Work on VideoCore V GPU Drivers Could Pave the Way for Raspberry Pi 4 Board

Orange Pi Development Boards

I’ve come across an article on Phoronix this morning, about VideoCore IV GPU used in Broadcom BCM283x “Raspberry Pi” processors, but part of the post also mentioned work related to VC5 drivers for the next generation VideoCore V GPU, written by Eric Anholt, working for Broadcom, and in charge of the open source code related to VideoCore IV GPU for Raspberry Pi. This led me Eric’s blog “This Week in VC4/VC5” and articles such as “2017-07-10: vc5, raspbian performance“, where he explains he committed Mesa drivers for VC5.

I’ve just pushed a “vc5” branch to my Mesa tree (https://github.com/anholt/mesa/commits/vc5). This is the culmination of a couple of months of work on building a new driver for Broadcom’s V3D 3.3. V3D 3.3 is a GLES3.1 part, though I’m nowhere near conformance yet. This driver is for BCM7268, a set-top-box SOC that boots an upstream Linux kernel. I’m really excited to be working on a modern GLES implementation, and one that has its core platform support upstream already.

Raspberry Pi 3 is a nice little board, but competition is building with features not found in the RPi foundation board such as Gigabit Ethernet, USB 3.0, 4K video playback and so on. So even though, Eben Upton has clearly said he was in no rush in releasing Raspberry Pi 4, this will happen at some time, and in order to leverage part of existing software it would make sense to use an upgrade to VideoCore IV GPU like VideoCore V.

So I’ve tried to find more information about BCM7268 SoC, and information is quite limited. We know it’s designed for 4K Ultra HD set-top boxes, and features four cores delivering up to 14,000 DMIPS. Some speculated that the processor could be used in Raspberry Pi 4, but Jamesh – Raspberry Pi Engineer & Forum Moderator – poured cold water on the idea:

That’s a set top box chip, doesn’t cover the requirements for the next gen Pi.

So I tried to specifically find more details about the GPU, but again information is limited. The GPU supports OpenGL ES 3.1, and VideoCore V (V3D-530) is about 2.4 times faster than VideoCore IV (V3D-435) in T-Rex GFXBench benchmark on dual core development boards. We don’t know the GPU frequency however, so that’s just for reference.

Anyway, there’s still plenty of time to find out since the “official” timeline for Raspberry Pi 4 is sometimes in 2019.

Support CNX Software - Donate via PayPal or become a Patron on Patreon

30
Leave a Reply

avatar
25 Comment threads
5 Thread replies
9 Followers
 
Most reacted comment
Hottest comment thread
14 Comment authors
Jean-Luc Aufranc (CNXSoft)513DavidJimbobcrashoverride Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
geokon
Guest
geokon

I think this is wishful thinking. A more powerful computer doesn’t seem at all in line with the goals of RPi foundation or anything that Eben Upton has hinted at. Their goals are not in line with the goals of most of the people buying their stuff! 🙂

Things like “Gigabit Ethernet, USB 3.0, 4K video” are not relevant to their goal of introducing computing to more people (esp. children). Eben Upton has been clear that the current focus is on the software side and lowering the barrier for people to get start learning on a RPi. In all his interviews he always stresses that the Rpi3 is the “delux” version and that their baby are the lower end devices like the RPi zero. The ones that are accessible to everyone.

The way I see it, what could bring about the RPi4 would be only features that are ease-of-use. Like easier to run from a battery, easier to to connect to a screen/TV, etc.

I wish they pushed to lower the barrier to OS development. A person learning on the RPi is really disconnected from how the system actually works. If the RPi ran NetBSD or something simpler and if this video driver mumbo jumbo got cleared up it’d be a real game changer. A fully open, simple, accessible and modifiable system stack would make the RPi a unique device again. However they’re so closely tied to Broadcom that I’m afraid it might never happen.

Bruce
Guest
Bruce

@geokon
can you elaborate why NetBSD would make the people “more” connected on how the system actually works ?

Mihai
Guest

RPi was never friendly to the user, but is relatively easy to use. I do not think it will ever use NetBSD, because is is definitely not user friendly. There are a lot of OS alternatives, if it runs on some type of hardware, it also runs on a RPi. I would love to see a RPi with Gigabit ethernet and maybe an USB 3.0, but that is not going to happen. Just as other users stated, they want RPi to be accessible to everyone. RPi never wanted to be a TV box (the community made it possible). Cheap and 4K do not mix well.

