Replies to post #148545 on Intel Corp (INTC)

[Unkwn]

think the big problem was that Intel massively overhyped this technology and the final product was nothing near what Intel initially promised. Maybe it gets better with the next revision.

Personally, I don't think Intel, together with Micron, did go all that way just to provide some niche application such as database acceleration in servers. Just consider how much money had to be put in order to develop this technology. I am pretty sure they have seen a path for this product to become ready for mass market, probably even replacing flash at some stage as the better memory technology.

How could that look like? Currently, it is too expensive, pretty clear. It is good stuff, but simply too expansive for mass memory. Maybe, at the time of developing 3D Xpoint, they didn't see 3D Flash coming. Not really probable in my opinion. Maybe they have seen a path in the distant future to make this technology price competitive with flash or even cheaper. The closest explanation to me seems stackability. By that I mean the increase of layers of 3D Xpoint and the costs involved with doing so. Today we already know that 3D flash runs into issues when stacked at above 64 layers. I am not saying those hurdles cannot be overcome, but what if all this is much more "easy" and "straightforward" to do with 3D Xpoint (hence the name) than 3D flash? What if, at a certain point, 3D Xpoint scales a lot better in the third dimension than 3D flash? This could mean, in the far future, that this technology may actually become cheaper than flash and hence replacing it.

The problem I have with this, already since the initial announcement of 3D Xpoint, is/was that we are simply not told. Neither Intel nor Micron actually disclose their plans to shareholders in this regard and therefore you simply can't know. The other thing is: Plans are plans and reality often is different. In addition, it doesn't necessarily mean that Intel and Micron will actually benefit from this move. Just have a look at who invented flash memory. That was Toshiba, but it didn't prevent them from having serious financial issues (though, not related to its memory business, to be fair). Inventing a technology and dominating it are two very different things. More often than not, the actual inventor is left with empty pockets.

Still, I wouldn't expect Intel and Micron to have gone through all this without having a serious application in mind. Database acceleration it is definitely not.

There have been many promising memory technologies in the past (PCM, FRAM, MRAM etc.) that all were not taken anywhere as far as 3D Xpoint by any big memory company. That in itself is telling you something, I guess. Still doesn't mean they are going to succeed, but it tells you they both were thinking it was worth a shot.

[DavidA2]

3D XPoint is an intrinsically better device than NAND for caching. Look at the first graph of this review:

http://www.tomshardware.com/reviews/intel-optane-3d-xpoint-p4800x,5030-5.html

Ever since NAND SSDs, there's this constant turmoil with "steady-state" and "clean vs dirty" performance. And these are the best-in-class SSDs. The graph illustrates that's simply not true with Optane.

Caching devices for SSHD and SRT is worse than the good SSDs. You know how features like TRIM were so desired by people with SSDs? Because without it the performance tanks. That's ok, if you have it as a main drive. But I'd have to guess doing all that cleaning work wreaks havoc on the caching algorithms.

Here's 1st run tests I've found:

http://www.digitaltrends.com/hard-drive-reviews/intel-optane-32gb-review/
http://hothardware.com/reviews/intel-optane-memory-with-3d-xpoint-review-and-performance?page=3
https://www.gamespot.com/articles/intel-optane-memory-review/1100-6449530/

That's pretty decent. I don't see the Optane SSDs as compelling as enthusiasts make it out to be because the absolute costs are too high.

The only ones I see making any impact is Optane Memory, and the eventual DIMM versions. The reality in terms of adoption to make it realistic for consumers, and revenue for Intel is just those two. The DIMM version and the SSDs are too expensive for anyone other than those buying HEDT setups. While Optane Memory is expensive per GB, its cheap in an absolute sense, and it fills a void that will exist for, another decade if not more.

Best Buy and Costco already sells systems with Optane Memory, and you'll see a more massive rollout throughout the year. On Costco, same system with one having 16GB Optane + 2TB versus 120GB SSD + 1TB is $100 cheaper. The volume impact and practical impact to consumers will be there.