Startup Blurbs

Out of the blue two major semiconductor startup acquisitions occurred during the last week. First, Apple acquired PowerPC chip developer P. A. Semi, and then a day later Sun Microsystems snapped up Montalvo who we just recently wrote about as being in need of some serious funding. Let’s talk about the latter acquisition first. If what the The Register is reporting is correct, then Sun got quite a bargain in acquiring Montalvo for a mere five million dollars, as venture capital firms poured about $73 million into the low-power startup over the last few years. It has been widely speculated that Sun is mostly interested in some of the patents which we discussed here, as well as the engineering team, rather than the actual product that the company was working on. Since the name of the game has changed from a pure MHz race to performance per Watt race, any patents that will enable Sun to produce low-power versions of their Rock and Niagara processors seem like a good investment. In any case, it truly seems that Montalvo simply could not find additional funding and thus opted for being acquired as an exit strategy rather than shutting down shop completely.

The Apple acquisition sent shock waves throughout the industry and was followed by rampant speculation. Initially, when Apple dropped Freescale as the chip supplier for their notebooks, many expected for the company to announce that they were going with P. A. Semi. As a matter of fact, when Apple announced that they were going with Intel, the news even caught the P. A. Semi CEO off guard. But P. A. Semi marched on and delivered a PWRficient 64-Bit dual-core processor earlier this year that consumes somewhere between 5 and 13 Watts with a maximum frequency of 2 GHz. Needless to say, this power envelope is way too high for a mobile part, and is more targeted at the networking and telecommunication space. Indeed, it seems that the processor has been selected by several OEMs for their designs. I would find it highly unlikely for Apple to keep the company alive as is though, and thus it is unlikely that there will be a future roadmap for the PWRficient line which might become a problem for some of P. A. Semi’s existent customers. This in turn could play into the hands of Freescale, which has PowerPC based products in a similar power envelope. Additionally, Broadcom and Qualcomm have been looking into expanding past their existing markets, so this might prove a boon to them as well. It also seems unlikely hat Apple would intend to use the PWEficient processor to go against Intel’s latest Atom processor, which has a lower power envelope and has been well received in the press since being introduced earlier this year. Maybe Apple’s acquisition was along the same lines as Sun’s, mostly for patents and the engineering team. Anyhow, I would not hold my breath for Apple to announce what they intend to do with the acquisition any time soon. In the meantime, for some interesting speculation and comments, take a look at Beyond3D’s article here.

We wrote about Montalvo about a couple months, at which point the company raised a total of $73 million in funding and we were looking forward for it to emerge from stealth-mode and reveal more information about their upcoming low-power x86 processor. A couple of patent applications that we ran across gave us a small glimpse about what the company might be up to but nothing substantial. Now it seems, that there is a possibility that we may never find out, for VentureBeat.com is reporting that Montalvo Systems is very close to running out of money and possibly shutting down altogether. According to their post, the company has grown to over 300 people between Santa Clara and India, and as such began burning through the raised money at a very fast rate. One thing is for certain, developing an x86 chip be it lower-power or even something less complicated that the latest Intel and AMD offerings is not an simple matter. It requires plenty of time and as such money as well, both of which many startups and established companies often times underestimate. VentureBeat.com found that the Montalvo has been shopping around for an additional $100 million but has so far come up empty handed. There have not been any updates or announcements on Montalvo’s website since the last time we checked it, so no final conclusions can be made. However, should things not work out for Montalvo, those not happy with the regular x86 providers have another alternative to fall back onto, namely Centaur Technology. While not exactly a startup, Centaur has been operating a small team out of Austin, TX as an independent subsidiary of VIA over the last few years and has just recently revealed information about the upcoming low-power Isaiah architecture.

When considering the consumer handheld market, one might not usually think of it as a segment that might be well served by a programmable device, such as a field-programmable gate array (FPGA). One reason being that FPGAs are usually produced using processes that are about a couple of generations behind the leading edge. This in turn makes these chips large and expensive, and thus not very suitable for markets were margins are razor thin. However, assuming that this problem can be overcome, suddenly FPGAs become a viable option, especially for applications that might require frequent upgrades due to ever changing standards. Enter SiliconBlue Technologies, a Sunnyvale, California based startup which claims to be the first company that has managed to combine non-volatile memory (NVM) and static RAM (SRAM) at the 65nm process node on a single chip. Unlike purely SRAM based FPGAs that require an external PROM to store the programming bit-stream, and have a configuration penalty, the SiliconBlue iCE family of devices can be turned on instantly. Additionally, since no bit-stream programming is necessary, there is also no danger of bit-stream snooping during the configuration process.  On the other hand, compared to purely NVM based devices, the iCE family is significantly smaller due to smaller SRAM geometries.

The current iCE 65nm family consists for four devices: iCE65L02, iCE65L04, iCE65L08, and iCE65L16, which contain from 1,792 to 15,260 logic cells. The current for the iCE chip family ranges from 25 to 250-uA at 32-KHz, and from 5 to 40-mA at 32MHz. The BGA packages for these devices range from 3x4 to 12x12 mm. Programming is performed in a similar manner to SRAM based FPGAs. The SiliconBlue chips are also available in DiCE packages to allow vendors to combine them with existing chips through multi-chip system-in-package (SiP) technology. Samples are available now and SiliconBlue also offers an iCEMAN evaluation board that features the iCE65L04 and iCECUBE development software.

The combination of NVM and SRAM definitely seems appealing, as well as the low-power consumption and package size. However, the FPGA space is a tough business and it is hard to tell whether SiliconBlue will flourish or fold. Many companies including Quicksilver, Adaptive Silicon, PlusLogic, and Chameleon, just to name a few, have tried taking on the industry big boys Altera and Xilinx, and have either failed or been absorbed. In SiliconBlue’s defense, the fact that company has been able to combine NVM and SRAM on a standard low-power CMOS process in 65nm, and claims to have a solid roadmap towards 45nm, might just give it the edge to stay ahead of the competition.