AMD Publishes New Chipset Documents

[phoronix]

Phoronix: AMD Publishes New Chipset Documents

Last month the engineers at AMD managed to put out public, NDA-free documentation that covered the SB700/710/750 Chipsets. This south-bridge documentation is not nearly as exciting as seeing a new ATI graphics processors be documented in the public, but it does greatly help out the CoreBoot developers in enabling support for their BIOS project to run on systems with such hardware...

http://www.phoronix.com/vr.php?view=NzQzOQ

[Louise]

I wonder if Coreboot are working on adding GUID Partition Table support?

http://en.wikipedia.org/wiki/GUID_Partition_Table

As I found out lately, harddisks can not exceed 2TiB with the classic MBR. GUID is needed.

Or does these Southbridges support GUID perhaps?

[Ex-Cyber]

I wonder if Coreboot are working on adding GUID Partition Table support?That's a job for the payload, not for coreboot itself. Apparently, GRUB 2 supports GPT and can be built as a coreboot payload.

[bugmenot]

http://www.coreboot.org/pipermail/coreboot/2009-August/051132.html

There is nothing to say but: Awesome, thank you so much AMD! :)

[Kano]

Nice, would somebody send me a test board with bios chip in a socket and a replacement chip then i would test it... My intel systems are fast but do not support that.

[MaestroMaus]

Kudos for the AMD guys and volunteers. You guys rock the planet!

[Louise]

Kudos for the AMD guys and volunteers. You guys rock the planet!

Yes, it it amazing. I wonder if Bridgman is behind this push as well?

[d2kx]

and to Zheng Bao, who is an AMD engineer working on the CoreBoot support already in his spare time.

This is not entirely correct. Zheng Bao worked on coreboot support in his working time, but did some cleanups and stabilization work in his spare time.

[unimatrix]

Hey nVidia, can you guess which video card I'm going to buy next?

[AHSauge]

Nice! Sort of been waiting for this since I bought my newest mb. I guess I now can actually expect superfast booting sometime in the future. Kudos to AMD & co for making this happen :)

[kensai]

Hey nVidia, can you guess which video card I'm going to buy next?

Wait there, don't take that step too soon, still the oss drivers sucks for newer cards, no power management and a lot of useful features are nowhere to be seen. 3D sucks and all of that.

Maybe in a year or so they will be good oss drivers but not yet. If you switch you will be another of the AMD/ATI depressed users.

Now, I really like this announcement, because that documentation benefits my chipset, and I sure hope the oss drivers get better for me now, they are improving, and are getting there but still there is like a year or 2 until they are the best oss graphic drivers.

[L33F3R]

Hey nVidia, can you guess which video card I'm going to buy next?

S3 540gtx?

[Kano]

I don't see a reason to switch gfx card - you usually don't boot from a gfx card ;) A Phenom II as testsystem for Coreboot would be nice, i don't like outdated slow systems. But i guess OC would not possible with CB.

[bugmenot]

Of course we switch the manufacturer of the graphics card not because of the graphics card but we like the open source politics amd is doing! :)

[whizse]

What's the latest about CoreBoot on Intel platforms?

I seem to recall hearing they're nowhere near as open about this as they are when it comes to GPUs. Is this still the case today? Are there any petitions or lobbying going on to convince them otherwise?

[chithanh]

Be aware that buying a random 780G mobo now and expecting it to be perfectly supported in coreboot soon will probably lead to disappointment.

Coreboot supports only a select few mobos, namely the ones that are owned by the coreboot developers and/or where manufacturers have been cooperative enough. So if you are interested in coreboot it would be wise to wait which mobos are going to receive support at all.

[NeoBrain]

What's the latest about CoreBoot on Intel platforms?

I seem to recall hearing they're nowhere near as open about this as they are when it comes to GPUs. Is this still the case today? Are there any petitions or lobbying going on to convince them otherwise?

They are still as mostly unsupported at they were a year ago. Support has been added for some _mobile_ Intel Core 2 Duo chipset, but apart from that I haven't seen any change in the Socket 775 area.

I wonder how come Intel is such a nice player in OSS graphics but not for coreboot :(

[deanjo]

They are still as mostly unsupported at they were a year ago. Support has been added for some _mobile_ Intel Core 2 Duo chipset, but apart from that I haven't seen any change in the Socket 775 area.

I wonder how come Intel is such a nice player in OSS graphics but not for coreboot :(

3 letters why...... EFI

[Kano]

Now also SFI for Atom MIDs.

