[MinnowBoard] 4GB Ram need pcb rerouting?

[Krau, Michael P]

:: To the question of memory configuration in the UEFI Firmware for MinnowBoard MAX ::

I believe I have published this information before on this this list, so you may find it familiar. 

In general, the memory configuration is really not part of the UEFI, but part of the Platform Initialization code.  The Platform Initialization code comes from the hardware vendor (in the case of memory configuration on Intel platforms... Intel).   

Now in general, the Memory Reference code (MRC) from Intel is designed to read the SPD values from the DDR3 DIMMS, and configure the memory controller appropriately (with some fine adjustment to the device parameters along the way to optimize memory performance based on actual measurement/observation).  The MRC is supposed to be able to support all of the configurations approved for the processor.  (This pre-supposes the memory configuration is reasonable and supported).

However, for the MinnowBoard MAX there were a few modifications made to the MRC due to issues with the hardware in production.  First off the original boards have an EEPROM to store DIMM type SPD data (even with the soldered down chips), but I am not sure if the EEPROM was EVER programmed with the data.  Instead the MRC code was modified to behave as follows:

1) If the PCDs for memory configuration in the build DSC are set (see the release notes), these values will have priority over all other memory configuration mechanisms (hard coded). - a developer option
2) If the EEPROM for the SPD data is populated with data that appears to be valid (by cursory evaluation), or if the SPD data from a SO-DIMM is available, then that data will be used in the training algorithm.
3) Failing points 1 and 2, there are GPIOs on the MinnowBoard MAX, which define the platform as 1, 2 or 4 GB DDR3 memory configuration (per the system design - no ECC). 

So while it should be possible to get the configuration support you are looking for in mechanism #2 (above), it has not been tested on the MinnowBoard MAX.   

Note , per mechanism #2 - I believe if someone were to re-do the MinnowBoard MAX to have SO-DIMM sockets, the firmware should be able to read the SPD data directly for the SO-DIMMs (this has not been tested or verified, so please consider it an academic observation)


Sincerely,


Michael Krau
 
While I am an Intel employee, I do not represent Intel and am not authorized to speak for Intel. 
-----Original Message-----
From: elinux-MinnowBoard [mailto:elinux-minnowboard-bounces at lists.elinux.org] On Behalf Of John Hawley
Sent: Thursday, June 16, 2016 10:37 AM
To: elinux-minnowboard at lists.elinux.org
Subject: Re: [MinnowBoard] 4GB Ram need pcb rerouting?

Inline -JH

On 06/16/2016 04:06 AM, Joe Tom Collins wrote:
> I think this this thread alone should tell the MinnowBoard designers 
> there is a market for a board where the user has more choice for RAM.
> 
> Personally, if it's possible, I would like to see some way to add a 
> couple of SO-DIMM's for up to 32GB.

So the reason SODIMMs weren't used boils down to 2 things:

1) Cost
2) Size of the board

To the first point, we are trying to keep the board within a certain price range, and a lot of that has to deal with what we learned out of the old V1 (the old square board), and partly just based on what's available in the market and price ranges.  Folks are very cost sensitive, and we are trying to keep that in mind with the design decisions we've made.

To the second adding the connectors would push the size of the board up, at least, into the size of the old v1 if not actually closer to a mini-ITX.  We generally felt that doing that making a board that was too big.

Now to the ram size, with the Baytrail's - we aren't getting to 32G of RAM, and in fact we'd have to go up into the Atom server chips to get close to that.

http://ark.intel.com/products/78477/Intel-Atom-Processor-E3826-1M-Cache-1_46-GHz

http://ark.intel.com/products/78475/Intel-Atom-Processor-E3845-2M-Cache-1_91-GHz?q=e3845

Max out at 8G officially supported, kicking up into the Atom server chips would make the board a lot more power hungry, a lot bigger, and a LOT more expensive of a board.

> I realize the processor might not be capable of that, but my retro PC 
> history has me gun shy about any system that can't at least take the 
> full amount of RAM that the processor takes. No matter what, it always 
> ends up being an issue.

Depends on specific use cases, but yes.  We made specific design decisions, that may, or may not, work out for your use cases.  There are things in the works, like the board that B-O was talking about that they are working on, where the full amount of RAM is available.  Right now there's the Turbot and the MAX that are available, that's not to say that more boards won't be available with other options - that's why this is an Open Source Hardware platform, so folks can modify it as they need.

> Or how about some high speed SD RAM if power consumption is an issue.
> Maybe 2GB of SD RAM that can run as main memory, or also as a cache 
> for an SO-DDR3 DRAM hood that can take as much RAM as the processor 
> could handle.

Are you talking more like onboard eMMC?  Right now the best we can do is mSATA via something like the Silverjaw, or even an nvme based storage off of mPCI-E via the same.  I know that adding eMMC has been something we've discussed a lot, and I know some folks are already working to add (like B-O)


- John
_______________________________________________
elinux-MinnowBoard mailing list
elinux-MinnowBoard at lists.elinux.org
http://lists.elinux.org/mailman/listinfo/elinux-minnowboard