"Xpoint pretty much broken"

[h4rm0ny]

That's semiaccurate's take on it. I'm really not sure what to make of it as I find it hard to believe it could be this far off expectations. On the other hand, semiaccurate are pretty much the anti-Fudzilla: a professional high-level site with a narrow focus on this sector. They're traditionally very good with their news.

http://semiaccurate.com/2016/09/12/intels-xpoint-pretty-much-broken/

I could / should have put this in the memory forum but realistically hardly anyone will see it there and more legitimately, one of the major selling points of Kabylake is its support for Optane / Xpoint. If that's not as ground-breaking as expected, that puts Kabylake back to being more of just a Skylake bump.

I honestly don't know enough about the figures to say how much value Xpoint retains if this article is accurate, but the article seems to be pretty damning on the subject.

 

[Yutang13]

That's semiaccurate's take on it. I'm really not sure what to make of it as I find it hard to believe it could be this far off expectations. On the other hand, semiaccurate are pretty much the anti-Fudzilla: a professional high-level site with a narrow focus on this sector. They're traditionally very good with their news.

http://semiaccurate.com/2016/09/12/intels-xpoint-pretty-much-broken/

I could / should have put this in the memory forum but realistically hardly anyone will see it there and more legitimately, one of the major selling points of Kabylake is its support for Optane / Xpoint. If that's not as ground-breaking as expected, that puts Kabylake back to being more of just a Skylake bump.

I honestly don't know enough about the figures to say how much value Xpoint retains if this article is accurate, but the article seems to be pretty damning on the subject.

Wow, how very unimpressive. I thought "1000x" was unachieveable at any rate. How many times have we gotten a 1000x jump in performance? Or even a 10x jump?

Just another product from Intel- the I stands for iterative

 

[drunkenmaster]

The claims have dropped off massively but with new techs it's perfectly possible to get massive changes.

4kb random reads going from hdd to ssd did indeed go up a several magnitudes. Good hdds at the time were doing 0.6MB/s in random 4kb reads where the first ssds did 15MB/s or so in single queue depth 4kb random reads. With higher queue depths HDDs didn't improve at all where SSDs could improve significantly. Latency certainly dropped a couple of magnitudes from 8ms to 0.08ms.

Iterations of the same types of tech rarely achieve anything over say 20-30% gains sometimes at best, entirely new technologies are often completely game changing.

The most important thing I think Charlie noted wasn't so much Intel's dropping of specs as much as Micron who make the chips haven't announced a single product to be available this year while Intel are saying they'll have several things done. Probably hints that Intel will paper launch some stuff with no general availability at all and next year(or maybe even later) when Micron actually announced products is when we'll see it get real availability.

 

[Rroff]

Charlie is often pretty down on Intel and nVidia - not the first time he has posted how something is "simply broken", late, unfixable, hot, etc. and turned out its just a few teething issues and nothing that isn't unfixable or its just a little less than optimal but still mostly ok.

 

[AllBodies]

I mean, maybe it is theoretical max vs first gen products? Not that I'm trying to defend them.

Also it only really needs to be 10x faster than NAND to do its job, as long as that NAND is Samsung's 3D vNAND in their M.2 drives.

If you could make something 10x faster, with 3x the endurance, of Samsung's fastest Polaris drive, you could absolutely use that as a non-volatile replacement for RAM.

And that's the true point/job of this new memory in the long run. It doesn't matter too much how fast it goes at the top-end. What matter is if it's fast enough to replace DDR4 memory for CPUs.

 

[Concrete]

Why do I get the horrible feeling that someone from Intel went "Hey guys, you didn't get the memo? Drip feed is the name of the game now, look at our CPU division!"

 

[h4rm0ny]

EDIT: Misread the reported latency on the XPoint memory. Post therefore wrong.

