it support main10 as well, wich is not the case.
As for HEVC 8 bit even Kaveri can (software) decode it albeit at 60% CPU utilisation with a A8 7600 but with pre Crimson drivers, dont know if they updated the drivers in this respect..

Don't you get it? Skylake 8 bit HEVC is in hardware, who cares about the software solution of Kaveri? Your goalposts are constantly shifting from post to post.

So it doesnt work well even in KBL, not even sure that it s full hardware decdoding given the CPU utilisation of 54%, so who is catching up to whom, or rather, who is actually far behind.???.

Nonsense.

Kaby Lake changes that: Our demanding 4K trailer (HEVC Main10, 50 Mbps, 60 fps) is handled smoothly by the i7-7500U at an average power consumption of just 3.2 Watts (CPU Package Power), while the video stuttered noticeably on the i7-6600U despite a consumption of 16.5 Watts.

Neither of the two chips has problems with a simpler 8-bit 4K video (HEVC, 6 Mbps, 60 fps), but Kaby Lake (3.4 Watts) is still more frugal than its predecessor (4.1 Watts).

Open yor eyes and take attention to the last. Some posts above you even denied the hardware support of 8 bit HEVC videos in Skylake.

 

[Sweepr]

Looking at NBC tests you are just in full denial mode and spreading innaccuracies as a poor smokescreen by implying that AMD would be late when it s the other way around, Intel are more than late, as for AMD all they need is to catch up with themselves and keep being innovative.

No, you are the only one in denial mode. Even Polaris lags behind Kaby Lake in terms of media support, 'smokescreen slides' prove nothing. Some basic research would save you from more embarrassment:

 

[mikk]

Correct, note that their drivers still doesn't support VP9, did it change recently? Real question is what about 10 bit HEVC for Bristol Bridge, why is there no mention in slides unlike Polaris? Is there a proof in any Bristol Bridge review?

 

[Abwx]

[

Nonsense.

Full denial mode as answer..?..

So you are saying that NBC are blatantly liying when they state that :

The added VP9 support is particularly interesting for 4K videos on YouTube, because the videos use this codec by default in Google's Chrome browser. It unfortunately looks like the hardware acceleration does not really work yet: The average CPU load slightly drops from 66 (i7-6600U) to 54 percent (i7-7500U) and therefore also the power consumption (11.0 instead of 16.6 Watts), but it is still not very efficient. We probably need a driver or browser update to activate full hardware support.

http://www.notebookcheck.net/Kaby-Lake-Core-i7-7500U-Review-Skylake-on-Steroids.172692.0.html


54% CPU utilisation instead of 66% means that it s not a hardware decoding, quite possible that it was all talks from Intel to give the impression that they were up to date, curiously the 10% difference amount to the difference in CPU perf and an eventualy slightly better driver, really curious isnt it...

So before looking what is supported or not by AMD you should start by checking if the brand you re eagerly supporting does even manage to fullfill her promises..

No, you are the only one in denial mode. Even Polaris lags behind Kaby Lake in terms of media support, 'smokescreen slides' prove nothing. Some basic research would save you from more embarrassment:

You can talk as much as you want, NBC numbers prove that there s no better hardware support in KBL than in SKL, numbers are numbers and the 54% CPU utilisation need no more argumentation, that s all good to put non existent features in the control panel, when the tests are done we can check the rubbishness of the thing..

 

[mikk]

[

Full denial mode as answer..?..

So you are saying that NBC are blatantly liying when they state that :.

You are unable to read properly or just trolling. This refers to VP9, don't you get it? Another goalpost shift from HEVC to VP9 and a different matter. VP9 seems supported from the Intel driver according to DXVA Checker, it may require a specific Chrome version. There is no detailed info. Intel demonstrated the advantage already.

http://www.notebookcheck.net/filead...e/7th_Generation_Intel_Core_Processors_13.jpg


https://en.wikipedia.org/wiki/Unified_Video_Decoder#UVD_6

According to Wikipedia Bristol Bridge only has UVD 6.0 build in, means no 10 bit HEVC in hardware.

 

[superstition]

or, a laptop.

Yeah. A laptop is a small form factor. : )

For general use, an apu (or i3) makes a lot more sense. For general use you dont need eight cores, and you get an igpu that is more than good enough for general use. An eight core FX only makes sense if you are doing heavily multi threaded productivity tasks or are going to pair it with a fairly powerful dgpu.

You get four FPU units with an 8 core chip and more cache. That outweighs the benefit of a minor GPU.

