V G wrote:
> Hi all, just throwing this out on the table here:
>
> A PCI-e RAM disk that could make use of your (old) RAM, sit in your PCI-e
> slot, and act as a mass storage unit. A few years ago, there was this uni=
t (
> http://www.newegg.com/Product/Product.aspx?Item=3DN82E16815168001), but t=
hey
> no longer make it. It only DDR2 and had capacity limitations.
>
> Most consumer motherboards have a 32GB RAM limit,
Intel LGA1155 and AMD AM3 boards do indeed have 32GB as the practical limit=
..
(I beleive theoretical limits may be higher but if so the modules needed=20
don't exist
yet).
>  and it would be nice to
> make a 512GB RAM disk for special tasks (like compiling in RAM, virtual
> machines in RAM, other disk intensive things).
>  =20
If you are so concerned about build speed that you want to do your=20
compiling on
a ramdrive you probablly want a lot of processing power too which means you
would be looking at intel LGA2011 or maybe AMD G34.

Desktop LGA2011 supports 64GB with current modules and server LGA2011
supports even more.

How many applications are there that really need more than 64GB of ram and
yet don't also need a correspondingly crazy ammount of CPU.
> There have long been FPGAs with built-in DDR3 controllers as well as PCI-=
e
> controllers. To build a similar device, for example, something with
> multiple slots that can address more than 256GB of RAM would be nice.
>
> How expensive and difficult would this be? Most of the logic is already
> there, the only thing really is the electrical engineering (routing,
> layout, etc.) considerations and the memory addressing logic which doesn'=
t
> seem too hard from what I know (I've built a simple memory controller on =
a
> Spartan 3E for my PIC32 in the past).
The problem is that both the speeds and the number of lines involved are=20
very
large. For signal integrity reasons with modern memory modules you don't=20
really
want to be putting more than two modules per channel and each channel is
going to have a lot of lines (the data bus along is 128 bits wide on PC=20
memory
modules).

That makes the PCB design decidedly difficult and a lot of layers will=20
be needed.
That combined with the big BGAs means that prototyping costs will be very
high.
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.