 

[Rroff]

IIRC the latency of RAM is in ns and considerably less than 0.013ms. I think he is saying 10x SSD speed which is around DDR4 speeds if we are talking the very top end fastest SSD stuff or about 25-50% of it if we were talking run of the mill SSD.

 

[h4rm0ny]

IIRC the latency of RAM is in ns and considerably less than 0.013ms. I think he is saying 10x SSD speed which is around DDR4 speeds if we are talking the very top end fastest SSD stuff or about 25-50% of it if we were talking run of the mill SSD.

Yes, I misread their post and have since edited out what I wrote. Too slowly for you, it seems!

The article says Xpoint is now looking at latencies of around 10x SSD (I'm going to kindly assume they mean 1/10th and guess marketing people put the slides together rather than engineers ). If we generously take something very fast like the PCI-E x4 SSDs from Samsung, they have latencies in the order of 0.3ms. DDR4 2400 has something in the order of 13ns. I.e. four orders of magnitude difference. Meaning whilst this stuff is interesting as a fast intermediary layer between SSD and RAM (or a replacement for SSD if close in cost which I don't expect it to be), it's problematic as a drop in replacement for RAM. It can hit the bandwidth, but not the latency.

So I could see this stuff being useful for where you want large contiguous blocks of data read - maybe textures on a graphics card for example - but the question is how important is it to have non-volatile memory? Where does this help?

 

[Rroff]

^^ I'm taking his 10x performance to mean bandwidth not latency (which puts the transfer rates potentially into RAM ballpark though more last gen than current gen), latency claims are between 3x and 10x which is considerably short of RAM even assuming a bleeding edge SSD.

EDIT: Be careful with comparing RAM "latency" with SSD access latency as on RAM it is often rated in clock ticks not ns as it first appears and you'd have to do further calculations to make an equivalent comparison between the two - suffice to say RAM has a significant advantage that would need more than 10x SSD capabilities to catch up.

 

[h4rm0ny]

^^ I'm taking his 10x performance to mean bandwidth not latency (which puts the transfer rates potentially into RAM ballpark though more last gen than current gen), latency claims are between 3x and 10x which is considerably short of RAM even assuming a bleeding edge SSD.

Yes, that's what I misread initially before I deleted my earlier post. They were referring to bandwidth and they're right. XPoint at these figures (10x SSD) can match low-end DDR4 RAM assuming we're talking 10x a fast PCI-Ex4 SSD. But latency is a very important factor.

EDIT: Again, I think what I'm posting is correct but am open to someone with more subject knowledge catching it if I've made an error.

 

[DragonQ]

So basically they're going to be next-gen SSDs rather than any kind of RAM replacement. Just as everyone assumed, right?

 

[Rroff]

Yes, that's what I misread initially before I deleted my earlier post. They were referring to bandwidth and they're right. XPoint at these figures (10x SSD) can match low-end DDR4 RAM assuming we're talking 10x a fast PCI-Ex4 SSD. But latency is a very important factor.

EDIT: Again, I think what I'm posting is correct but am open to someone with more subject knowledge catching it if I've made an error.

Your first post was on the right lines, just using the 3x figure when comparing bandwidth rather than 10x as he was suggesting. Latency wise you are correct as far as it makes any odds for a general comparison - I suspect it is a bit more complex for an equivalent comparison though as the controllers for each are doing a little bit different job and not necessarily optimised for like for like comparisons.

 

[AllBodies]

Intel did release a graph recently suggesting a lot of the latency in their first gen Xpoint stuff comes from the PCI and/or SATA interface, and also the way the CPU communicates with it, since it sees the Xpoint as 'a hard drive'.

They claimed if the Xpoint was put into RAM slots, in a RAM formfactor, and the CPU was made to communicate with it in the most efficient manner, the latency would drop considerably.

Also bear in mind I suppose we should be comparing this with the first gen SSDs, to get an idea of how much room for improvement there may be across generations. I don't want to overhype it, if it does just turn out to be 'very fast SSD', but on paper it looks to have the potential to be a more fundamental advancement. Even if that improvement comes a couple of years down the line, in the 3rd generation or something, rather than the first products in 6 months time.


