AMD Navi Product Reviews and Previews: (5500, 5600 XT, 5700, 5700 XT)

Option 1 and specifically targeted at Apple would sound reasonable.

 

Just putting an idea out there - Microsoft surface family, specifically studio 2 is running old hardware. A Ryzen 3700 Pro with 5600 XT or Navi12 for NUC form factor. Sounds almost like a devkit for Project Scarlet

 

It'd be interesting to see Apple release an all AMD product as part of their lineup and I'd love to see it. But I remain somewhat sceptical. Unless it's time for Apple to renegotiate its contract with Intel. Renoir would be a good fit for lower end MacBooks perhaps? A redesigned Air and/or new MacBook SKU for the 13 inch segment? But again, I remain somewhat sceptical. Mainly because I'm sure Intel would give quite a bit to keep Apple as a high profile customer.

 

Van Gogh is already in the 10.15.2 beta 3.

 

Really? It's the first time I've seen it mentioned. Ming-Chi Kuo et.al. are usually rather quick with reporting on suspected supplier changes. Interesting times ahead then!

 

Problem is, Intel can't offer Apple a competetive part.
TGL CPU part straight out sucks versus what will be the market in but a day.

 

Huh? Sunny Cove (Ice Lake) is great core with notably higher IPC over Skylake (including all the latest Skylake-iterations), there's no reason to think Willow Cove (Tiger Lake) would be any worse. The only issue is the 10nm process itself which is still worse than their 14nm even in mobile space.
If they can't fix that, they could be in trouble but TGL still wouldn't suck compared to what's on the market - it just might not be the best one. If they can fix it, there's no reason TGL couldn't be the best mobile solution out there.

 

Apple, to my chagrin, doesn't mind waiting if they feel they have to. Parts of their lineup have had much more than a year between updates historically, even when newer and better Intel components have been available to them. If Apple truly cared about having the latest and greatest I'm sure they would have migrated over to AMD with the new Mac Pro for example. Besides which they most likely have massive sunk costs, inventories, and business efficiencies predicated on Intel hardware I bet.

 

What are the odds of RDNA2/Big Navi coming with AV1 decoding?

 

I wonder if they'll do so with Renoir vs Ice Lake and Comet Lake.

 

Btw why is this no hardware decoded yet ?

 

Hen & egg? Last year, the first dedicated IP was announced (edit: and there is also this Android Box with this chip strangely enough not mentioning AV1, and an SoC from Broadcom ), now at CES the first TVs with HW support for AV1 are coming. That means, streaming services can VERY SLOWLY start to use it (apart from testclips on YT which you actively have to select). Depending on how much die space it will cost, maybe we are going to see dedicated AV1 IP in the next generation of GPUs.

 

Edit- See above, oops

 

Is there an advantage of AV1 over HEVC, other than royalties?

 

I had the same question as well. Looks to be more efficient than HEVC and VP9 for bandwidth on playback, but it seems thousands of times slower on encoding.

Here's a snippet from one of the first articles I found on it:

https://www.macxdvd.com/mac-dvd-video-converter-how-to/aom-av1-vs-hevc-h265.htm

Bitmovin tested the MPEG-DASH/HLS playback of AV1 in Firefox Nightly by ensuring encoder and decoder use the same bitstream version, proving that the new coding format is 25%-35% more efficient than HEVC and VP9.

AV1 aims to be 30% more efficient than HEVC and VP9. Tests performed on a wide range of bitrate and resolutions demonstrate that the average bitrate of AV1 is reduced by 17%, compared to HEVC/H.265. When it comes to 4K Ultra HD 2160p video, AV1 outperforms HEVC up to 43.90%, compared to 720p of 37.81%. More results are shown in the following HEVC and AV1 dataset pictures from Bitmovin. Facebook shows that AV1 encoder is roughly bitrate reduction of 40% compared to VP9 and up to 50% while compared to H.264 (x264). That also means AV1 codec cuts down bandwidth demands and compresses 8K/4K file size by up to 30%.

Compare AV1 and HEVC/H.265, the newer format loses in terms of speed. Studies suggest that AV1 encoder is extremely slow - "2500-3000 times lower than competitors". At 2018 NAB Show, AOM developers told AV1 was about 50-200 times slower than HEVC and VP9. Currently, it takes nearly 100 seconds to encode a single 4K frame. The later development focuses on AV1 encoding speed. Considering its deadly slow speed, it still takes a long way to go to be in wide use.

HEVC encode and decode require 4-10 times the compute power at the same speed as H.264 due to its complexity. It's a CPU and memory-intensive task to deal with HEVC.​

 

AV1 will compete with h.266, not h.265. It's next generation codec developed for 8k video - focused at higher compression ratio (than h.265/vp9) at comparable visual quality. The problem is that AV1 is late and h.266 extremely late. 8k LCD panels are ready, but decoders aren't (AV1 decoders starts to emerger, h.266 isn't ready yet). Current "8k" TVs don't support codecs developed for 8k resolution (only h.265/vp9), so once real 8k content (AV1 / h.266) appears, current 8k TVs will be obsolete / incompatible.

 

Unless you buy a cheap streaming box down the line.

 

Why can't they use HEVC?

It's not like bitrates on HEVC 8K are astronomical, given today's home internet connections in a typical developed country that isn't the U.S.
Netflix does 4K HDR @ ~15Mbit/s, so even if they make 8K as 4*4K then it's still only 60Mbit/s, or 7.5 MB/s.


Or are the embedded codecs present in current 8K TVs just not powerful enough for 60Mb/s HEVC?

 

Exactly. What I would personally require though, is full featured hdmi 2.1 interfaces for debices to plug in to.

 

It isn't that simple. For 8k @60Hz @10bit (HDR) you need DisplayPort 1.4 (most TVs don't have DisplayPort) or HDMI 2.1 (only some 8k TVs support HDMI 2.1). Customers can buy so called 8k TV, but it won't be able to play 8k video compressed in native codecs (AV1/h.266) nor use external decoder, because of the lack of fast enough interface.