The LSI SandForce Acquisition: Anand's Thoughts
by Anand Lal Shimpi on November 2, 2011 12:11 PM ESTI'd almost given up hope that someone would acquire SandForce. After taking so long to fix the infamous BSOD issue and our own interests beginning to shift away to other vendors, the future of SandForce seemed somehow less important. Just last week however LSI announced its intention to acquire SandForce, a deal that should close in early 2012.
At first glance, the LSI acquisition didn't make a whole lot of sense to me. The one thing SandForce needed was a partner that understood validation and could elevate SandForce's standards in that department. I wasn't thinking broad enough.
Insufficient validation may be what has troubled SandForce's designs in the consumer space for so long, but there was a much bigger problem. SandForce's current customers are companies like Corsair, Kingston, OCZ and OWC. The client list doesn't include traditional hard drive manufacturers, companies like Seagate or Western Digital.
When asked about why they are timid about entering the SSD market, every hard drive manufacturer I've spoken to has pointed at the more conservative nature of their businesses. The hard drive guys apparently like certainty and predictability in their products. The current state of SSD controllers, firmware and designs is anything but that. To date, there is no turnkey solution that a company like Seagate or Western Digital could implement that would give them hard drive-like compatibility and dependability, with all of the benefits of an SSD.
LSI has the ability to fix this.
A quick look at Seagate's Barracuda XT or even the recently announced 1TB-platter Barracuda and you'll find a controller from a single company: LSI. In fact, LSI's semiconductor business is responsible for over 70% of its revenues. LSI wants to be a semiconductor company that plays in the storage space, and when viewed through those glasses - the SandForce acquisition makes a lot of sense.
There are a lot of companies that will work with LSI thanks to its reputation, but not SandForce. That's what LSI is banking on to help grow SandForce's business.
Obviously to do so, LSI needs to address some of SandForce's shortcomings - particularly in the validation department. While LSI wouldn't commit to a timeframe to do so, having been a customer of SandForce itself I do believe that LSI knows what needs to be done moving forward.
LSI's interests in SandForce aren't purely altruistic of course. If the world does shift away from hard drives and towards SSDs, LSI's semiconductor business would suffer. By acquiring one of the highest performing SSD controller manufacturers in the business, LSI has a good chance of being able to compete in the SSD space should it see considerable growth at the expense of hard drives.
What's going to happen to SandForce's existing customers? As a silicon company, LSI is completely fine with being a controller and firmware supplier to SSD manufacturers. LSI reaffirmed it has no desire to enter the standalone consumer or enterprise SSD business.
In fact, the only area where LSI plans to deliver a branded solution driven by SandForce hardware is in the enterprise PCIe space - although once again, LSI says it has no issues supplying SF controllers for competing PCIe SSDs such as the OCZ Z-Drive.
I see the acquisition as a positive for SandForce. I believe that SandForce has the best technology in the industry but suffers from inferior validation. With existing design wins in the consumer hard drive and enterprise markets, LSI clearly understands the customer requirements for delivering a robust storage controller solution. If LSI can extend those learnings to SandForce, this acquisition will have a significant impact on the SSD industry going forward.
22 Comments
View All Comments
skroh - Wednesday, November 2, 2011 - link
Hold on, here! I have been reading SSD reviews at AT for years now. "...our own interests beginning to shift away to other vendors" ? Time and time again, reviews posit SandForce drives as the desirable performance leader, both in its first generation, where it dethroned Intel, and now in its second, where it ups its game and fends off other vendors' attempts to make a comeback. Where are the articles that express this supposed loss of interest in SandForce drives, and what other vendors have allegedly taken SandForce's place in your affections?As far as reliability, in the articles I have seen, concerns and incidents have been mentioned from time to time but their significance has always been minimized. If SandForce is horribly unstable and fatally unreliable, you have utterly failed to communicate that to your audience, but that is what the offhanded comments in THIS article seem to imply.
I don't have a dog in this fight--I bought an Intel G2 on sale back when SandForce 1.0 was just too rich for my blood. But the tone of this article seems completely inconsistent with previous drive comparisons. Which is it? Is SandForce the enthusiast's choice, or a fast Italian sportscar that breaks down twice a month? Seems like there's a real discontinuity between the narrative of the formal reviews over time and the cussing and sweating over the test bench in the back room, if that makes any sense! :)
Beenthere - Wednesday, November 2, 2011 - link
The hardware review sites love to hype new tech, especially when it offers a faster operating environment, but they are not quick to "bite the hand that feeds them" even when there are known defects in the products.Ad revenues pay for most commercial websites and that's why the writers "walk softly" around product defects. It's also why they get hand picked hardware to review, get special mobo BIOS or firmware sent to them overnight - that consumers are unable to get, etc. Try getting tech support on a defective product and see if the company calls you within hours of getting your e-mail or writes a new BIOS for you or does simulation work to try and replicate a defect you found as a unpaid beta tester.
