News Intel's Optane Business Haemorrhaged Over Half a Billion Dollars in 2020

[Admin]

Intel launched Optane with great fanfare in 2015 as a paradigm-shifting bridge between DRAM and NAND. However, it might only still exist as a loss leader to sell Xeon processors.

Intel's Optane Business Haemorrhaged Over Half a Billion Dollars in 2020 : Read more

 

[Deleted member 431422]

Everyone writing about Optane forgets about the limitations Intel emposed on it. Had to have n-th generation CPU, same for chipset. It wasn't really the price, I'd buy one if I could. It was "you won't be able to use Optane because your Intel processor is to old and we don't care" logic that prevented me.

 

[maestro0428]

Intel really missed the boat with pricing on the Optane drives and the timing as well. People were already dumping HDs for SSDs and the accelerator version of Optane was finicky to set up or recover data from. That said, I LOVE OPTANE. I have five of them in my workstation. Two PCIE 900p and two U2 900p and one 800p M2 stick. Everything thing is faster as all my data is on Optane. They get filled up but never slow down and they are far more durable than nvme drives. Sure, sequential is a bit slower, but in real life, you don't notice it. I came from Western Digital Blacks SN750s which are no slouch either. Just wish I could afford the new 5800x PCIE 4 versions of the drives.

 

[Kamen Rider Blade]

As long as Intel makes it "Proprietary" with all these Gotcha's, it'll die on the vine.

ThunderBolt was going that route until they opened it up to USB-IF.

If Intel wants Optane to take-off, it needs to hand over the specs, API's, tech to JEDEC and let everybody play without licensing costs.

Otherwise it'll be doomed.

 

[mdd1963]

"
Haemorrhaged"

Fascinating!

 

[JayNor]

The most recent Sapphire Rapids chip slides show Gen3 Optane as supported. It was supposed to be used in the Aurora project.

Leaked slides for Emerald Rapids show "CXL-MEM" as an addition. Someone at Intel gave a presentation, stating that Optane would be available in different package configurations via a memory controller that will be on the CXL bus.

 

[rluker5]

Everyone writing about Optane forgets about the limitations Intel emposed on it. Had to have n-th generation CPU, same for chipset. It wasn't really the price, I'd buy one if I could. It was "you won't be able to use Optane because your Intel processor is to old and we don't care" logic that prevented me.

The limitations depend on what you buy. I ran my pcie 900p as a boot drive on a few different Z97 motherboards with no issues. My Z97 boards didn't want to boot from nvme m.2 for some reason, but that same 800p would work fine for boot in a m.2 to pcie adapter. Of course my Z97s couldn't run the 16,32GB memory sticks to their potential. Just as tiny non boot drives so not worth it on that platform.

It is a shame it never caught on enough to scale down costs. It probably won't be tried again. I use it for OS and drag over my most played games. It mostly just helps load times. If you want to see how much check out 3dmark's ssd benchmark.

 

[watzupken]

This is a self inflicted injury in my opinion. Intel wanted to keep Optane within its ecosystem, and even within that ecosystem, they dictate that only certain chipset and processor will be able to use it. The high cost and high power consumption limited its use to mostly desktop or just a small cache in mobile device. So from the consumer standpoint, it is not surprising it failed.
From an enterprise standpoint, I feel those restrictions are not the main issue, but I feel the cost and high power draw may have done it in. And as enterprise starts to use AMD and ARM based systems, it will further limit sales.
Overall my opinion is that Intel dropped the ball on Optane. I don't always follow the news on Optane, but I was quite excited about it at launch. It was a promising replacement for NAND based SSDs, but it is not without its problems. Intel could have worked on improving it, but I don't think much was done to address the shortfalls.

 

[TerryLaze]

Everyone writing about Optane forgets about the limitations Intel emposed on it. Had to have n-th generation CPU, same for chipset. It wasn't really the price, I'd buy one if I could. It was "you won't be able to use Optane because your Intel processor is to old and we don't care" logic that prevented me.

