- It is no secret that ARM is looking to replace Intel x86 chips wherever it can.
- In mobile this has been easy as ARM’s superior power consumption has meant that devices had have had much longer battery life which is a major factor in the user experience.
- Furthermore, all the software has been written for ARM chips meaning that there have been no software compatibility problems.
- Now that the mobile handset market is slowing, ARM needs other avenues of growth and the PC and server markets where Intel has 90% market share look particularly appealing.
- The biggest problem is software.
- Almost all PC applications and all server software are written for the Intel x86 processor and as a result won’t run on an ARM powered device.
- The arrival of Windows 8 on ARM has made a big step towards correcting this problem, but still all of the applications that are not Microsoft in house, need to be re-written.
- This is the biggest factor that keeps ARM from really making an impact in this space as companies like Calxeda, Marvell and Applied Micro Circuits have found.
- 18 months ago the gap between Intel and ARM in terms of power consumption was so huge that it made sense to consider re-writing all of the software.
- For a server farm, power consumption is important because excess energy is dissipated as heat and more energy needs to be expended to cool the machine.
- Hence a power saving helps both in terms of the energy consumed by the device and the energy expended to cool it down.
- These figures are very meaningful and if it could be made to work with an ARM processor then the savings could very substantial.
- However with the advent of Intel’s new transistor structure (trigate), and its mad dash through the geometries, I think that the gap has closed somewhat.
- This makes it less exciting to go through the agony of a huge software port and why I suspect that ideas to replace Intel with ARM have so far not gotten off the ground.
- But wait! What if there was a way to run Intel software on an ARM processor without re-writing it?
- It turns out that there is.
- A Russian development lab (Elbrus) has written an emulator that runs on an ARM processor creating a virtual x86 processor upon which x86 software can run.
- Sounds great but the reality is not quite so hot. (Or in fact much too hot!).
- An emulator is effectively a translator. If the hardware speaks English (ARM) and the software speaks French (x86) then he two can be made to work together with commands and results being translated backwards and forwards by the emulator.
- Anyone who has conducted a meeting through a translator will know what a slow and arduous process this can be and sadly emulators are no different.
- They consume vast amounts of resources.
- In Elbrus’ case the emulator currently produces just 40% of the native ARM performance but the company hopes that that will be up to 80% by the end of next year.
- Given that the power gap between ARM and Intel has closed somewhat neither of these figures cut it in my opinion.
- To me, this means that an ARM powered server running x86 software and the Elbrus emulator, would probably consume more power than a server with an Intel chip. Pointless.
- As a result it does not look that this is silver bullet to solve the ARM software problem in servers and I suspect that the idea will continue to languish with no one wanting to port the software to ARM.
- Hence, I fear that Calxeda, Marvell and Applied Micro Circuits attempts to crack the server market with ARM will probably remain stillborn.
- I would be looking for another application of the technology and chips that they have developed.
Blog Comments
Paul Werp
October 14, 2012 at 3:07 am
I agree with the conclusions