[patch] DDR3 support of AMD Family 10

[Bao, Zheng]

DDR3 supporting is added.

More testing need to be done. But we don't have boards that cover
all the options.

Features that have been tested.
The socket AM3 and ASB2.
Single rank and double rank DIMM.

Features that need to be tested.
Socket C32 and G34, I don't know if it is DDR3.
Quad rank.
Registered DIMM, which is quite complicated and I am not confident
about.

The tilapia board is the demo that shows the way to use DDR3.
It is quite close to mahogany, except tilapia uses AM3.
For a new board, just specify the socket type in Kconfig the
devicetree.cb.

Signed-off-by: Zheng Bao

[Carl-Daniel Hailfinger]

On 23.04.2010 03:24, Bao, Zheng wrote:
> DDR3 supporting is added.

Very nice, thank you!
I'm sure this will get reviewed quickly.

Regards,
Carl-Daniel

[Harald Gutmann]

Wow, that's once again great news!

Thanks!
Harald

On Friday 23 April 2010 03:24:43 Bao, Zheng wrote:
> DDR3 supporting is added.
>
> More testing need to be done. But we don't have boards that cover
> all the options.
>
> Features that have been tested.
> The socket AM3 and ASB2.
> Single rank and double rank DIMM.
>
> Features that need to be tested.
> Socket C32 and G34, I don't know if it is DDR3.
> Quad rank.
> Registered DIMM, which is quite complicated and I am not confident
> about.
>
> The tilapia board is the demo that shows the way to use DDR3.
> It is quite close to mahogany, except tilapia uses AM3.
> For a new board, just specify the socket type in Kconfig the
> devicetree.cb.
>
> Signed-off-by: Zheng Bao

[Stefan Reinauer]

Dear Zheng!

A big "thank you" to you and AMD from me and on behalf of the coreboot
community.

I am very happy to see how great AMD's coreboot support is and has been
for all these years!

I commited your patches as r5481, r5482 and r5483 to the coreboot
repository.

Stefan

On 4/23/10 3:24 AM, Bao, Zheng wrote:
> DDR3 supporting is added.
>
> More testing need to be done. But we don't have boards that cover
> all the options.
>
> Features that have been tested.
> The socket AM3 and ASB2.
> Single rank and double rank DIMM.
>
> Features that need to be tested.
> Socket C32 and G34, I don't know if it is DDR3.
> Quad rank.
> Registered DIMM, which is quite complicated and I am not confident
> about.
>
> The tilapia board is the demo that shows the way to use DDR3.
> It is quite close to mahogany, except tilapia uses AM3.
> For a new board, just specify the socket type in Kconfig the
> devicetree.cb.
>
> Signed-off-by: Zheng Bao

[xdrudis]

On Fri, Apr 23, 2010 at 09:24:43AM +0800, Bao, Zheng wrote:
> DDR3 supporting is added.

Thank you very much. I would thank you more but I'm too busy looking
at what to buy from AMD...

[bari]

Thank you very much for the DDR3 coreboot support.

Are there any coreboot plans at AMD to support G34 with Opteron 6000's, 
SR56x0 and SP5100, Dinar or Drachma, etc.?

http://www.amd.com/us/products/server/processors/6000-series-platform/pages/...

-Bari

Bao, Zheng wrote:
> DDR3 supporting is added.
>
> More testing need to be done. But we don't have boards that cover
> all the options.
>
> Features that have been tested.
> The socket AM3 and ASB2.
> Single rank and double rank DIMM.
>
> Features that need to be tested.
> Socket C32 and G34, I don't know if it is DDR3.
> Quad rank.
> Registered DIMM, which is quite complicated and I am not confident
> about.
>
> The tilapia board is the demo that shows the way to use DDR3.
> It is quite close to mahogany, except tilapia uses AM3.
> For a new board, just specify the socket type in Kconfig the
> devicetree.cb.
>
> Signed-off-by: Zheng Bao

[Greg Scantlen]

Hello Zheng Bao,

I'm Greg at Creative Consultants in Albuquerque, NM. We build clusters, 
mainly for LANL.I'm hoping to bid on a RFQ that involves coreboot for 
some custom cluster nodes. I've been benchmarking a supermicro *H8QGi-F* 
motherboard with (4)x 6128 CPUs and (16)x 4GB ECC UnBuff DDR****. Would 
this be a good candidate for core boot? I'm just starting reading the 
Howto on coreboot now. I do have experience with flashing BIOS - is that 
technique sufficient or should I have special equipment to get started? 
Thanks for your guidance in advance.

