JMicron NAND flash controller could lead to significantly lower SSD prices

Truth be told, SSD prices have been declining at a noticeable tick since way back in 2007, but the reality is that the average consumer still can't afford one -- or, at least they aren't willing to pay the lofty premium for the decent increase in speed. If a bold claim from JMicron is to be believed, all that could change in the run-up to CES 2010. A new report has it that the aforesaid company will be demonstrating its new NAND flash controller next week at Computex, with the JMF612 aimed specifically at a "new generation of NAND flash chips built using smaller process geometries that will be entering the market soon." If all goes well, the cheap single-chip controller could lead to SSD prices falling by around 50 percent by Christmas, but after years of waiting for these things to really get priced for Joe Sixpack and his gaggle of siblings, we're still cautiously skeptical. Not that we wouldn't love to be proven wrong or anything.

[Image courtesy of HotHardware]

[youngstunna @ May 31st 2009 8:33AM]

isnt the current jmicron controller the vastly inferior one that has all the stuttering issues?

[Rhino @ May 31st 2009 8:52AM]

Yeah~ So we can buy cheap and shuttering SSDs by X'mas for cheap and those great ones (say Intel or OCZ Vertex) will remain at current price tag. What a great marketing idea!

[Astro @ May 31st 2009 9:01AM]

The article doesn't say about that vast improvements to the controller (which apparently JMICRON have done to it)

[r3loaded @ May 31st 2009 9:06AM]

Yep, tho I presume this new controller will solve those problems. Other controller manufacturers to keep your eyes on are Samsung (found in the Samsung PB22-J and OCZ Summit) and Indilinx (found in the Intel X25/X18 and OCZ Vertex)

[ryan @ May 31st 2009 9:33AM]

The Intel drives don't use Indilinx controllers, they have Intel controllers in them.

[Wwhat @ May 31st 2009 10:02AM]

jmicron has several generations of controllers, their most recent generation isn't that bad, it's slightly outdone by intel, but intel has their annoying slowdown issue making them end up even.
(I know intel updated their firmware, but it's not 100% effective I think I read.)

The problem with SSD's is that you have to check what controller (generation) is used, and the thing is that most shops/boxes don't supply that info, so you need to research before buying, which most consumers find annoying upto a point of just giving up, or worse they just buy on chance and then are forever sour about SSD's if they end up with a 1st generation buggy controller in theirs.

[loosely_coupled @ May 31st 2009 1:49PM]

Do not listen to "Wwhat"

Although this new one promises to be a vast improvement, the CURRENT JMICRON CONTROLLER IS HORRIBLE! I am not talking about a little less performance others, I am talking about 100X slower random write performance, and very poor performance in most metrics. So bad that the majority of reviewers like Anandtech, Bit-Tech, Pcperspective, etc recommend that you don't even consider buying one!

I'm not an expert, but I have done a lot of research. Here are some tips on SSDs that I have learned:

1) NEVER buy an SSD without seeing a review of the exact model from websites like those I mentioned above. SSD performance differs drastically among manufacturers and even within manufacturers different product lines and generations. Even SSDs with the same controller can have somewhat different performance characteristics due to firmware tuning.

2) The controller situation is actually pretty easy to follow. There are only four main companies producing them:

# JMicron - strongly advised to avoid. Many drives using their current controller use two of them in internal RAID 0 which helps a bit, but doesn't solve the primary problem of terrible write performance.

# Indilinx - Their current controllers are very good and competitive with Intel and Samsung. A very popular and highly-rated drive using their controller is the OCZ Vertex which is by far the best performing drive the money. This drive shows none of the rapid slow-down effect of the Intel drives. (although that will probably be fixes soon).

# Samsung - They used to be the only game in town. Their first generation controllers are still far better than any JMicron, but their new controller --- in drives like OCZ's Summit series and Corsair P256 -- is blazingly fast and obliterates older Samsung-based drives.! The sequenital write performance blows away Intels X25-M, although Intel has the advantage with random writes. An overall excellent new chip that readily outperforms Indilinx in most situations.

[loosely_coupled @ May 31st 2009 2:03PM]

(continued from above - got cut off)

# Intel's controller in their X25-M is very good, offering incredibly the industry fastest MLC read speeds and most importantly the fastest random write performance by far. (sequential write performance has been sacrificed for the more-important-to-most-non-server-situations random write performance). The problem is that Intel's controller is only available on Intel's own drives for now, which are very expensive compared to alternatives. Intel's X25M goes for $315 (80GB) or $620 (160GB) where as OCZ's Vertex goes for $220 (60GB), $375 (120GB) or $830 (250GB).

