Intel plans 600GB SSDs for end of 2010

[Jowy Atreides]

lol, you don't understand the point of SSDs

You dont understand business economics.

People want high capacity. Why buy SSD when a TB of space is cheap on regular discs.

And before you refute me. I am kinda a big deal in business. :p

[KB24LA]

Please improve the write speeds. It's really the only significant weakness

agree, that will make overall perfomance much better

[snoro]

dont need a 600gb ssd, give me 2 30gb ssd with good speed at a nice pirce and i will put them in raid 0 and laugh at everyone face, sdd are not for capacity, they are used for their speed. SO when someone is going to make a 30gb ssd with 400mb/s write and read speed. That will be real news for me.

[N19h7m4r3]

Capacity and price is what I want.

SSD's can slowly have increases in speed as time goes on, same for mechanical drives.

I've tired of seeing 128GB drives going for well over 250. It's bloody pathetic.

We need 600GB drives and bigger, and we need prices to drop.

[Smartidiot89]

33nm?
seriously, what node is that supposed to be? 24nm?

I think the next node will be 28nm tbh but don't quote me on that

Also read that many in the industry are afraid that NAND modules won't be able to scale sub-20nm so we might reach an halt soon :/

EDIT: As the feature size of flash memory cells reach the minimum limit (currently estimated ~20 nm), further Flash density increases will be driven by greater levels of MLC, possibly 3-D stacking of transistors, and process improvements. Even with these advances, it may be impossible to economically scale Flash to smaller and smaller dimensions. Many promising new technologies (such as FeRAM, MRAM, PMC, PCM, and others) are under investigation and development as possible more scalable replacements for Flash.

http://en.wikipedia.org/wiki/Flash_m...sh_scalability

[Machinus]

You dont understand business economics.

People want high capacity. Why buy SSD when a TB of space is cheap on regular discs.

And before you refute me. I am kinda a big deal in business. :p

It's not nearly that simple. SSDs are more expensive and have different characteristics. People don't need a lot of HD space for applications, it's mostly for storage, which is a waste of a SSD. So they wouldn't be buying SSDs anyway. They will never be as cheap as HDDs anyway, so it doesn't matter.

The controller is the weakest part of SSDs, and is the part that needs to be worked on. Intel quadrupling the number of chips and then quadrupling the price isn't really useful unless they will have a better controller.

[sdsdv10]

You dont understand business economics.

I may not understand business economics, but to me that is not the point. I do understand my economics. At this point, I would be perfectly happy with a fast OS/boot SSD drive (~150-300GB) and can store all my media on mechanical drives. Sequentially write speeds of regular harddrives isn't that bad and I don't write large files all that often. As for reading them, it is usually when I'm watching a video or listening to music, where the speed of current mechanical drives is just fine. It isn't like I'm going to watch a movie at warp velocity just because a SSD can read data at incredible speeds. For now, I fine with a nice speed boost on boot times or launching applications. It's easier for me as I don't play games, so I can make do with a 160GB OS/application drive, for others this might not be the case. I'm just waiting for the Intel 2gen 160GB drives to drop below ~$400 (from the current $550).

While I may not think the same as the typical consumer, I believe I'm more akin to the typical reader of these forums (where I see many people spending large dollars on graphic cards or RAID controllers). As such, I think I'm one of the target audience for the current SSD, as well as many others on these boards.

[Xoulz]

capacity isn't fine, but why not improve both?

Really..? What are you filling a 120GB SSD with..?

[zalbard]

You're a HEAVY media user, studio production with terabytes of uncompressed videos

Why use a SSD for storage purposes?..

[kadozer]

600gb already? sheesh I only just bought a measley 80gb for Win 7.

[gumballguy]

Intel wont touch this for a while, last time I saw their roadmap it went to 320gb max and that was actually like 2011 or something. It blew my mind how far down their roadmap it was when their current drives allow room for enough chips to take the 160gb -> 320gb. It was quite clear Intel didn't think their was any real competition, but had allowed in their design to make their drives bigger, just in case. The ONLY thing that'll get intel to move to that size is competition. And honestly if its 600 flat I dont believe its true... try 640.

Ill make do with a non gimped 80gb (I dont like that the 160 is faster via firmware) but would love a fast 120...

[Tech Dav]

Why use a SSD for storage purposes?..

