Vega-Based Renoir APU Has The Same VCN Video Encode/Decode Block As Navi

[phoronix]

Phoronix: Vega-Based Renoir APU Has The Same VCN Video Encode/Decode Block As Navi

The next-generation AMD "Renoir" APU is turning into being an interesting successor over the existing Picasso APUs. While at first it was a letdown finding out that the APU is based on Vega and not their newer Navi architecture, follow-on open-source Linux patches have continued to show that it's more than a facsimile and in some areas like display and multimedia has blocks in common with Navi...

http://www.phoronix.com/scan.php?pag...ame-Navi-Block

[coder]

why we can not just agree that "non-free" is bad for humanity.

I'm not trying to take a position on the issue, though I respect yours.

One can argue that governments should fund innovation at universities, and then we can have unencumbered implementations, or that the private sector should fund innovation at companies. There are cases to be made for both options.

I'm just trying to draw the distinction that "open source" is merely a means of collaboration, where as FLOSS includes the ethos you describe.

[bridgman]

So, a more accurate phrase would be "non-free open source", because it's open but not free to use.

Or "free and open source implementing a non-free standard".

[coder]

for me this is closed source

You can call it whatever you want, but for communication to work, words need to have some mutually agreed meaning. Your definition of closed-source is non-standard, because the source, in this case, is literally open.

because FLOSS is for me: Free to use and free to share without any payment to patent holders.

So, a more accurate phrase would be "non-free open source", because it's open but not free to use.

[coder]

yes and i really mean x265 because GPL x265 with Patents means in fact closed source

WTF?

H.265 is patented. x265 implements H.265 and is GPL. So, patents do not mean closed source!

However, in order to use a GPL implementation of a patented technique, you might be legally required to pay a license fee to the patent holder. But the implementation can still be open source.

That said, perhaps the patent holder can also order the open source developer(s) to cease distributing their (unlicensed) implementation. I'm not sure about that, but I've always assumed they could shut it down (or at least try).

open-source but patented is really idiotic ...

Open Source is just a collaboration model. It's an entirely legitimate way for people or companies to cooperatively develop software that might still be non-free to use. Sure, it might not attract as much interest and contributions as a project that's not encumbered by patents, but that doesn't make it idiotic.

in fact the software patents where in fact only made by corrupt politicians
to make sure it sabotage the FOSS/FLOSS movement.

Just because you don't like software patents doesn't make it okay to spin lies about them. Software patents long predate the open source movement.

https://en.wikipedia.org/wiki/Softwa...oftware_patent

Patents on software constructs were likely embraced for the same reasons as patents in other domains, which is not to say that they're not abused.

in my point of view there is no real difference between H.265 and X.265

One is a standard and one is an implementation.

[coder]

Really? People still use the ancient MPEG-2?

Yeah, this is somewhat dubious. I mean, if you're working with a legacy system, like ATSC, then it's the only option. Otherwise, there's no reason to prefer it over newer codecs.

I agree. I feel it's mostly designed with hardware accelerators in mind, and some are inconsistent. I tried decoding HEVC video with my AMD card, and saw a few differences when compared to CPU decoding.

That's weird, because the decoders are exactly standardized. Since H.264, the standards leave no ambiguity about exactly how a given bitstream should be decoded. The only thing I don't know is if H.265 specifies error recovery so rigorously.

Anyway, what happens if decoders are allowed to produce a different output is that the differences between the expected output & actual output will accumulate between decoder refreshes, leading to some sort of creeping picture corruption.

Apparently professionals prefer proprietary formats... I don't see the reason though...

JPEG2000 is an ISO standard, just like MPEG-4 - not proprietary. And they are open standards, meaning anyone can access the specification. However, open standard doesn't imply free-to-use.

Anyway, here's what wikipedia says about JPEG 2000:

JPEG 2000 is covered by patents, but the contributing companies and organizations agreed that licenses for its first part—the core coding system—can be obtained free of charge from all contributors.

...

However, the JPEG committee has acknowledged that undeclared submarine patents may still present a hazard

More, at: https://en.wikipedia.org/wiki/JPEG_2000#Legal_status

[tildearrow]

You do realize that x265 is GPL'd software, do you not?

He probably means the patent-filled codec.

I would disagree that people want AV1, people want something that is compatible across a wide range of devices and uses, which is why MPEG-2 is still used in many places, and why H264 is still so popular.

Really? People still use the ancient MPEG-2?

HEVC is still a beast to work with, in terms of editing and hardware playback support is still not ubiquitous.

I agree. I feel it's mostly designed with hardware accelerators in mind, and some are inconsistent. I tried decoding HEVC video with my AMD card, and saw a few differences when compared to CPU decoding.

VP9 is completely open, with open source implementations and is very high quality, yet it hasn't taken over.

I use lots of VP9 in YouTube, especially on mobile.
Maybe it hasn't taken over due to its slower encoding?

JPEG2000 came out in, wait for

No, I'm not waiting. It's super obvious

RED cameras capture in JPEG2000 natively and ProRes and DNxHD with third party add ons.

Apparently professionals prefer proprietary formats... I don't see the reason though...

[coder]

You do realize that x265 is GPL'd software, do you not?

IMO, it's obvious Qaridarium meant H.265, which is patented.

I would disagree that people want AV1, people want something that is compatible across a wide range of devices and uses, which is why MPEG-2 is still used in many places, and why H264 is still so popular.

Well, perhaps AV1 was too ambitious, by targeting superior compression ratios and being royalty-free. Perhaps they thought their only chance of displacing H.264/H.265 was by leapfrogging it/them. Anytime a new compression standard comes out, it always takes a few years for hardware support to arrive - AV1 is no different.

HEVC is still a beast to work with, in terms of editing and hardware playback support is still not ubiquitous.

If the licensing terms had been better, I think we'd see more widespread support & usage, by now.

VP9 is completely open, with open source implementations and is very high quality, yet it hasn't taken over.

And so maybe the AV1 folks were right...

JPEG2000 came out in, wait for it, 2000, and it's still widely used as both an acquisition format, RED cameras capture in JPEG2000 natively and ProRes and DNxHD with third party add ons.

Because its patents (and any looming submarine patents) have expired, and it relies on 19ish-year-old tech. That's not so big an ask, on modern CPUs.

[Spooktra]

in my point of view the people want AV1 and Open-source firmware/microcode/bios

what we get is more x265 and more Closed source shit.

You do realize that x265 is GPL'd software, do you not?

I would disagree that people want AV1, people want something that is compatible across a wide range of devices and uses, which is why MPEG-2 is still used in many places, and why H264 is still so popular.

HEVC is still a beast to work with, in terms of editing and hardware playback support is still not ubiquitous.

VP9 is completely open, with open source implementations and is very high quality, yet it hasn't taken over.

JPEG2000 came out in, wait for it, 2000, and it's still widely used as both an acquisition format, RED cameras capture in JPEG2000 natively and ProRes and DNxHD with third party add ons.

[chithanh]

Is that yet shipping in any product?

No, it was demonstrated at Computex only. Anandtech writes that it may take 1-2 years for consumer devices to reach the market. But the chip itself exists and can be ordered now I think.
https://www.anandtech.com/show/14560...8k-ultrahd-tvs