Greg

[bari]

Greg Scantlen wrote:
> Hello Zheng Bao,
>
> I'm Greg at Creative Consultants in Albuquerque, NM. We build clusters, 
> mainly for LANL.I'm hoping to bid on a RFQ that involves coreboot for 
> some custom cluster nodes. I've been benchmarking a supermicro *H8QGi-F* 
> motherboard with (4)x 6128 CPUs and (16)x 4GB ECC UnBuff DDR****. Would 
> this be a good candidate for core boot? I'm just starting reading the 
> Howto on coreboot now. I do have experience with flashing BIOS - is that 
> technique sufficient or should I have special equipment to get started? 
> Thanks for your guidance in advance.
>
> Greg

The last I heard the 5650 was working with coreboot. The 5690 is still a 
work-in-progress. Zheng Bao should know the latest.

-Bari

[Bao, Zheng]

The releasing of 5650 is in the progress. It is about 1 month to go. If
you can not wait, please contact AMD to get the code.

Zheng
> -----Original Message-----
> From: [e-mail]

[mailto:[e-mail]]
> On Behalf Of bari
> Sent: Thursday, July 22, 2010 5:16 AM
> To: Greg Scantlen
> Cc: [e-mail]
> Subject: Re: [coreboot] [patch] DDR3 support of AMD Family 10
>
> Greg Scantlen wrote:
> > Hello Zheng Bao,
> >
> > I'm Greg at Creative Consultants in Albuquerque, NM. We build

clusters,
> > mainly for LANL.I'm hoping to bid on a RFQ that involves coreboot

for
> > some custom cluster nodes. I've been benchmarking a supermicro

*H8QGi-F*
> > motherboard with (4)x 6128 CPUs and (16)x 4GB ECC UnBuff DDR****.

Would
> > this be a good candidate for core boot? I'm just starting reading

the
> > Howto on coreboot now. I do have experience with flashing BIOS - is

that
> > technique sufficient or should I have special equipment to get

started?
> > Thanks for your guidance in advance.
> >
> > Greg
>
> The last I heard the 5650 was working with coreboot. The 5690 is still

a
> work-in-progress. Zheng Bao should know the latest.
>
> -Bari

[Carl-Daniel Hailfinger]

Hi Greg,

On 21.07.2010 22:04, Greg Scantlen wrote:
> We build clusters, mainly for LANL.I'm hoping to bid on a RFQ that
> involves coreboot for some custom cluster nodes. I've been
> benchmarking a supermicro *H8QGi-F* motherboard with (4)x 6128 CPUs
> and (16)x 4GB ECC UnBuff DDR****.Would this be a good candidate for
> core boot?

AFAICS it looks like a good candidate, but you'll have to do a bit of
development work. If you're lucky, you can get this done in a few days.
The good news is that the chipset looks supportable (may even be
supported already in part), and the board has a serial port which will
help you tremendously for debugging.
IPMI may be a bit of a challenge because some IPMI controllers screw up
booting of any custom firmware unless you know how to silence those IPMI
controllers.
> I'm just starting reading the Howto on coreboot now. I do have
> experience with flashing BIOS - is that technique sufficient or should
> I have special equipment to get started? Thanks for your guidance in
> advance.