[Cairo_BR]

Hi lads, it would be good if ati linux users who have "old" graphic cards make a undersigned(I don't know if this is the correct word(google translate)) to AMD to they make patches to the old 9.3 catalyst driver be compatible with xserver 1.6 and, possibly, to new kernels. This could be just for while radeon/radeonhd opensource drivers aren't ready enough to manage video-out ports and 3d resources fully.
I confess the impression AMD gave to me is that they don't care for their product users, and I've my notebook with radeon x1250/690M for less than one year. I, and I believe a lot of people, want to have the lastests resources that new distros, apps, games and bugfixes that depends of newer kernels(not only of the graphic card world) could give, if the situation won't change the "neigbhor's garden will always be greener"....

[AdrenalineJunky]

Hi lads, it would be good if ati linux users who have "old" graphic cards make a undersigned(I don't know if this is the correct word(google translate))

the word your looking for is petition.

and AMD has actually done alot for linux users of late - but between documentation of chipsets, drivers, open/closed source driver developement, etc, dropping the support for older graphics hadware was something they felt would help them give the best support to the largest amount of people.

that said, go for it - i certianly think bringing kernel/xorg updates to legacy drivers would be nice.

[cutterjohn]

I wonder if Coreboot are working on adding GUID Partition Table support?

http://en.wikipedia.org/wiki/GUID_Partition_Table

As I found out lately, harddisks can not exceed 2TiB with the classic MBR. GUID is needed.

Or does these Southbridges support GUID perhaps?IIRC that is a function of the BIOS, as it is the BIOS that is first started once the machine is turned on and is responsible for initial booting of the OS as well as basic hardware config beyond the powerup defaults...

Even if the BIOS doesn't support it I imagine that a bootloader might be able to work around the limitation...

AFAIK the only system that uses GUID ATM is OSX, and they use EFI (Extensible Firmware Interface from Intel, which they used as a replacement for OpenFirmware when they switched from powerpc. Either one of those are much nicer to work than traditional PC BIOSes as they are much mroe malleable and useful at boot-time...) Oh, yes, either of those replace the traditional BIOS, so machines with those do NOT have what most here would consider a BIOS although they provide much the same function in addition to MANY other things and are, well, extensible by design...

[Kano]

Well it has a "protective" mbr, so in theory you could put a bootloader in the first bytes too...

[hargut]

IIRC that is a function of the BIOS, as it is the BIOS that is first started once the machine is turned on and is responsible for initial booting of the OS as well as basic hardware config beyond the powerup defaults...

Even if the BIOS doesn't support it I imagine that a bootloader might be able to work around the limitation...

It seems that you misunderstand something about coreboot + payloads.
Coreboot itself does "just" the job the hardware initialisation.
The payload is the part of software which is called by coreboot when the hardware is "ready to go".

A "BIOS" image with coreboot needs to contain one or more payloads, otherwise your machine won't boot anything.
http://www.coreboot.org/Anatomy_of_a_Failover_coreboot_v2_Image


I think that it is not a good idea to look at coreboot with the classical/typical bios approach. The way things are done in coreboot is quite different compared to a classical bios.

When grub2 does support the "GUID Partition Table" it will work with coreboot, as grub2 is a vaild payload.

You can also use for example a Linux vmlinux image as coreboot payload, and when you compile the kernel without harddisk drivers, you won't have access to a harddisk, but maybe you don't need it. There are hardly limitations for realising your ideas on "what should the boot progress do for my use/needs" on using coreboot.
If you like tetris and you don't want to do something else with your computer then check out the tint payload. (You won't need a harddisk in that case to play a tetris-style-game.)

http://www.coreboot.org/Payloads

When you say "a function of BIOS" it sounds for me like "what the bios isn't able to do nobody can do" - but as coreboot replaces the bios nearly everything what hardware can do could be done. And therefore good hardware documentation as AMD provides is really nessecary for getting coreboot running on a system.
Thanks AMD for providing hardware documentation for recent mainboard-chipsets! (Hopefully the frist AM3 boards can be supported within a few weeks/months.)

Over-/Underclocking is also possible with coreboot (and a few more things which can be realized with a processor - e.g. using the cache as ram (CAR) before initializing the ram (romcc)).
But right now there is no "fancy-bios-gui" for changeing settings for overclocking/underclocking. (changing coreboot code & recompiling would be needed)

Kind regards,
Harald

[chithanh]

In practice too, using syslinux gptmbr.bin which allows booting GPT disks from non-EFI systems. (Syslinux has no GPT chainloading support yet, though)

[highlandsun]

Well, those docs certainly made for interesting reading. It highlighted a fact that I hadn't realized until now - a lot of the BIOS comes from AMD as a blob, and you're just suppose to call into it or provide callbacks to it from your own main BIOS code. So in some respects the hardware initialization is still a black box...

