Gartner on flash SSDs and HDD price trends, no price parity

Discussion of Features for the News Log or Forum
Post Reply
Dolby Digital
Posts: 7
Joined: Sat May 24, 2008 12:00 am
Location: Belleair Bluffs, Fl

Gartner on flash SSDs and HDD price trends, no price parity

Post by TheDaddy » Thu May 29, 2008 10:03 pm

Gartner on flash SSDs and HDD price trends, no price parity by end of 2010

John Monroe, Research Vice President, Storage Markets Technology and Service Provider Research, at Gartner, sent in his views on flash solid state drives (SSD) and hard disk drive (HDD) pricing trends. It was written by him with regard to questions B&F asked him following an article on the topic entitled EMC versus Gartner. His reply follows:

I have been constantly amused in recent years to see so many supposedly objective and intelligent folks bite on “the bait of the hype” and subsequently “story board” these technology developments as a simple-minded, single-faceted “war of the worlds” in which there will be clear losers (apparently, HDDs) and clear winners (apparently SSDs), a war in which one “winner” will “take all.” The chips are down…the die is cast…”Our system [Communism] will bury yours [Capitalism]!” (Khrushchev, banging his shoe on the table at the UN back in the early 1960s)…hogwash…

Fact is, both technologies will continue to win, and win big, in different markets in different configurations for different reasons.

I think EMC may be coming at this issue from the "angle" of presumed MSRP end-user cost for xxx GB of highest-end, fully integrated FC-interface-enterprise-grade HDD vs. presumed retail MSRP end-user cost for (perhaps) non-integrated (or minimally integrated) xxx GB of enterprise-grade SSD. I did not hear the presentation or read the presentation materials, I read only your article @, and I cannot agree with the statement above regarding SSD:HDD price/GB parity in 2010 (it may never come, and both technologies will face enormously difficult and costly design and production hurdles in 2011-2012).

Nor can I in any way agree with your statement that “taken at face value the import of Donatelli's presentation is that enterprise SAS drive arrays are doomed. Flash will take over as it gradually trickles down the disk tiers in a drive array, heading towards the SATA capacity tier.”

To begin: All of the top-tier storage system suppliers have for many years charged users a substantial premium for their FC-HDD-based enterprise systems, but the end-user price differentials between SCSI/SAS-based storage systems and FC-based storage systems have nothing to do with the OEM cost of the bare drives – the top-tier OEM cost for FC vs. SCSI/SAS bare HDDs is, in fact, “equivalent” or “nearly equivalent” (<$25 per drive, and zero in some quarters for some capacities).

I will reply to your assertions from the "angle" of prices paid to the drive makers by a large OEM for HDDs vs. SSDs. What the OEMs mark up to end-user price for a “storage system” (or “replacement drive”) is another matter. For example, top-tier enterprise OEMs are currently paying $250 or less for 300GB/15K-rpm/FC drives, but I believe the "MSRP" for bare replacement drives may well be over $2,500 in some cases for some systems; it is curious to see how the retail price for a 300GB drive that costs the OEM the same amount of money can vary greatly from system to system offered by the same OEM – “Well, it’s different microcode, you have to pay more for that version”…

In 4Q/2010, it is my belief that a large, strategic OEM (such as EMC or HP or IBM) will be paying the following approximate prices for 2.5-inch 7200-rpm 500GB "mobile" HDDs (cherry-picked to possibly be "enterprise-grade," capable of 7x24 operation); 3.5-inch 7,200-rpm 1TB "desktop" HDDs ("enterprise-grade" not "consumer-grade," e.g. WD's "RE" ["RAID edition"] or equivalent); and 600GB 10,000-rpm/15,000-rpm SAS 2.5-inch SFF HDDs:

2.5/500GB: $85 (2-disc design @ 7,200-rpm) = $0.17 per GB.

3.5/1TB: $85 (2-disc design @ 7,200-rpm) = $0.085 per GB.

SFF/600GB: $145 (3-disc design @ 10,000-rpm; there should be a 1TB/3-disc version in 2011 … Will 1TB of packaged/enterprise-grade SSD cost EMC or anyone else $145 in 2011? I seriously doubt that; even if there is some stumbling in the areal density advancement, you will still certainly see both 10K-rpm and 15K-rpm 2-disc 600GB SFF designs in 2011, which may cost $109-$130 at that time, figure that in to your forward thinking) = $0.48 per GB.

I cannot speak for the enterprise SSD side, I'll let Joe (Joe Unsworth, who covers NAND flash technology for Gartner) do that (we are both jointly assuming research responsibility for SSDs, but Joe has had more briefings and has more contacts on the enterprise SSD side of things). Both Joe and I agree that a 64GB "PC-grade/packaged” NAND flash SSD, with “adequate” controller and interface electronics, will cost a major OEM about $75 in 4Q/2010 (about the same cost for a 32GB SSD in 2009); unlike HDDs, where you add a platter to double capacity at a fraction of doubling the cost, NAND flash simply scales, so a 256GB SSD would cost about $300 (4 x $75 for a 64GB SSD). This equates to approximately $1.17 per GB. I would imagine that a true enterprise-grade SSD would be much more expensive. But even for a "consumer-grade" SSD, there is no SSD/HDD price/GB “equivalence" – not at the OEM level.

