backtop


Print 14 comment(s) - last by Lugaidster.. on May 18 at 1:07 AM

Veteran architecture shares some with ARM, but features an even lighter instruction set

The Microprocessor without Interlocked Pipeline Stages (MIPS) reduced instruction set computer (RISC) architecture has undergone a lot of quiet evolution since 1981 when Stanford University Professor John L. Hennessy (today president of the university) began to develop it.  MIPS got serious as a startup in 1984, and then underwent a brief period of ownership by defunct Silicon Graphics (SGI), before a spinoff left it fully independent as MIPS Technologies, Inc. (MIPS).

But MIPS is still far from a household name.

Meanwhile the industry has watched another RISC architecture -- ARM -- rise to prominence, courtesy of ARM Holdings plc (LON:ARMdominance of the mobile and embedded market.

In comparison to ARM, MIPS has some similarities.  Its most recent evolutions have added multiple cores, superscalar processing, out-of-order execution, and other niceties.  However, MIPS features a far leaner instruction set which would make even ARM look bloated.  And unlike ARM, it already has 64-bit processors.

The veteran chipmaker is looking to shake up the market with its new Aptiv family.  High-end chips are dubbed "proAptiv", mid-range "interAptiv", and low-end "microAptiv".
MIPS Aptiv
MIPS Technologies has unleashed a new family of RISC cores. [Image Source: New Electronics]

MIPS Technologies is looking to take the fight directly to ARM, showing off benchmarks that indicate that in the CoreMark CPU-centric benchmark, its proAptiv cores scored 50 percent higher than Cortex-A9 cores.   That would put the proAptiv cores in the same league as ARM's new architecture, Cortex-A15.

Like ARM Holdings, MIPS Technologies hopes to lure in system-on-a-chip designers like Qualcomm, Inc. (QCOM) and NVIDIA Corp. (NVDA) to license its intellectual property cores for use in embedded and mobile device chips.  Unlike ARM Holdings, MIPS Technologies has virtually no such contracts in place in the mobile market.

That said, with a fine historic pedigree and many advanced features MIPS could be the latest architecture to make a splash as the market warms to RISC designs.  As ARM steps up its campaign with Windows PC support and higher clocked chips, is it possible we might see Microsoft Corp. (MSFT) adding MIPS support as yet another x86 alternative?  It's possible, but MIPS Technologies has a long way to go first.

Source: MIPS Technologies



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Bring it on!
By bug77 on 5/16/2012 6:12:09 PM , Rating: 3
The more, the merrier. Even though some will cry "market fragmentation" and such, it's good to see more competitors pushing innovation.




RE: Bring it on!
By JediJeb on 5/16/2012 8:55:52 PM , Rating: 2
I wonder if we will see a return of the early to mid 90s where hardware was diverse and OSes were written for several platforms. Back then WinNT was written for the DEC Alpha which was a 64 bit RISC chip well before 64bit was mainstream.

I even remember one piece of equipment we had in college in the late 80s made by Nicolet that used a 20 bit processor, it was much faster than 8088 and 80286 computers in the rest of the lab at the time. It was the Model 680 listed here http://www.versci.com/1080evolves/index.html now that brings back some memories.


RE: Bring it on!
By nafhan on 5/17/2012 10:12:57 AM , Rating: 2
That would be great, and it would be less of a big deal these days, as there are generally several more abstraction layers between the software the user is running and the actual CPU. They also wouldn't really need to do "versions" of the OS. It'd be more like make a few very small changes and recompile for the different architecture. For Android based phones, at least, it'd likely be similar to the Android on Intel situation: most stuff would just run without any changes, but some games might not work that great.


RE: Bring it on!
By rs2 on 5/17/2012 2:58:48 AM , Rating: 2
Except...none of these chips can be called "x86 alternatives" until they can run all the software than an x86 chip can.


RE: Bring it on!
By bug77 on 5/17/2012 6:02:02 AM , Rating: 2
quote:
Except...none of these chips can be called "x86 alternatives"


Except... it doesn't have to be.


RE: Bring it on!
By maugrimtr on 5/17/2012 8:29:50 AM , Rating: 2
quote:
Except...none of these chips can be called "x86 alternatives" until they can run all the software than an x86 chip can.


They can run the same software - but it's up to the software authors to go and compile versions for RISC architectures. Software is written in such a way to be largely abstracted from CPU specifics. x86 has blunted that practice but it's still very possible for developers to invest some time and release versions for both architectures. It's done in the Linux world all the time with very little hassle.


RE: Bring it on!
By amanojaku on 5/17/2012 10:12:19 AM , Rating: 2
quote:
Except...none of these chips can be called "x86 alternatives" until they can run all the software than an x86 chip can.
Computationally, most CPUs are capable of the same functions. What an x86 chip offers is multimedia capabilities, which are being added to today's ARM chips. The minute a smartphone or tablet started playing music, video and games the feature advantages of x86 disappeared.

