Hi,
I contacted some UK custom PC shops and only one got back to me - nsys.co.uk - and I really recommend their customer service! :-)
The machine they have put together totals £1018.27
The motherboard is a "Tyan Thunder K8SE (S2892G3NR)" with a single Opteron 250 2.4GHz Processor. Since this is a server-grade motherboard, it requires an expensive (£200) PSU - a "24Pin 8 Pin Enermax Server Grade".
The "128MB Asus Radeon X300SE Extreme" appears to use the r300 chipset which is being steadily more supported by X.org
I am considering "Seagate 250Gb NCQ" harddisk, and the NCQ features are not yet supported - http://en.wikipedia.org/wiki/Native_Command_Queuing#Operating_system_support - but should be soon.
I wonder about firmware status of CD/DVD Optical disk drives and sound cards.
Does this list have any recommendations and comments on the above? :-)
When I get this machine, I'll write up a report about what I did and why as per http://www.fsf.org/campaigns/supportlinuxbios.html
Thanks, dave.
One thing I'd like to work with you on come build time for the bios -- setting up a 'buildrom target for it. Buildrom is a real stress reliever. So let me know when you get to that point.
ron
On Mon, Nov 20, 2006 at 10:57:46PM +0000, Dave Crossland wrote:
Since this is a server-grade motherboard, it requires an expensive (£200) PSU - a "24Pin 8 Pin Enermax Server Grade".
This may not be true at all. Someone up the foodchain at your dealer may have decided that customers with cash for a server-grade mobo will also have cash for the expensive PSU and hence teach their customer service that it is required - with the reality being that no motherboard needs any special PSU as long as the PSU works with the board - ie. supplies enough power (Ampere) of enough quality (voltage tolerance) on each voltage rail to meet the requirements of the whole system. I think £200 for a single PSU is over the top.
On the other hand it may be true in a worst-case scenario because the motherboard has 8 SATA ports and 2 PATA ports and a built-in SCSI controller that could connect to 15 SCSI drives and 8+4+15=27 hard drives draw a LOT of power.
I suggest that you look into power requirements for the system, possibly with a few upgrade scenarios to make the investment more future proof.
I am considering "Seagate 250Gb NCQ" harddisk, and the NCQ features are not yet supported - http://en.wikipedia.org/wiki/Native_Command_Queuing#Operating_system_support
- but should be soon.
The disk will still work, just that any benefit from NCQ is lost.
I wonder about firmware status of CD/DVD Optical disk drives and sound cards.
Hard drives have firmware too.
I don't expect to see free firmware for any of these devices very soon, but will be glad the day I'm proven wrong. :)
There's been an increase in interest for optical drive firmware with the Xbox 360 though, so I guess it could happen sooner than I think.
//Peter
On Tue, Nov 21, 2006 at 04:39:58AM +0100, Peter Stuge wrote:
I wonder about firmware status of CD/DVD Optical disk drives and sound cards.
Hard drives have firmware too.
I don't expect to see free firmware for any of these devices very soon, but will be glad the day I'm proven wrong. :)
Yeah, I _so_ hope you are or will be wrong :) I don't know of any such firmwares at the moment (I didn't search a lot, though). But that would sure be a great thing to have (and support). Any pointers from someone who might know about efforts in this direction?
Uwe.
I wonder about firmware status of CD/DVD Optical disk drives and sound cards.
Hard drives have firmware too.
I don't expect to see free firmware for any of these devices very soon, but will be glad the day I'm proven wrong. :)
Yeah, I _so_ hope you are or will be wrong :) I don't know of any such firmwares at the moment (I didn't search a lot, though). But that would sure be a great thing to have (and support).
What good would it do to replace the known-good firmware of a hard drive with something home grown? What's the point?
Segher
On Sun, Nov 26, 2006 at 06:47:37PM +0100, Segher Boessenkool wrote:
Yeah, I _so_ hope you are or will be wrong :) I don't know of any such firmwares at the moment (I didn't search a lot, though). But that would sure be a great thing to have (and support).
