-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
ARM vs x86?
- Thread starter Orcvader
- Start date
Hrm, ARM is a long way off competing yet.
It may have a place in mobile/low energy platforms like Netbooks/Netops/"Internet appliances" etc.
Where ARM stuff is generally more efficient in terms of power, it may have a place for servers etc. as well, where energy is the largest cost factor in a datacentre, but then it shoots its self in the foot still being 32bit (well, afaik anyway).
It may have a place in mobile/low energy platforms like Netbooks/Netops/"Internet appliances" etc.
Where ARM stuff is generally more efficient in terms of power, it may have a place for servers etc. as well, where energy is the largest cost factor in a datacentre, but then it shoots its self in the foot still being 32bit (well, afaik anyway).
ARM are a company that design RISC processors. So your question is probably "I'm just curious to know what benefits RISC CPUs have over x86?"
The answer is less instructions. x86 was made for the 8086 of the 70's. Back then a lot of assembler programming(where you typethe instructions by hand) was done so it was advantageous to have a processor understand loads of instructions many of which are amalgamation of simpler ones. This saves programming time in Assembler and gives you a smaller program (a useful thing when a few KB's were a big deal).
Now days nobody programs Assembler and nobody cares about a few KB and a reduced instruction set where by the processor only understands the basic instructions is more *efficient*.
What this means to end users is you will have EXE files that are bigger on a RISC CPU OS and you can have a processor with less microcode built in and the 'instruction decode' area of the chip has less work to do.
Since you are working with simpler instruction set it could be easier to implement advanced features that sort the code and perhaps 'paralleling' it for better scaling on multicores and pipe-lining it better. Who knows?
I think ARMs net gen are going to have 'out of order instructions' this might be quite the boon on a RISC design compared to x86 design. Prob why Microsoft has announced a RISC compatible windows and what you said about Nvida.
RISC at the moment is mainly used by specialist systems and low powered systems like phones and some sub-netbooks.
The answer is less instructions. x86 was made for the 8086 of the 70's. Back then a lot of assembler programming(where you typethe instructions by hand) was done so it was advantageous to have a processor understand loads of instructions many of which are amalgamation of simpler ones. This saves programming time in Assembler and gives you a smaller program (a useful thing when a few KB's were a big deal).
Now days nobody programs Assembler and nobody cares about a few KB and a reduced instruction set where by the processor only understands the basic instructions is more *efficient*.
What this means to end users is you will have EXE files that are bigger on a RISC CPU OS and you can have a processor with less microcode built in and the 'instruction decode' area of the chip has less work to do.
Since you are working with simpler instruction set it could be easier to implement advanced features that sort the code and perhaps 'paralleling' it for better scaling on multicores and pipe-lining it better. Who knows?
I think ARMs net gen are going to have 'out of order instructions' this might be quite the boon on a RISC design compared to x86 design. Prob why Microsoft has announced a RISC compatible windows and what you said about Nvida.
RISC at the moment is mainly used by specialist systems and low powered systems like phones and some sub-netbooks.
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,208
- Location
- Billericay, UK
We shoukd all get behind a company like ARM, there British after all and god knows how much of the total processor market they could ocupie in the 10 years.
Yep. They are used in all sorts of devices where you need some basic computing functions but without the cost/size/heat of traditional PC components.
Associate
- Joined
- 27 Jan 2009
- Posts
- 321
- Location
- Cambridge, UK
The ARM (Cambridge-based) Global HQ is just down the road from me. Hmmm wants job!!
@Orcvader
There are lots of difference between the RISC architecture ARM use and the x86 architecture Intel pioneered. Big picture style though... The x86/x64 architecture is an ugly kludge that wins through massive R & D effort in the implementation of the Instruction Set Architecture (ISA) in silicon and a (daily lengthening) binary compatibility treadmill.
I can't see MS Windows running to well on ARM CPUs because only the MS software will be recompiled to run natively - 3rd party apps. (TM) would need to be re-compiled or run in an x86 emulator.