SSD is perfect for storage particularly in a server environment. It has everything right now but capacity and capcities and needs are growing all the time with multimedia.
SSD will eventually replace mechanical drive or make it way too cheap as it is becoming more and more.

[ethernal]

Capacity and price is what I want.

SSD's can slowly have increases in speed as time goes on, same for mechanical drives.

I've tired of seeing 128GB drives going for well over 250. It's bloody pathetic.

We need 600GB drives and bigger, and we need prices to drop.

Well of course we want prices to drop but to call a 128GB SSD going for going over $250 "pathetic" is a somewhat offensive statement towards the industry. Any kind of modern silicon chip production is extremely capital intensive and costly. The rapid advancement of fabrication technology means that (physical) capital has to continue to be created to produce higher performance and higher capacity chips because old capital that is more than two or three process generations behind becomes noncompetitive in high performance/capacity applications (imagine buying a 130nm Intel/AMD chip today). While old fabs (90nm, 130nm, etc) still find use in embedded and similar applications these tend to be low margin products and soon the marginal cost of running the fab rapidly rise above marginal revenue meaning that the fab must be decommissioned or upgraded.

Because of the rapidly declining marginal revenue of fabs (over time, not quantity) there must be some degree of "profit taking" during the early life of a fab where the marginal revenue is greatly above marginal cost. However, this profit taking isn't really profit taking so much as accounting for the high financial expenditures in capital which raises the average cost significantly. In other words the fab produces only slightly above its long run average cost over its life with the marginal cost above the long run average cost relating to the firm's eventual profit (obviously). So while yes - the 128GB (which is, remember, over 64 billion transistors) does not cost $200 to produce (assuming the other $50 in the price is packaging, marketing, shipping, etc) ignoring capital costs, if you include capital costs it is likely relatively close to that amount.

You may point out Intel's high margins as a counterpoint but remember that Intel is rather the exception to the rule in the fabrication industry because of their strong market position (relating to market dominance and intellectual property) allowing them to charge far above long run manufacturing costs. Most other semiconductor companies - especially those doing simple production (DRAM, Flash, etc) - do not have this luxury and tend to run on a relatively low long run margin. Flash memory production is complicated; flash memory design is simple.

While it is true that the Flash market is something of an oligopoly (with two or three firms producing the vast majority of flash memory on the market, with a few fringe firms as well) one still generally comes close to a competitive market's equilibrium price and quantity in oligopolies due either Bertrand competition, or in this case given the generally fixed capacity of semiconductor fabs, Cournot competition. In other words (barring collusion, which, while not impossible, the numbers - at least those post 2007 (and the investigation by the DoJ has, to the best of my knowledge, been inconclusive at best) - do not support) the market price of raw flash memory is relatively close to what perfect competition should produce. Given the ease of packaging the raw flash memory into a SSD (see: the dozens of different suppliers of SSD products) the main price driver for SSD is the raw flash memory and not the specific packaging so if the raw flash memory market is (relatively) competitive - which, again, using rudimentary Courtnot competition as a model indicates that it's likely "close enough" to perfect competition, especially considering the elastic demand of SSD given HDD substitutes - then the final packaged product is likely competitive as well. Meaning that the price is about as "cheap" as the market can make it all things considered.

In short it's not pathetic that it costs that much. 64 billion transistors on the latest nodes (which are needed to create the appropriate performance and capacity) costs a good bit. If you don't like it go build your own $2.5 billion fab and do a better job. Just be thankful that transistor fabrication technology advances as rapidly as it does that you know that prices will come down in the future in a rather quick way.

Yes, IAAECTHWITSI (I Am An Economic Consultant That Has Worked in the Semiconductor Industry).

[zads]

This should be a 3-bit MLC device.

As for the SSD debate, since I am the lead engineer of SSD devices at my company.. here's my breakdown of the computer storage heirarchy:

Price/GB (highest)
Access Speed (highest)
.
.
.
CPU Cache
.
DRAM
.
SSD
.
HDD
.
.
.
Price/GB (lowest)
Access Speed (lowest)


Find your storage capacity and speed/bandwidth requirements, factor in your budget, and buy your storage accordingly.

Some customers need high storage capacity at the expense of speed (we sell HDDs as well), some need even faster storage (we sell DRAM as well).

[Unoid]

I'll take 300gb SSDwith amazing trasnfer speeds, and thats it. I would buy/build my own 2+TB NAS with 5400rpm WD green HDD's

[NapalmV5]

bullocks!! come on intel.. just say it! youre getting out of ssd aswell

no ?!? wheres the new slc ssd @ ??

[Wesker]

...In short it's not pathetic that it costs that much. 64 billion transistors on the latest nodes (which are needed to create the appropriate performance and capacity) costs a good bit. If you don't like it go build your own $2.5 billion fab and do a better job. Just be thankful that transistor fabrication technology advances as rapidly as it does that you know that prices will come down in the future in a rather quick way.

Yes, IAAECTHWITSI (I Am An Economic Consultant That Has Worked in the Semiconductor Industry).

Excellent post.

I think people are kidding themselves when they expect SSD's to approach HDD capacity in the next few years. I have doubts that SSD's will ever match HDD's when it comes to capacity or cost per gigabyte.

SSD's will be used for their speed, HDD's will be used for their capacity. I see the two being in harmony in many systems. Maybe we might see a hybrid of SSD's and HDD's from the storage manufacturers in the future--A drive with both NAND flash cells and platters that can be used and address as a single drive.

[aka1nas]

Excellent post.

I think people are kidding themselves when they expect SSD's to approach HDD capacity in the next few years. I have doubts that SSD's will ever match HDD's when it comes to capacity or cost per gigabyte.

SSD's will be used for their speed, HDD's will be used for their capacity. I see the two being in harmony in many systems. Maybe we might see a hybrid of SSD's and HDD's from the storage manufacturers in the future--A drive with both NAND flash cells and platters that can be used and address as a single drive.

Conventional disk drives capacities are stagnating as they are again reaching physical limits with magnetic storage. NAND-based SSDs at least have a few more generations of capacity doubling due to process shrinks. There are also denser/cheaper solid-state storage technologies coming up from most of the major players.

I do expect SSDs to reach hard drive capacities within 5 years, though I doubt they will reach price parities at those sizes with conventional drives. If they get to the point that they are less than, say, twice the cost/GB of magnetic drives than the performance advantages of an SSD start to outweigh the marginal cost increase in many applications.

I already have a pair of 128GB SSDs, I have little trouble filling them up with OS/Apps/Games. Media storage still goes on a conventional drive for now, but ideally you would want everything on the SSD.

[Rob94hawk]

I'm not even that impressed with the speed of these SSD's. Yeah the speed is good but c'mon. You want to impress me make reads and writes 1 TB/s. Enough with this slow ass milking of the computer enthusiasts. These companies are doing with these SSD's what Intel and AMD are doing with their cpu's. Releasing the latest product at only a hundred or two hundred mhz higher than the previous one. Bah.

[N19h7m4r3]

Well of course we want prices to drop but to call a 128GB SSD going for going over $250 "pathetic" is a somewhat offensive statement towards the industry. Any kind of modern silicon chip production is extremely capital intensive and costly. The rapid advancement of fabrication technology means that (physical) capital has to continue to be created to produce higher performance and higher capacity chips because old capital that is more than two or three process generations behind becomes noncompetitive in high performance/capacity applications (imagine buying a 130nm Intel/AMD chip today). While old fabs (90nm, 130nm, etc) still find use in embedded and similar applications these tend to be low margin products and soon the marginal cost of running the fab rapidly rise above marginal revenue meaning that the fab must be decommissioned or upgraded.

Because of the rapidly declining marginal revenue of fabs (over time, not quantity) there must be some degree of "profit taking" during the early life of a fab where the marginal revenue is greatly above marginal cost. However, this profit taking isn't really profit taking so much as accounting for the high financial expenditures in capital which raises the average cost significantly. In other words the fab produces only slightly above its long run average cost over its life with the marginal cost above the long run average cost relating to the firm's eventual profit (obviously). So while yes - the 128GB (which is, remember, over 64 billion transistors) does not cost $200 to produce (assuming the other $50 in the price is packaging, marketing, shipping, etc) ignoring capital costs, if you include capital costs it is likely relatively close to that amount.

You may point out Intel's high margins as a counterpoint but remember that Intel is rather the exception to the rule in the fabrication industry because of their strong market position (relating to market dominance and intellectual property) allowing them to charge far above long run manufacturing costs. Most other semiconductor companies - especially those doing simple production (DRAM, Flash, etc) - do not have this luxury and tend to run on a relatively low long run margin. Flash memory production is complicated; flash memory design is simple.

