VIA's Open Linux Graphics Driver Has Been Defenestrated

[Kivada]

Wondoze on arm.... ??? You actually believe that this is their real objective? Or you think that maybe they are simply trying to threaten the x86 chip makers, who have a lot invested in x86 APU stuff now...?

Initially no, it won't be Win7 on ARM, it'll be Server 2008 on ARM based servers, from the looks of it they are aiming for a release to coincide with the release of the Cortex A15 chips http://www.arm.com/products/processo...cortex-a15.php Keep in mind that nVidia has stated that they aren't going to support anything other then Windows on ARM http://arstechnica.com/gadgets/news/...oses-wince.ars and are using the A15 as the basis for the next gen Tegra chips http://www.anandtech.com/show/3905/a...ks-and-servers

So expect to see something like the aforementioned server but with 8 2.5Ghz nVidia Tegra3 quad core chips running Windows Server 2012.

In other words, maybe there was something to those Mayan prophecys...

Why would this threaten X86 chip makers though? The APUs still beat ARM on desktop performance, to match X86 you need allot more cores but most code can't yet make use of it all, meaning that ARM is currently still no threat to X86 on anything other then small tablets, cheap netbooks and low performance servers.

As for the netbooks, see the difference between a crappy ARM netbook like the $100 Augen Genbook7 vs something decent like the $330 Open Pandora

MS is just covering their ass on all fronts to ensure that ARM servers still run MS products.

[Kivada]

I'd question the CPU performance edge Intel has on them in this space, but the rest is dead on. And that gets only into Atom machines. If you've got a bit of ventilation capability, you could come up with a Mini-ITX using a Core i5 with an Nvidia or a Phenom II with an AMD on-board for slightly more money that blows the doors off of even the Atom and Nano boards out there. And...you CAN make a quiet machine doing this- it just takes a bit of ingenuity.

You can do allot more then you'd think, You can get a top end GPU(I've seen this done with the 8800GTX) running fanless with a set of Thermalright coolers like the HR-03GT with the HR-11 backplate cooler, though this will take up 3-4 PCI card slots, CPU wise you can get a decent quad under one of the massive fanless blocks and you can power this behemoth with a mostly fanless PSUs from the Seasonic X- series. Now you can do this, but you'd better have a cool house as you are going to push your luck with something like a half hour of Unigine's OilRush to get your box cooking.

[devius]

Keep in mind that nVidia has stated that they aren't going to support anything other then Windows on ARM http://arstechnica.com/gadgets/news/...oses-wince.ars and are using the A15 as the basis for the next gen Tegra chips... So expect to see something like the aforementioned server but with 8 2.5Ghz nVidia Tegra3 quad core chips running Windows Server 2012.

Geez... nvidia sure likes to not support linux lately (1). A server that only runs windows? That's bound to be a success considering the current trends :P </sarcasm>

(1) except for its geforce and quadro lines, unless it's in a laptop with optimus, because those aren't supported

[deanjo]

Geez... nvidia sure likes to not support linux lately (1). A server that only runs windows? That's bound to be a success considering the current trends :P </sarcasm>

(1) except for its geforce and quadro lines, unless it's in a laptop with optimus, because those aren't supported

That's is an old and outdated article.

Here was the response to that article, the comments were taken out of context.

http://www.theinquirer.net/inquirer/...sition-android

[schlobinux]

How convenient to rewrite the story to your liking and repeat it again and again until you've convinced yourself this is the holy truth....
Do you remember why openchrome forked of unichrome ? Mostly because you are unable to behave in a socially acceptable manner. Calling people name doesn't make up for a healthy developers and users community, whatever your other skills might be. And it certainly didn't help to get a good contact with VIA either. This is the only reason openchrome people had to sometime use private mails and private channels, only because even after the fork, you kept poisoning all conversations. If you hate openchrome so much, please stay off our mailing lists and chans, as we are staying off yours, thank you very much.

Now, let me write some facts :
- VIA is providing an opensource driver for their hardware.
- VIA is trying to get their work upstream'ed (admitedly without much success but also without too much help either).
- VIA is releasing some documentation regularly. More doc and code are available earlier to developers willing to sign an NDA.
- VIA is providing sample hardware to FOSS developers.
- openchrome driver might not be full featured, but it at least basically supports all of the VIA IGPs, ranging from CLE266 to VX900.