What good would it do to replace the known-good firmware of a hard drive with something home grown? What's the point?
Well - sometimes it's not known-good. I remember having a high-end server a couple of years back that *required a firmware upgrade* for the hard drives because re-syncing software raid partitions would just crash the machines.
Also, we might find that lower end drives that are 'slower' might only be artificially so, because the firmware is not as good as the higher end drives.
And there is the whole Free as in Freedom aspect of course. I think 'because we want Free software' is a really good reason. Even if having Free firmware for things like hard drives seems somewhat far fetched and difficult to achieve now, think of how the whole concept of a Free OS sounded back in the 80s. Or a Free BIOS 10 years ago.
On top of that, think DRM. Say LinuxBIOS becomes really successful, and we basically kill EFI (and therefore, DRM in the bios). The next thing they'll try is putting it in video cards, audio cards, hard drives etc. In fact, they tried that 5 or 6 years ago - remember the whole CPRM fiasco? http://www.theregister.co.uk/2001/01/10/everything_you_ever_wanted/
Also, just look at those Linksys wireless routers (WRT54G). There's a whole ecosystem out there - people are doing things with them that were *never* anticipated by Linksys. Basically, having Free firmware for things like hard drives could allow some amazing innovation.
And finally, 'because we can' is one heck of a good reason too.
Thanks, Ward.
Yeah, I _so_ hope you are or will be wrong :) I don't know of any such firmwares at the moment (I didn't search a lot, though). But that would sure be a great thing to have (and support).
What good would it do to replace the known-good firmware of a hard drive with something home grown? What's the point?
Well - sometimes it's not known-good. I remember having a high-end server a couple of years back that *required a firmware upgrade* for the hard drives because re-syncing software raid partitions would just crash the machines.
So, did you get a (small) update from the vendor, or did you pull off the semi-impossible task of rewriting the thing from scratch :-)
Also, we might find that lower end drives that are 'slower' might only be artificially so, because the firmware is not as good as the higher end drives.
This doesn't solve a _problem_. Oh, and there are plenty of binary patches you can find around the web that do such things. They typically void your warranty *for a good reason* though.
And there is the whole Free as in Freedom aspect of course. I think 'because we want Free software' is a really good reason.
It would be nice to have, yes, but I think right now we have much bigger problems to solve first.
Even if having Free firmware for things like hard drives seems somewhat far fetched and difficult to achieve now, think of how the whole concept of a Free OS sounded back in the 80s.
Huh, there were plenty of free OSes back then. OSes were simpler of course.
Or a Free BIOS 10 years ago.
Same thing. But point taken (on x86 hardware).
On top of that, think DRM.
Yes, that thing again. With properly implemented DRM, you *cannot* replace the firmware, there is just *no way* to do that. Also, DRM is not a technical problem, and so cannot be fixed with a technical solution. It's a political and economical thing.
Say LinuxBIOS becomes really successful, and we basically kill EFI (and therefore, DRM in the bios).
EFI is not the only thing with DRM in the firmware -- not by a long shot.
Also, just look at those Linksys wireless routers (WRT54G). There's a whole ecosystem out there - people are doing things with them that were *never* anticipated by Linksys.
Yes. And none of those new things have anything to do (directly) with firmware changes.
Basically, having Free firmware for things like hard drives could allow some amazing innovation.
...and will lead to *lots* of bricked drives ;-)
And finally, 'because we can' is one heck of a good reason too.
I'm not convinced you can really; esp. not legally. But yeah, that "geek factor" would make me want to do it, sure -- except I see a HDD as 100% a black box with no internals that I care about (or want to care about). I also don't feel like reprogramming the ucode on CPUs, or even the ucode on a flash chip's internal controller, etc.
Segher
On Sun, Nov 26, 2006 at 11:05:27PM +0100, Segher Boessenkool wrote:
So, did you get a (small) update from the vendor, or did you pull off the semi-impossible task of rewriting the thing from scratch :-)
Sure, we got the vendor patch. But I'm glad we were not the *first* people to run into this problem. It would have taken days or weeks for the vendor to do something, I'm sure.