While it is true that the Flash market is something of an oligopoly (with two or three firms producing the vast majority of flash memory on the market, with a few fringe firms as well) one still generally comes close to a competitive market's equilibrium price and quantity in oligopolies due either Bertrand competition, or in this case given the generally fixed capacity of semiconductor fabs, Cournot competition. In other words (barring collusion, which, while not impossible, the numbers - at least those post 2007 (and the investigation by the DoJ has, to the best of my knowledge, been inconclusive at best) - do not support) the market price of raw flash memory is relatively close to what perfect competition should produce. Given the ease of packaging the raw flash memory into a SSD (see: the dozens of different suppliers of SSD products) the main price driver for SSD is the raw flash memory and not the specific packaging so if the raw flash memory market is (relatively) competitive - which, again, using rudimentary Courtnot competition as a model indicates that it's likely "close enough" to perfect competition, especially considering the elastic demand of SSD given HDD substitutes - then the final packaged product is likely competitive as well. Meaning that the price is about as "cheap" as the market can make it all things considered.

In short it's not pathetic that it costs that much. 64 billion transistors on the latest nodes (which are needed to create the appropriate performance and capacity) costs a good bit. If you don't like it go build your own $2.5 billion fab and do a better job. Just be thankful that transistor fabrication technology advances as rapidly as it does that you know that prices will come down in the future in a rather quick way.

Yes, IAAECTHWITSI (I Am An Economic Consultant That Has Worked in the Semiconductor Industry).

An extremely well worded and well put arguement.

I retract my claim that it is pathetic and will now state I am displeased with the prices.

Although the bolded part was a bit unneccesary I can understand your frustration.

I find that I am just unimpressed by the size to cost ration of SSDs compared to that of when HDDs were expanding in size and speed, and as such might have gone a bit over the top.

[Xoulz]

SSD is perfect for storage particularly in a server environment. It has everything right now but capacity and capcities and needs are growing all the time with multimedia.
SSD will eventually replace mechanical drive or make it way too cheap as it is becoming more and more.

Honestly, do you run a server out of your house? If a business decides SSD capacity is NEEDED, then 2 or 3 SSDs shouldn't be a problem.

But, for end user's, high capacity SSD are unnecessary.

[ethernal]

An extremely well worded and well put arguement.

I retract my claim that it is pathetic and will now state I am displeased with the prices.

Although the bolded part was a bit unneccesary I can understand your frustration.

I find that I am just unimpressed by the size to cost ration of SSDs compared to that of when HDDs were expanding in size and speed, and as such might have gone a bit over the top.

Thank you for your kind words - and you are quite correct - that was a little unnecessary and I apologize. I was (to put it lightly) rather inebriated when I wrote that post. I apologize if I sounded a little rough around the edges.

I understand where you are coming from regarding the cost/size relative to hard drives. I think one of the driving factors in that perception is the underlying difference in how SSD's and HDD's are created. HDD's have a relatively fixed cost to create - sure there are some differences based upon the number of platters, etc, but once the technology to make a drive is there it costs around the same to make a 320GB drive as a 2TB drive: all it is is a magnetic coating on a couple of spinning discs. Much of the cost in making the hard drive is in the universal components (the precision spindle, etc) that is unattached to the drive's capacity. Because of this there is a rather quick migration to higher end drives because producers want to maximize what they can sell their (relatively fixed cost) item for: a 320GB single platter drive costs them maybe $38 to make; a 2TB drive might cost $49. So obviously there is a push to place the highest capacity drive at a market penetration price point. That is why there is so little variance in the hard drive market: a 320GB drive costs about half as much as a 2TB drive. This has always been the case in the hard drive market (except on the extreme high end and recently released products).

Unlike the hard drive market, however, SSD's costs are directly related to their capacity: it costs twice as much to produce 128GB of Flash over 64GB of Flash. There may be some small, small savings in the packaging of the device (per unit storage) but the packaging stage is a relatively small cost compared to the raw flash; in addition the higher capacities tend to require greater interconnect costs (stronger controllers, more complex circuitry, etc) that dim even this "scale" factor.