This consumer-grade SSD, on a cost/GB OEM basis, is 6.9 times as expensive as the 500GB 2.5-inch HDD, 13.8 times as expensive as the 1TB 3.5-inch HDD, and 2.4 times as expensive as the 600GB SFF HDD. I doubt the “enterprise-grade” SSD is anywhere close to these numbers, since the enterprise-grade SSD would involve far costlier controller and interface electronics to provide 7x24 functionality and rigorous and exacting data integrity over time (the bare chip costs would be the same, the real magic will be in the controller electronics and attendant microcode). The most competitive recent end-user prices we’ve heard for enterprise-grade HDDs are in the $280/GB range ($28,000 for 1TB).

Say that cost gets cut in half every year (as the NAND flash vendors claim), that’s still $7,000 for 1TB of SSD storage in 2010. How widely integratable this kind of inexpensive enterprise-grade SSD will be in high-end applications is surely open to question (and remember, the enterprise changes slowly: “If it is not necessary to change, it is necessary not to change,” is the mantra of most enterprise CIOs, who demand reams of “proof” in order to make any significant technological change – especially one that may not be backward-compatible with their installed system base -- and are inordinately sensitive to fear, uncertainty and doubt). But I can’t imagine spending $7,000 (even at MSRP) for 1.2TB of enterprise-grade SAS 10K- or 15K-rpm HDDs (qty 2 600GB SFF drives at $3,500/each? that’s too much money!).

SSD and SAS HDD price/GB “equivalence” is just not a likely scenario from any well-reasoned perspective in 2010 (and probably not for 2015 – some brilliant discovery of an effective and inexpensively manufacturable solid-state hatchling of the mind yet to be imagined will displace HDDs, not NAND-flash-based SSDs).

Our scenario is that enterprise-grade SSDs, either as stand-alone units, or in conjunction with traditional HDDs or new enterprise-grade hybrid HHDs, will be widely used as strategic "storage network accelerators” (“SNAs,” our neologism), but will hardly “displace” HDDs in any storage tier.

The enterprise SSDs will provide in some cases "transformational" performance benefits combined with less power -- very compelling features -- but they will certainly not be competitive on a cost/GB basis with SAS HDDs. (We are already forecasting the death of all FC-interface HDDs by 2012/2013, but that’s another topic.) There are also some profound manufacturing and cost issues that will affect the evolution of these technology trends. A technology may be “perfect” as a replacement for some other technology (NAND flash isn’t, let’s ignore that for the moment), but surely its adoption rate will depend upon its availability.

It will be far too costly for the flash industry to take over very many of the world’s HDD terabytes (and how will the flash makers balance enterprise-grade vs. PC-grade vs. consumer-grade flash production?). Who will foot the bill for $50-$100 billion in new factories to support an industry that we project to be fairly flat in revenue at $25 billion/year?

(Ironically, debilitating price wars leading to precipitous price declines are currently far more prevalent and pervasive in the flash industry than in the HDD industry, but these price wars cannot long endure, because they are costing the flash industry too much of its potential profits.) All the Wall St. analysts I’ve spoken with immediately “get” this: “No one.” Never mind the technology and price/GB issues, no one is going to write that check for $50 billion or $100 billion. This is our overall “scenario”:

Questions of Capacity and Cost

-In 2006, the HDD industry shipped a total of 52.186 million terabytes.
- In 2006, the NAND flash industry shipped a total of 0.737 million terabytes.
- In 2009, we predict the HDD industry will ship a total of 197.431 million terabytes.
- In 2011, we predict the HDD industry will ship a total of 514.374 million terabytes.
- Can the NAND flash industry afford to take on even 10%, much less 2%, of the HDD industry's terabytes? Probably not.
- How much fab capacity and at what cost would have to come online in 2009 for the NAND flash industry to ship 20 million terabytes of storage? $60 billion?
- How much fab capacity, and at what cost, would have to come online in 2011 for the NAND flash industry to ship 50 million terabytes of storage? $100 billion?

Chart 3:

-On the way to "solid state" from "moving parts," expect to see diverse "hybrid" solutions.

- NAND flash will be the "tip of the iceberg" - HDDs will power the massive bulk of storage content hidden in the "ocean" of available data.

-Technology providers and end users must realize that a spectrum of opportunity in these emerging applications is far more compelling than any kind of disruptive battlefield.

- Alliances between HDD and NAND flash technology providers should serve to enlarge the library of forms in which storage is a necessary element, thereby enhancing the variety and value of storage products and services.

Posts: 420
Joined: Sat Feb 23, 2008 2:48 am
Location: Santa Clara, CA

Post by DAve » Fri May 30, 2008 2:13 am

Flash is going to become more important for sure.

BDR media is still quite a bit more expensive than HDD for the capacity - it needs to come down considerably to make sense as a storage medium.

Post Reply