All that's saving x86 right now is high performance and an existing library of compiled software. But most users don't care about high performance, as they're switching to phones and tablets. And today's developers are moving to apps. Hell, today's programmers can't even write functional code in a real language without using an IDE. Apple was smart to push apps: you can have a few good programmers selling a few expensive programs, and collect nothing from it, or you can have millions of idiots selling junk for a $1, and collect 30%.


RE: Bring it on!
By Lugaidster on 5/17/2012 10:44:09 AM , Rating: 2
quote:
Hell, today's programmers can't even write functional code in a real language without using an IDE.


I will have to shamelessly agree. I'm the kind of programer that feels like in the dark ages without my IDE by my side and my knowledge and experience in managed (C#, Java, etc.) and scripting languages far outweights the one on "real" languages. I've grown used to them.

I'm not defending either approach but as long as we have both types of programmer then we can have applications of all kinds (I don't pretend to be a kernel hacker in the near future and I'm pretty sure some other programmers have no interest in programming farmvile-esque games). To each their own I say, the wider the market, the more jobs for all of us.


RE: Bring it on!
By Gondor on 5/17/2012 3:44:14 PM , Rating: 2
quote:
Except...none of these chips can be called "x86 alternatives" until they can run all the software than an x86 chip can.


Actually there are MIPS derivatives which have extra instructions added to speed up x86 code execution (virtualized, of course), much like FX!32 on Digital's AXP.

Performance penalty becomes much less significant with those but native code always runs faster, of course:

http://en.wikipedia.org/wiki/Loongson#Hardware-ass...


Networking equipment.
By Trisped on 5/16/2012 6:49:18 PM , Rating: 3
quote:
Unlike ARM Holdings, MIPS Technologies has virtually no such contracts in place in the mobile market.
From what I have been told MIPS is very popular in dedicated networking equipment.

Strangely, I learned both x86 and MIPS assembly in school. MIPS was taught as a platform for processor architecture. Funny thing is that when I started the class a number of other students were implying MIPS was dead.

I like the idea of MIPS coming in and challenging the existing architectures. Sure it makes OS and software development harder, but it is nice to see some variety.

That being said, the trade off between a standard RISC architecture and a more feature enabled architecture (like x86/x64) is that some commands take 2-3x more cycles more on a RISC CPU.




RE: Networking equipment.
By Lugaidster on 5/17/2012 10:49:57 AM , Rating: 2
Unless something drastically changed from my time at computer architecture classes (or I'm not reading your comment correctly as English isn't my first language) I disagree. RISC architectures have simpler ISAa and because of that each instruction takes less time (usually one cycle).

Now, reality is more complex than that, but I'd still say is appreciation is inaccurate at best. Tradeoffs exist but at this point but they aren't as clear because all those architectures stray quite abit from the original definition of RISC and CISC.


RE: Networking equipment.
By Chudilo on 5/17/2012 10:51:26 AM , Rating: 2
Instructions taking multiple cycles to complete is actually a CISC (x86)problem. In an x86 architecture, executing a complex instruction blocks the pipeline until it finished. On a RISK architecture it is broken up into smaller pieces which eliminates the deep pipes . Which means that instructions can be swapped in and out at any cycle. That means none of the pipes are ever blocked for more then one cycle. That results in a consistent response time for most workloads. and better response times.
The only way CISC was able to address this inherent architecture problem is to add more pipes (cores).

Multi-core RISC is why ARM is so much more efficient(per clock cycle).
Bottom line: it's x86 CPUs that take longer to do complex instructions not RISC, because CISC instructions are solid unbreakable multi-cycle instructions.


RE: Networking equipment.
By Lugaidster on 5/18/2012 1:07:36 AM , Rating: 2
As I said in my post above, the general idea is what you posted. But none of the existing designs from MIPS, ARM and Intel are fully RISC or CISC at this point.

Current x86 processors are CISC in their frontend and RISC in their backends as the x86 instructions get decoded into small 1 cycle instructions and sent to the backend to be processed in a pipeline just like a regular RISC machine.

The only issue with x86 these days is that they need a huge frontend (compared to MIPS or RISC designs) to achieve this levels of performance. On the other hand, there are several things that current RISC designs have adopted that aren't very RISCesque.

Things aren't as clear-cut this days. However, ARM and MIPS designs should always have a power advantage over x86 just because of the front-end. Furthermore, with the same real-estate in die area, they should have a performance advantage (but look at AMD vs Intel, clearly that also depends on your engineers' ability to make use of that real-estate for something useful).


By Arsynic on 5/17/2012 10:41:46 AM , Rating: 2
IMO, AMD needs to jump on this quick and buy them out and rebrand the tech.




"Nowadays you can buy a CPU cheaper than the CPU fan." -- Unnamed AMD executive














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki