While industry analysts have touted IP licensors such as ARM, Imagination and CEVA as possible buyers of processor IP licensor MIPS Technologies, we have explored the possibility that a buyer might come out of China (see Would the Chinese buy MIPS?).
Both academic and private sectors in China have long been familiar with MIPS architecture. Loongson, for example, is a family of general-purpose MIPS64 CPU developed at China’s Institute of Computing Technology (ICT) and Chinese Academy of Sciences. The early implementations of the family, though, did not use four instructions patented by MIPS in order to avoid legal issues.
In recent years, ICT, Loogson Technology and Ingenic Semiconductor licensed MIPS architecture directly from MIPS.
Some in the industry, including The Linley Group’s senior analyst J. Scott Gardner, believe that China's move to define a national CPU architecture specification could be boon to MIPS. “MIPS is a strong candidate,” Gardner noted. Others, however, aren’t so sure. An alternative interpretation for China’s CPU initiative is a sign that Chinese officials perceive the uncertainty around MIPS’s future as a problem.
Although acquiring MIPS and its patent portfolio would be a way for a Chinese company to increase its presence in the semiconductor industry, the biggest hurdle to a deal is cultural since China is new to the IP licensing business model and Chinese companies would no doubt struggle to administer it.
– Peter Clarke
Ms. Yoshida - I read through all your articles and it reminds me of "fool me once shame on you, fool me twice shame on me" proverb. MIPS was a good company when founded by the pioneers of computer architecture. It improved the versatility of Intel microprocessors by giving it stiff competition. But now it is in the hands of one of the most incapable management team. But your characterization, "I wouldn't discount Sandeep so quickly either. Much of the difficulties MIPS had gone through are due to what happened even before Sandeep arrived at MIPS. " is terrible. CEO is a total and utter failure. I have seen it first hand. He is off on several vacations and even when around discussion his vacations. Under his leadership, MIPS stock and MIPS is a disaster.
People use other architectures so that they do not have to deal with Intel for non-desktop devices. Intel is smart enough not to have to buy anyone when their margins are so good. If you are not buying MIPS now, why would you buy MIPS from Intel? x86 CISC overtook RISC long ago due to pricing from DEC, Sun, Silicon Graphics and other misdirected (now historical) companies. Intel courted the Asian board manufacturers, the others ate their young.
Why can't the Chinese get MIPS? The DLX/MIPS-I architecture is essentially in the public domain. There is nothing special about the MIPS architecture. Any group can take it and create a chip, like the x86 competitors did a few years ago (TI, ST, IDT, etc). If you want to sell, the ecosystem is important, for in-house use, it does not matter. Power is determined by technology and the number of transistors, not the ISA.
Yes, and Intel once dominated the market for processors.
Processors are different than cell phones. I won't even comment on something as unimportant as social networking on the internet.
And no, I don't work for ARM. I wish I did.
I think the biggest problem for ARM is that they are an IP provider. I think it's hard to get the maximum money out of that. I think the purchase that should be made is for Intel to buy ARM.
Then they could sell tablet and phone chips on the dominant process technology.
No news on an acquisition. My guess is nobody wants to buy it. Why would you need to buy it? Just switch to ARM. It seems like the ploy to generate interest has failed with an X. X = don't care!
The only reason to buy MIPS is to buy the design team. 100 R&D engineers at 2M each would be a 200M sale of the company.
That's the only reason to buy it.
I'll add an observation that a many changes to a License Core, can actually penalize the licensees.
Much of the Cortex M3 business, merely cannibalized ARM7 Microcontrollers - so those ARM7s reached EOL sooner than they would have.
Even the Cortex M3V2 release, devalues those first out the gate, with first Gen M3's - and the Cortex M0+ is another re-spin...
["Good and interesting calls on Microchip and Intel, which we did not include in our runners and riders list.
Perhaps we should have."]
Amazing it never occurred to you ?
If I was Intel, I could buy MIPS, merely to annoy ARM. Intel is already acting as a Fab, so a good question is have they already fabbed the latest MIPS Aptiv cores ? What did the numbers say ?
If I was Microchip, I would buy MIPS to 'secure the road map' and get a design team. They already made the call, to avoid ARM royalties, and the Aptiv cores do fit well.
["But terminating those licensing agreements is not necessarily easily done."]
Who needs to terminate ? You simply let them run their course.
The real issue here is Road maps - the same long-term decision Apple made, in jumping to Intel.
The actual Core matters less and less; what matters more, is engineering reference examples in the latest FAB process.
FABs need choose only one Core to benchmark their latest shrinks, and the natural 'first cab of the rank' is the larger-user-base one.
["Switching to new cores is tricky since it requires an extensive rework of the internal software infrastructure."]
Yes and no.
As a stunning example, note what Apple/Intel did, when they decided Freescale's Road Map, was simply not up to Intel's.
Also note the comment about Google including MIPS, which shows that adding a new core is not as hard as many imagine.
Yes china could leverage the IP the most most likely. But what of the defense products that use MIPs? Is this a security muddled issue also?
Can anybody desribe what is alcking in the MIPs archetechture compared to ARM?
I do not yet understand why they did not improve and what is deficient for their products compared to ARM. How much is hurd mentality and how much is a performance/cost issue?
Blog Doing Math in FPGAs Tom Burke 14 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...