Intel Turbo Memory

Intel Turbo Memory

[Russell Fleitman]

Hello, I have searched the archives of this list and have found no references to Intel Turbo Memory or Robson.

Has anyone heard of these chips working?

I have the output from dmesg and lspci ready to post but thought I would ask if I am barking up the wrong tree first.

Thanks in advance for any help,

Russell


      

Re: Intel Turbo Memory

[Jörn Engel]

On Wed, 4 June 2008 15:58:34 -0700, Russell Fleitman wrote:
> 
> Hello, I have searched the archives of this list and have found no references to Intel Turbo Memory or Robson.
> 
> Has anyone heard of these chips working?
> 
> I have the output from dmesg and lspci ready to post but thought I would ask if I am barking up the wrong tree first.

It appears as if those devices are raw flash and could be accessed
through mtd - provided someone either has documentation or creates some
by reverse engineering.

I have tried to obtain documentation and failed.  My best guess is that
it is as dead as hybrid disks - those simply don't perform.  It would be
neat to use for different purposes, but Intel doesn't seem to see a
business case anymore and won't ship either a Linux driver or
documentation.

In short: forget about it. :)

Jörn

-- 
When in doubt, punt.  When somebody actually complains, go back and fix it...
The 90% solution is a good thing.
-- Rob Landley

Re: Intel Turbo Memory

[Markus Schiltknecht]

Hi,

Jörn Engel wrote:
> It appears as if those devices are raw flash and could be accessed
> through mtd - provided someone either has documentation or creates some
> by reverse engineering.

Isn't that the first - and so far only - device which adheres to their 
NVMHCI specification [1]?

Markus


[1]: Non-Volatile Memory Host Controller Interface Specification:
http://www.intel.com/standards/nvmhci/index.htm

Re: Intel Turbo Memory

[Jörn Engel]

On Mon, 9 June 2008 09:20:32 +0200, Markus Schiltknecht wrote:
> Jörn Engel wrote:
> >It appears as if those devices are raw flash and could be accessed
> >through mtd - provided someone either has documentation or creates some
> >by reverse engineering.
> 
> Isn't that the first - and so far only - device which adheres to their
> NVMHCI specification [1]?

Interesting.  If that is the case then we're dealing with a weird device
indeed.  "Erases ... are abstracted", OOB is 12 bytes per 512.  So they
do some sort of FTL stuff, but still leave 12 out of 16 bytes of OOB
available for the user.  Quite odd.

Anyway, thanks for the spec!

Jörn

-- 
You can take my soul, but not my lack of enthusiasm.
-- Wally