The only problem is that Intel's MLC drive apparently has an issue that causes it to rapidly lose performance after multiple re-writes of the same sector. Although all SSDs experience a slight slow-down as their drives fill up, other drives do not experience the dramatic Intel problem. Most expect this to be fixed with new firmware. Additionally, future SSD firmware from all manufacturers are supposed to use new drive instructions that will essentially completely eliminate the slowdown that well-used SSDs experience.

[Wwhat @ May 31st 2009 6:33PM]

I certainly admit I did not delve into this one and just repeated what I picked up in passing, but uhm, didn't jmicron release new generations twice already? Or am I wrong on that one too? I could be and you seem to be in the position to tell me.

[bebop @ May 31st 2009 8:13PM]

I can relate, having just bought a couple of 64GB Samsung SSDs, the original SLC version and the latest MCL PB22J. Looselycoupled pretty much gives you the straight dope. The vast majority of the SSDs you find for sale are pretty much junk as a boot drive. You can list the good ones on one hand.

My Samsung SLC drive is a smooth operator. Even though the flat out max read and write are "only" 100 and 80 MB/s, respectively, it stays fast no pretty much matter what you throw at it. There is little slowdown when the block sizes get smaller or the IO queue starts to stack up. My subjective experience is the MLC (I have the slower 64GB version, the 128GB and 256Gb are faster to write) drive isn't quite as good under pressure, but given a clear run at something it burns rubber of course.

The benchmarks would indicate that the new MLC is a clean win, but for the 64GB version at any rate I don't think it's cut and dry like that.

[msalivar @ Jun 1st 2009 1:33AM]

A couple quick points. Corsair hasn't released the new MLC drives in 128GB or 64GB yet, so if you got a 64GB Corsair drive, it's last gen and it's performance won't be all that great. You shouldn't have stuttering, or very little, but it won't be the best jack of all trades SSD that the P256 is.

Second, the 'Intel slowdown' is, while reproducible on a test bench, nearly impossible to encounter by chance in the wild. Don't worry about it. The SATA TRIM command will be a good thing, as there is a very real performance hit as cells are used, but it's not a huge deal for consumer use. Here's some good coverage of normal aging on the current top contenders:
http://www.techreport.com/articles.x/16979

[Max @ May 31st 2009 8:42AM]

Why is there a copy of win 7 in the background. Also who buys hp discs.

[superhobo @ May 31st 2009 9:32AM]

I think these broke off leaving the dots behind:
ʔ
ʔ

[Max @ May 31st 2009 9:47AM]

Now I feel sad, I looked at my comment twice before posting to check for spelling mistakes.

[Wwhat @ May 31st 2009 9:57AM]

Lol

[zioncat @ May 31st 2009 8:43AM]

Yes you are a superior idiot.

[Blacky @ May 31st 2009 8:51AM]

Excellent news! Can't wait to get a 250 GB SDD for 150 USD. - one day... one day.

[Beastage @ May 31st 2009 10:52AM]

That's not exactly cheap when you can get a 1.5TB for that.

[yopladas @ May 31st 2009 3:00PM]

And yet there is a difference; namely power consumption, speed and loudness.

[sven vollstag @ May 31st 2009 8:53AM]

...if you were, you would know to write "superior TO everyone"...

[superhobo @ May 31st 2009 9:29AM]

So the controller costs as much as the flash?

[Wwhat @ May 31st 2009 9:55AM]

The power of monopoly eh, if you make the only controller you are sitting pretty.

[Wwhat @ May 31st 2009 9:54AM]

Here's another issue I found, when you use windows it often for no special reason revs up all drives causing a wait and noise surge, which means that if you have a SSD and a secondary classical drive for the big stuff you'll STILL be waiting on a regular basis for the rev-up to have finished, so to compliment SSD either windows7 should not have that stupid behaviour (unlikely that that is true knowing MS), or the classical spinning HD makers should somehow design a larger more clever cache that helps fool windows by saying 'ready' with the cached data while it revs up in the background, although you would still have the noise until it went idle again.
Note that I haven't tried windows7 so I can't say it also does the wake up/rev up of all drives for stupid reasons, but since vista is based on XP and w7 based on vista (and XP) I fear the worse.
It's anyway odd that HD makers didn't incorporate deeper intelligence into drives to try to fix flaws in OS's I think.

[superhobo @ May 31st 2009 12:30PM]

Windows 7 doesn't thrash the disk as much as Vista. Especially if you disable automatic defragging and backup, and most importantly, have enough RAM.

Unless you're talking about something else.

