Beyond his own CPU ambitions, Yassaie appears to see three upsides in purchasing MIPS’ operations.
is the Android factor. MIPS is one of the only three CPU architectures
directly supported by the Android OS. The other two are ARM and Intel.
He expects the deal to settle the current uncertainty surrounding MIPS, giving Imagination a chance to pitch MIPS to the industry as
a legitimate CPU choice – “a real alternative to ARM and Intel” –
especially in the Android world. That, by extension, means the global
Second is the China factor. Imagination will
inherit MIPS licensees, including the China-based fabless company Ingenic.
Gardner explained, “Ingenic’s CEO learned to design MIPS-based
processors at the Chinese Academy of Sciences, which has spawned several
MIPS designs, including the Loongson processor based on MIPS64.” He
added, “Imagination may be able to use its MIPS acquisition to gain
traction for its GPUs in the companies that favor the MIPS
Third is MIPS’ current product portfolio. Gardner said Imagination, in continuing to develop MIPS cores, will “start in a good position with the three-pronged Aptiv family of
cores, including the high-performance ProAptiv line.” He noted that
ARM’s recently announced Cortex-A57 “will achieve 3.9 CoreMark/MHz,
which is below the MIPS ProAptiv score of 4.5 CoreMark/MHz.” Gardner
added, “The MIPS cores should also consume less die area and power than
the high-end ARM CPUs.”
Yassaie stressed that the MIPS deal is “not an asset
acquisition,” but “a strategic acquisition.” Imagination
has a strong interest in “MIPS’ business, people and prospects,” he concluded.
82 patents Imgaination is buying are strictly related to MIPS architecture, essential for Imagination to develop MIPS core further.
While 498 patents ARM-led consortium is buying is more on fundamental processing.
It's important to note that Imagination is granted with "royalty-free, perpetual licence" to all of the remaining 498 patents it did not purchase.
So Imagination seems to be replacing its own home-grown CPU for an industry standard one. Where does Imagination sell it's CPUs - do they expect to compete with ARM with their graphics-cpu combination - which markets?
Rick, don't you and Junko work at the same company? I see UBM editors writing to each other often in article comments. Maybe you could peer-review or read each others' stories before publishing them rather than asking the questions with the rest of the readers.
As for the topic of this article, the MIPS architecture has had very limited success outside of networking and is not in the broad spectrum of markets that ARM continues to seep into. Outside of networking, MIPS's primary value is in its patent portfolio. Indeed, that is something that both ARM and Imagination should be able to take advantage of in their own designs and/or monetize through licensing - a fundamental part of their businesses.
We editors like to join and spark the online conversation in the "open source" world rather than do it privately. We get more crowd sourcing smarts from the engineering world that way.
Question for you, Tom: Does this deal significantly upset the balance of processor patents in the ARM vs. Intel camps?
I'm guessing ARM bought MIPS for their Simultaneous Multithreading (SMT or "Hyper-threading") know-how. ARM has steadfastly refused to implement SMT (and made some grossly inaccurate statements about SMT's power/perf) despite the fact that practically every other new CPU on the market (including Intel Medfield) has it.
When ARM eventually does implement SMT they'll probably refer to it by something like "Argon Mist" and claim that it's unique to the industry.
ARM's interest squarely rests on the general patent protection. By being a part of the consortium which bought 498 patents out of MIPS' large patent portfolio (580 patents to be exact), they seek for the protection from any future law suits.
From what I understand, AST -- the consortium -- is not in the business of litigation, but rather, it exists to make sure these essential patents, such as those by MIPS, won't fall into the hands of patent trolls.
How is MIPS right move in the 3.5G-4G as reported by you going? Value of MIPS patents was value of MIPS patents when Sandeep Vij took over. Business sold for $60M - what a debacle. Blind trust in a mediocre executive by a reporter and being a free advertising vehicle is probably not considered good reporting.
MIPS will not be completely dead, more like a zombie or the walking dead. It will still be used like PowerPC and SH are still used but no longer one of the mainstream high profile application processors like Intel and ARM. It may once again come back to life but it would require backing by a major player with deep pockets. China could have run with MIPS but they have a weak track record with software.
MIPS died because it lost the critical mass software support like Intel and ARM retain. If you don't have the software ecosystem and application developers your hardware is useless. Yes, MIPS did have software support but it was declining badly and getting left behind by ARM and Intel. Worse MIPS was eating majority of that software development cost themselves. Their biggest customer Broadcom was not helping them at all in this area.
Meanwhile ARM and Intel have built armies of software developers supporting their ecosystems and even better they are subsidized by their deep pocket customers. For example, ARM has Google and Linaro (i.e. TI, Samsung, IBM, Freescale), and Intel which doesn't even need subsidizing has enterprise server guys (e.g. IBM, RedHat, Novell, Oracle, Dell, ..). And that's not including all the open source guys and smaller companies working for free.
Also, MIPS didn't have a cheap development board which open source software hackers could support. All the Linaro guys have cheap development boards for Android and Linux hackers to work on. Cheap Intel boxes are everywhere. What did MIPS software guys have to work on that cost less than $300? Nothing!!! MIPS's customers like Broadcom don't offer cheap development boards only expensive closed proprietary reference designs. Yes, MicroChip has the PIC32 but that's a very low end MIPS core which requires their proprietary toolchain. Now compare that to a $35 Raspberry Pi (ARM-based) that is powerful enough to run Linux.
Moral of the story is its all about the software people.
Join our online Radio Show on Friday 11th July starting at 2:00pm Eastern, when EETimes editor of all things fun and interesting, Max Maxfield, and embedded systems expert, Jack Ganssle, will debate as to just what is, and is not, and embedded system.