Intel to Shut Down Optane Memory Business, Retire 3D XPoint Memory

[Chrispy_]

I follow a lot of tech news, and I could never figure out what Optane was supposed to do, I just knew it was stupid expensive.

It was supposed to bridge the performance gap between SSD and RAM.

The problem with that is that you also need your application to be designed around this SSD-RAM hybrid layer to see real benefit, otherwise it's just like running a faster SSD; SSD's are already fast enough that they're no longer the bottleneck in load times and application performance. And here's the real kicker; If you're re-writing your application to run on this hybrid layer, it's a very costly process that requires both time and money that would be better spent just throwing more RAM in the systems in the first place.

 

[Tomorrow]

Not only cost, but
P900 series, idle power 6W, active power 16,4W. Thank you - no.

So like upcoming PCIe 5.0 SSD's then?

 

[Crackong]

Good product with a worse than stupidity price tag

 

[Airisom]

I might be crazy, but it's kind of a happy coincidence that this news is released shortly after their poor earnings report and after they announced fabbing mediatek chips.

 

[80251]

I have windows installed on my 260GB Optane SSD, and that's the best part of my current rig.
It will be missed.


"Consumers" have spent way more for less useful things, especially in the last 2 years.

Telstar do Optane SSD's use any overprovisioning? Or can they achieve those huge TBW figures without it?

 

[Wirko]

Telstar do Optane SSD's use any overprovisioning? Or can they achieve those huge TBW figures without it?

Intel says they don't need it.

 

[ReallyBigMistake]

I just hope Intel doesn't bail on discrete gaming GPUs.

 

[RedBear]

Well looking at their financials I won't be surprised if they have to spin off their fabs as well, that chips subsidy can't come soon enough!

According to Moore's Law is Dead, first they might end up shutting down ARC, so we can return to the good old dupoly that we've learned to love and appreciate. Eventually I suppose they might go down IBM's way, if they keep it up.

 

[80251]

Aside from the amazing petabytes written figures for Optane SSD's do they outperform Samsung's best in IOP's, sequential writes and sequential reads?

 

[Tomorrow]

Aside from the amazing petabytes written figures for Optane SSD's do they outperform Samsung's best in IOP's, sequential writes and sequential reads?

Yes. Both use Gen4 so sequentials are capped at around 7GB/s. Where Optane crushes regular NAND is in latency and randoms where it has 1-3x advantage.

 

[80251]

Considering what Tomorrow wrote it is sad that Optane isn't going to be developed further.

 

[Wirko]

Considering what Tomorrow wrote it is sad that Optane isn't going to be developed further.

Will find its place at the cemetery right beside SED. It could have been the greatest high-end display technology, probably not to this day but at least for some time.

 

[bug]

Yes. Both use Gen4 so sequentials are capped at around 7GB/s. Where Optane crushes regular NAND is in latency and randoms where it has 1-3x advantage.

Optane also had better endurance, which is nothing to sneeze at.

 

[regs]

They should've tried to license it to SK Hynix

They actually sold 3DXPoint manucaturing plant to SK hynix. Wonder if hynix can now manufacture own 3DXPoint now.

 

[kapone32]

In consumer space Optane as a solution looking for problem to solve.

I still wish has bought one when they were expensive but not stupid.

They actually sold 3DXPoint manucaturing plant to SK hynix. Wonder if hynix can now manufacture own 3DXPoint now.

Well not really anything to do with it but the SK Hynix top end NVME is much cheaper than it's peers for the same basic performance.

 

[Count von Schwalbe]

They actually sold 3DXPoint manucaturing plant to SK hynix. Wonder if hynix can now manufacture own 3DXPoint now.

SSD yes. Optane no.

I used an Optane H20. I also am probably the only person to get the Optane side of it usable on a B550 motherboard - by accident.
Apparently the QLC side was the second set of lanes, the Optane side was the first. The M.2 slot did not support bifurcation. I was planning on using it as a 512gb SSD only, as a boot drive, and I could only see the Optane side - at 32gb.

Anyways, I replaced it and it is now in my laptop on which I type this, as an SSD only as it is not compatible with the chipset! Almost as bad as Apple's NAND only SSD.