If you want to do any serious development, I recommend to make sure the
BIOS flash chip is socketed (solder a socket on the board if not), or at
least check for a SPI recovery connector. You absolutely want a fast SPI
flash programmer or ROM emulator which works under the operating system
you're using for coreboot development (which will probably be Linux).
Make sure to compare prices (and consider the option to run the
programmer software in VMware if you can't get a Linux version). You
will likely reflash the chip a few hundred times during development, and
you want flashing to be fast and reliable.
Oh, and make sure the flash chip you're using has fast programming time
(vendors usually ship cheap flash, not fast flash, and your time is
worth a lot more than a few cents for a faster chip).

A nullmodem cable for the serial port will be crucial for debugging, and
a POST card is recommended as well.

I usually recommend to get your feet wet with a desktop board that is
already supported and cheap, and to move on to the real target only
after that board runs fine. That way, you get a feeling for working with
coreboot, and you know what to expect. One of the supported AMD 780
boards might be a really good starting point.

Side note: I'd be really happy if you could test flashrom
http://www.flashrom.org/ on those boards and submit test reports to
[e-mail].

Regards,
Carl-Daniel

[Greg Scantlen]

Hi Dale,

Do you want to see the Statement of Work, Instruction for this Quote. If
Creative were to get the award, I would need your (or someone's) help with
setting up the serial console server, and online power management for PDUs.
The timelines are rediculus, it could only be done by somewith a lot of
experience. We're getting lot's of help from FlashRom.org, and CoreBoot.org
already. I'm only interested in the H8QGi-G (or maybe H8DGU) motherboards.

Greg

On Wed, Jul 21, 2010 at 4:04 PM, Greg Scantlen  wrote:
> Hello Zheng Bao,
>
> I'm Greg at Creative Consultants in Albuquerque, NM. We build clusters,
> mainly for LANL.I'm hoping to bid on a RFQ that involves coreboot for some
> custom cluster nodes. I've been benchmarking a supermicro *H8QGi-Fmotherboard with (4)x 6128 CPUs and (16)x 4GB ECC UnBuff DDR
> ***. Would this be a good candidate for core boot? I'm just starting
> reading the Howto on coreboot now. I do have experience with flashing BIOS -
> is that technique sufficient or should I have special equipment to get
> started? Thanks for your guidance in advance.
>
> Greg

[bari]

Greg Scantlen wrote:
> Hi Dale,
>
> Do you want to see the Statement of Work, Instruction for this Quote. If 
> Creative were to get the award, I would need your (or someone's) help 
> with setting up the serial console server, and online power management 
> for PDUs. The timelines are rediculus, it could only be done by somewith 
> a lot of experience. We're getting lot's of help from FlashRom.org, and 
> CoreBoot.org already. I'm only interested in the H8QGi-G (or maybe 
> H8DGU) motherboards.

Greg,

"The sr5650 is in the progress. It is about 1 month to go." as Zheng Bao 
mentioned.

It's going to take a few more months to get coreboot up on sr5670 or 
sr5690. They definitely will get coreboot support since there are funded 
projects for new coreboot mainboard designs using the sr5690 for HPC 
applications. Some boards will support multiple sr5690's and 8/12 core 
6100 series G34 sockets along with drip free water cooled chassis and 
sub microsecond latency >2.5Gb/s network interconnects.

-Bari

[Arne Georg Gleditsch]

bari  writes:
> "The sr5650 is in the progress. It is about 1 month to go." as Zheng
> Bao mentioned.
>
> It's going to take a few more months to get coreboot up on sr5670 or
> sr5690. They definitely will get coreboot support since there are
> funded projects for new coreboot mainboard designs using the sr5690
> for HPC applications. Some boards will support multiple sr5690's and
> 8/12 core 6100 series G34 sockets along with drip free water cooled
> chassis and sub microsecond latency >2.5Gb/s network interconnects.

Hi,

I'm interested in the progress of the sr5690 -- I would like to port
Coreboot to the SuperMicro H8QIi+-F, where this is used.  I'm willing to
do work and/or testing myself, but if work is underway it would be good
to avoid duplication of effort.

Thanks,