 

[Sweepr]

You can talk as much as you want, NBC numbers prove that there s no better hardware support in KBL than in SKL, numbers are numbers and the 54% CPU utilisation need no more argumentation, that s all good to put non existent features in the control panel, when the tests are done we can check the rubbishness of the thing..

There is a huge difference between not working yet - to be fixed with a driver/browser update (what NBC said), and no hardware decoding for VP9 (trolling from you). Kaby Lake has full hardware support as DXVA shows above, while Bristol Ridge only got VP9 'acceleration' according to AMD's own slides you posted. Even Polaris 10 (ahead of Bristol Ridge) lags behind HD 620 in terms of media support. AMD won't have a similar solution till Raven Ridge a year from now, at the earliest.

 

[Blue_Max]

Any word on coming improvements to the IGP? (The whole point of the APU, right?)

I want to build a very mini system which means no room for a video card at all, even single-slot half-height. IGP only. Lot of fun builds to be had with that and ITX (and the new format smaller than ITX - what was it?)

...but AMD really should bump up that IGP performance - it's the whole purpose of an APU in the first place!
*shrug*

 

[VirtualLarry]

And here I thought that Bristol Ridge would be out by now. Ah well, more waiting.

 

[Dresdenboy]

And here I thought that Bristol Ridge would be out by now. Ah well, more waiting.

According to Crashtest@P3DNow, they secretly launched on 8/30:
http://products.amd.com/Lists/APU/A...&&View={C155402C-EFC9-4976-98DD-79FDC5CF7F7B}

 

[hojnikb]

SKL has no support other than some hybrid solution that do not work well, heck at some point you even said that it has main10 support while Hardawre.F stated that there s none and that they had no answer from Intel when they asked for some clues.



You had to end your post with a blatant lie, i guess that you dont even read in the thread, i posted this today just above your post..:

Actually Intel just caught up with KBL, as for SKL it has no adequate support and is stuttering according to Notebookcheck in their KBL/SKL comparison, so much for the actually disadvantaged one..



http://www.phoronix.com/scan.php?page=news_item&px=HEVC-Main-10-UVD

For Bristol Ridge and Stoney Ridge...

It is dated 8 March 2016, so it looks that you dont even bother about AMD capabilities, you re just here to mislead the public and vaunt non existent features on your beloved brand CPUs while negating what AMD did better..

Actually, stuttering only happens on higher resolutions and FPS IF you're not using dual channel ram. With dualchannel ram, it works just fine.
I've tested 4k@hevc@60fps with dual channel ram on i3 6100 and it works just fine. With only one stick of ram however i can only hit around 35fps

 

[Blue_Max]

According to Crashtest@P3DNow, they secretly launched on 8/30:
http://products.amd.com/Lists/APU/AllItems.aspx?Paged=TRUE&p_ID=199&PageFirstRow=96751&&View={C155402C-EFC9-4976-98DD-79FDC5CF7F7B}

I got all excited... only to be let down again. Better IPC, same 512 graphic cores only clocked 12% faster. *sigh*
I guess they just don't want to cannibalize their own low-end graphic card division.

 

[leoneazzurro]

That, and the fact that with more CU they will be even more bandwidth limited than now.

 

[Dresdenboy]

I got all excited... only to be let down again. Better IPC, same 512 graphic cores only clocked 12% faster. *sigh*
I guess they just don't want to cannibalize their own low-end graphic card division.

There is some improved power management in BR, which might allow for somewhat higher CPU and GPU clocks, than with CZ.

That, and the fact that with more CU they will be even more bandwidth limited than now.

One poster says: same number, the other says: more CUs.... ^^
BTW, there is DDR4 instead of DDR3.

 

[coffeemonster]

I got all excited... only to be let down again. Better IPC, same 512 graphic cores only clocked 12% faster. *sigh*
I guess they just don't want to cannibalize their own low-end graphic card division.

it's been known for some time now that Bristol Ridge is basically just full Carrizo on AM4. Those 512 SP's were never able to be fully utilized by APUs before anyway. What were you getting all excited about?

 

[The Stilt]

BTW, there is DDR4 instead of DDR3.

Which of course makes no (positive) difference performance wise, since it cannot be clocked any further than DDR3 (without increasing the BCLK).
The increased latency of DDR4 meanwhile makes the already abysmal memory latency on Excavator even worse. It is already so high on DDR3 that it kills the performance in certain workloads (can be seen in Anand review too).