Honestly, though, the only real advantages of Optane are random WRITES and latency, at a higher cost. Tom's Hardware did a test comparing it to an enterprise-grade Samsung SLC drive, and the Samsung had better random reads and better sequentials. It was also cheaper.

This is in contrast to the NAND vs HDD comparison - NAND was so much better that it didn't go away, but was iterated upon until it became cheap.

There is a reason people buy TLC or even QLC instead of MLC and SLC drives. People don't really need the performance and endurance of the better drives.

 

[Tomorrow]

Will find its place at the cemetery right beside SED. It could have been the greatest high-end display technology, probably not to this day but at least for some time.

Wow. First time hearing about this. Reading further it looks interesting combining strengths of CRT's and LCD's.

Also supposedly there is XL-FLASH that is fancy name for SLC based SCM (Storage Class Memory) enterprise drives.

Dapustor X2900P SCM SSD Review

The Dapustor X2900P is the company’s newest storage-class memory (SCM) data center SSD with PCIe Gen4, DPU600 controller and KIOXIA XL-Flash.

www.storagereview.com

Kioxia FL6 SCM SSD Released

The Kioxia FL6 SCM SSD is the company's first SSD to use XL-Flash. It will be available in 800GB-3200GB capacities.

www.storagereview.com

Unfortunately availability is even more nonexistant that Optane itself.
So for now it seems the only way is to snag one of the U.2 Optane variant or by some miracle and SLC drive from Toshiba etc.

 

[TheinsanegamerN]

According to Moore's Law is Dead, first they might end up shutting down ARC, so we can return to the good old dupoly that we've learned to love and appreciate. Eventually I suppose they might go down IBM's way, if they keep it up.

If they keep losing money their investors are going to insist on trimming the fat sooner or later, and given that ARC is barely competitive with pascal GPUs....well......

 

[80251]

Would it make sense to use an optane SSD as a cache for slower SATA SSD's?

 

[bug]

Would it make sense to use an optane SSD as a cache for slower SATA SSD's?

Intel certainly seems to think so. It just depends on the hoops you have to jump through, I guess.

 

[Wirko]

It was supposed to bridge the performance gap between SSD and RAM.

The problem with that is that you also need your application to be designed around this SSD-RAM hybrid layer to see real benefit, otherwise it's just like running a faster SSD; SSD's are already fast enough that they're no longer the bottleneck in load times and application performance. And here's the real kicker; If you're re-writing your application to run on this hybrid layer, it's a very costly process that requires both time and money that would be better spent just throwing more RAM in the systems in the first place.

The future of servers is more tiers of messy tiered storage, not fewer. I expect CXL memory to become a big thing, and Sapphire Rapids with HBM too, whenever it arrives. OS and application vendors will have to adapt and redesign many memory- and storage-related things. If Intel persisted and kept pushing (and supporting!) Optane in the enterprise, while bringing prices lower too, they would eventually see more success.
But with prices like $8,500 for 512 GB Optane on DIMM, they couldn't possibly get anywhere.

 

[Tomorrow]

Would it make sense to use an optane SSD as a cache for slower SATA SSD's?

They did that with Optane H10 etc. No one cared and these were the worst Optane products.

 

[Count von Schwalbe]

They did that with Optane H10 etc. No one cared and these were the worst Optane products.

Technically it was a PCIe 3.0x2 QLC drive. It would have been better if they had worked with someone like Marvell to come up with an SSD controller that could run both sides, instead of bifurcating the bus. PCIe 4.0 (or 5.0) and TLC would have made it a high performer (target enthusiasts and gamers) instead of trying for a weird balance of price and performance on OEM systems with compatible chipsets.

 

[Tomorrow]

Technically it was a PCIe 3.0x2 QLC drive. It would have been better if they had worked with someone like Marvell to come up with an SSD controller that could run both sides, instead of bifurcating the bus. PCIe 4.0 (or 5.0) and TLC would have made it a high performer (target enthusiasts and gamers) instead of trying for a weird balance of price and performance on OEM systems with compatible chipsets.

Yep. And this was at a time where QLC was not yet ready for prime time. I would argue it's still not ready with TLC being alsmost as cheap.