Yes. Now nvidia sucks even more. [LWN.net]

Yes. Now nvidia sucks even more.

Posted Nov 24, 2008 14:41 UTC (Mon) by jch (guest, #51929)
In reply to: Yes. Now nvidia sucks even more. by sylware
Parent article: VIA releases chipset documentation

Intel still hasn't released firmware-level programming documentation. We're still stuck with binary blobs with the Intel drivers.

to post comments

Yes. Now nvidia sucks even more.

Posted Nov 24, 2008 16:15 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (7 responses)

Which graphics card vendors have released firmware-level programming documentation? Heck, which wireless vendors?

wireless

Posted Nov 24, 2008 22:01 UTC (Mon) by xaoc (guest, #54140) [Link] (4 responses)

I using ath9k right now which work quite nicely. I doesn't require binary crap i think.

wireless

Posted Nov 24, 2008 22:12 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (2 responses)

No, because Atheros's (admittedly thin) firmware is in a ROM on the card instead of being loaded at runtime.

wireless

Posted Nov 28, 2008 14:58 UTC (Fri) by jch (guest, #51929) [Link] (1 responses)

There's a big difference between Atheros' and Intel's firmware.

Atheros' firmware is a very thin layer that doesn't change. All of the interesting stuff (association, link quality computation, rate control etc.) happens in the driver.

Intel's firmware is a huge piece of software, larger than most drivers. It implements a lot of stuff that any self-respecting free software developer would like to be able to modify.

wireless

Posted Nov 28, 2008 15:18 UTC (Fri) by mjg59 (subscriber, #23239) [Link]

I don't disagree, but reliance on closed firmware doesn't magically vanish just because the firmware is in the chip rather than in ram.

It does require binary blob

Posted Nov 26, 2008 23:04 UTC (Wed) by khim (subscriber, #9252) [Link]

I always wondered why so many people are insisting on having source for firmware of videocard or wireless chip while happily using binary blobs embedded in HDD, CPU (yes, both Intel and AMD CPUs require binary blobs to operate) and other components. Why 10 or more binary blobs embedded here in there in your system don't bother you, but single one sitting on your CD is such a big deal? And why trivial move of said blob from CD to embedded ROM suddenly make hardware more acceptable?

I can see why binary drivers embedded in kernel are bad idea: there are no "walls" in the kernel and so any driver can bring the whole system down (and DMA gurantees that even microkernel will not solve this problem), but firmware for Intel wireless device works on different CPU - like firmware for HDD, CD-ROM or Ethernet card! The host system just uploads it, nothing more...

Yes. Now nvidia sucks even more.

Posted Nov 28, 2008 14:54 UTC (Fri) by jch (guest, #51929) [Link] (1 responses)

> Which graphics card vendors have released firmware-level programming documentation?

As far as I know, none. Your point being?

There's a lot of neat stuff that could be done if we could hack the firmware. Intel are trying to advertise themselves as being the do-gooder of Free Software, so please let them release firmware development tools.

Yes. Now nvidia sucks even more.

Posted Nov 28, 2008 15:17 UTC (Fri) by mjg59 (subscriber, #23239) [Link]

Well, other than the minor problem that Intel's current graphics hardware doesn't have any downloadable firmware, AMD and VIA are both trying to position themselves as open-source friendly companies. Holding Intel to different standards seems somewhat odd.

Yes. Now nvidia sucks even more.

Posted Nov 24, 2008 18:30 UTC (Mon) by iabervon (subscriber, #722) [Link] (2 responses)

If only they were willing to freeze their firmware in immutable silicon on their products, we'd be satisfied! Or, rather, we wouldn't know there was more to ask for.

(Of course, it would be nice in an operational sense if Intel devices as shipped would work without needing something else to reconnect them with their binary blobs; that way, operating system distributors wouldn't have to distribute the blob that they can't modify. It might happen if demand for SSDs drives down the price of flash enough that it's cost effective to ship reasonable firmware blobs for on-board hardware in on-board flash instead of putting them on user-data block storage devices or driver CDs.)

Yes. Now nvidia sucks even more.

Posted Nov 28, 2008 14:55 UTC (Fri) by jch (guest, #51929) [Link] (1 responses)

> If only they were willing to freeze their firmware in immutable silicon on their products, we'd be satisfied! Or, rather, we wouldn't know there was more to ask for.

Who ever said that?

The original IBM PC's BIOS was frozen in silicon. That didn't prevent smart people from reverse-engineering it and other smart people from providing improved replacements.

Yes. Now nvidia sucks even more.

Posted Nov 28, 2008 16:54 UTC (Fri) by dlang (guest, #313) [Link]

many people (including RMS) have said that they have no problem with sourceless firmware if it's in ROM, they only have a problem with sourceless firmware if it's in flash or loaded by the driver.