Gionatan Danti (g.danti.delete@this.assyoma.it) on August 1, 2022 6:07 am wrote:
> Adrian (a.delete@this.acm.org) on August 1, 2022 2:15 am wrote:
> > AFAIK, Intel has never given any information about the materials used.
>
> I doubt that material availability or production difficulty were
> the main motivations after Intel's decision to cancel Optane.
>
> From a technical standpoint, even relatively low capacity 3DXPoint can be very useful for non-volatile
> read or write cache (ie: think using it for storing incoming writes on HDD or SSD).


With endurance of just few millions write cycles?
I think, for that to work the cache has to be rather big, probably many times bigger than DRAM caches that are used now. Say, where today they have 256MB of DRAM cache, 3DXP will have to be 4-8 GB.
Which is several times more expensive than DRAM.
Write (i.e. non-read) cache can be smaller, but still I don't see how it could be as small as DRAM cache.

>
> I suppose Intel's decision roots in the fatal reason for many products cancellation: lower
> than expected margins. It seems a recurring problem for Intel, destroying many products
> which, albeit very profitable, show less margins than high-priced server CPU.
>
> Its NAND production had basically the same fate some years ago, and now even enterprise
> SSDs are outside Intel's offer (with client SSDs being cancelled ~10 years ago).
>
> Maybe it is a fine thing to do from management standpoint, but very sad at a technical level.
>
> Just my 2c.
> Regards.