Maybe, but Optane is fairly strange. For most cases its not much faster than the existing PCIe SSDs like those that Apple has been using.
Not in latency. Most of the tech porn benchmarks don't trying to measure it.
The intent of Optane is not to replace SSDs but to replace RAM. ( or not so much replace RAM, but to expand tasks associated with RAM at more cost effective price).
Almost all of the tech porn sites miss that. The comment sections even more so.
Optane expanding VRAM an 3x-8x is a bit impact. Optane expanding the file cache buffer of a file system out 3x-8x is going to be big impact in HDD context.
( mainstream users typically want just one volume to dump all their stuff into. The argument that $128-256GB SSD are now 'cheap' doesn't really fly if have 300GB of stuff to store. For example, your collection of photos and 4K video off your smartphone. )
The downside is that it consumes more power at idle and is much more expensive in large capacities.
Relative to RAM, it is a actually much larger capacity per $ . Again this is
not trying to replace HDD capacity sizes at all. Nor it is out to eliminate SSD ( which are trying to limbo down to HDD $/GB levels from 3-6 years ago. )
Likewise on power .... RAM is powered at idle.
In the future Intel imagines this as technology that could potentially blend RAM and SSDs, but I'm not sure its being sold as such yet.
It already is a blend in terms of $/GB of RAM and SSD. it is about in the middle. The preplexing part is that folks mindset is only looking at the SSD half and not the RAM half of applications.
[/quote]
Right now Intel is targeting Optane at enterprise applications that need to read and write lots of small information at a time (think databases) or as the SSD portion of a fusion drive. [/quote]
No. Databases targeting performance spend large amount of effort implement a file buffer cache. Major DB's basically spend tons of effort trying to completely bypass the file system's file buffer cache because they typically think it is 'wrong'. ( e.g., Oracle DBMS , DB2 , etc. ). The RAM based DBs (memcache , etc. ) even more so.
DBs are not working at byte level. There are tons of full blocks that are being shuffled in out. Even the logs aren't largely a couple of 'bytes' like increments ( but yes will get multiple updates to larger blocks in rapid succession. )
I'm not really sure where this would fit in with the mac pro and I don't think its worth the extra cost over traditional PCIe drives on a consumer desktop.
These initial products largely don't . The M.2 product is extremely targeted to speeding up HDDs. ( e.g., better Fusion Drive). The Mac Pro has no Fusion drive. Unless the Mac Pro is revised and provisioned to have two full bandwidth SSD NVMe slots then is no room. This is not a "replacement" for a single primary SSD.
The "enterprise" product is out of price range.
A x4 PCIe M.2 sized product that was used to expand vRAM for a GPU card probably work well.
http://www.anandtech.com/show/10518/amd-announces-radeon-pro-ssg-fiji-with-m2-ssds-onboard
Vega has "SSD" support built in. While AMD and Intel sometimes don't get along it would be pretty boneheaded for AMD to ignore Optane class devices for this kind of product. That is exactly what should be doing; not the NAND based stuff (especially as the bulk of that transitions over to TLC (triple 'bits' per cell) and higher. The latency is going to get worse. )
Optane would be useful as Apple Fusion Drive applications for 500GB-1TB like drives on the lowest entry level systems. Only makes sense if Apple can get the right balance on pricing of HDDs and Optanes. Apple appears to be walking away from 'Fusion Drive' ( at least at the file system level) so they are still betting on users will just pay for SSD $/GB even at the entry level.
[doublepost=1490715150][/doublepost]
I didn't read the whole thread (obviously, 181 pages ;-) ) but could this be one possible reason why the new MacPro isn't out yet?
Other than a possible Vega class GPU coupled to yet another variant of the M.2 format
There is no spinner HDD or SATA device in the Mac Pro so the M.2 "SATA data cache" product doesn't have any traction. A next generation Mac Pro about as equally less likely to have a SATA device in it also.
A x2 PCIe M.2 card isn't up to snuff as being a VRAM supplement for a GPU. The x2 is aimed more so at staying out in front of a HDD on SATA or trying to extend the endurance life of a cheap SSD.
Intel's enterprise card is probably just too expensive for the workloads typically aimed at a Mac Pro. Apple could move to having a configuration where one GPU is replaced by x16 PCIe. worth of storage bandwidth (i.e. drop 2 - 4 M.2 slots on a board ), It is a bit of 'overkill' for a scratch drive for most folks (their read/write workloads aren't at the very top end range and aren't typically mid triple digit GB RAM workloads either. ) .
I don't want to get up hope too much but this would really fit a future MacPro well.
Apple's bug up their butt about the SSD card slot had some merit when they were rolling out before M.2 got established. As M.2 gets more widely distributed and mature their approach is a bit lame for the desktop Mac models. The laptops are moving toward integrated SSD assemblies, but the desktops don't have anywhere near those kind of "thinness" requirements.
Throw on top Apple finally enabling TRIM with a macOS built in command the 'blocking 3rd parties' by having a physically proprietary slot is even more lame.
If Apple expands to having a second SSD socket then it should be M.2 ... whether get a 2nd or even an update is up in the air.
[doublepost=1490716548][/doublepost]
IHMO, Largely misses the point. I think whether this works or not depends upon where the HDD prices go. They have kind of plateaued at $/GB. If can get the 1-2TB HDDs down to $20-40 then coupled to a $40-70 Optane M.2 gets you are 1-2TB storage system in the $60-110 that:
does boot fast.
does hibernate-sleep/wake fast.
does VM paging fast enough.
is more responsive over a fix set of modest, daily tasks ( Grandma who uses the same 2-3 apps every day).
can store just everything in one simple simple "c: drive" just like the last 10+ years.
This is aimed at $400-600 PCs...... not some gamer, 'drag racer' box. If that PC feels more responsive than an 5-10 year old 5400-7200K spinning HDD based system then it is a win. Just as fast as the bleeding edge SSD is not a targeted metric ( outside of the tech porn sites).
Yeah it is restricted to the latest Gen 7 computers but over the next 3-4 years there will be more than a decent number of those sold ( and HDD $/GB probably will keep falling if HDD's want to remain competitive).
It isn't a "make money fast" mechanism for Intel/Micro but can probably sell a decent number at some steady rate ( some baseline load to keep the fab going. )
As much as Intel is overselling their product, this "sky is falling" response is about just as oversold.