[Wwhat @ May 31st 2009 6:45PM]

I mean that when HD go in idle and power down and then you for instance use the start menu or a rightclick and get the sendto option or whatever event you have, suddenly windows needs to first wake up all drives, for a while I thought it was to gather the icons scattered about but it wakes up drives that are not in use and not linked to, probably because for instance the rightclick on a file then shows the sendto which insists on listing all drives for you to send the selected file to, which incidentally on XP variants is not optional, you cannot exclude drives from that listing the MS sites tells me.
Anyway a drive that went idle and after a while spun down needs to first rev up again, which takes time, in that time windows happily messes up the multitasking and stalls everything.
That is one of the things that annoy me a lot about windows and which attracts me to the concept of SSD's since they don't spin down, but a system with just SSD's either has way too little space or becomes so expensive only a handful of people could afford it.

[Wwhat @ May 31st 2009 10:05AM]

Perhaps he thought that saying 'to' was pushing it beyond the credible :)

Although he was obviously joking about himself there so it wouldn't have mattered.

[statucfish @ May 31st 2009 10:11AM]

Has anyone thought about the differences between NAND and NOR. NAND is inherently terrible for large scale read write operations (anything other than thumb drives)

From Wikipedia: (NOR allows random-access for reading, NAND allows only page access)

This means that not only do you get all of the disadvantages of flash (wear spanned across the blocks after RW operations,) but you forgo the benefits of flash itself, massively faster Read Write operations

Using NAND in this way just doesn't seem worth the risk

[Dracorius @ May 31st 2009 10:34AM]

For as far as i know, all the larger flash memories out there are nand flash. Even the high end SSD's. And thats good, since Nand flash is waaaaaay cheaper than nor flash, And OS'es usually access disk in a per-page fashion. Nor flash is huge, expensive, and only used when per-bit access is really needed.

Cheerz

[iofthestorm @ May 31st 2009 2:20PM]

Uh, I've thought about them a little bit, but it seems like you don't really understand the difference between NOR and NAND. I think with NOR you can randomly access any block easily, but with NAND you only have random access to each page (which I think is around 2kb). NAND is a lot cheaper though, and NOR hasn't really been cost effective for use in larger flash memories. Random page access is usually good enough anyway, and then caches make up for the speed difference.

[why not the LS2LS7? @ May 31st 2009 9:13PM]

NOR costs too much and isn't dense. The largest NOR I've ever heard of is 32MB.

NOR is random access for reads, but to be honest, that's not an issue. Reads are so fast on NAND that reading an entire page (8K) is really quick. Besides, most OSes gather up disk accesses into increments of 2K, 4K or 8K anyway.

NOR is slow to write, because it isn't random access on writes.

The biggest value of NOR is that it is reliable enough that you usually don't need to do wear-leveling.

NOR is a specialty market now. It may all but disappear within 5 years.

[Alvaro @ May 31st 2009 10:41AM]

Isn't the price of SSD more linked to the price of flash memory itself? and that's why n-times more capacity is about n-times the price (for a given brand)?

[PuBeLeSs @ May 31st 2009 11:25AM]

would you rather have a raptor? or an SSD? im not big into SSD's but am building a computer soon.

[J @ May 31st 2009 11:33AM]

Required reading material for anyone serious about dropping any kind of money on an SSD:

http://www.anandtech.com/printarticle.aspx?i=3531

It's long, it's relatively tedious, but you will know more than almost everyone at Engadget if you read it.

[PuBeLeSs @ May 31st 2009 12:44PM]

thank you kind sir

[tikiwk @ May 31st 2009 3:04PM]

J,

Thanks for the link.

[ph0bi0s @ May 31st 2009 10:52PM]

I concur. The article was quite helpful, but not tedious in the slightest. Though I will likely never buy a Jmicron product without some intense scrutiny first.

[Joe @ May 31st 2009 12:15PM]

Just say "no" to jmicron controllers. With the right chipsets, they're fine, but otherwise you are screwed.

I have an older OCZ SSD that was completely unusable on my nVIDIA 780i chipset, but dropped it into my laptop, and it flies.

I will certainly steer clear of anything that says "jmicron controller" in the future.

[BigD145 @ May 31st 2009 4:45PM]

Oh, please yes. Laptop prices will drop with this.

[Sizer @ May 31st 2009 4:59PM]

I splurged for a X25-M 80GB model after the recent firmware upgrade, and it was worth every penny - Win7 x64 boots in 20 seconds even with all my crap loaded onto it. It's just zoom zoom zoom, and so tiny. I'll certainly never get a platter drive on a laptop again.

So it'll be great if prices drop further, though JMicron itself sucks.

[Ian @ May 31st 2009 6:02PM]

Will it suck radioactive monkey balls like the current jmicron controller? Everyone who follows SSD knows their current controller is complete garbage.

[badhack @ Jun 1st 2009 4:50PM]

Is this really news? Every six months for the last two years SSD have doubled in capacity effectively cutting the $/GB in half.