Intel's optane technology in financial trouble?

[80251]

This is an old article from February 2022, but I never saw it posted here.

https://www.tomshardware.com/news/intel-optane-massive-losses

Is the biggest drawback to optane (aka. 3D XPoint) that it isn't easily scalable to large capacities? The largest Intel DC Optane SSD I see is the DC P5800X 3.2 TiB. Micron has already given up on it.

How can it be so fast if it's DRAMless?

 

[Shrek]

The more chips one uses the more you can run them in parallel.

 

[kapone32]

Optane was a compelling product that I really wanted to try, unfortunately the price of Optane is ridiculous for what you get.

 

[80251]

Why is optane so expensive though? Because according to the article cited Intel is losing money on optane! Is the manufacturing process that complex that there are a lot of losses?

 

[dgianstefani]

Low volume.

 

[kapone32]

Why is optane so expensive though? Because according to the article cited Intel is losing money on optane! Is the manufacturing process that complex that there are a lot of losses?

Because of the Greed co-efficient Intel applied to it. They could have actually taken a huge bite out of NVME but they priced Optane to the moon which inhibited adoption. $500 for 48GB does not make any common sense.

 

[Count von Schwalbe]

Why is optane so expensive though? Because according to the article cited Intel is losing money on optane! Is the manufacturing process that complex that there are a lot of losses?

Low volume.

This + astronomical R&D costs. Remember how expensive SSD's used to be, not that long ago? And that was with generations of development already behind them. Intel and Micron were developing 3DXpoint based flash from scratch. IIRC Micron has pulled out of the joint venture and Intel is shutting it down.

I believe most of what they sold were NVDIMM's for the server market, allowing for a much better version of Fast Boot, where an image of the RAM is saved to disk (or DIMM in this case) and loaded again on startup.

Optane SSD's with QLC NAND and 3DXpoint on the same board are still around, and a cool concept if they didn't require specific chipset and software support to work properly. Also, the bus is cut in half (3.0x2 3DXpoint, 3.0x2 NAND) meaning my B550 board was only able to view the 3DXpoint side. Not too useful for a boot drive.

 

[80251]

The Optane NVDIMM's also allowed for higher capacities than traditional DDR4, not so sure about DDR5 though.

 

[DavidC1]

This is an old article from February 2022, but I never saw it posted here.

https://www.tomshardware.com/news/intel-optane-massive-losses

Is the biggest drawback to optane (aka. 3D XPoint) that it isn't easily scalable to large capacities? The largest Intel DC Optane SSD I see is the DC P5800X 3.2 TiB. Micron has already given up on it.

How can it be so fast if it's DRAMless?

It's fast because the base media is fast. Regular SSD uses the slow NAND media, so it needs DRAM to do the so-called garbage collection. That's why the DRAMless ones suck. NAND needs to delete data in it's storage cells before it can be re-written and the delete process is dog slow. DRAM helps doing that so it doesn't feel uselessly slow. Optane doesn't need deletion - it can write in place. The base media is indeed about 1000x faster.

It's expensive mainly because it doesn't have enough volume. It's said by those close to the industry you need a brand new technology to be within order of magnitude(so 10x) before it becomes competitive in price. DRAM and SSD volumes are probably several hundred times larger in volume. So once it gets to say 1/10th of DRAM volume then it'll achieve the promised costs.

That's why despite them selling for really high, Intel is losing money. They said the high end 900P SSDs they were barely making money on it. Imagine how much they were losing on the tiny $50 Optane Memory devices!

Although I think they could have done a better job with Optane Memory. On the server side they had something called Memory Drive, so you can use those SSDs as part of slow RAM and extend capacities. I think if you could have done that with Optane "Memory" the reception would have been better with enthusiasts. Unfortunately the Memory Drive server versions were like a $400 adder so you can see why they didn't bring that to consumer. Short sighted decision IMO.

I like the tech too, and I think the DRAM versions hold the key. I'd have opened it up to be fully compatible with DDR5 so AMD/ARM/Power can use it. The future doesn't look bright at all, especially since Pat Gelsinger doesn't like the memory business.

 

[80251]

@DavidC1
So optane, even though it is a superior technology for SSD's, is probably going to die off?

 

[dragontamer5788]

The Optane NVDIMM's also allowed for higher capacities than traditional DDR4, not so sure about DDR5 though.

DDR4 / DDR5 is all DRAM-technology. Optane is cheaper than DDR4 / DDR5 per TB. The problem is that Optane's competitor is Flash, which is even cheaper.

So optane, even though it is a superior technology for SSD's, is probably going to die off?

Define "superior". Optane is faster and has more durability than SSDs. But its not "faster enough" given its high costs.

DRAM / DDR4 / DDR5 is the other competitor here, which is even faster than Optane, and modern CPUs can support 4TB of RAM easily. Given the high costs of Optane, it looks like major companies are deciding to just "go all the way" with all-DRAM / DDR4 systems and make 4TB MongoDB instances or Redis instances, rather than experimenting with Optane.

 

[Assimilator]

@DavidC1
So optane, even though it is a superior technology for SSD's, is probably going to die off?

Yes. Just like how Firewire/IEEE 1394 died to USB because although it was superior in terms of speed, it was also prohibitively expensive. NAND flash, just like USB, is good enough and cheap enough for consumers to make it really difficult for any similar product, especially a more expensive one, to compete with it.

In order to get people to buy Optane SSDs and thus take marketshare away from NAND SSDs, Intel would need to sell Optane drives for a price very close to a NAND drive of the same capacity. The only way Intel could accomplish that is by selling Optane at a large loss for a long time - and with the price of NAND flash dropping constantly, they'd have to keep ploughing more and more cash into Optane in the hope that it would potentially get to the critical mass needed to allow them to make a consistent profit on it. Since Intel has never been particularly interested in making a loss on anything, the likelihood they'll do that is pretty close to zero.

Ultimately Optane will go down in history as an innovative technology that deserved better than Intel's treatment of it, but also wasn't superior enough to its competitors to justify its success. At the end of the day economics talks louder than innovation, unfortunately.

 

[defaultluser]

it's not fast enough to replace dram, and it's way less dense than NAND - it was always destined for failure.

 

[80251]

it's not fast enough to replace dram, and it's way less dense than NAND - it was always destined for failure.

It's not fast enough to replace DRAM but it does have MUCH higher capacities than DDR4 DIMM's and it's non-volatile.

Is optane faster and/or have higher capacity than SLC NAND? Wasn't Intel using it as a cache for their QLC NAND products? Would that be a useful niche for it?

 

[defaultluser]

It's not fast enough to replace DRAM but it does have MUCH higher capacities than DDR4 DIMM's and it's non-volatile.

Is optane faster and/or have higher capacity than SLC NAND? Wasn't Intel using it as a cache for their QLC NAND products? Would that be a useful niche for it?

rEMENMBER THE END GOAL FOR ANY MEMRISTOR-ALIKE IS TO ALLOW FOR isingle-memory compute

HP Enterprise unveils The Machine, a single-memory computer capable of addressing 160 terabytes

Join Gen AI enterprise leaders in Boston on March 27 for an exclusive night of networking, insights, and conversations surrounding data integrity. Request an invite here. Hewlett Packard Enterprise announced what it is calling a big breakthrough — creating a prototype of a computer with a single...

venturebeat.com

Otherwise,what exactly is the point of redesigning a SYSTEM FOR FULL PERSISTENCE,WITHOUT COPIES,IF THE PEAK-PERFORMANCE IS SO SLOW?