Disappointing is not the right word. Total heartbreak is more appropriate.
"Heartbreak" is even generous. Total devastating disappointment. This was one of the big features of Kaby/200 series chipsets... native "special" support for optane/xpoint products. Kaby was already on life support being 4 core and modest default clocks, but now this? The 960 is looking like it will be most enthusiasts go to choice for the next year.
Pff, all the sheep are posting here repeating the same conclusion the article had that this is disappointing (except OneCosmic's insightful post which is accurate). It's pretty obvious those numbers are after any cache on the Optane is depleted.
While it's true Intel has failed at media relations lately, this leak was without context and without thoughtful analysis by guru3d. Think about those 950 Pro speeds. They're inflated, after the DDR cache is full the speeds drop dramatically. 950 Pro out of cache can be as low as 200MB writes and this 32GB Optane lowend is apparently 500MB.
With the right controller and the cache (which it will have), Optane will match and crush the 950/960 Pro drives at every level and this is the beta..
This is also an example of an intel sheep response.
There's not a single review of this, we can only go by the advertised specs which are disappointing..
Your 'thoughtful analysis' is hearsay. Where do they state 500mb/s is the minimum speeds.
Everything is rated at a best case scenario. SM961 destroys their rated specs using a beta firmware and default windows 10 driver and it only has 2 flash packages.
The 960 pro is also going to be much faster than the sm961 in specific workloads. The 2TB 960 pro will have 4 flash packages, that will bring a huge improvement in certain workloads as well.
Your tidbit "match and crush the 950/960 Pro drives at every level"
proves that you are the biggest intel fanboy around.
BTW you can expect to pay $500 for this 32GB drive..
Talk about money well spent.
High random write IOPS figures, really small capacities, decent but not so great sequential speeds -> the only thing comes to my mind is ZFS SLOG (the theoretically ideal SLOG device is small [16Gb should be plenty for most], pretty fast with synchronous random writes and looses minimal user data [implying it also keeps it's internal storage structure, including it's own ECC data consistent] upon a power failure [including a kernel panic induced watchdog-kind reset]).
Although, I believe most of the actual ZFS users don't really need a SLOG device (they don't even do meaningful amount of synchronous writes to begin with or these aren't really important to be handled synchronously, so sync=disabled is a viable alternative to SLOG if they have a UPS and troubled by ZIL performance), let alone they shouldn't pay for a "theoretically ideal" device (they can plug in any kind of decent SSD at hand, anything with some backup capacitors, a mature firmware and not unusually low write performance or durability rating).
Other than that, I don't know what Intel thinks the end-user might use these devices for, especially since you have to upgrade your platform with little to no real benefits (if you would go from >=Haswell) and I imagine these won't be handed out for cheap.
Now with the capacity thing aside, there is another important thing to bring up. First gen XPoint dies are 128 Gbit, which works out to 16 GB. That means the product specs for the 16GB part are turning in those specs *WITH ONE DIE*. NAND based SSDs can only reach these sorts of figures by spreading the IO's across four, eight, or more dies operating in parallel. This is just one die, and it is nearly saturating two lanes of PCIe 3.0!
Reading between lines you seem to be suggesting that these products will only have a single or two chips, producing that measly amount of space that won't satisfy anybody anymore. Why is that, then? Because even a 16GB model will cost as much as 500GB NAND SSD?
http://www.*******.com/intel-8000p-3d-xpoint-drives-leaked/
This is Intels roadmap for their new technology. Speeds will be much more ramped up when fully released as a FINISHED product. Way too much is being extrapolated from an engineer model that is not even fact checked. I'm pretty sure Intel isn't going to release a new technology that is slower than anything they already have out there. Give it a little time, I'm sure they will come through with a good product. And no I am by no means an Intel fan boy.
*Insert e t e k n I x where the asterisk are it won't let me show link.
I'm pretty sure Intel isn't going to release a new technology that is slower than anything they already have out there. Give it a little time, I'm sure they will come through with a good product. And no I am by no means an Intel fan boy. *Insert e t e k n I x where the asterisk are it won't let me show link.