Google VP9 "Next Generation Open Video" information posted

[Jamaika]

2017-01-09 v1.6.1 "Long Tailed Duck"
  This release improves upon the VP9 encoder and speeds up the encoding and
  decoding processes.

  - Upgrading:
    This release is ABI compatible with 1.6.0.

  - Enhancements:
    Faster VP9 encoding and decoding.
    High bit depth builds now provide similar speed for 8 bit encode and decode
    for x86 targets. Other platforms and higher bit depth improvements are in
    progress.

  - Bug Fixes:
    A variety of fuzzing issues.

https://chromium.googlesource.com/webm/libvpx/+/v1.6.1
Patches are added to ffmpeg.

[Leeloo Minaï]

Can someone provide a vpxenc 1.6.1 build for Windows x64 ?

[LigH]

I can try to run jb-alvarado's media-autobuild_suite tonight...

[wiak]

Originally Posted by Leeloo Minaï

Can someone provide a vpxenc 1.6.1 build for Windows x64 ?

lets hope this works
https://awesome.nwgat.ninja/misc/libvpx-1.6.1-win64.7z

[LigH]

vpx v1.6.1-95-g9efc42f4f (static builds, MSYS2, GCC 6.3.0, Win32+Win64)

[Leeloo Minaï]

Thanks for your help !
I asked for the latest standalone version in order to compare with older revisions and FFmpeg builds.

In fact, i get slower encoding for revision 1.6.0 and 1.6.1 than with previous revisions 1.4.0 and 1.5.0 using vpxenc at same parameters (the same appears with FFmpeg too) :

Video properties :

384x256 @ 55fps, 400 frames, YUV420 (game capture from the arcade game "R-Type" (1987) using MAME emulator)

Encoding parameters :

vpxenc test.y4m -o vp9.webm --codec=vp9 -p 2 --pass=1 --fpf=stats.log --good --cpu-used=1 --target-bitrate=400 --end-usage=vbr --auto-alt-ref=1 --minsection-pct=5 --maxsection-pct=800 --lag-in-frames=16 --kf-min-dist=0 --kf-max-dist=360 --static-thresh=0 --drop-frame=0 --min-q=0 --max-q=63
vpxenc test.y4m -o vp9.webm --codec=vp9 -p 2 --pass=2 --fpf=stats.log --good --cpu-used=1 --target-bitrate=400 --end-usage=vbr --auto-alt-ref=1 --minsection-pct=5 --maxsection-pct=800 --lag-in-frames=16 --kf-min-dist=0 --kf-max-dist=360 --static-thresh=0 --drop-frame=0 --min-q=0 --max-q=63

Encoding times :
- 1.4.0 : 1:38,56 (~4.06 fps)
- 1.5.0 : 1:26,25 (~4.64 fps)
- 1.6.1 : 1:38,78 (~4.04 fps)

My laptop is rather old and low-powered (Pentium Dual-Core @ 2.2Ghz under Windows 7 x64), but i still laugh when i read "Faster encoding"

[Selur]

my guess is the 'faster encoding' would come from some assembler optimizations which might not help on older systems,...

[wiak]

Originally Posted by Selur

my guess is the 'faster encoding' would come from some assembler optimizations which might not help on older systems,...

that was my first idea too, newer systems has alot of new instructions and moar cores!

[Nintendo Maniac 64]

Originally Posted by wiak

moar cores!

You know it's becoming mainstream when even Intel is supposedly hopping on this bandwagon with Coffee Lake!

AMD was just too ahead of their time.

[wiak]

Originally Posted by Nintendo Maniac 64

You know it's becoming mainstream when even Intel is supposedly hopping on this bandwagon with Coffee Lake!

AMD was just too ahead of their time.

they have been ahead of the time for a while, case in point gcn

[CruNcher]

Not only GCN remember Fusion/HSA as well which every console has become part of and where Mantle arose from

[LigH]

cries in German

My best available CPU is still a Phenom-II X6. No AVX to boost modern codecs. Part-time paid full-time jobs. Globalization FTW.

[mandarinka]

Originally Posted by LigH

cries in German

My best available CPU is still a Phenom-II X6. No AVX to boost modern codecs. Part-time paid full-time jobs. Globalization FTW.

With Phenom, your main problem is lack of SSSE3 and SSE4, not lack of AVX. If you need more recent SIMD instructions tho, I see lots of refurbished office desktops with the likes of Core i5-2400 with prices starting at circa 175 € at the etailers that deal in such machines. Maybe that could be a viable way. Also Xeon E5-2600 chips of the "v1" (Sandy Bridge) category are cheap on ebay, but in this case it is hard to get the needed motherboard for cheap.

(And I bet we would all have even shittier computers and living standards were worse, if this "globalization" didn't occur. It's just scapegoat and/or red herring in somebody's agenda, as it happens.)

Sent from my second hand AMD A10-6800K...

[Nintendo Maniac 64]

I wouldn't be surprised if the Ryzen CPUs launching at the end of this month cause the prices on Sandy Bridge CPUs and LGA1155 systems to completely crater.

[mandarinka]

Originally Posted by Nintendo Maniac 64

I wouldn't be surprised if the Ryzen CPUs launching at the end of this month cause the prices on Sandy Bridge CPUs and LGA1155 systems to completely crater.

That would be nice, but the prices of those second-hand office desktops are already low if you consider that they have 4 GB of DDR3 (which got more expensive since 2012 instead of getting cheaper) and W7 license sticker and you can use those serials to activate clean installs of W10 on the machine, AFAIK.

Ryzens will launch at prices of i7-7700K and higher, so they might lower prices of second hand gaming computers, motherboards and CPUs like i5-2500K, i7-2600K and other quadcores that gamers or "enthusiasts" have been buying.

[Nintendo Maniac 64]

Originally Posted by mandarinka

Ryzens will launch at prices of i7-7700K and higher

You realize there will be fewer-core versions of Ryzen and not just the 8core/16thread variant...

[CruNcher]

Most probably though the most expensive ones 8 core will be launched first for pre adopters and then later the avg consumer 4 core ones following which would be the opposite from the GPU Strategy

[mandarinka]

Originally Posted by Nintendo Maniac 64

You realize there will be fewer-core versions of Ryzen and not just the 8core/16thread variant...

I know (https://s23.postimg.org/tqb06wo0p/9d...4c300b9ea9.jpg). My estimate is that they will get to about 150-175 USD price level ( 4C/4T). But currently it seems that such SKUs won't be part of the initial launch, it's unknown when will they start to sell.

[CruNcher]

Interesting now it seems we have a standard for Performance indexing with a Number from Intel and AMD

Ryzen 3,5,7
Core 3,5,7


which respectively points to 2,4,8 and 4,6,8 cores with a balancing in between thread executions and clock

So in the middle Intel fights vs 6 cores now with 4 and its AVX2

The most interesting compare point between both at those 65W

[nevcairiel]

Originally Posted by CruNcher

which respectively points to 2,4,8 and 4,6,8 cores with a balancing in between thread executions

But it doesn't, not for Intel anyway.
There is i7s with 4, 6, 8 and 10 cores.