New Jmicron Controller to be released soon

[geoffry]

http://www.dailytech.com/SSD+P...Flash/article14176.htm

Interesting read, I definitely wouldn't mind a proper Jmicron controller putting some extra pressure on the Intel and Indilinx camps.


Edit:

Little more info on it

http://www.bit-tech.net/news/h...ler-to-feature-cache/1

 

[magreen]

If this pans out it would herald the beginning of truly affordable consumer level ssds that actually work. I'm excited.

 

[Idontcare]

magreen we need cheap high-density flash to make affordable SSD's, the controllers need to be cheap too but right now the real cost component to SSD's are the flash chips themselves.

The NAND industry is getting ready to produce 3-bit and 4-bit MLC flash which will increase density another 2x over current 2-bit MLC flash.

If you want to read more about it, grab your favorite web translator and head on over to here and here.

 

[taltamir]

Originally posted by: Idontcare
magreen we need cheap high-density flash to make affordable SSD's, the controllers need to be cheap too but right now the real cost component to SSD's are the flash chips themselves.

The NAND industry is getting ready to produce 3-bit and 4-bit MLC flash which will increase density another 2x over current 2-bit MLC flash.

If you want to read more about it, grab your favorite web translator and head on over to here and here.

bingo! its the chips that make it expensive...

that being said, 3 and 4 bit chips would most likely be even SLOWER and even shorter lived than the 2 bit ones are compared to SLC...

 

[Old Hippie]

It looks to me like the controllers and their firmware are responsible for the read/write speeds (within the capacity of the flash chips) of SSD drives.

Would it be possible for a company to market a drive with different firmware "modes" and the user could flash to whatever suits their purpose?

You could probably have a gaming flash, screaming read flash, last-a-long-time-but-slower flash..........

 

[taltamir]

It looks to me like the controllers and their firmware are responsible for the read/write speeds (within the capacity of the flash chips) of SSD drives.

Well, its more like the intel and indilinx controllers compensate for horrible write speeds by doing an internal 10 way raid 0 combined with other such complicated algorithms.

Would it be possible for a company to market a drive with different firmware "modes" and the user could flash to whatever suits their purpose?

You could probably have a gaming flash, screaming read flash, last-a-long-time-but-slower flash..........

Well, it might be possible, but usually improving one of those doesn't come at the expense of the others
And even if it did, "slightly less awesome" performance in one field for unacceptable level in another is not really desirable. right now they are also focusing on meeting the bare minimum in all fields.

It will also add a lot of cost and complexity while undercutting their SLC business... you want greater speeds of all kinds and longevity? get SLC. MLC is a budget drive so multiple firmwares with slightly different focuses make less sense...

plus intel is the only one with really solid firmware and even they had a major bug (rare, related to slowdowns).

 

[magreen]

Idc: I agree that the flash should be the most expensive item in an ssd drive. But due to current market forces, namely that intel and ocz's vertex are the only viable products on the market right now, we are being charged a premium for intel's and ocz's flash. Jmicron's new controller ought to level things out so we can start just paying a fair market price for the flash + inexpensive controller.

 

[taltamir]

We aren't saying that it SHOULD be the most expensive part...
AFAIK the most expensive component of the drive is the MLC chips, not the controller.

 

[Forumpanda]

Given the performance of the Intel controller on current MLC chips I would say that even if 3 or 4 bit chips are slower individually there is probably a very large market out there for them still.
We are still talking drivers that are from another world compared to spindle drives, and also higher density means the drives can have more and more 'hidden space' to improve write performance.

I wouldn't be surprised if at some point we start seeing drives with so much dedicated space internally to improve write performance that essentially the write speed of the individual blocks doesn't matter. (think 10+ GB).

At least it is my understanding that sequential write speed is not particularly slower on MLC memory, so if a drive can just keep enough extra room internally to always write the first 10, 20 or 50 GB sequentially, and rearrange the data when there is a lower/no write load then from an end users perspective the write performance between MLC and SLC is near equal.

 

[magreen]

Ok, that wasn't exactly what I was trying to say. I'll give it another shot. Even if the MLC flash is currently the most expensive part, we're still having to pay a pretty penny at a premium for the flash drives we're getting, since there's only 2 viable products on the market. I'm not saying that the controller is currently the most expensive component in flash drives. I am saying that due to controller issues with most ssd drives that had been brought to market, we are forced to pay higher for the complete (flash+controller) ssd drives that are functionable.

That is why I'm saying a new working jmicron controller could lower prices on all quality SSDs. And not because the controller is the limiting factor in producing an ssd cheaply.

 

[Idontcare]

Originally posted by: magreen
Ok, that wasn't exactly what I was trying to say. I'll give it another shot. Even if the MLC flash is currently the most expensive part, we're still having to pay a pretty penny at a premium for the flash drives we're getting, since there's only 2 viable products on the market. I'm not saying that the controller is currently the most expensive component in flash drives. I am saying that due to controller issues with most ssd drives that had been brought to market, we are forced to pay higher for the complete (flash+controller) ssd drives that are functionable.

That is why I'm saying a new working jmicron controller could lower prices on all quality SSDs. And not because the controller is the limiting factor in producing an ssd cheaply.

magreen, totally understand what you are saying now, thanks for clarifying.

Yep, you are absolutely right. The way to frame this is to simply compare the $/GB for leading MLC drives as the underlying flash costs are commoditized so the delta's are almost entirely premiums paid for the controller and reliability/brand-name/etc.

Compare X25M $/GB to Vertex to GSkill to a Core or supertalent or some other cheap-ass-dirt Jmicron-based SSD. I think Anand had such a table somewhere.

Now Intel's flash is slightly cheaper for it as they get their flash from IM flash, so it is subsidized meaning if anything Intel's gross margins are even higher than the $/GB analysis would lead you to suspect.

 

[magreen]

Originally posted by: Idontcare

Originally posted by: magreen
Ok, that wasn't exactly what I was trying to say. I'll give it another shot. Even if the MLC flash is currently the most expensive part, we're still having to pay a pretty penny at a premium for the flash drives we're getting, since there's only 2 viable products on the market. I'm not saying that the controller is currently the most expensive component in flash drives. I am saying that due to controller issues with most ssd drives that had been brought to market, we are forced to pay higher for the complete (flash+controller) ssd drives that are functionable.

That is why I'm saying a new working jmicron controller could lower prices on all quality SSDs. And not because the controller is the limiting factor in producing an ssd cheaply.

magreen, totally understand what you are saying now, thanks for clarifying.

Yep, you are absolutely right. The way to frame this is to simply compare the $/GB for leading MLC drives as the underlying flash costs are commoditized so the delta's are almost entirely premiums paid for the controller and reliability/brand-name/etc.

Compare X25M $/GB to Vertex to GSkill to a Core or supertalent or some other cheap-ass-dirt Jmicron-based SSD. I think Anand had such a table somewhere.

Now Intel's flash is slightly cheaper for it as they get their flash from IM flash, so it is subsidized meaning if anything Intel's gross margins are even higher than the $/GB analysis would lead you to suspect.

Exactly. There's a reason the intel costs triple what a supertalent does, and it's not because intel flash is significantly more expensive to produce.

Now, you have to temper this comparison a little bit when comparing the $/GB of an intel or vertex to the $/GB of an el-cheapo OCZ core or supertalent or something. Remember, the prices of the el-cheapo drives are also being modified by market forces. They're often sold at firesale pricing, much less than the price they'd garner if they were functionable, due to their (well-deserved) bad reputation.

So we shouldn't expect new quality drives using a good jmicron controller to bring prices of quality drives down to the level of the latest hot deal on the POS ssd they're trying to pawn off on you, which is below commodity pricing. Unless there's additional reasons that prices would fall.

But hopefully the new pricing will split the difference.

EDIT: whew, I'm sure glad I finally got my point across. I was about to resort to a car analogy, and I know you don't want to go there!

 

[Old Hippie]

whew, I'm sure glad I finally got my point across.

I got it on the first post, but I liked the conversation! :beer:

 

[Phynaz]

Originally posted by: Idontcare
magreen we need cheap high-density flash to make affordable SSD's, the controllers need to be cheap too but right now the real cost component to SSD's are the flash chips themselves.

Let me extend that to cheap flash that the producers can make a profit on so they can stay in business.

 

[Idontcare]

Originally posted by: Phynaz

Originally posted by: Idontcare
magreen we need cheap high-density flash to make affordable SSD's, the controllers need to be cheap too but right now the real cost component to SSD's are the flash chips themselves.

Let me extend that to cheap flash that the producers can make a profit on so they can stay in business.

Damn straight! They've been operating at negative margins (losses) since early 2008.

http://pc.watch.impress.co.jp/...9/930/html/06.jpg.html

 

[aka1nas]

Originally posted by: taltamir

Originally posted by: Idontcare
magreen we need cheap high-density flash to make affordable SSD's, the controllers need to be cheap too but right now the real cost component to SSD's are the flash chips themselves.