This doesn't solve a _problem_. Oh, and there are plenty of binary patches you can find around the web that do such things. They typically void your warranty *for a good reason* though.
We don't really care about that, though, right :) Installing LinuxBIOS voids the board vendor's warranty too, I'm sure.
And there is the whole Free as in Freedom aspect of course. I think 'because we want Free software' is a really good reason.
It would be nice to have, yes, but I think right now we have much bigger problems to solve first.
Sure. This whole thread is pretty hypothetical. I'm just answering your question, pointing out a few reasons why free firmware for devices might be a good idea.
Also, just look at those Linksys wireless routers (WRT54G). There's a whole ecosystem out there - people are doing things with them that were *never* anticipated by Linksys.
Yes. And none of those new things have anything to do (directly) with firmware changes.
Sure - my point was that opening up a black-box 'device' can lead to some pretty amazing innovation.
Basically, having Free firmware for things like hard drives could allow some amazing innovation.
...and will lead to *lots* of bricked drives ;-)
Maybe video cards or network cards would be a better example here. But, yes.
I'm not convinced you can really; esp. not legally. But yeah, that "geek factor" would make me want to do it, sure -- except I see a HDD as 100% a black box with no internals that I care about (or want to care about). I also don't feel like reprogramming the ucode on CPUs, or even the ucode on a flash chip's internal controller, etc.
But your network interface? Video card?
Thanks, Ward.
Ward Vandewege wrote:
On Sun, Nov 26, 2006 at 11:05:27PM +0100, Segher Boessenkool wrote:
So, did you get a (small) update from the vendor, or did you pull off the semi-impossible task of rewriting the thing from scratch :-)
Sure, we got the vendor patch. But I'm glad we were not the *first* people to run into this problem. It would have taken days or weeks for the vendor to do something, I'm sure.
True, vendors might take a little time to make a patch, but in the long run, who would have done a better job of it? The guys who've read the docs, and are rushing to make something work, or the guys who built the drives themselves, and make full-time jobs of doing it? Besides, if you'd just set up these hard drives, you must've been working without them in the past, so I can't imagine that another couple weeks would have resulted in a catastrophe.
This doesn't solve a _problem_. Oh, and there are plenty of binary patches you can find around the web that do such things. They typically void your warranty *for a good reason* though.
We don't really care about that, though, right :) Installing LinuxBIOS voids the board vendor's warranty too, I'm sure.
LinuxBIOS is based off usually well-documented chipsets, and the BIOS itself is designed to be built by the vendor. Hard disks and ATAPI drives are usually assembled and programmed by the same people, so providing anyone outside the company with documentation on how they work isn't necessary. Video cards and NICs might be a different story.
And there is the whole Free as in Freedom aspect of course. I think 'because we want Free software' is a really good reason.
It would be nice to have, yes, but I think right now we have much bigger problems to solve first.
Sure. This whole thread is pretty hypothetical. I'm just answering your question, pointing out a few reasons why free firmware for devices might be a good idea.
No offense or anything, but I don't think you really have. Aside from the reference to the server, you haven't given much for reasons.
Also, just look at those Linksys wireless routers (WRT54G). There's a whole ecosystem out there - people are doing things with them that were *never* anticipated by Linksys.
Yes. And none of those new things have anything to do (directly) with firmware changes.
Sure - my point was that opening up a black-box 'device' can lead to some pretty amazing innovation.
A linksys wrt54g is basically an embedded computer, with a TI AR7 (MIPS) cpu and a very specific purpose in life, it was inevitable that someone would figure that out and create more software to run on it. A PC's internal hardware doesn't have a dedicated CPU, and it's not designed to run software...it just, works.
Basically, having Free firmware for things like hard drives could allow some amazing innovation.
...and will lead to *lots* of bricked drives ;-)
Maybe video cards or network cards would be a better example here. But, yes.
