Last week Intel killed Optane. Today, Kioxia and Everspin announced comparable tech Last week Intel killed its Optane storage-class memory product, because it just couldn't sell enough of it to make a difference to the bottom line. But this week, two Intel rivals revealed advances in their own Optane-like tech. Japan's Kioxia teased the second generation of its XL-FLASH Storage Class Memory. In an …
While the lion was hunting, the jackals kept their heads down. Now that the lion is gone, the jackals can feast.
So, if these new attempts (that seem interesting, viewed from my chair at least) succeed, does that mean that Intel will buy these companies out ?
Or will Intel just eat its humble pie in the corner and skulk ?
They'd have to have clue where they are going first.
Remember folks, Intel has sold it's flash and memory divisions before. It may yet do it again.
What killed Optane? Over hyped performance compounded endless delays to market, that were then coupled with eye watering prices that only the delusional could imagine businesses buying at scale, and no user level sales to speak of. They weren't the only ones, Dell kept trying to suggest 7000$ enterprise SSDs to me and swore that the crappy sata drives would never work. There was never going to be a viable market in that segment. Especially as the Optane memory was going to be Intel only, and only on certain motherboards, in certain verticals. They utter hubris in the CPU and Optane offerings was as stunning as the marketing was bad. When they renumbered everything again and introduced "Silver", "Gold", and "Diamond plated bullshit" tiers it was clear they didn't have a clue where they were going, what they were offering, and expected large piles of money because they were Intel.
Instead we bought piles of WD reds and blues, both as NVME and Sata. AMD drove tanks on their lawn and stole donuts from their break room. And TSMC is now serving as a contract fab and fabbing them humble pie.
Intel's failures are mostly their own, they should have either cancelled Optane years ago when they couldn't deliver on the performance promises, or eaten the losses if the delays were caused by the wider headwinds their fabs had during those years. They would have had better chances going wide and standardized. They could also have set down the "Then Intel takes over the world" playbook and go back to collaborating with the other fabs and companies like they did in the old days. They proved over the last decade that nobody in the top tier of the semiconductor world can succeed by going it totally alone. Lucky for them both AMD and TSMC have to keep them around to avoid the same antitrust issues that Intel used have so many problems with.
Gelsinger said he didn’t want to be in the memory business - pretty clearly defines the trajectory and direction of travel. Had Intel focused on persistent memory packaging only for the last three + years, the outcome would be significantly different in my opinion.
Based on the emotional distress permeating yesterday's Analysis, you probably just made someone very happy with this news.
https://www.theregister.com/2022/08/01/optane_intel_cancellation/?td=rt-3a
Why the end of Optane is bad news for all IT
"The future was here, but when viewed through the blurry scratched old lenses of 1960s minicomputer OS design, well – if everything is a file, this Optane was just a sort of really fast disk drive, right?
No, it wasn't. It was the biggest step forward since the minicomputer. But we blew it.
Goodbye, Optane. We hardly knew you. ®"
As yesterday's article pointed out, this should be great but it requires a whole new operating system design. Forcing it to behave like a file isn't going to get the best out of it.
I'd love to see that happen, but inertia is a bitch.
" this should be great but it requires a whole new operating system design."
Multics did it 50 years ago. It's a different way of looking at things, but very workable.
Multics and an emulator to run it are freely available, for anyone interested:
https://multics-wiki.swenson.org/index.php/Main_Page
50 years ago, the market was small and diverse enough that new innovative designs could carve out a share of the market. Most of the market is dominated by Windows today and, as another poster put it, inertia is an issue now. If MS can't make it work, it'll remain at best a small niche that, like Optane, will wither on the vine or be a very slow growth.
> If MS can't make it work
Except that a simple search engine query shows that Windows SERVER (and SQL and a few other MS stuff) can leverage Optane for good. So MS made it work. It was Intel who insisted to block it from all normal Server CPUs and limit it to the expensive ones which no one actually buys.
Flash is having a good run, but there are plenty of other candidates. But while as a result of hitting the walls with existing flash memory we got massive layer stacking working in the fabs(kudos to Samsung especially there), it's still on it's last gasp.
The base technology for Optane wasn't the big bang win that Intel wanted, they pursued a defective market segmentation strategy, and it was late. Now the other memory makers that had been trying to catch up on their next gen persistent memory tech are going to slingshot past Intel weather they cancel Optane or not. So it's not a surprise that Intels management repeated their earlier decisions to scuttle their earlier memory divisions in the face of rising competition and falling prices. It may actually turn out better for the market, as smaller memory startups could have gotten starved out in the oversupply glut that is likely a few years down the road as some of this new fab capacity comes online.
As a persistent memory architecture it was just not fast enough and useless in large virtual clusters where you have no idea where the VM is likely to restart. So is was really just a bare metal solution only.
Memory hungry applications tend to be memory bandwidth and latency dependent, so to get the required performance you ended up having to run in a 1:1 DIMM to Optane configuration.
As a storage solution (sitting on the PCIE BUS) is was far too expensive and again was local to a physical machine, so at best it was a good caching mechanism before writing to a fast remote storage device, but then again, an array of fast nvme drives would have been just as fast and cheaper for the same capacity.
So riddle me this one. What was the compelling thing about Optane? What was the one thing Intel continued to hammer home and the likes of some storage vendors who based their company solely on this product initially vamped on constantly?
BYTE LEVEL ADDRESSABILITY
You can only get that, according to Intel, from DRAM, but because of how Intel allows you to programmatically access their SCM SSD, you can get this type of access versus the traditional block or page access of NAND. Right?
So the news, I believe, is, what is the strategy for those companies who have made it part of their operational and functional business to operate at this level for its deduplication and compression to continue to maintain their claims of performance and efficiency?
1. It sounds to me like these companies are going the Data Domain way and increasing memory capacity (DRAM which is supposed to be so much more expensive than Optane SSD - oh, but it can serve so many other purposes too…so, maybe the cost balance is worth it?).
2. Could it be that they (Kioxia, Everspin, etc) will be looking at a CXL partner to bundle their SCM with? You infer that Kioxia is looking at CXL, but as far as I know right now CXL is focused on memory not SCM - if I’m wrong, then correct me.
3. Could Kioxia and Everspin just decide this market isn’t big enough and with CXL, anyone with NAND based flash can do the same thing without having the heavy burden of layering “SCM” on top of it?
I don’t know, but those are the things I’m interested in finding out. Today Kioxia and Everspin have time and money to invest in testing the market - why after both Micron and Intel bid farewell, I don’t know - perhaps they need a good tax write-off or they have contractual obligations to the one vendor who absolutely depends on this type of technology.
There is a lot more we don’t see or know - and I think that’s where the news really lies in this bit.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