Linux/GNU for the win!!
Hey don't call me biased or nothing!
Off topic I remember ARM were working on a clock-less, asynchronous CPU while I was at uni (~1990)... Wonder what happened to that!
Bob
@Orcvader
There are lots of difference between the RISC architecture ARM use and the x86 architecture Intel pioneered. Big picture style though... The x86/x64 architecture is an ugly kludge that wins through massive R & D effort in the implementation of the Instruction Set Architecture (ISA) in silicon and a (daily lengthening) binary compatibility treadmill.
I can't see MS Windows running to well on ARM CPUs because only the MS software will be recompiled to run natively - 3rd party apps. (TM) would need to be re-compiled or run in an x86 emulator.
Linux/GNU for the win!!
Hey don't call me biased or nothing!
Off topic I remember ARM were working on a clock-less, asynchronous CPU while I was at uni (~1990)... Wonder what happened to that!
Bob
Well it's true that having to support x86 is a limitation but Intel and AMD have been throwing billion after billion into R&D for years now which has helped them keep up with RISC chips. There are many high powered RISC chips such as IBM's PowerPC, which were used in iMac's until a few years ago, but none of them seem to offer a sizeable advantage. If RISC was so much better than CISC why did Apple dump IBM and switch to Intel?
However speed won't decide this. If x86 is still important then ARM won't be able to get a foothold. If x86 isn't important then it's open season and Intel themselves could start making RISC CPU's!
However speed won't decide this. If x86 is still important then ARM won't be able to get a foothold. If x86 isn't important then it's open season and Intel themselves could start making RISC CPU's!
Associate
- Joined
- 27 Jan 2009
- Posts
- 321
- Location
- Cambridge, UK
You are confusing the theoretical with the implementation.
A RISC ISA will scale to smaller transistor count while still supporting out-of-order execution models, etc. It will support a simplier compiler model. It depends largely on an instruction cache to support it's (typically) more modular/simpler instructions = higher bitrate instruction stream (load, store, operate on any 2 registers - in a large bank of registers).
However in practice you only see the advantage at the die size of the ATOM vs. ARM mobile chips (where the ATOM is pants).
People don't seem to appreciate just how large the transistor count of modern CPUs has become:
http://en.wikipedia.org/wiki/File:Transistor_Count_and_Moore%27s_Law_-_2008.svg
That gives Intel a hell of a lot of real estate to play with!!
When Intel can fit shoe horn a truck-load of cache and clever decoding circuits in their modern CPUs... Plus R & D budget in the billions... Plus the need to maintain x86 compatibility with the (predominant) MS Windows OS... Uhmmm I think you see the catch-22 situation we have gotten into!
IBM/Motorola failed to deliver a PowerPC chip which was power efficient enough for Apples long (battery) life laptop range. That had more to do with the 'Process'/implementation of the ISA (die size, transistor leakage, etc.) than the design of the ISA. In modern times it takes massive investment to maintain the constant, year-on-year process feature size shrinking, improving/maintaining transistor leakage, defect management (process features are now many times smaller than a spec of dust), etc. (So large an investment is required that PowerPC was initially developed by a consortium of firms.)
ARM have been clever. They license their ISA and (generally) leave other firms to get down and dirty implementing this ISA.
Bob
The main advantage ARM has for Nvidia is that it in a massive growth market. While the X86 market is very mature market it growth and it growth been slow compared to portable market (mobiles, tablets, games consoles). The other reason is that Nvidia can actual get a licence easily while they had problem with Intel in the past and giving money to ATI for a licence probably does seem a good idea to them.
With regards to comments about ARM process being slower it depends what you measure i'm, sure if you looked at performance per watt they are probably quicker. In the long run they are likely to be a real challenge for Intel and AMD.
With regards to comments about ARM process being slower it depends what you measure i'm, sure if you looked at performance per watt they are probably quicker. In the long run they are likely to be a real challenge for Intel and AMD.