Intel's Poulsbo Driver A Bloody Mess?
Phoronix: Intel's Poulsbo Driver A Bloody Mess?
Most of the Intel Atom netbooks currently on the market (like the Lenovo IdeaPad S10, Samsung NC10, and ASUS Eee PC) use the GMA 950 Chipset for their integrated graphics, but some of the newer models are using the Intel GMA 500. The GMA 500 doesn't share many traits with other mobile Intel IGPs since much of the technology was licensed from PowerVR, which means a different X.Org display driver is required. Rather than using the xf86-video-intel driver that receives active development from Intel and is often leading the other open-source X.Org drivers when it comes to features like the Graphics Execution Manager, kernel mode-setting, and DRI2, a new driver had to be created for the GMA 500...
Finally someone speaks up about this mess!
First of all, I think very highly of intel's open source efforts, but the way this GMA 500 business was dealt with sucks.
If "we" have to restart the whole "no blob, please publish documentation" stuff every time some new hardware comes around, things will never get anywhere... Intel could have handled this better, instead there already are devices out there, and as usual, microsoft has the leg up.
When I first heard about the Dell Mini 12 I thought it was pretty cool, especially the relation between the screen size and the weight; but then I heard about the "GMA 500"-not-really-intel, and all poulsbo devices were off my possible purchase list instantly.
At least I think it's interesting to see the first driver implementing VA-API that I know of.
Hopefully the intel guys come together and fix the damn thing and also make it completely open source. Did not know about this, yet.
Note that the PowerVR SGX stuff is used in much more then just the Dell Mini-12.
It's used in other architectures. For example the OMAP3 being used in the OpenPandora Handheld and the developer oriented BeagleBoard has the PowerVR SGX chipset.
It's what is going to be showing up in handhelds all over the place. Any powerful ARM 'smartphone' or any stripped down Intel MID or PDA or smartphone device is going to be utilizing the PowerVR technology.
So it's not just a question of the Atom platform.. this bad boy is going to be showing up all over the place in all sorts of different devices.
So it's especially bad because the proprietary x86 driver won't work in x86-64, nor will it work in the ARM platform. People shipping PowerVR-based graphics in ARM will need their own, seperate, proprietary Linux driver.
It's a clusterf*k alright. With a open source driver you'd be able to update your code as well as created a unified driver that will be easily made to work across lots of different devices irregardless of the actual architecture.
Phoronix, Thanks SO much for shining a light on this mess!
Intel, are you listening? Anyone?
I spent $2800 Cdn on a Poulsbo based Panasonic CF-U1 MID ( my story here: http://forums.fedoraforum.org/showthread.php?t=201018 ) with no idea what I was getting myself into.
Intel has been doing such a great job maintaining the desktop stuff (i810 driver, etc), that I saw their name and just jumped in trustingly. I keep hoping someone on that team will come to my rescue and help clean this stuff up? I don't know who else would possibly have the time or knowledge to keep this stuff updated.
I think the problem here is more Imagination Technologies than Intel, and Intel is far from being their only customer. Without convincing them, the options for proper support are going to be limited. The PowerVR variant in Dreamcast has been pretty well reverse-engineered, but I don't know how similar it is to the current stuff.
I still have unfond memories of the PowerVR Kyro. After lots of requests from the community, a proprietary Linux driver was released by Imagination Technologies (the company behind PowerVR). But after some time, they stopped updating it for newer kernels and so the driver and the cards became next to useless for Linux.
It would not surprise me if the reason for the blobs lies with the IP policy at ImgTec. I hope that Intel can sort this out soon, after all they have a good reputation with the open source community to lose.
EDIT: Heh, Ex-Cyber beat me to it. Anyway, here is the freedesktop.org bug stating that they have no interest in open source PowerVR drivers http://bugs.freedesktop.org/show_bug.cgi?id=1652
Last edited by chithanh; 01-31-2009 at 02:52 PM.
I don't think it was a good idea to use a 3rd party chip under the Intel brand. Intel devs don't know it and the current Linux code does not look ready for prime time...
Well this isn't the first time that Intel has used PowerVR stuff. I don't know for sure, but I beleive that the old Pre-GMA Intel IGP (extreme blaster 3d stuff found on 8xx series chipset) is derived from PowerVR's cores. I think.
But anyways... it was probably a time-to-market and cost effectiveness issues why Intel decided to move away from their own IGP designs for the Atom MID stuff.
It would be nice to get the thing more open, of course. Your going to see PowerVR stuff cropping up more and more now. Especially in Android-style devices and other netbooks.
This is because most netbooks are using the GMA 950 stuff which is designed for desktops and 'value' laptops. Bettery life and features are just not there... the chipsets in those Atom netbooks are going to be using much more power then the actual CPU does and is probably the only reason why most those devices ship with fans and have reduced battery life compared to other Atom-based devices.
It's to bad that AMD sold off their mobile graphics to Quallcom.
The only way to convince a company like 'Imagination Technologies' to open up is through competition. That's it, really. So unless there is a compelling open source graphics platform for embedded 3D graphics then convincing them to open up is going to be very difficult.
with Embedded Linux this sort of thing is very difficult.
Typically everything they do is one-off. You sell a product and it ships with whatever firmware you ship. Barring that then there is no reason to upgrade that firmware. If customers want more features or fixes then they simply can purchase a newer version.
So Linux OS is now the 'firmware'. Embedded developers are still going to be in the same mindset and they are not going to understand what is the point to openning up drivers so that people can install a 2.6.24 or newer kernel.
Until those guys understand the importance of being open and sending changes back to upstream then there is going to be very little force acting on "Imagination Technologies" to convince them to open up.
SGX reverse engineering
I believe there are ongoing efforts to reverse engineer the PowerVR SGX 530 (Intel's GMA500 integrates the SGX 535 but they should be similar) core, according to this OpenPandora development forum thread. Unfortunately, there haven't been any status updates in over a month...
One thing is for sure, my acquisition of an MSI Wind U115 netbook depends on the availability of a free and open source GMA500 Xorg driver.