[Louise]

But right now there is no "fancy-bios-gui" for changeing settings for overclocking/underclocking. (changing coreboot code & recompiling would be needed)


As for myself atleast the only options I have ever used are:

* change boot order
* enable Cool'n'Quite
* enable QFAN (ASUS feature)
* change allocated IGP memory
* PXE
* date and time

Are this kind of features something that the developers have planned?

And how do they feel about all the other options modern BIOSes have?

[Kano]

@chithanh

Did you ever look at the doc/gpt.txt? Usually somebody would like to use GPT when a hd or raid is bigger than 2 TB. But syslinux is not able to boot from a partition > 2 TB, then you can use a standard mbr too as you gain nothing.

[hargut]

As for myself atleast the only options I have ever used are:

* change boot order
* enable Cool'n'Quite
* enable QFAN (ASUS feature)
* change allocated IGP memory
* PXE
* date and time

Are this kind of features something that the developers have planned?

And how do they feel about all the other options modern BIOSes have?

* change boot order
use grub2/filo/or another bootloader as payload and you can change the boot order
* enable Cool'n'Quiet
Is supported, but needs ACPI support in your target motherboard.
* enable QFAN (ASUS feature)
I've no idea about what QFAN is, but using fancontrol from lm-sensors package should do the same job as I imagine.
* PXE
Use an etherboot/gpxe Payload.
* date and time
There are no setting fields for that, but hwclock --systohc does the job.

So in fact, all the stuff you'd need is supported. ;)
But maybe not directly via modifying the values with a "BIOS-UI".

Kind regards,
Harald

[chithanh]

syslinux is not able to boot from a partition > 2 TB, then you can use a standard mbr too as you gain nothing.That is a limitation but it is IMO not so severe to have /boot restricted to the first 2 TB. With MBR you cannot use >2 TB at all.

[Kano]

You can use more than 2 TB, but one partition may not exceed 2 TB.

[deanjo]

You can use more than 2 TB, but one partition may not exceed 2 TB.

The restriction only applies on the boot partiition. A simple workaround is creating a separate /boot partition and as long as you have CONFIG_LBD enabled in the kernel you should be able to use partitions > 2TB easily.

[Kano]

And you verified that on your system already?

[deanjo]

And you verified that on your system already?

Yes, got a fileserver at work already setup like that.

[mtippett]

Hi lads, it would be good if ati linux users who have "old" graphic cards make a undersigned(I don't know if this is the correct word(google translate)) to AMD to they make patches to the old 9.3 catalyst driver be compatible with xserver 1.6 and, possibly, to new kernels. This could be just for while radeon/radeonhd opensource drivers aren't ready enough to manage video-out ports and 3d resources fully.
I confess the impression AMD gave to me is that they don't care for their product users, and I've my notebook with radeon x1250/690M for less than one year. I, and I believe a lot of people, want to have the lastests resources that new distros, apps, games and bugfixes that depends of newer kernels(not only of the graphic card world) could give, if the situation won't change the "neigbhor's garden will always be greener"....

Just to be clear.

The actions taken of removing the legacy ASIC Ids from the Proprietary driver was for both Windows and Linux, we share code between the drivers and so what happens for Linux, happens for Windows, and what happens for Windows, happens for Linux. The split was taken primarily along technology lines, and not time based lines.

Regarding the age of the product, we cannot control when an OEM starts or stops shipping a particular ASIC. The RS690 has been superceded by the RS740, RS780 and the recently released RS880.

Due to AMD's efforts in seeding the community with specifications as well as investing engineering effort to progressing the support for the older ASICs, the issues to the Linux community should be minimized due to the existance of the OSS RS690 driver.

Regards,

Matthew

[highlandsun]

So in fact, all the stuff you'd need is supported. ;)
But maybe not directly via modifying the values with a "BIOS-UI".

Kind regards,
Harald

My interest here is in setting a lower voltage for DRAM; e.g. installing 1.6v DDR2 SODIMMs (instead of stock 1.8V) in my AMD Puma notebook to get lower temperatures and longer operating life. I would expect that this is something you'd want to do quite early in a boot process. Dunno of any Linux packages that would enable this just yet.