It is somewhat a paradigm shift in thinking when compared to hard drives but it brings non-volatile storage more into the realms of the rest of the tech industry: generally, per unit of performance, the marginal price per unit of performance increases as overall performance increases. For example - an E8600 may outperform an E8400 but the marginal cost of that little extra bit of performance was much greater than the average cost of the E8400's performance. This is different from the hard drive industry where, in general (especially the recent past), the marginal cost of extra space actually falls as you increase in storage space. The storage industry will never be anything like the processor industry (where you pay ridiculous premiums for a couple of extra stock hertz) because flash memory producers do not have the same market position that Intel et al have so their marginal costs at increasing performance units (storage) will be more constant with some variation in either direction. But the prime point remains that SSD production is an entirely different game than HDD production when it comes to market dynamics, cost per unit storage, and market penetration of certain capacities.

[prava]

@ethernal: for what it's worth not only is nice to see a well put post (even when its so technical that non-native speakers like me have to read it twice) but also is nice to have an inside POV as the ones outside doesn't have such knowledge nor (normally) ways to get it.

On topic: how much speculation there is at the moment? I can't believe that the 34nm process hasn't brought any price-drop yet. Ok, intel drivers are not only the fastest (for destop-usage) but, at least here, they are the cheapest gb/$...but if we compare the price drop between 50nm and 34nm drives (without talking about intel) within the same company there is no significant differente...even though they even have slower controllers, just as in the G.Skill Falcon vs Falcon II case (where the difference is about 20% and the performance is not even the same althouth they are done on different nodes). Thats why I wonder If I'm missing something or if there is so much demand that prices just wont drop.

[Levish]

~ three more years and the capacity / $ issue won't be so sore, we'll probably-hopefully need a next gen SATA also
SATA1200GB/sec?

Although SSD's are already priced so that they are within reason for those wanting to make the switch and don't need large amounts of storage space. I'm very eager to see the performance of this 600GB SSD

[ethernal]

@prava - Thanks - writing diatribes about random stuff is how I avoid doing (real) work so I'm glad you approve of my procrastination habits.

Disclaimer: I don't work in flash memory production planning or anything - the last work I did was early this year with an embedded chip designer - so it's not like I'm on the forefront of the latest juicy fab gossip or anything.

To answer your question - there isn't ever any speculation on a rapidly depreciating asset or anything (no one is hording a bunch of 34nm flash chips hoping for a global flash shortage next year ). To answer your "why haven't prices come down?" question - it goes back to what I mentioned in my first post. Fabs cost a lot and rapidly lose their ability to generate revenue. Flash producers (Intel or otherwise) will sell their products at a price as high as the market will bear at their production capacity. The fact that it is 34nm does not prescribe it any inherent cost decrease in what is provided at the market - and indeed it could provide a cost benefit over older process generations if high capacity per unit volume is important or if the transistor performance is superior. Again - fabs are extremely capital intensive and operate at a specific capacity. You want to operate at 100% capacity because marginal cost is decreasing rapidly and variable costs are low, but you can't really pay your machines overtime if you need more: chip production is already a 24/7/365 procedure. You sell your chips for as high as you can when they first come out to help pay for the massive capital investment related to the fab creation.

The Flash memory market is not like the x86 market where we are accustomed to the price drop of older products when a new process tech comes out (which is only because AMD and Intel are price setters and they do it for product differentiation purposes). While Flash memory is relatively differentiated in some aspects (performance, cell lifespan, etc) it can be viewed as somewhat of a commodity for practical modeling. Price decreases will be much more gradual and sloping (which may be why people don't realize how much Flash prices have already come down - they're used to obvious drops in prices on x86 chips or graphics cards) because the price will be whatever the market supports. The x86 market and graphics card market has price setters and not price takers like the Flash market. As more and more flash fabs move to smaller nodes supply will increase and the prices will drop; the older generations of fabs will have paid off much of their capital through the high price they received during their heyday and so the older generation fabs can cope with the lower price of their product while the new fabs milk their technology advantage to pay off their capital investment.

Obviously the SSD market has created additional demand for Flash - although if I had to take a guess I would wager that the Flash market for cell phones and iPods is much larger in overall total capacity. The simple fact of the matter is that if anything there is a slight overcapacity issue in the market. See this article just as an example. Flash prices continue to fall at an exponential rate - see here, here (2008), and here (2007). In short.. be patient. They're coming down - but it just seems slow when you're waiting to get a SSD.