Open-Source Raspberry Pi Graphics Drivers Add Double Buffer Mode

[coder]

Now hopefully this new SoC if it comes out this year will have A76 cores and at least 4x the GPU cores on a 12nm process, but who knows what the restrictions are for the design.

No, they're still building it to a price. That means 28 nm for at least another generation. And that means not A76 cores, nor anything like a 4x GPU increase.

[Quackdoc]

Quackdoc ffmpeg, vlc, mpv, firefox, chrome, gstreamer, Kodi, Kdenlive, OBS, etc. do all support VAAPI, so yes, on Linux it is the de-facto standard as in "that's what the most relevant programs/libraries support". Kodi has dropped all Pi specific code because the project doesn't want the additional maintenance burden.
OMX support is poor and often requires out-of-tree patches, the same goes for decoding via V4L2/mmal. Since the patches are often tied to specific versions of these libraries, the situation effectively limits the number of distros, which are able to provide some form of video decoding acceleration at all, and even then it works rather poorly because it's a hackjob.

again, VAAPI only works on linux, BSD, and somewhat on android, although in a fairly fragile state. openmax is similar, but is fragile on linux, and more supported on android.

vaapi is no defacto standard, on linux it may be, but that is a narrow view.

kodi doesn't need pi specific code anymore.

ffmpeg, vlc, mpv, gstreamer support openmax too. I can't say I've had issues using it myself

if thats not good enough go pester broadcom, or hope RPI gets vulkan video support

[hotnuma]

Raspberry Pi can do a lot of things but will do everything badly.

Any Raspberry Pi user can go to YouTube and search "screen tearing test", then you'll see with you own eyes the result.

Same for electronic projects, an Arduino will do a better job, at least it will be open source hardware, it will consume less power and programs will be C++.
On a Raspberry Pi you will need sudo to execute your own programs which is the most ridiculous thing I saw in computing. :-D

Not to mention the proprietary firmware, the non disclosure agreements with Broadcom, which is far from an ethical practice in my humble opinion.

Now some reading.

https://wiki.debian.org/RaspberryPi#Should_I_buy_a_Raspberry_Pi.3F

https://www.fsf.org/resources/hw/single-board-computers

Enjoy. :-)

[s_j_newbury]

No, they're still building it to a price. That means 28 nm for at least another generation. And that means not A76 cores, nor anything like a 4x GPU increase.

28nm will always be the process targeted for design to cost. It's the sweet spot, and it's been obvious for the last decade. Larger processes give you less chips per wafer, but newer processes are subject to diminishing returns and rising costs. This is why capacity for newer processes is taken up with high margin products [it's part of the reason for the chip shortage], for low price, low margin products it's just not economic, and it's questionable for anything else. This is why foundries are now expanding 28nm manufacturing again after a lot of capacity was lost chasing Moore's Law on the semiconductor industry treadmill.

Expect higher end chips to remain expensive, but at least the price for commodity chips should come back down.

https://min.news/en/economy/870f2fff0332b177b7584030218422e8.html?__cf_chl_f_tk=gl1hsYYhq3uHTwozOT8PCBNK_1xcKb3SpsIFA2b7M_c-1642255390-0-gaNycGzNCH0

[coder]

Greetings, new user!

Raspberry Pi can do a lot of things but will do everything badly.

As much as I complain, it's still not bad value for money. There are tons of things where a Pi is "good enough". It just happens to be that "desktop replacement" isn't quite one of them.

Same for electronic projects, an Arduino will do a better job,

Depends on how much compute, memory, and supporting capabilities you need. The Pi has the advantage of being a fully general-purpose computer, with a fully general-purpose OS. For some projects, that's hugely valuable.

programs will be C++.

Um, the Pi can do C++.

On a Raspberry Pi you will need sudo to execute your own programs which is the most ridiculous thing I saw in computing. :-D

That's only to access special resources, like devices or system config files. If you know what you're doing, you can nearly always get around that. I'd advise you to take a look at "group" permissions and ownership.

On a multi-user system, root ownership of system-wide resources keeps users from accidentally or maliciously stepping on each other. On a single-user system, it serves to reduce the chances of shooting yourself in the foot. It also helps with malware, as Microsoft eventually figured out (maybe you've seen those pop-up dialogs on Windows that ask you to authenticate when you want to install a program or do other privileged operations?).

Anyway, if you just write a "hello world" program (or most other normal apps), you don't need sudo to run it.

Not to mention the proprietary firmware, the non disclosure agreements with Broadcom, which is far from an ethical practice in my humble opinion.

As bad as the Pi's GPU is, I think there's no better one they could use that doesn't have binary firmware. The Pi foundation started out as a small, little hardware project and were stuck using whatever SoC they could get. That's how they ended up signing on with Broadcom. Pi is now so big that I've heard Broadcom is maintaining the VideoCore GPU purely for that project. I think they're not even using it for anything else.