I've been talking with Bruce Chang very regularly since a few years and I'm confident VIA better understands the FOSS ecosystem and is slowly drifting to be a better FOSS citizen. We all admit the current situation is far from optimal. Having a fragmented community and fragmented codebase is a waste of time and resources and we would all love to see everyone gathered under the same roof. Unfortunately, this will not, and did not, happen overnight as both VIA and openchrome are seriously understaffed. I too am disappointed that VIA and openchrome did not delivered more, but the situation is way better than it was a few years ago. If you want to rage against a manufacturer that truly doesn't give a damn about Linux, go after NVidia, not VIA. At least, VIA is trying.
So now, for all developers willing to be constructive and not destructive as Luc loves so much to be, please join the openchrome mailing lists. If you have a genuine will to help out, there is plenty of work and plenty of ways to dedicate some time and skills to improve support for VIA IGPs within the FOSS ecosystem.

Regards,
Xavier

[Adarion]

See, Schlobinux, I don't want VIA to come up from now to next second with a complete driver set, but

a) I doubt they are sooo understaffed, it still was/is a large international enterprise (I mean some years ago a felt 50% of mainboard chipsets came from VIA)

b) if they lack manpower or Linux knowledge (I guess the second might me a more valid point) then this would not be a problem if they would communicate and release sufficient (!) specifications to the devs.

You see there were a lot of drivers with a share of developers around, and it seems they all gave up over time because work with VIA was frustrating.
Moreover I also do not expect a team of spare time devs to come up with a driver from now to next. Also in AMD/ATIs case it took some time to establish a driver base but today new chips are added "on the fly".

And: Seeing openchrome source tree, and seeing some people correcting typos in code commenst... this is a sign of desparation to me. Fixing typos in comments is the thing you do when nothing else is left and you're totally bored and or desparate. (At least that is my impression.)

[libv]

How convenient to rewrite the story to your liking and repeat it again and again until you've convinced yourself this is the holy truth....
Do you remember why openchrome forked of unichrome ? Mostly because you are unable to behave in a socially acceptable manner. Calling people name doesn't make up for a healthy developers and users community, whatever your other skills might be. And it certainly didn't help to get a good contact with VIA either. This is the only reason openchrome people had to sometime use private mails and private channels, only because even after the fork, you kept poisoning all conversations. If you hate openchrome so much, please stay off our mailing lists and chans, as we are staying off yours, thank you very much.

Now, let me write some facts :
- VIA is providing an opensource driver for their hardware.
- VIA is trying to get their work upstream'ed (admitedly without much success but also without too much help either).
- VIA is releasing some documentation regularly. More doc and code are available earlier to developers willing to sign an NDA.
- VIA is providing sample hardware to FOSS developers.
- openchrome driver might not be full featured, but it at least basically supports all of the VIA IGPs, ranging from CLE266 to VX900.

I've been talking with Bruce Chang very regularly since a few years and I'm confident VIA better understands the FOSS ecosystem and is slowly drifting to be a better FOSS citizen. We all admit the current situation is far from optimal. Having a fragmented community and fragmented codebase is a waste of time and resources and we would all love to see everyone gathered under the same roof. Unfortunately, this will not, and did not, happen overnight as both VIA and openchrome are seriously understaffed. I too am disappointed that VIA and openchrome did not delivered more, but the situation is way better than it was a few years ago. If you want to rage against a manufacturer that truly doesn't give a damn about Linux, go after NVidia, not VIA. At least, VIA is trying.
So now, for all developers willing to be constructive and not destructive as Luc loves so much to be, please join the openchrome mailing lists. If you have a genuine will to help out, there is plenty of work and plenty of ways to dedicate some time and skills to improve support for VIA IGPs within the FOSS ecosystem.

Regards,
Xavier

Openchrome forked away because modesetting was deemed useless, xvideo and xvmc was way more important for some people, and i didn't accept that. Guess who was right there.

VIA is trying, but they have been listening to the wrong advice. Instead of listening to a bleak story by a highly experienced graphics driver developer, albeit a cynical one (comes with experience), they instead listened to those people who told them that everything is roses and is going to be so great and fantastic.

Fact is, the first gets results, the second gets just noise, and history catches up on both of them.

As for not being constructive, in words, this is true. But I also believe that i more than made up for that in the past, and will continue to do so in future, by proper code and proper solutions, and actually freed hardware.

