Intel Announced That It Is Winding Down Its Optane Memory Business

Peter_Brosdahl

Moderator
Staff member
Joined
May 28, 2019
Messages
8,899
Points
113
Intel announced in its Q2 2022 earnings release that it's winding down its Optane memory business and ceasing its future product development. The earnings report detailed that Intel...

Go to post
 
I tought they killed it of years ago.
They killed off the consumer side but where still selling and supporting enterprise. Reading through the release though, it looks like they were just doing that out of surplus stock - they hadn't manufactured anything for several years. Looks like they just wanted to mitigate the write down from severely overestimating the market years ago, and were working out of a warehouse full of the stuff for as long as they had it.
 
They killed off the consumer side but where still selling and supporting enterprise. Reading through the release though, it looks like they were just doing that out of surplus stock - they hadn't manufactured anything for several years. Looks like they just wanted to mitigate the write down from severely overestimating the market years ago, and were working out of a warehouse full of the stuff for as long as they had it.
I didn't want to state that in the event that I was wrong but I thought the same.
 
That's a crying shame because Optane is an amazing product. I was hoping they would get around to releasing Gen 4 or 5 versions.

I didn't realize that the Optane business was separate from the SSD business they recently sold to SK Hynix, i had kind of assumed that SK Hynix would continue to carry the Optane torch.

Even today, if you have tasks that benefit from high IOPS or need lots of write endurance or low write latency Optane is the best choice, despite it still being a Gen3 product. For the last several years Optanes have been just about the best you can buy for a ZIL/SLOG drive for ZFS. I wonder what will replace them in that spot.

I heard that Intel was losing money on their 3D X NAND tech, and I can't quite understand why. I don't claim to know how it differs from Samsungs 3D V-NAND tech, but Samsung hasn't been losing money on theirs. Maybe Intel's mistake was to go the route of low volume high end Enterprise products, instead of mass producing consumer models like Samsung did with theirs. Or maybe they had terrible yields?

I'm at least glad I bought the two 900p's for mirrored SLOGs when I did. I just hope they last a long time.

They seem to have amazing write endurance, considering they are constantly being hammered with writes, and have yet to register any wear in the SMART stats.

I can't calculate a predicted life span in this role yet. I have 8,992 power on hours and 7.35TB of writes on each, but because I have registered zero percent wear, I would get a divide by zero error.

At this rate, I know the write endurance will last at least 100 years in this capacity as I am about a year in, and if it were to flip to 1% wear right now, that would be a hundred years...

Intel's ARK page lists the write endurance as 5.11PB, which is crazy compared to other SSD products. Using that and my 7.35TB writes in 8,992 hours, I predict the write endurance will last ~730.8 years.

I think that might just enough to last until they are obsolete.

I'll let you know if that holds up in November 2751. :p
 
I wonder if part of the issue with optaine is Intel’s insistence on 60% margin. I know they have killed products and even left markets like mobile processors at “only” 30% margin.
 
I wonder if part of the issue with optaine is Intel’s insistence on 60% margin. I know they have killed products and even left markets like mobile processors at “only” 30% margin.

Heh,

Only a monopolist can maintain that kind of crazy margins. I guess they still deep down want to be a monopoly, and have learned nothing.
 
Before I saw that Anandtech article about this, I had seen this from elsewhere: Which then led me to that article, cuz I was like "wait... what the f*ck?". Yeah I was kinda sad about this too. I was hoping to one day eventually get a product with 3D XPoint. The article says: "Despite being designed for scalability via layer stacking, 3D XPoint manufacturing costs have continued to be higher than NAND on a per-bit basis, making the tech significantly more expensive than even higher-performance SSDs" and "Intel has been losing money on its Optane business for most (if not all) of its lifetime." This type of memory has a great deal of potential, and it would be terrible to see it go to waste.

I heard that Intel was losing money on their 3D X NAND tech
Not to be nitpicky, but 3D XPoint isn't any kind of flash memory. Not sure how true this is, but I heard 3D XPoint is also the first new type of non-volatile memory to be developed since NOR/NAND in the 80s.

I'm at least glad I bought the two 900p's for mirrored SLOGs when I did. I just hope they last a long time.
Dude that's f*cking awesome. Glad you got your hands on some.

Intel's ARK page lists the write endurance as 5.11PB, which is crazy compared to other SSD products.
3D XPoint is no joke.
 
Sounds similar to our favorite uarch punching bag, Itanium.

Though we need to keep in mind that Intel's goal with 3D Xpoint / Optane was that it would be used as non-volatile main memory, and their premiere products using Optane came as DIMMs for servers - they weren't really trying to compete with SSDs directly. They were trying to compete with DRAM.

And as far as I can tell, they more or less succeeded - but their technology, like enterprise-grade VLIW CPUs, is perhaps 'too early', and simultaneously too late. Too early in the sense that the technology needs supporting developments in hardware and software stacks, and too late because a theoretically inferior technology managed to better satisfy market demand.
 
Truth in regards to it being targeted at a higher capacity DRAM. I just got done evaluating optaine memory at work for an 80TB database that is to be run entirely in memory. It’s perfect for that kind of load, but I’m not sure how many companies actually want a 80tb in memory database.
 
Truth in regards to it being targeted at a higher capacity DRAM. I just got done evaluating optaine memory at work for an 80TB database that is to be run entirely in memory. It’s perfect for that kind of load, but I’m not sure how many companies actually want a 80tb in memory database.
Well, for me it seems like it's the future of enterprise - dedicated compute nodes with specialized hardware (be it AVX, GPGPU, AI/ML-focused etc.), but most other workloads being backed by 'general-processing nodes' full of ARM cores and cargo-ship loads of RAM. The fewer nodes that queries have to touch, the faster the enterprise can run, etc.
 
I wonder if part of the issue with optaine is Intel’s insistence on 60% margin. I know they have killed products and even left markets like mobile processors at “only” 30% margin.
What!?
Thats mental.
 
I had looked into an optane drive as a boot drive, but prices were/are so bad, modern GPU's are cheaper so that idea went out of the window pretty fast.
 
Become a Patron!
Back
Top