The NAND industry is getting ready to produce 3-bit and 4-bit MLC flash which will increase density another 2x over current 2-bit MLC flash.

If you want to read more about it, grab your favorite web translator and head on over to here and here.

bingo! its the chips that make it expensive...

that being said, 3 and 4 bit chips would most likely be even SLOWER and even shorter lived than the 2 bit ones are compared to SLC...

If the controllers improve enough, we should at least get similar speeds but larger capacities at similar price points. Until we have faster interfaces available, SSDs significantly faster than the X-25M and Vertex won't make much sense because we'll hit the bandwidth limits on SATA II. PCI-E based SSDs are one option, but most PCs don't have the extra slots available for those, either (x1s won't cut it, might as well use SATA at that point).

 

[Nathelion]

You know, with RAM prices at what they are, I've been wondering for a long time why there aren't viable PCI-E RAM-drives out there. I know, I know, there is the one DDR one, but DDR is expensive and the unit is like 500 bucks, too. One would think that it'd be perfect for servers or other applications where they are currently running 15k SAS-drives that are very expensive and very limited capacity as it is. If you think about it, a graphics-card-sized unit could probably accommodate 20 sticks of ram fairly easily. To make it reasonably priced, let's assume 4 GB sticks. That would make 80 GB of super-fast storage at a price of roughly 1600 + cost of expansion card, assuming that the RAM needs to be both buffered and ECC. Less otherwise. 80 GB 15K rpm SAS-drives are like 150 bucks apiece, so yeah this would be more expensive, but there should be situations where it would be worth it..?

Sorry for taking the threadjacking, I wasn't planning to make such a lengthy post when I first started writing it.

 

[taltamir]

magreen, I have to agree with what you are saying... there is definitely a lot of room to go down in price on both components.

say... who else besides intel and samsung makes MLC chips?

And aka1nas, yes its true that awesome controllers can certainly compensate... it will be interesting if we see 4 bit chips (QLC?) resulting in drives so cheap and plentiful that they replace platter storage as a main storage medium... Since it should be double the size of an MLC drive for the same price... I wonder if they will go beyond 4 bits per cell anytime soon.

 

[Idontcare]

Originally posted by: taltamir
magreen, I have to agree with what you are saying... there is definitely a lot of room to go down in price on both components.

say... who else besides intel and samsung makes MLC chips?

Are you joking Toshiba? I heard they had something to do with the creation of NAND :laugh:

(Hynix, STMicro, Renesas, Nanya, ProMos)

 

[taltamir]

well... its just that all the reviews where they cracked open the drives i have seen all contained intel and samsung chips... never anything else.

could it be due to chip densities?

 

[aka1nas]

Originally posted by: taltamir
well... its just that all the reviews where they cracked open the drives i have seen all contained intel and samsung chips... never anything else.

could it be due to chip densities?

Might be, I keep seeing all the press releases with Samsung being first to market with denser chips.

I'm sure the other flash manufacturers are tooling up for this though. The slower flash used in thumbdrives and flash cards is so commoditized now that I would imagine they would be jumping all over this as a chance to get into a higher-margin market.

 

[taltamir]

mmm... actually to be specific... aside from intels own x25 SSD... EVERY other SSD review i have ever seen was ALWAYS samsung...

We have indilinx and intel competing for the top controller (With a samsung and jmicron and a few others at the bottom)...

If we are correct about samsung having the highest density chips at any given time, than AFAIK we only have samsung and intel at the high end... and intel is the only one i know of to use thier own chips ... (of course, they might be more common than that... my evidence is purely anecdotal)

 

[aka1nas]

Originally posted by: taltamir
mmm... actually to be specific... aside from intels own x25 SSD... EVERY other SSD review i have ever seen was ALWAYS samsung...

We have indilinx and intel competing for the top controller (With a samsung and jmicron and a few others at the bottom)...

If we are correct about samsung having the highest density chips at any given time, than AFAIK we only have samsung and intel at the high end... and intel is the only one i know of to use thier own chips ... (of course, they might be more common than that... my evidence is purely anecdotal)

It also might just be that the Jmicron, Indilinx and (of course) Samsung controllers only validated Samsung chips as they were ready first and everyone is basically building a reference model.

 

[taltamir]

a highly likely situation.... mmm.. i wonder if samsung paid them to not validate their competitors?

 

[Nathelion]

Is Intel even selling their flash on the open market? Their flash drives are selling pretty well, I could imagine them using it all themselves. Also, who would want to buy their flash from a direct competitor?