3D XPoint Debuts, Intel Announces Optane SSD DC P4800X And Pricing

Status
Not open for further replies.

Elixon

Prominent
Mar 19, 2017
1
0
510
This is fake news. How could the author forget to mention that Intel/Micron promised "Game-Changing 3D XPoint Memory, 1000x Faster Than NAND"? You lay down "10x faster" fact in front of readers without even checking what it was supposed to be and what Intel realy delivered?

That is unprofessional and I smell the ad revenue from Intel.
 

memadmax

Distinguished
Mar 25, 2011
2,492
0
19,960
SCREW YOU INTEL AND MS!

Optane/Xpoint are gonna be "Win10 Only"

You feel the noose getting tighter yet? You have been warned, again(and many times before!)...

They might as well rename it to: "MicroApple"....
 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
858
315
19,360

We've already covered the disparity between the initial speed claims, which apply to the media only, and not the device level performance.

Here is some fairly critical coverage of that, which we linked to in this piece.

See the sections named "original claims" and "walking it back":

http://www.tomshardware.com/reviews/3d-xpoint-guide,4747.html


 

Tanyac

Reputable
Unless my maths is faulty isn't this about $4.00 per gigabyte? Will enterprises really be willing to pay that much for space? 375GB is a small drive so it's not a good choice for large storage applications. The performance doesn't seem that groundbreaking.

In AUD that thing will probably cost $2200, about $5.85 AUD per gigabyte. That's got to be hard to justify, surely?

I must admit I'm no guru when it comes to comparing SSD performance, so how does this compare to say, the Samsung 960 Pro, which is about $0.80 AUD per gigabyte for the 1TB drive?
 
G

Guest

Guest
Ok, this is impressive. After a disappointing start, the "word on the street" was the first cards would be 16 GB, 32 and 64. Now it's 375 GB from the get go. It's still not enough, but it's a good start. And the price point is fine for now. They will sell a lot of these.
 

ssdpro

Honorable
Apr 10, 2013
162
0
10,680

It isn't impressive at all. It is not a consumer product and Intel said there would be a consumer product on the way. Those are still slated to be 16-64GB and will be expensive still.

And as for data centers, yes they will pay for this. Intel has likely already sampled to MS, Netapp, Amazon, Brocade, etc and has buyers in place for the tech.

 

alextheblue

Distinguished

I really hope they'll be affordable enough to at least use as a boot drive/cache in a couple of years. For a cache drive in particular it wouldn't need to be that large. If they're fast enough it might be worth it to run something like a ~100GB cache in tandem with a larger conventional SSD drive, for example.


Today this is useful for professional/business systems. In the near future, it (or other post-NAND storage tech) will be available in common consumer form factors at a more affordable price. So making a statement like that just seems a bit silly on a tech site.
 

CRO5513Y

Expert
Ambassador


^ This, in a couple of years should make for good boot and priority app/game drives. Just like SSDs it will take time to become affordable for consumers but that's just how it is not much we can do about it but whinge at Intel and storage manufactures lol. Additionally, people still complain about Windows 10 exclusivity? Just get used to it they aren't going to just stop with DirectX 12 and Kaby Lake. Should have rolled with the free update when it was out. Windows 10 has been the most convenient and fast Windows OS for me since 7 and before that XP. Anyone on a modern gaming PC should have Windows 10 unless you dont want modern hardware support, better drivers, faster boot and DirectX 12 support among other things... Just for reference i have moved to every new Windows since 2000 and although some have been rough (Vista and Windows 8 for sure) 10 has been exceptional so far, too many people on the "Windows 10 spies on you" bandwagon and if you think Windows 10 spies on you then why are you on the internet to begin with... your ISP is spying on you, your phone manufacturer is spying on you, CIA is spying on you.... list could go on so not much point squalling at Microsoft for being a business trying to make money and nothing wrong with ending support for a OS that is now over 7 years old. :p
 

DavidC1

Distinguished
May 18, 2006
493
67
18,860
/Optane/Xpoint are gonna be "Win10 Only"

No, that's for Optane Memory branded devices, which are caching setups. If you just want to use it as a tiny 16GB SSD you'll be able to do so.

/After a disappointing start, the "word on the street" was the first cards would be 16 GB, 32 and 64.

Those aren't really "cards" or SSDs. Those are caching devices. Several articles regarding P4800X state consumer equivalents are coming later.
 

bit_user

Polypheme
Ambassador
The 12-14W active power consumption is also within expectations for a normal NAND SSD.
Um, close to normal for enterprise PCIe SSDs, that is. Performance per watt is definitely in the normal range, if not a bit better than average.

I'm definitely impressed at the low queue depth performance. That's where it counts. I can see myself with one of these, in a few years (once the price comes down to Earth).

