Micron-Intel 3D XPoint split: It's not you, it's m... nah, it is totally you Micron's commercial discussions with Intel over 3D XPoint have concluded that the tech partnership will be dissolved once second-gen development is completed next year. Intel's Optane XPoint product sales have been disappointing and Micron is facing a collapse in sales of its XPoint chip output to Intel. Intel is, of course, …
"Optane XPoint drives do not have sufficiently faster performance compared to NVMe SSDs to make their cost worthwhile."
Isn't the point of Optane to add as a chunky NVRAM cache for high capacity low speed storage so that initial load times stay the same but certain data gets loaded into the Optane module so subsequent loading is much improved?
Realistically speaking, it's not as if one day NAND Flash got released and conquered the world by storm.It's only over the past years that an SSD has become a standard feature of new laptops, and even then at rather pitiful capacities at exorbitant prices.
Of course, it's not as if new technologies and products cannot be fumbled so badly by Marketing that they die on the vine. Considering XPoint's promises and performance so far, I really hope that this split creates the competition and variety of products that XPoint needs to find its course.
One big benefit of XPoint has been that it has given Flash manufacturers a reason to look nervously over their shoulders. That alone makes it a net positive for us consumers :)
"Realistically speaking, it's not as if one day NAND Flash got released and conquered the world by storm.It's only over the past years that an SSD has become a standard feature of new laptops, and even then at rather pitiful capacities at exorbitant prices."
My first hard drive was a whole 20MB of storage for an Amiga 600 back in the mid-1990s. Every hard drive from then has looked the same: the circular thing underneath, a metal plate on top, about the same size, the same weight, just capacities have increased. Initially SSDs were a change, but the cost was a bit crap, and they still looked roughly the same.
The first time I installed an M.2 SSD I felt excited. I couldn't decide whether this was finally the future or I was stepping back in time and plugging in a SNES cartridge. Either way, here was something different.
7. It's utterly proprietary: Intel won't say how it works. Without this information, there can be no rigorous independent evaluation of its endurance and of its failure modes. A guarantee is all very well, but one doesn't want it to fail in the first place.
8. It's single-sourced, meaning there's no alternative if Intel/Micron should have production problems or give up on it, and no prospect of meaningful price competition.
NAND tech is unreliable and it does matter a lot!
NAND SSD don't last. If you don't use your PC then yes they will last for a long time. They can last a long time on smartphones and tablets too if you don't install/uninstall many apps and you don't use the apps to make content.
A NAND SSD drive can reach the end of its expected life even in a few days of heavy usage. You surely can't expect to use it like an Hard Disk , it wouldn't last more than some hundreds hours at best. Only very expensive Enterprise class SLC NAND SSD units could survive. Consumer MLC and now the worse TLC and QLC NAND just couldn't. Even because the expected usage is by statistics on samples and a critical failure usually happens way sooner. That is the reason even few Hard Disk drives are able to reach their expected life without going into a critical failure losing all spare sectors in multiple clusters. Still Hard Disk drives on average last months to years more than any NAND SSD.
Hard Disk drives too shouldn't be used without a good UPS but for NAND SSD that is even worse and Enterprise class SSD drives usually have a very small UPS built in the package to give 2-5 seconds to the controller to shutdown properly otherwise there would be cells corruption.
Intel and Micron must start selling 3D XPoint SSD at a loss. Selling them at the same price of cheap NAND SSD would give them a huge advantage pointing out that 3D XPoint drives don't fail quickly while NAND ones do, and that is an absolute truth.
I wondered if you'd poke your head in. I've been running SSDs in all classes of hardware from the datacenter to the desktop, and I haven't seen the reliability issues you allude to. Care to back your FUD up with some statistics and citations?
>>NAND SSD don't last. If you don't use your PC then yes they will last for a long time. They can last a long time on smartphones and tablets too if you don't install/uninstall many apps and you don't use the apps to make content.