I remember when I got my first new video card, a Radeon 9500 (to replace my Voodoo 5). There was a whole fiasco about how you could simply flash this card with a 9700's bios and open up the 4 pipelines that were locked by the manufacturer for 8 total, like the 9700. OTOH, in many cases, those 4 pipelines were locked for a good reason: at least one of the pipelines was no good. Now, in some cases (roughly 50% when I got mine) people lucked out and got a 9700 Pro for a 9500's price. Had the BIOS been open source though, it would have made the lock mechanism much easier to spot, and so ATI would have found another way to lock the pipelines, like the design change in later models. So, no free beer, for anyone.
I'm not convinced you can really; esp. not legally. But yeah, that "geek factor" would make me want to do it, sure -- except I see a HDD as 100% a black box with no internals that I care about (or want to care about). I also don't feel like reprogramming the ucode on CPUs, or even the ucode on a flash chip's internal controller, etc.
But your network interface? Video card?
Thanks, Ward.
I agree with Segher. LinuxBIOS offers much faster boot times and greater flexability than a propriety BIOS. I can't imagine getting nearly the same gain from a hard drive, cdrom, or even NIC. Yes, I might be able to make my cd burner run at 32x instead of 24x...whoop-de-doo. Heck, maybe I could even make my 10/100NIC work at 1000MBit, I'm sure that would make my 3MBit DSL so much faster. And all these menial little improvements for creating firmware for something that is designed and programmed internally, so public docs are all but nonexistant.
I agree with Segher.
Interesting, since I don't think you do. Maybe I didn't explain myself well enough. But let's drop this thread now.
Segher
Segher Boessenkool wrote:
I agree with Segher.
Interesting, since I don't think you do. Maybe I didn't explain myself well enough. But let's drop this thread now.
Segher
I meant in that hard drives are 100% black boxes that I don't really care how work, just that they work. And good idea.
-Corey
This doesn't solve a _problem_. Oh, and there are plenty of binary patches you can find around the web that do such things. They typically void your warranty *for a good reason* though.
We don't really care about that, though, right :) Installing LinuxBIOS voids the board vendor's warranty too, I'm sure.
Of course. Note the "*for a good reason*" though.
It would be nice to have, yes, but I think right now we have much bigger problems to solve first.
Sure. This whole thread is pretty hypothetical.
Ah good :-)
Basically, having Free firmware for things like hard drives could allow some amazing innovation.
...and will lead to *lots* of bricked drives ;-)
Maybe video cards or network cards would be a better example here. But, yes.
Most video cards already *are* programmed by (free) software. Maybe not very well, but that's due to lack of specs.
Many (high-end and wireless) network cards can get the same treatment, we just need more datasheets ;-)
I'm not convinced you can really; esp. not legally. But yeah, that "geek factor" would make me want to do it, sure -- except I see a HDD as 100% a black box with no internals that I care about (or want to care about). I also don't feel like reprogramming the ucode on CPUs, or even the ucode on a flash chip's internal controller, etc.
But your network interface? Video card?
Yeah. Those devices directly access main memory so there's fun stuff to do with their embedded CPUs :-)
Segher
On 26/11/06, Ward Vandewege ward@gnu.org wrote:
On top of that, think DRM. Say LinuxBIOS becomes really successful, and we basically kill EFI (and therefore, DRM in the bios). The next thing they'll try is putting it in video cards, audio cards, hard drives etc. In fact, they tried that 5 or 6 years ago - remember the whole CPRM fiasco? http://www.theregister.co.uk/2001/01/10/everything_you_ever_wanted/
Isn't this what EFI already does though, by shifting drivers into firmware thats located inside hardware?
I'd love to read more about how LinuxBIOS compared to EFI in this way - is there any recommended reading? :-)
-
Corey Osgood -
Dave Crossland -
Peter Stuge -
ron minnich -
Segher Boessenkool -
Uwe Hermann -
Ward Vandewege