News & Analysis
Comment
markhahn
regarding the (mistaken) claim that ARM has already won: observe that pretty ...
junko.yoshida
HI, Bill-MIPS. Thanks for chiming in. It's easy for anyone outside MIPS to ...
Four reasons why MIPS new cores may make it relevant again
Junko Yoshida
5/11/2012 11:54 AM EDT
NEW YORK – MIPS Technologies rolled out this week a new generation of microprocessor cores called Aptiv.
With the new cores’ smaller die size and reduced energy consumption compared to ARM’s midrange core like A15, MIPS is hoping that the new family of cores can put the company back on track. “We will gain leadership again” on the competitive microprocessor IP market, according to Gideon Intrater, vice president of marketing at MIPS. “We believe that Aptiv can beat A15 by a large margin.”
MIPS is at a crossroads, however. Some analysts see Aptiv arriving in the market a little too late.
Indeed, until this week, MIPS had not introduced new cores for two years. Meanwhile, its rival ARM busily pumped out a series of new microprocessor cores.
Why did MIPS, once a technology leader, fall behind? One explanation is MIPS’s acquisition of ChipIdea, an analog and mixed-signal IP supplier, in 2007. This deal apparently posed a huge distraction. MIPS’ plan then was to offer its customers a ‘virtual SoC’ by creating a strong IP arsenal – going beyond microprocessor cores – to mount a challenge to ARM. The marriage ended, however, after 18 months. “Once we decided to sell ChipIdea in 2009, we were able to put an emphasis back on microprocessor core development,” said Intrater.
MIPS is also facing an even bigger upheaval: the company’s potential sale. Recent speculation that “MIPS is up for sale” has not died down, and MIPS has neither confirmed nor denied the reports. Gary Mobley, senior research analyst at The Benchmark Company, calls the sale “a distinct possibility,” as MIPS has been shopped around at various times in the past 10 years. But he added: “It is important to note that if the company intends to sell, it better do so soon before potential licensees grow more concerned regarding MIPS’ long-term processor roadmap.”
Against this backdrop, MIPS is betting on Aptiv to turn things around. The Aptiv family, consisting of high-performance proAptiv, multi-threaded interAptiv and highly-efficient microAptiv, is the fruit of the company’s substantial investment and development efforts over the last three years. While the Aptiv family addresses distinct performance levels for applications across mobile, home entertainment, networking and embedded, MIPS’s focus is squarely on mobile.
“We see the mobile market the most important, because of sheer size,” said Intrater. “Even if we get a small share, it makes a big difference.”
But then, the $64,000 question is: If you are already an ARM licensee (most likely you are, if you are in the mobile market), why even consider licensing MIPS now?
MIPS’ Intrater said, “Companies are upgrading their cores all the time. We think that our Aptiv can give them compelling reasons to switch to MIPS core.”
EE Times asked a few industry analysts for their opinions.
1. Need to keep the competition healthy
J. Scott Gardner, a senior analyst at The Linley Group explained: “SoC developers need healthy competition among the suppliers of licensable CPU cores. MIPS is the most viable alternative to ARM for most of these applications.
“Even though ARM has grown much larger than MIPS, the ARM licensees will undoubtedly give some consideration to MIPS Aptiv as a way to differentiate products and perhaps gain negotiating leverage with ARM.”
Jon Peddie, President of Jon Peddie Research, said, “You need to have a second source.” But he added: “It depends on the platform. Mobile phones will be a tough sell, tablets are better, Game consoles, STBs, industrial controls, routers, automotive, etc. are opportunities for MIPS.”
2. It’s all about power efficiency and cost
Based on published EEMBC CoreMark performance estimates, the Aptiv family “should deliver better performance per MHz than any of the cores currently offered by ARM,” according to Gardner. “The ProAptiv has 50% higher CoreMark/MHz than Cortext-A9, and Microprocessor Report expects that ProAptiv will have significantly-higher CoreMark/MHz scores than Cortex-A15. The higher IPC should allow operation at a lower frequency and power than competing ARM cores. For many of these applications, power efficiency and cost have become the primary differentiators.”
In other words, “When compared to systems based on Cortex-A15, an SoC designer using MIPS Aptiv should be able to integrate almost twice as many CPU cores into the same silicon area and power budget,” Gardner explained.