Those limitations are imposed by the technology itself.
The whole idea was for the CPU to be able to directly read and write to the hard disk as if it where ram, a CPU that doesn't know it can do that...well, can't do that.
If your CPU can't do it you are left with using a software driver which defeats the purpose of circumventing the software to get higher speeds.

https://www.intel.com/content/www/u...faster-access-to-more-data-article-brief.html

And the Intel® Optane™ SSD has been designed to provide that advantage to the rest of the computing system through a hardware-only SSD read/write path through the SSD controller, unlike the firmware-involved path found in NAND SSD controllers.

Intel® Optane™ memory technology is byte addressable and allows rapid, in-place writes. NAND requires large block reads and a very large block erases before relatively slow writes. Data returned faster means less time waiting and faster application execution time.

 

[Krotow]

Intel launched Optane for consumer market pretty much too late and for too high price. SSDs was already here and SSD price dropped very fast. Optane drives in end turned out as small system drives, often even slower than SSDs added by user later.

Although because Optane is byte-addressable, it may work very well as unified RAM/storage substitute in embedded devices. So do not close coffin for Optane yet. It will be useful somewhere else.

 

[JayNor]

"...The high cost and high power consumption..."
I read an article on the BSC supercomputer on nextplatform. They mention that Optane is 1/10 the power of same sized DDR. See "Building An Ecosystem for Heterogeneous Memory Supercomputing"

 

[cryoburner]

"...The high cost and high power consumption..."
I read an article on the BSC supercomputer on nextplatform. They mention that Optane is 1/10 the power of same sized DDR.

It should be noted that RAM is more than 10 times as fast as Optane. So, lower power draw probably makes sense given Optane's much slower speeds. If you don't need the speed of RAM for some application, then the power savings could make sense. Not so much in applications directly affected by memory performance though. And compared to an SSD, the power draw tends to be significantly higher.

And the comment you were quoting was referring to the failure of Optane on the consumer side of things. If you have supercomputing software that's built to directly access a large dataset stored in XPoint memory, then it has the potential to be great for certain workloads. Consumer and workstation software isn't designed that way though, so it isn't being used like a slower, non-volatile RAM, but rather as a lower-latency SSD (or HDD cache).

But the vast majority of today's software isn't going to perceptibly benefit from that, since it isn't designed with XPoint's characteristics in mind. It's typically designed with the expectation that data will be accessed off a hard drive, or maybe an SSD for some recent software. And while SSDs offer a massive latency improvement over hard drives, the additional latency reduction from XPoint isn't going to be nearly as noticeable. Latency-critical tasks will tend to be performed on data loaded into RAM, and not directly from storage. For the most part, even lower-end SSDs are fast enough where the performance differences between them and Optane storage will be hard to notice in a consumer/workstation setting. So, the incentive just isn't there to pay multiple times as much for an additional latency reduction that software generally isn't going to make use of.

 

[mtrantalainen]

Intel launched Optane with great fanfare in 2015 as a paradigm-shifting bridge between DRAM and NAND.

I think the problem is that they were over-promising and fixed the price to high.

Intel 905P is still hands down the best drive I've ever used but they should still market it as a really fast SSD only. They current keep saying that it's not an SSD so it will not be included in most SSD reviews, which obviously makes it harder to sell. The only issues I see with 905P is too high price and a bit too high power consumption.

QD1 random read 4K of Intel 905P is so much better than anything else it's not even funny. And that reflects the real world response time of the storage device a lot more than many people are willing to say. And the really good write endurance is just a bonus.

Of course, QD1 random read 4K is the only metric that cannot be fixed just by adding more drivers to a RAID, which is why the Optane is so special. If you need more bandwidth than a single Optane can offer, just get multiple and use software RAID on top. Trust me, it's fast!

Intel just needs to release M.2 version of 905P with sensible pricing and they could take the market.

Of course, if manufacturing 905P is too hard to make it any cheaper even with mass production, then Intel has a big problem.

 

[2Be_or_Not2Be]

I also wonder where the money losses really came from. Was it all R&D, or did a good chunk of it get lost in other avenues? Was money lost on trying to make Optane "DIMMs" that could have been focused on the storage side?

I also wish they had concentrated on making it cheaper & more available. Keep it as a SSD product, and branch into Optane memory later when mfg costs have normalized and you're making a profit.