I think they they will not even produce a RPi4 any time soon. I bought an oDroid XU4 for USB 3, Gigabit Ethernet and the extra power. I know there are other alternatives to oDroid, but I am very satisfied with it now (running OMV 3.x based on Armbian). I would love to get a RPi3, but I do not have any usage for it now.

A little offtopic, but RPi Zero W is so hard to find that a Romanian seller started importing them from a US seller with a price so high I am better off with a RPi2 or 3. They would better make it easier to buy than ever. I would love a RPi Zero W because of its form factor. I could use it in a lot of ways.

tkaiser
Guest
tkaiser

geokon :
their baby are the lower end devices like the RPi zero. The ones that are accessible to everyone.

Huh? Have you ever tried to buy an RPi Zero? This device is most probably a loss leader for this mythical RPi foundation so they limit everyone having easy access to it by not selling this device in volume.

Wrt ‘feature wish lists’ and RPi models. It’s still the same old boring and incapable VideoCore IV design dating back from 2011 where they exchanged ARM CPU cores twice and allowed the SoC to access more than 512MB DRAM with some tricks still sharing the same ‘only one IO connection to the outside’ (one single USB2 OTG port). Not being innovative on the hardware front at all allowed software support to become mature over time and to establish a big homogen community around these boards (the crippled software support for the Raspberries with more advanced CPU cores prevented fragmentation successfully so why should they introduce an innovative but incompatible new RPi that would destroy their only asset: all Raspberries being able to run the same software and not differing that much from each other).

Demetris
Guest
Demetris

Bcm7268 is the gpu core used on bcm7251S/52S enigma 2 devices like mutant hd51. Cpu can reach 1.7ghz without overclocking(default freq 1.5ghz) and is beyond compare to current
Dev boards plain simple it just flies

tkaiser
Guest
tkaiser

@cnxsoft
I would believe @Demetris meant that both BCM7251 and BCM7252 are based on the same BCM7268 VideoCore 5 VPU/GPU just like BCM2835, BCM2836 and BCM2837 are all based on the BCM2763 (VideoCore 4) just with exchanged CPU cores running the secondary OS (Linux in this case). BCM2763 products without any ARM core are for example Nokia 600, 700, 701, 603 and 808 (there only the primary OS called ‘firmware’ by RPi Foundation is running).

geokon
Guest
geokon

Sorry, I didn’t mean to name drop NetBSD – but some kind of simpler OS that is more easily modifiable and understandable. There are plenty of more “learner friendly” OSs – you just need to strike the right balance and have something that will also run enough software for people to use it as an educational platform.

To me the people behind Linux flavors for these minicomputer (like tkaiser and others at ARMBIAN) are engaged in some black magic I’ll never fully understand – most of which seems to not involve actual code and more about modifying config files and toggling compilation flags – which isn’t the most pedagogic activity 😛

Maybe this is a naiive outlook

Member

RaspPi ecosystem is just messed up. They make these nice, easy to use boards. Then they get everyone trained on them. But then when you actually want to produce something with it to sell you run straight into a brick wall. Either cough up $1M for chips from Broadcom or buy $50 modules. This totally ruins anyone from using it as the basis for a small to medium volume product. So if you have any plans of producing something for sale, skip the RaspPi and move on to something else.

theguyuk
Guest
theguyuk

@Jon Smirl
I always wonder does having a device based on a widely used education board make it more open to hacking?

Peter
Guest
Peter

Wish they could get basic BT sound working properly on RPI’s

Tommi
Guest
Tommi

“That’s a set top box chip, doesn’t cover the requirements for the next gen Pi.”

Doesn’t cover the “requirements” for the next gen Pi, hmm they want improve it even better? I just want SoC upgrade (higher clocks on CPU + GPU VC5), 2Gb RAM (LPDDR3!), usb 3.0 ports (maybe 2 of 4?) and improved sd card reader for better write/read + support for bigger cards than currently supported.

crashoverride
Guest
crashoverride

@Tommi
You have to understand what is meant by “requirements”. The VC5, unlike the VC4, is a 3D accelerator only. It does exactly what Mali does. The VC4, by contrast, is a SoC. It does what S905, RK3328, etc does.

The “requirement” is that a future “Pi4” chip also be a SoC with ARM attached co-processors. This is what prevents RPi “clones”. Nobody else can procure the chip and, even if they did, the firmware to use it is licensed preventing 3rd party products (VAR, OEM, makers) from using it. Its a business constraint, not a technical one.

tkaiser
Guest
tkaiser

@crashoverride
VC5 is part of Broadcom’s ARM SoCs like BCM7251 and BCM7252. These SoCs use VC5 as GPU/VPU and have own Brahma15 ARM CPU cores (the begin of Broadcom’s Broadband group moving away from MIPS to ARM).

