Intel overhyping flash-killer XPoint? Shocked, we're totally shocked XPoint will substantially undershoot the 1,000-times-faster and 1,000-times-longer-lived-than-flash claims made by Intel when it was first announced – with just a 10-times speed boost and 2.5-times longer endurance in reality. This became clear when Micron presented details of its QuantX-branded XPoint memory technology at the …
For many consumer and business applications the raw speed of flash is not the limiting factor so increasing the hardware speed will not gain anything. This is a common fallacy when looking at quoted speeds of ignoring what is the actual limiting factor which may be something as mundane as the user's typing speed.
The fact that a given memory technology is ten times faster than another is certainly not limited by the user's typing speed. It will just store each keypress ten times faster. Of course, the user will not notice any difference whatsoever.
Where the user will see a difference is when the user transfers a large block of data (or a large quantity of small blocks) to that memory. That is where the speed will be significant and where the notion of limiting factors will come into account, such as CPU power, number of threads allotted to the task and available bandwidth between the memory modules and the source/destination of the data.
If you're just waiting on user keyboard operations, a good ol' PC XT could do the job nicely.
You're confusing servers with workstations. Workstations spend most of their lives waiting for users to tell them what to do. Servers can get thousands of requests per second, in which case the time it takes to transfer the requested data from disk to memory is important.
In case you hadn't figured, this is a server class product (read expensive).
Not really. First, I think QuantX is shared business with Intel (i.e. not in competition), secondly we are talking about generation 1 of the device. As long as HP and others delay delivery of alternative memory technologies, XPoint has a fighting chance. Still, at this moment it is far from impressive.
Specs on DC P3700, table 7 on page 11, state sequential read/write latency 20us and random read/write latency 115us/25us (yes, writes are faster - ideal for ZFS SLOG device). This is far from 200us used for comparison in the article. Yes, DC P3700 is expensive, but I suspect that comparable capacity QuantX would be more expensive anyway.
Also if I read the article right, that latency for a PCIe interface to the memory. PCIe latency penalties are a minimum of 200ns per packet - and packet size is typically quite small, typically no more than 1KB. This and other communication overheads mean that the quoted 20ns is not too shabby - for a PCIe interface.
Also worth bearing in mind is that being a 1st gen device the controller is probably implemented in an FPGA, so potentially a clock speed of 200-400 MHz - with a heavily pipelined architecture.
So what we are saying is that 3D X-Point, instead of being a pants-shaking revelation, might instead just be Quite Good.
Quite Good is nothing to sneeze at, when most things are, ipso facto, Fairly Average. If you can do nothing but reliably distinguish those two categories, you are of more use than all the marketing people and telephone sanitizers on the planet, collectively.
Rather than expound further, I would refer you to the Blackadder episode that exists to explain this phenomenon with uncommon precision. But I don't know which one it is.
""Quite Good is nothing to sneeze at, when most things are, ipso facto, Fairly Average. "
The question is "Will you pay ten times more for it?" and the answer in 90% of cases is "NO""
Agreeing violently !
They've done pretty well already with Gen.1, and they are at the start of XPoint's development curve so there is likely a lot of room for improvement on price and performance, I think there is reason to be optimistic - particularly if other big vendors license it. On the other hand Flash has had a couple of decades of competitive development invested in it, there is much less margin for improvement with Flash and much thinner margins.
Although it's not matching up to the initial claim of being akin to slow, non volatile memory, this is exceptional.
GPUs? The GTX1080 averages 20-30% faster than a 980Ti, and this is seem as exceptional.
CPUs? Using Intel's own (optimistic) numbers, a Skylake 6700K is '30% faster than a three year old PC' (i7 3770K)
20-30% is seem as decent, and the article complains about a technology averaging 5x to 10x improvement, depending on area?
When you were promised 100000%, and think that a 5x price premium isn't worth it for most market segments - a further 10x reduction in the already massive 10000x latency reduction SSDs have delivered over hard drives won't make much difference for the large majority of applications. In many cases that extra money would be better spent on more RAM, faster CPUs, better networking, etc.
In fact, in many cases even NVMe is overall, and SATA SSDs are more than good enough.
Let's start with a very simple and basic block diagram:
CPU -----> chipset -----> storage subsystem (i.e.3D XPoint).
Try to visualize the CPU as a radio frequency transmitter:
4 cores x 64-bits per register @ 4 GHz is a lot of binary data
On the right is 3D XPoint.
As their measurements show,
Micron achieved "900" w/ PCIe 3.0 x4 lanes; and,
Micron achieved "1800" w/ PCIe 3.0 x8 lanes.
Read: almost perfect scaling.
And, the flat lines speak volumes:
in both cases, the storage subsystem
saturated the PCIe 3.0 bus.
Now, extrapolate to PCIe 3.0 x16 lanes:
wanna bet "3600"? My money says, "YES!"
Now, extrapolate to PCIe 4.0 x16 lanes:
my money says ~ "7200" -- flat line
(maybe not perfect scaling,
but you get the idea :)
Conclusion: 3D XPoint is FAAAST, and
Micron's measurements show that
the chipset is now the bottleneck --
all cynicism aside.
In other news ... Large corporation produces marketing material that highlights the absolute best attributes if its new product..
It can address a single bit 1000X faster than NAND Flash .... get over it and stop bitching.
Anandtech don't seem to have a problem with "1000X faster".
http://www.anandtech.com/show/9470/intel-and-micron-announce-3d-xpoint-nonvolatile-memory-technology-1000x-higher-performance-endurance-than-nand
You've been churning out this same crap for months and we're all tired of it.The only question is "What motivates Chris Mellor"??
http://www.theregister.co.uk/2016/04/15/intel_xpoint_emperor_has_no_clothes/
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