For backward-compatibility reasons, as well as other relationships between certain Pi folks and Broadcom, I think there's basically zero chance of them ever switching to another SoC vendor (unless Broadcom decides to terminate the arrangement).

As I've said before, I'm partial to the ODROID N2+, which uses an Allwinner SoC with a Mali GPU. I'm using the open source GPU driver, but certain things still require installing the binary driver. Do note that the N2+ is only available with up to 4 GB of memory. This is a limitation of the SoC.

[coder]

newer processes are subject to diminishing returns and rising costs.

Yields on some smaller process nodes are reportedly very good. Price-wise, I think it's an issue of supply/demand, rather than yield.

This is why foundries are now expanding 28nm manufacturing again

That's a logjam effect. Older nodes are going offline, but equipment for newer nodes is already back-ordered. And there's pent-up demand from people trying to find capacity wherever they can. So, what happens is you get a bulge in the middle. In such a high-demand market as the one we're in, supply will come online wherever it can. There are plenty of older chips that can be ported to 28 nm, if that's where the availability lies.

[hotnuma]

coder, you demonstrate perfectly another problem with the Raspberry PI.

The fanboyism. :-)

It always gives biased opinions, disconnected from reality. :-)

Anyway thanks to the ridiculous sudo, I ordered an Arduino and for video playback I'm simply waiting for prices to be reasonable to buy a Mini PC.

It seems more expensive at first glance, but comparing features, as RTC, SATA interfaces, included SSD, software compatibility and of course performance, the higher price of a Mini PC is way cheaper than any ARM gadget.

Raspberry Pi is a modern Commodore 64, it's very popular today, it sells million units but will soon be replaced by Mini PCs. :-)

[s_j_newbury]

Yields on some smaller process nodes are reportedly very good. Price-wise, I think it's an issue of supply/demand, rather than yield.

It's not just yield, power density scaling puts design constraints on newer nodes, it becomes necessary to have increasing areas of the chip power-gated or running at lower frequencies to handle thermal dissipation and current draw. Design costs have been increasing exponentially.

https://www.extremetech.com/computing/272096-3nm-process-node

I suspect this is the main reason why modern CPUs use "Turbo" frequency-boosting, not for more performance, but to manage the power envelope. It's all just marketing. This means more complex design, and extra overhead, so the advantage of the die-shrink doesn't scale in the same way as it did before 28nm.

Then there's the up front cost of the foundry, that has to be amortized, you're not going to do that with low margin production in a foundry that cost $20B!

That's a logjam effect. Older nodes are going offline, but equipment for newer nodes is already back-ordered. And there's pent-up demand from people trying to find capacity wherever they can. So, what happens is you get a bulge in the middle. In such a high-demand market as the one we're in, supply will come online wherever it can. There are plenty of older chips that can be ported to 28 nm, if that's where the availability lies.

[qarium]

It's not just yield, power density scaling puts design constraints on newer nodes, it becomes necessary to have increasing areas of the chip power-gated or running at lower frequencies to handle thermal dissipation and current draw. Design costs have been increasing exponentially.

https://www.extremetech.com/computing/272096-3nm-process-node

I suspect this is the main reason why modern CPUs use "Turbo" frequency-boosting, not for more performance, but to manage the power envelope. It's all just marketing. This means more complex design, and extra overhead, so the advantage of the die-shrink doesn't scale in the same way as it did before 28nm.
Then there's the up front cost of the foundry, that has to be amortized, you're not going to do that with low margin production in a foundry that cost $20B!

according to libre-soc the price if you go from 28nm to 22nm is doubled...

https://libre-soc.org/22nm_PowerPI/

"

• USD 2,000,000 for 22nm Production Masks (1,000,000 for 28nm)
• USD 200,000 per 22nm MPW Shuttle Service (test ASICs. 28nm is 100,000)

"

[direc85]

Of course it will be less performant even comparing to some older tech, the chips it's designed to use a fraction of the power. and are made to be extremely cheap to produce. with very low amounts of ram, 4-8gb shared between cpu and gpu which in turn cripples memory bandwidth.

age is NOT a good descriptor of performance, Raspberry pi 4 smokes my dual core celeron from 2015 which handily looses against many older cpus budget cpus from 2007 and before.

the issue with video decode is simple, it only supports 4k60 videos formatted in HEVC. which means you are forced to use local content, and not streaming. as long as you keep it within supported hwdec capabilities it works fine

Well said. RPi4 isn't a bad device, if it's in a fitting use. It's not a workhorse to play *any* FullHD (or better) video; just the format(s) it's designed to support. My RPi3 is a PiHole server amd DHCP, perfect for me! Light emulator use? Maybe. Desktop replacement? Not really. It's a $50 matchbox size 'puter, how could it possibly compete with an actual PC?