So both BCM7251 and BCM7252 are totally incompatible to VC4 based mobile Broadcom SoCs now used on Raspberries and can’t be ‘sold’ as an upgrade since the clueless crowd called ‘Raspberry Pi community’ has been told that much fairy tales already by RPi Foundation that it will never get a dual-core A15 ARMv7 SoC might be a performance upgrade for the slow quad-core A53 (ARMv8 CPU condemned to run ARMv6 code).

I would also assume that BCM7251 and BCM7252 are rather expensive compared to the outdated SoC now used on Raspberries.

Jimbob
Guest
Jimbob

About the only thing that makes sense in this post is the part about pricing. Everything else is just far enough from the actual truth to be rather misleading

Demetris
Guest
Demetris

@tkaiser correct, thats why raspi 4 will come in 2019 so price will be at
Minimum for a 4y old chip then days obsolete

tkaiser
Guest
tkaiser

@Demetris
Hmm… while I don’t care about features or release dates of an RPi 4 people who do for whatever reasons should IMO better focus on the commercial interest represented by ‘Raspberry Pi Trading Ltd’ than the fairy tales told by ‘RPi foundation’ (or the lack of stories/roadmaps told and playing lousily Apple’s mystery mongering game wrt future products/plans).

IMO the most important part of RPi Trading’s roadmap can be extracted from the comments here: http://www.cnx-software.com/2016/10/14/raspberry-pi-compute-module-3-to-be-launched-by-the-end-of-the-year-used-in-nec-displays/#comments –> future Compute Modules being eletrically compatible so still a single USB connection to the outside. So once the SoC they use on future Raspberries is USB3 capable RPi customers get something similar to an ODROID-XU4 today with all high bandwidth interfaces behind USB3 port(s).

And I would be really surprised if next SoC on any future Raspberry will be incompatible to the current ones since software compatibility is IMO one of the few assets in Raspberry country anyway.

crashoverride
Guest
crashoverride

> focus on the commercial interest represented by ‘Raspberry Pi Trading Ltd’ than the fairy tales told by ‘RPi foundation’
My *personal* belief is that the Pi Zero is a direct result of “Trading’s” interest. I am sure their contract for BCM23xx involves a minimum yearly purchase quantity. With RPi2 unable to meet the quota, they made a very prudent business decision to offer the Zero. This “loss leader” ensures availability of BCM23xx for their “compute module” customers. Simultaneously, the choice of “non standard” connectors and purchase limits provides a value to distributors that “up sale” connectors/kits or just hike up the shipping fee.

I hope the Pi is around for a long time. All the $25/$35/$45 boards are a direct result of market forces responding to it. While I have no use for a Pi, I do enjoy the variety of now affordable competitors. Unfortunately, as explained above, I feel the Zero heralds the beginning of the end rather than a “new life” for Pi.

Feel free to post flames below. 😉

Jimbob
Guest
Jimbob

Since I work for RPF trading, I’d just like to say that comments on internet sites are generally not particularly accurate, especially when it comes to predicting Pi roadmaps and future directions.

theguyuk
Guest
theguyuk

RPI merged https://techcrunch.com/2017/05/26/raspberry-pi-foundation-and-coderdojo-to-code-club-together/

They have added a WiFi , Bluetooth to zero and it already has camera. So where next, a single core A5, A7, A35 RPI Z ?

crashoverride
Guest
crashoverride

I don’t think its that difficult to predict. “PI next” will launch next year (2018).

The recent increase in Pi Zero production means a decrease in PI3 production since they share the same production line. Based on recent discount prices from various sellers, the PI3 is under-performing. When we start seeing Pi Zero supply dry up again, it means “Pi next” has hit the production line in anticipation of launch. All indications are it will be an incremental update: more Pi3+ than Pi4; a product “refresh”.

(My joke when Pi3 launched as 32bit with “N75” wifi was they would release “the other half” next year – 64bit, N150 wifi)

David
Guest
David

@Jimbob
Tell us more!

513
Guest
513

FYI, this STB uses a BCM7268 : https://sites.google.com/view/droid-tv/pay-tv-provider/cj-hellovision-uhd-red

It’s manufactured by Samsung for a Korean ISP : CJ HelloVision
This is the Geekbench, you can see it’s a 4 core SoC (the BCM7252S in only dual-core for example)
https://browser.geekbench.com/v4/cpu/2676643