Now, out of the whole unichrome story, two proper developers emerged, each of them is now employed for such work. That's me and Thomas Hellstrom. It should be very clear now who of those two changed the most for free software graphics drivers in the last 7 years.

[libv]

[QUOTE=schlobinux;164897]

Now, let me write some facts :
- VIA is providing an opensource driver for their hardware.

As i said, ever since 2003. Ever since then, they have been dumping their highly untransparent blob on everyone. With sometimes closed licenses, without dri code for the last 5 or so years, with the tv encoder and decoder and other external encoders support lacking in comparison to their binaries. There's no one who has spent more time with VIAs codedrops than me, and that probably includes the people who have been working on that code for VIA, as there has been quite a lot of variation in style and structure of that code.

- VIA is trying to get their work upstream'ed (admitedly without much success but also without too much help either).

Err, right. Getting GregKH to break SLE and openSuSE for all VIA users, whether they pulled in the openchrome package or not, for more than a year, that was a real achievement.

- VIA is releasing some documentation regularly. More doc and code are available earlier to developers willing to sign an NDA.

If it wasn't for the AMD thing, and OLPC...

I have never accepted this NDA, from day 1, not for code that i write in my spare time, and definitely not for things which should be very open indeed. This explains, to a large extent, why we at SuSE, at the time, thought "hey, let's try to convince AMD of freeing docs. It's now or never.".

I haven't looked at a recent NDA paper of VIAs, do they still demand that all GPLed code coming from VIA cannot be passed on by a third party?

- VIA is providing sample hardware to FOSS developers.

To actual developers, or to people who talk nice?

- openchrome driver might not be full featured, but it at least basically supports all of the VIA IGPs, ranging from CLE266 to VX900.

Yes, the wonders of sprinkling pci-ids all over the code base, and then either doing vbe, modesetting code #1 or modesetting code #2, and then disabling hw cursors most of the time.

I've been talking with Bruce Chang very regularly since a few years and I'm confident VIA better understands the FOSS ecosystem and is slowly drifting to be a better FOSS citizen.

It seems that this has failed, and that you should now own up to that too.

[Svartalf]

Why would this threaten X86 chip makers though? The APUs still beat ARM on desktop performance, to match X86 you need allot more cores but most code can't yet make use of it all, meaning that ARM is currently still no threat to X86 on anything other then small tablets, cheap netbooks and low performance servers.

As for the netbooks, see the difference between a crappy ARM netbook like the $100 Augen Genbook7 vs something decent like the $330 Open Pandora

The GenBook7 used an OLD ARM9 SoC for shaving pennies off of the BoM purposes.

The OpenPandora uses an OMAP3 running at 600 MHz. VAST difference.

Now, as to the rest of your remarks, Kivada, I'd contend that things are a bit narrower to threatening things on the X86 space than you'd think. Witness the latest gems from Motorola. A tablet that can give the iPad a run for it's money and a mobile phone that can be docked into a netbook/laptop body and transformed into a full-on device like a netbook/laptop. It should be noted that these are 1GHz devices and amazingly, they have a similar power profile to the Pandora you mention.

It should be noted that the Pandora is roughly the same class of machine as an X-Box in many ways. It can run 10 full hours doing 2D emulators that stress the CPU and run the backlit display- on a 13.5 watt-hour battery. It'll have about a 5-ish to 6-ish run time with the 3D being involved.

It should be noted that if you scale these devices up to the same clocks as the X86 parts they exhibit a similar performance profile to the X86 devices with similarly low power consumptions when compahttp://www.phoronix.com/forums/images/icons/icon8.gifred to the lower clocked devices. They're clocked low for the power consumption tradeoff.

At least one of the OEMs has taken on an ARM license to make their own bespoke cores for low-power, fairly high-performance 1U/2U server solutions.

There's a reason Microsoft's making the rumblings they are.

[paulsiu]

Woooo, a Linux user is angry. That's gonna hurt VIA a lot.

It's not just linux user. VIA have a reputation for crappy drivers under windows. Having experienced driver crashes under windows, I avoid them like the plague. May be having open source would have improve their quality, but we'll never know now.

Besides, there are really no Nano products out there any way, at least not in the US. Most of the S3 stuff are on older products. The point is rather moot.

Paul