http://www.intel.com/content/www/us/en/solid-state-drives/optane-solid-state-drives-dc-p4800x-series.html
http://www.anandtech.com/show/11208/intel-introduces-optane-ssd-dc-p4800x-with-3d-xpoint-memory
THANK YOU BASED INTEL
3 YEARS LATE
Expensive as fuck but it will only go down with time
That's nice but PCIe cards are shit, give me a U2/M2 one
why don't they write actual numbers though? why does it have to be xN faster?
>starts off with 375GB
Lol
>>59481767
oy goy.
it's faster.
that is all you need to know
>>59481148
this kills the ryzen right?
Intel Optane SSD DC P4800X has a write endurance rating of 30 Drive Writes Per Day,
>at 375GB
Ahaha, that's pathetic.
>>59481818
Heh it's not a graphics card dummy
>>59481831
Or a co-processor or whatever. Those Xeon Phi computer cards Intel has, it's not that.
>>59481148
weren't they supposed to be *literally* thousand times faster than regular SSDs?
what happened?
>>59481825
The 375GB drive consumes up to 15w~, get ready for absolute fucking housefire temps and power figures with their higher capacity drives.
This is nowhere near the endurance nor performance of DRAM to be useful as system memory.
>>59481915
are they selling them as it?
>>59481859
Intel massively revised their own marketing figures
http://semiaccurate.com/2016/09/12/intels-xpoint-pretty-much-broken/
>>59481915
That was known from the getgo. Its not a DRAM competitor.
>>59481937
Yes.
>Intel Memory Drive Technology
>Along with the DC P4800X, Intel is launching Intel Memory Drive Technology software as a paid add-on. The Memory Drive Technology software is a custom virtual memory system implemented as a minimal hypervisor. The guest operating system or general-purpose hypervisor is presented with the illusion that a portion of the 3D XPoint memory on the Optane SSD is directly addressable,
And they're asking for $2000 for the software for it too.
>>59481947
>Intel massively revised their own marketing figures
oh lol, it's Itanic all over again
>>59481947
Well it's still an improvement over NAND flash.
>>59481947
oh wow
>>59481958
I can't wait for GNU Memory Drive in 3 months
Is this the thread where consumerist whores shitpost about technology they'll never be able to afford because a company they like made it so obviously the competition is finished and bankrupt?
>>59482019
>then
Now:
2-4x faster
Actually less endurance than SLC drives lol
Density is possibly the only thing they managed to stick to, but even that's questionable with 256GB DIMMs around the corner
>>59481148
>30 Daily writes per day
What did they mean with this?
No really, what is this supposed to mean?
>>59482637
can write the entire drive capacity 30x per day, so for the 375gb model you could write 10tb to it per day for whatever the lifespan is meant to be
>>59481818
No, beacuse Ryzen commited suicide.
>>59482870
This this this this this
AMD is crap!
>>59481148
>Durability: 1000x faster (broken after a month of heavy use).
That's why you should never be an early adopter.
>>59482051
Here's the thing. If you are nowhere near RAM speeds, the memory might be great and scale well but it would be nowhere near a breakthrough and computers would work the same with some new protocols as what NAND SSDs did. The only memory that competes on DRAM speeds right now that have some chance of sucess that is shipping right now is magnetoresistive RAM (MRAM), ferroelectric RAM (FRAM), or non-volatile SRAM (nvSRAM).
Right off the bat, we can cross out nvSRAM, which is filling a niche but can't scale anywhere near the density required for storage and takes up even more space than regular SRAM.
So this is primarily a battle between MRAM and FRAM now, but we're going to discount a ton of in progress memory that hasn't shipped yet like racetrack memory or the memristor because time to market is a pretty huge hurdle that wouldn't allow any of the in progress memory type to gain on MRAM or FRAM, and the fact that they don't offer any huge advantage over them.
FRAM looks great where you can adapt DRAM technology and be non-volatile as well using ferroelectric instead of a dielectric material. The main problem is the fact that it has a destructive read process, where you have to write after reading to retain the memory, and that increases latency and power requirements, which is still not a lot but does affect performance. Scaling is around in the 1-10 Mbit range since there seems to be a scaling limit on current implementations.
MRAM by comparison works almost the same way current hard drives do. Flipping the magnetization of the material to a 0 or 1 but with the exception that this uses solid state materials to achieve that. Latency is faster than all but SRAM speeds and read/write endurance is unlimited because magnets. The main downside is the density scaling where it can only scale to the 1-10 GB range at the moment with one company shipping that amount where they were stuck like FRAM unable to scale but it should scale up pretty nicely.
Based what? It is much faster on latency, so they must be extremely responsive, and they're very durable, but they're late and expensive. They're not only late, their previous claims turned out false
When do we get the micron variant?