Sandisk and SK Hynix Publish First Open High Bandwidth Flash Standard (nerds.xyz) 19
BrianFagioli writes: Sandisk and SK hynix have published the first open technical specification for High Bandwidth Flash (HBF) through the Open Compute Project. HBF is designed to create a new memory tier between High Bandwidth Memory and traditional SSD storage, giving AI inference systems more near compute capacity without relying entirely on expensive HBM. The specification supports capacities up to 512GB using 8-high and 16-high NAND stacks, with bandwidth tiers ranging from about 0.4TB per second to 3.0TB per second. It also uses the UCIe chiplet interface, which could make HBF easier to integrate with CPUs, GPUs, and other accelerators. Google and Tenstorrent participated in the standardization effort, but commercial products and independent benchmarks are still missing.
An open standard... or a moat? (Score:4, Interesting)
Sandisk and SK Hynix are fleecing everybody - pc and phone makers, NVIDIA, the Mag7 - which means we are all paying a ton more for their products. They are essentially an oligopoly along with a small handful of memory makers. Did they do this out of the kindness of their hearts, or to create a standard they control to their own benefit?
Re:An open standard... or a moat? (Score:4, Informative)
Sandisk and SK Hynix are fleecing everybody - pc and phone makers, NVIDIA, the Mag7 - which means we are all paying a ton more for their products. They are essentially an oligopoly along with a small handful of memory makers. Did they do this out of the kindness of their hearts, or to create a standard they control to their own benefit?
Same with Dell with CAMM and nVIDIA+Micron with SOCAMM (similar name, similar idea, completely different use cases). Both of those presented the original stadard as a PoC, and then handed it over to JEDEC, so they became CAMM2 and SOCAMM2, completely neutral industry wide standards.
I suspect the same will happen here, a few demo/lab/PoC products will be produced, and then this will be handed over to JEDEC, so that it becomes an industry wide standard.
What do they (SKHynix, Sandisk and before them Dell, Micron and nVIDIA) win by doing this? IF (and only if) there are patents involved, they win FRAND revenue (just as an aside, neither CAMM2 nor SOCAMM2 have patents involved). And wether there are patents involved or not, these companies have a leg up on the competition in terms of know-how and practice making the stuff.
Re: (Score:2)
What do they (SKHynix, Sandisk and before them Dell, Micron and nVIDIA) win by doing this?
Well as you said FRAND is one reason private companies would fund research and then donate it to a standards body, and a “good” one.
Not the only reason though. A lot of things are driven by adopting standards. If you make your very own technology and other companies need to adopt it they tend to want to deal with standards not a single source product, so you get wider adoption by publishing “your thing” as a standard. It does basically burn most or all of the money you invested R
Re:An open standard... or a moat? (Score:4, Insightful)
Sandisk and SK Hynix are fleecing everybody
Supply and demand curve breakdown does not mean anyone is being fleeced. It just means someone with deeper pockets than you wants the limited thing more than you do. Their products are being sold at market rates.
They are essentially an oligopoly along with a small handful of memory makers.
Which is irrelevant. Being an oligopoly is just a state to describe how many companies are in a market. It doesn't have any impact on the market. Actions have impacts on markets, and while this oligopoly has in the past been found guilty of collusion, there's zero evidence that what is happening right now would be any different if there were 100 small manufacturers rather than a couple of large ones.
Did they do this out of the kindness of their hearts
No they did it to create a new market for a product they produce while offering benefits to potential customers. It's the reason why any product is created. It's literally how commerce works.
or to create a standard they control to their own benefit?
It's an open tech spec, not too different from how virtually every other memory and storage spec is created, except the resulting spec requires no membership with JEDEC.
Re: (Score:2)
There's nothing stopping you from starting your own memory manufacturer... except your lack of competence in the firld.
Optane (Score:2)
Re: (Score:2)
That guy who's running it it now is an nVidia puppet; a cuck to Jensen and his fucking cousin Lisa! The biggest chip designers are now all run by Taiwanese/Chinese. Intel is 10% owned by its competitors
Re:Optane (Score:4, Informative)
It's easy to blame Intel but that was a joint venture with Micron and Intel. IMFlash out of Lehi Utah. Micron was every bit as responsible and if I remember right still holds most of the patents.
You could even make the case it was Micron that killed it because they dumped the entire thing on Intel and sold off the Lehi fab. Without Micron sharing manufacturing costs and with scaling difficulties lagging behind traditional NAND bit density improvements, Optane became financially unviable.
Re: (Score:2)
It's on the line and wants to know if it can get out of the dumpster of failed tech
No it can't because it failed precisely for the reason that makes it incompatible with what is happening right now: It was a proprietary closed spec with no wider industry agreement or acceptance.
Intel tried to lock everyone into Intel. Instead here we have an open technical spec. Optane isn't welcome.
That and cost. Intel's fabs couldn't produce Optane at even remotely compelling prices.
Re: (Score:2)
Even if open, it just wasn't going to happen.
Problem was that it was awkwardly slower than DRAM and more expensive than NAND. Was it cheaper per GB than DRAM? Sure. Was it faster than NAND? Sure. Problem is there wasn't really space in the market for a tier between NAND and DRAM. Embarrassingly obvious as Intel ran around doing things like trying to force a whole different paradigm for application developers to do things awkwardly special to try to take advantage of this tier.
Now we have a scenario wher
Re: (Score:2)
Not to mention even within Intel's products, it was limited to specific motherboard chipsets (at least for the general PCs).
They reduce the chance for widespread acceptance by limiting it too much.
After that the rest is history .....
Re: (Score:2)
Optane was not 3D in the same sense as 3D-NAND/HBF, it was a dead end.
Not the First Offload Attempt (Score:2)
Re: (Score:2)
RAM and block-storage may, at some point converge, but that is in the future. Block-storage will require you design for it for quite a while still.
Re: (Score:2)
This is orders of magnitudes more performant. The highest tier is at roughly HBM bandwidth for reads, there is no equivalent in history. Though it remains to be seen if they can run it that fast without melting down.
Still waiting for HP to deliver "The Machine" (Score:2)
(Which was just before they entirely focused on selling overpriced ink.)
Re: (Score:2)
In 2017 they built a prototype. Subsequent to that they pivoted.
See https://newatlas.com/hewlett-p... [newatlas.com]
Also https://en.wikipedia.org/wiki/... [wikipedia.org]
that icon though ... (Score:2)
Am I the only one bothered that an article about really really fast storage is represented by an icon of a nine track tape, one of the slowest forms of storage?