3. Heritage and range
In Peddie’s mind, MIPS’ “powerful instruction set with a heritage going back to the first 64-bit workstations and servers” should still be one of the big reasons to consider MIPS even today. Further, “MIPS is the only 128-bit processor – Sony’s Emotion engine in the PS2,” he said. MIPS also offers “a good range of products with flexibility from 8-bit to 64 bit,” Peddie added.
4. China factor
Not to be missed in considering MIPS is perhaps MIPS cores’ impact in China. Peddie noted that “Chinese home grown Loongson processor is based on MIPS, and if you think having instruction-set compatibility with it is important (and I do) then it’s a slam dunk no brainer.”
|
|
|
|
|
Navigate to related information
junko.yoshida
5/11/2012 1:58 PM EDT
A couple of things I forgot to mention in this story, which should matter if you are looking into Aptiv.
1. MIPS did tell us that they already picked up three licensees for the Aptiv family.
2. MIPS is making a headway in the so-called "eco-system," accorindg to MIPS. Philips recently launched its first tablet in China, using MIPS core running Android 4. Philips' tablet is fully certified as it passed Compatible Test Suite (CTS).
Sign in to Reply
jakmal
5/11/2012 6:39 PM EDT
While it might be easy for new chip companies to go in for MIPS, I think one of the main issues with existing ones doing the shift is the extensive re-work of the internal software infrastructure. Current day SoCs need a fast turn around time, and the time needed to rework the software / debug with a new processor IP is just not there.
Getting licensees is one thing, having them tape out production silicon with that IP is another. MIPS needs its customers to show some working 1074K silicon in products fast. More than 18 months have passed since the 1074K was released for licensing. I haven't seen any SoCs announced with that processor.
I want MIPS to succeed against ARM as much as I want AMD to put up a good fight against Intel in the x86 space. Having choices with equivalent or better effectiveness and efficiency will be good for the ecosystem (chip companies and end consumers) as a whole.
Sign in to Reply
hm
5/12/2012 8:03 AM EDT
@Junko: Very lucid story and feel very nice to read. Need more of similar coverage.
It is never too late. With these achievements, MIPS has potential for big market capture. Along with performance boost, they will also offer better customer support, reference designs, design reviews for customer and lower cost.
Sign in to Reply
moronda
5/12/2012 3:13 PM EDT
Sorry, but this is simply laughable. ARM has won the battle decisively.
It's really a sad story. Sandeep Vij was brought in as the savior. I think he is just basically getting his face rubbed into the pavement.
That's why he wants to sell. Only a fool would design in a processor from a company being sold.
Sign in to Reply
Bert22306
5/12/2012 5:00 PM EDT
The four reasons are trumpted, however, by this intention to sell the operation. It might be nice for the industry as a whole to have this second source, but who's going to commit to MIPS if they don't know whether it will survive, or in what form it will re-emerge after the presumed sale?
My thought upon reading the article? "Swan song."
Sign in to Reply
fergie65
5/12/2012 5:33 PM EDT
Junko, nice piece. As you mention, competition is good. It is certainly not appropriate for me to provide specific feedback on the competitive positioning MIPS has shared as part of this product announcement. Suffice to say, we at ARM should not, and will not, disregard any new announcement of rival technology. With the Aptiv family spelling "M", "I" and "P" as you go up(With the Cortex family, we spelt A, R, M going down) , I assume MIPS is leaving a placeholder for an "S" family. Name guesses anyone? I am somewhat surprised at the timing and nature of the release......I had assumed that MIPS would be releasing more details about their 64-bit "Prodigy" core, announced in early 2011. But, like ARM, MIPS clearly see a contined strong market for 32-bit processors
Sign in to Reply
peter.clarke
5/13/2012 2:25 PM EDT
@fergie
Good catch on the M-I-P family naming.
Makes the intention to go head-to-head against A-R-M even more obvious. But intention is not execution.
Sign in to Reply
tb1
5/13/2012 2:40 PM EDT
"ARM has won the battle decisively."
They've won the battle, but have they won the war? With smartphones, the instruction set doesn't matter because you don't run compiled code, so it is ALL about processor power and battery life.