Then please, feel free to explain why my 3D NAND SSD (a non-enterprise Samsung 850 Pro) still has 98% of its remaining life and I've been using it for almost 3 years non-stop now on my gaming PC that has very heavy utilization? I've also had a number of power failures (because PNM aren't reliable at all) where my UPS didn't kick on, which according to you should have killed my drive. It didn't. Don't believe me? Take a look yourself here
I've had far better reliability out of my SSD than out of my HDDs. To the point where I'd like to get rid of the spinning rust completely because its unreliable and probably will on my next build.
> It looks like an ideal ram & cache replacement for 3 or 4 bit per cell nvme ssd.
Sure, but how many applications actually benefit from this additional layer of cache?
Most datasets are skewed so a very large proportion is "hot" and the rest is not - this is how caching works in the first place. So increasing the size of your cache has a proportionately lower return. If you have (say) 64GB of RAM then your hottest data will be there already. If you add another tier with (say) 256GB of Xpoint, then only warm data is there. Your cache hit rate might go from say 80% to 85%. That translates into a very low actual improvement in overall application performance.
Furthermore, if this cache is fronting some slower primary storage, i.e. spinning disk, then you could add 256GB of NAND, at a fraction of the cost, and get almost identical performance improvement.
Another option might be to try to use Xpoint to reduce cost: reduce your RAM to (say) 16GB whilst adding 64GB of Xpoint. The overall cost saving will be low, and the performance will actually decrease, since the Xpoint is not as fast as RAM, and you are pushing hotter data into a slower tier, and there will be increased movement of hot data between the tiers. On top of that, you've increased the complexity and failure modes of your server.
So the only use case for Xpoint I can see is when your main storage tier is NAND, and NAND is not fast enough, *and* your dataset has so much hot data that it would be uneconomical to put enough RAM in cache it effectively.
This to me sounds like a very niche area indeed.
"So the only use case for Xpoint I can see is when your main storage tier is NAND, and NAND is not fast enough, *and* your dataset has so much hot data that it would be uneconomical to put enough RAM in cache it effectively."
There are plenty of large capacity disk-based servers which could benefit from a single Optane card caching in front of the spinning monsters - it's just Intel is ultra-greedy, as always, pricing it as high as a real NVMe with higher capacity, the very thing they are trying to replace with Optane, not to mention making the caching function restricted to their latest chipset (so it won't work on your couple years old server or workstation, nothing AMD etc.)
It's idiotic, classic stupid arrogance from Intel; they really deserve to be caught burning few billions on this blunder. Hopefully the people decided the positioning of this product line get all fired.
I don't understand point (6) - 3DXP crushes the competition in everything but sequential, but are we going to quibble over 'only' 2GB/s throughput and that could be solved with more channels? Endurance is an order of magnitude higher which is a big deal for write intensive applications, and 3DXP trounces NAND at lower queue depths.
The only problem with 3DXP is it's too expensive to make. Intel is eating a loss with each Optane drive sold and the consumers are already nonplussed about it's current price. If they double the layers as well as double the lateral density as I've heard/read about elsewhere, it'll be a lot more competitive.
Micron has the option to call Intel's share of the Lehi fab in January (meaning Intel has to sell it to them should Micron choose to exercise the option). If Micron does not do so, that will be prophetic of 3DXP's future.
The performance issue is
Optane SSDs are much faster in benchmarks but other than high end datacenters, no one else can tell the difference (no one runs storage at 90% utilization).
the price is 5x as much at TLC NAND SSDs
DIMMS might be a benefit but again that is only high end datacenter
So its a great product for 1% of the applications. otherwise TLC NAND SSDs are much better
THAT is why micron stopped paying for development and isnt running any production in Fab and why Micron is done working with Intel.
If they had been able to match DDR3 in performance and released it to other SoC vendors then there was a chance to make big inroads with embedded devices like STBs and tablets because they would quite quickly adapt to a unified memory architecture and their storage needs are relatively small. A TV manufacturer would easily buy +1m 1GB RAM sets and they don't have to be the fastest tech on the market.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