EddyKilowatt - Wednesday, November 2, 2011 - link
Noobish question, but I don't get why bandwidth-hungry SSDs are being disguised as mechanical drives, placed in mechanical drive bays, and bottlenecked by mechanical-legacy cables and serial transfer rates?They're just boards full of chips (the "pick-and-place" allusion above). Why doesn't everyone get rid of the packaging, plug them into a 16-lane PCIe slot, and give them all the bandwidth they need?
Okay, I realize that not everyone in the consumer space has a spare PCIe x16 slot lying around, at least not yet, and also that SSDs don't need 64 Mb/sec speeds... at least not yet.
But still, why aren't we moving in that direction? Is there a big cost add for PCIe that isn't obvious from looking at the boards?
neotiger - Wednesday, November 2, 2011 - link
PCIe isn't enterprise only. See RevoDrive, a consumer PCIe SSD.But most PCIe SSD are for enterprise. Why? Because with a consumer SSD, WAY before you hit bandwidth bottleneck, you're hitting IOPS bottleneck. You don't need to worry about bandwidth unless your drive is doing something like 200K IOPS or more. None of the consumer SSD get anywhere close to that.
So unless you're using a $10,000 SSD like FusionIO that gives you 200k IOPS, PCIe is pointless.
jjj - Wednesday, November 2, 2011 - link
"To date, there is no turnkey solution that a company like Seagate or Western Digital could implement that would give them hard drive-like compatibility and dependability, with all of the benefits of an SSD.LSI has the ability to fix this"
I would remind you that:
- Marvell is the leader in HDD controller and also has a pretty good SSD controller that they tend to sell to others (every so often).
- Seagate has a certain agreement with Samsung
- Hitachi colaborates with Intel
So LSI is not saving the day here,they are just trying to compete.
chiddy - Thursday, November 3, 2011 - link
Absolutely!In addition:
- Western Digital has been shipping SSDs under their SiliconEdge brand for quite some time now.
ShieTar - Friday, November 4, 2011 - link
And let's not forget that Samsung itself is very much "a company like Seagate or Western Digital" when it comes to hard drives.I would assume the main reason why Seagate and WD show no interest in the SSD market would rather be the fact that the profit margins that can be had by assembling bought NAND with a bought controller do not meet the expectations of such huge companies.
Let's face it, todays SSD market is split into Intel and Samsung on the one hand, that basically can afford to give you a free controller with your order of NAND, or smaller companies that make do with the small profit margins to be had.
Personally, I am kind of hoping that LSI's reason for the SandForce acquisition are aiming at the development of a Hybrid-Controller. Let's keep in mind that OCZs first two RevoDrive generations did consist of a combination of LSI Raid and Sandforce controllers, coupled via SATA. I imagine that a decent amount of silicium as well as latencies may be saved if the full functionality of both controllers is included in a single chip, without the need to communicate through a serial link.
Powerlurker - Friday, November 4, 2011 - link
Let's not forget that they'd also be competing against companies like Kingston that have been making memory products for years already and have tons of experience in that realm.dealcorn - Friday, November 4, 2011 - link
Your analysis of the strategic business case for the acquisition sounds spot on. Also, recall LSI's great success with its SAS/SATA controllers. The writing is on the wall that Intel will substantively trash that business by migrating SAS/SATA functionality to it's Xeon chip-sets at discount prices (soon coming to Atom Xeon?). This acquisition may enable a LSI response that goes beyond "roll over and play dead". Perhaps, there is a sweet spot with precisely the right combination of SAS IP, Flash, and silicon pixie dust to create the "Wow" from some segment of the storage business.glad2meetu - Saturday, November 5, 2011 - link
LSI failed in its internal development of a SSD controller. Instead it decided to buy a solution with this acquisition of SandForce. LSI has had to buy or merge with companies in order to get customers. For example, Seagate came from its merger with Agere.The main profit margins for SSD are for enterprise. And PCIe is not a great solution for enterprise since it is not very scalable. The consumer market will have higher volumes but at much lower margins.
Many companies are working on SSD controllers. The main differentiating factor is the firmware rather than the controller. A good SSD design will have a simple controller and good firmware. As most consumers are aware, SSD designs have reliability issues due to firmware.
Overall, I am skeptical about this aquistion. Even with my skepticism, there is a decent chance that this aquisition may prove to be beneficial since LSI doesn't have much internal R&D nor many customers.