If Samsung/Nokia/HTC/whoever could come out with a phone that was faster than anyone else's yet the battery lasted longer, then that phone would own the market.
Because of this, companies would spare no expense to solve the 'reworking of internal software infrastructure' issue.
But ARM is a moving target. It remains to be seen whether MIPS can truly have enough of a technical lead for long enough to get the design wins.
Sign in to Reply
jeremybirch
5/14/2012 9:05 AM EDT
"you don't run compiled code" - sorry?
On iPhone you are generally running compiled objectiveC. I know some platforms are aimed at running java/javascript - but underlying both of those is compiled code too (eg in the browser). Ok - with Android you might hope the heavy lifting s done by Google - but you cannot get away from a certain amount of compiled code and hence instruction set dependency.
And if the main objective is high performance with low power consumption you are not going to optimise that whilst having multiple layers of interpreter and virtualization, because you will be running more code than necessary.
Sign in to Reply
tb1
5/14/2012 12:02 PM EDT
I was thinking of Android, not iPhone. Android apps are not compiled code--you can even run them on an Android simulator running on your PC (and the simulator isn't simulating Arm assembly code).
In fact, Motorola is coming out with an Android phone that uses an Intel Atom processor later this year.
Certainly Android itself has to be compiled to run on the processor. But from a customer point of view, any processor will run the apps, so they will tend to prefer the phone that runs the fastest with the longest battery life.
Sign in to Reply
WW Thinker
5/14/2012 9:53 PM EDT
Are you sure that no Android app uses compiled code? You should talk to Ingenic and its customers or to Philips China. I bet that if you try any of these 7" tablets in US or in Europe, you are going to see something broken. In the ideal world, there is no compiled code, in the real world, optimization will inevitably lead to the necessity of compiled codes.
Sign in to Reply
moronda
5/13/2012 3:43 PM EDT
ARM is stable and things work. Porting to a different processor has huge risk. Anyone that tells me that switching processors is a breeze I do not believe.
Anyway, I'm sure the moral there is completely in the toilet. What are the MIPS employees doing? Brushing up the resume to go to ARM, Intel, Freescale, Cavium, Broadcom, etc... If they are smart.
Sign in to Reply
Bill-MIPS
5/15/2012 1:23 AM EDT
As a member of the MIPS engineering team, I can tell you that I am working on some of the most exciting and challenging work of my career. These are not minor updates to old products, but completely new architectures built upon MIPS’ deep expertise in processor architecture. Some of the engineers I work with daily are from the original Silicon Graphics MIPS team, and others are from well-known processor companies. The work here is intense, exciting and continually intellectually challenging. Sitting next to me is a compiler team working on sophisticated projects, down the hall is a CAD team working on advanced nodes and on the other side of me are engineers working on OoOE pipelines. Our software team is doing amazing work, having recently completed the work necessary to get our NDK into the official Android source tree. In this highly competitive market, we have agility, expertise, and a fast and exciting work environment.
Sign in to Reply
junko.yoshida
5/15/2012 10:33 AM EDT
HI, Bill-MIPS. Thanks for chiming in. It's easy for anyone outside MIPS to second guess what MIPS engineers are doing. What we need is more voice from people like you!
Sign in to Reply
KC9
5/14/2012 2:03 PM EDT
We can also see the fact ARM vs. MIPS from the silicon process point of view. ARM is now moving forward to the 22nm process. MIPS is far behind ARM in terms of the silicon process but MIPS still has some advantage for cost and energy consumption if they start using a 32nm process.
Sign in to Reply
WW Thinker
5/14/2012 9:55 PM EDT
In this regard, the release of hardened A15 core on TSMC 28nm process is definitely an advantage at the moment.
Sign in to Reply
prabhakar_deosthali
5/15/2012 1:30 AM EDT
Are MIPS new release processors code compatible to ARMs? Like those AMD and Intel?. In that case at least MIPS has a chance to be the second source to ARM
Sign in to Reply
markhahn
6/15/2012 11:35 PM EDT
regarding the (mistaken) claim that ARM has already won: observe that pretty much all wifi routers are MIPS, not to mention a lot of NAS and other embedded networking. those are a lot less salient as the ARM in your phone, but at least as significant in ecosystem.
Sign in to Reply