I don't like the transparent swapping layer, though. It really belongs in the OS, but I think Intel just didn't want to be stuck with products on the market and still lacking or buggy OS support. I'd expect going through the OS will be a bit slower, in some cases, but more consistent over all.
 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
858
315
19,360


I guess I could have been more specific. For PCIe SSDs its within range of most of them, though the smaller capacity muddies the waters a bit. DC P3700 400GB is 12/9 W typ, Micron MAX 800GB is 7-16. Some, like the Mangstor, are up to 40W, albeit with much higher capacity. I guess we would have to figure up a Watts-per-addressable-GB metric to even the field. I don't think it sticks out as unreasonable, though.
The IOPS-per-Watt measurements would be pretty good, but where it would really shine in the transactions-per-watts metrics. The low-QD performance just murders everything out there, so applications will record a much higher delta than we see on the surface.
 
Despite the current enterprise pricing, this will be a huge benefit to consumer workstations especially (well, as soon as it gets cheap) due to it's very large advantages at low queue depths (QDs greater than 4 are are rarely if ever encountered when loading Windows and apps or games). 5-8 times the performance at low QDs will be VERY noticeable in day to day tasks.
 

John Fak

Reputable
May 28, 2014
6
0
4,510
Intel "launches" again 3D Xpoint. For the 5th time in the last 2 years. Pahleez!
Total silence about sequential is because it sucks balls.
500k IOPS is nice but not revolutionary, its barely better than other existing solutions. The probably 1.5 GB/sec sequential will be horrible.

Just fire the guy behind it and move on. This is clearly a failure.
How dare you compare it to X25!!!
 

DavidC1

Distinguished
May 18, 2006
493
67
18,860
I have doubts if you are the most qualified person to criticize the product because,

"500k IOPS"
"1.5 GB/sec sequential"

500K IOPS indicates 2GB/s sequential. If you were actually paying attention to 3D XPoint news you'll realize this particular product had a leak and it does 2.4GB/s read and 2GB/s write sequential throughput.

But sequential isn't the key point here. Whatever they promised with this NVMe product Intel delivered. News reports say the Datacenter guys are excited about this product.
 

computerguy72

Distinguished
Sep 22, 2011
190
1
18,690
Wow those performance numbers are great for the first purchasable device. I remember people hammering hard disks compared to floppy's saying HD's would never take off as they had similar claims as this. Optane is very much like that and you can already see just what an epic leap this is over NAND. Almost infinite endurance, blazing speed at low queue depth. A lot of people don't seem to realize that for many SSD's you often don't experience anywhere near their best theoretical performance. Optane actually moves the needle in it's field a great deal and could impact all PC's in fundamental ways. Up to Intel and Micron to handle this opportunity well.
 

IndignantSkeptic

Distinguished
Apr 19, 2011
507
0
18,980
I have an interesting question about endurance. Do these devices stop writing data when they measure that they have actually failed to do so, or when some guy, ages ago, predicted that they would probably fail to do so?
 

Discorama

Commendable
Aug 25, 2016
7
0
1,510
"Surprisingly, Intel still isn't revealing exactly what 3D XPoint is..."

OMG, it's crossbar memory. I can't believe no one has figured this out. What do you think the "X" in XPoint stands for? Fricken HP failed to bring it to market like they promised 3-4 years ago and now Intel has beaten them to their own punch. This would have been the one tech that could have saved HP too. What a shame...
 

bit_user

Polypheme
Ambassador
Well, it's also nonvolatile. So, how is a bit stored?
 

Discorama

Commendable
Aug 25, 2016
7
0
1,510
What?!
Nonvolatile means it can store memory.
"Crossbar’s patented built-in select feature allows thousands of Crossbar’s ReRAM cells to be inter-connected in true cross-point memory arrays. This extremely dense memory arrays with the capability to scale below 10nm, storing multiple bits per cell and stacking 3D layers provides a path towards TeraBytes on a single die.
The erase-free architecture with small page granularity that can be re-programmed without a block erase provide impressive performance boost over Flash-based Solid-State solutions. The reduced complexity of the storage controller and reduced number of background memory accesses enable a new era of Solid-State data storage solutions."
Performance sounds similar to what Intel was originally touting?
"delivers 100x lower read latency and 20x faster write performance compared to NAND Flash..."
Intel probably had to side step some patents and came up a little short on the performance side. Of course Crossbar and HP have been unable to market so might just be some inherent issues with the actual implementation limiting performance.
 

bit_user

Polypheme
Ambassador
You used lowercase "crossbar", which I thought was referring just to the access topology, or something like that.

Okay, thanks. Now it's clear to me that you were referring to a company called Crossbar. That still doesn't tell me how it retains information (the complaint you voiced about Intel), but at least I have something to go on.
 
Status
Not open for further replies.