EDIT: Found it -> http://www1.pcmag.com/media/images/...m-interconnect-idf-2016.jpg?thumb=y&width=480

From this article: http://forwardthinking.pcmag.com/ch...y-are-the-present-and-future-of-memory-at-idf

 

[Perfect_Chaos]

So basically they're going to be next-gen SSDs rather than any kind of RAM replacement. Just as everyone assumed, right?

From what i have read a while back, i already figured Xpoint was a much more exciting SSD replacement rather than a RAM replacement.

Xpoint is somewhere between SSD and RAM, speed wise, and apparently cost wise it will be cheaper than regular RAM.

So it's no replacement for RAM in performance terms, but could be a big thing for SSD's if they work as well as they were hyped up to be i guess.

 

[Orangey]

Normally when a big company is pushing this hard to get something out before it's ready it means there's an open competitor either yet to break cover or being birthed by a group of rivals.

 

[AllBodies]

Normally when a big company is pushing this hard to get something out before it's ready it means there's an open competitor either yet to break cover or being birthed by a group of rivals.

Could be whatever the 'Next Gen' memory is that AMD mentioned with be on their Navi GPUs?

Or it could just be they want to get out the door before Micron, since it was a joint development.

 

[Burr]

That's semiaccurate's take on it. I'm really not sure what to make of it as I find it hard to believe it could be this far off expectations. On the other hand, semiaccurate are pretty much the anti-Fudzilla: a professional high-level site with a narrow focus on this sector. They're traditionally very good with their news.

http://semiaccurate.com/2016/09/12/intels-xpoint-pretty-much-broken/

I could / should have put this in the memory forum but realistically hardly anyone will see it there and more legitimately, one of the major selling points of Kabylake is its support for Optane / Xpoint. If that's not as ground-breaking as expected, that puts Kabylake back to being more of just a Skylake bump.

I honestly don't know enough about the figures to say how much value Xpoint retains if this article is accurate, but the article seems to be pretty damning on the subject.

I read this not long ago myself, pretty damning stuff. Of course a lot could happen between now and it's release but if true, well... Oh dear. This part: "one of the major selling points of Kabylake is its support for Optane / Xpoint". I have a feeling 'support' for Kaby Lake amounts to having x4 PCI-E lanes dedicated to a U.2 connector from the PCH - nothing exotic really.

 

[h4rm0ny]

Could be whatever the 'Next Gen' memory is that AMD mentioned with be on their Navi GPUs?

Or it could just be they want to get out the door before Micron, since it was a joint development.

Well Navi will have HBM and more significantly, the architecture is designed for HBM rather than attached to an older architecture as with the Fury. If you're making an architecture with much higher memory bandwidth, you design it differently. HBM and Xpoint are not the same thing, but they both have applications in the GPU area. So maybe XPoint was intended to be a "we have a big new thing too" idea.

I read this not long ago myself, pretty damning stuff. Of course a lot could happen between now and it's release but if true, well... Oh dear. This part: "one of the major selling points of Kabylake is its support for Optane / Xpoint". I have a feeling 'support' for Kaby Lake amounts to having x4 PCI-E lanes dedicated to a U.2 connector from the PCH - nothing exotic really.

If that's all it is and XPoint does "just" turn into SSD 2.0 then maybe we can see support for it on AMD platforms as well. Presumably Intel can block that legally, though?

 

[Burr]

If that's all it is and XPoint does "just" turn into SSD 2.0 then maybe we can see support for it on AMD platforms as well. Presumably Intel can block that legally, though?

Hard to say with any real confidence but I doubt Intel would try to lock AMD out of using XPoint. Inevitably Samsung + others will offer competing tech. Intel locking out AMD just gives them a big leg up on the AMD platform. Of course i'm sure the idea has gone through Intel's head.