I hope a RISC offering is finally successful in delivering a cheaper server. All those promises in the mid 1980s about RISC is cheaper never materialised. I went through the Sun, DEC and SGI workstations, and they were anything but cheaper. They are now at the bottom of some landfill. If 64-bit is really the issue, then we can only see how poorly SPARC, PowerPC and MIPS have executed in competing with the x86-64. Bit of a pity that ARM decided to only release the architecture specs to silicon partners, as it will take some time to develop simulators and debuggers, and even ARM had to buy compiler vendor Keil to make headway when they had a compiler product that was marketed like some middle eastern market. It started off at $4500 and eventually letters arrived with a $2500 price. Thankfully I never bought at $4500. This is going to take at least two years to unfold, but getting some standard will be like herding cats.
I am not sure anyone noticed that AMCC implemented entire X-Gene core into 2 of 6 Xilinx V6 FPGAs from presentation. Although it is 64-bit processor, I am suspecting its performance might be below ARM 32-bit Cortex-A15 processor, at least per MHz performance. Xilinx just released Virtex-2000T device. If AMCC could get hold of those devices, it will be their dream comes true. They could fit entire chip into one 2000T device.
AMCC started to work with ARM really early. nVidia is probably behind since there was rumor that it was working x86 chip design. I guessed it changed course when MS announced Windows ARM support at beginning of this year. Although lack of vision, nVidia has immense amount of execution power. I will not feel surprised it will catch up from behind.
If I am Intel, I will be worried. I think Intel should seriously consider to license v8 Architecture as they did many years ago to create Xscale. Combining with its processing technology and server experience, it will leave no room for other v8 licensees to grow in this sector.
They haven't said, but I suspect they will try to have something sampling in 2012 and in production in 2013--probably initially leveraging their graphics cores and thus more focused on clients than servers
It was a matter of time. Well done. IBM charged AMCC a wack for the PowerPC 4xx cores, and MIPS did the same to their customers, while ARM charged modest licencing fees (see annual statements). Don't worry too much about the Intel/ ARM competition, the competition between the various ARM vendors will be fierce. It will take a while before you can walk into a store and assemble a motherboard, some RAM, an Intel or AMD 64-bit quad core and cheap hard drive. As for porting software, memory mapped I/O on the ARM and C/C++ software favours RISC; just look at the output of a compiler, and anyway, if you need to write in assembler, anything is better than the curse of the x86 instruction set. There is one Intel, the pie for ARM is cut into over a hundred pieces.
Not at all surprising really.
Power consumption and cooling (flip sides of the same coin) have always been a problem for servers.
The only thing really keeping ARM out of server space has been the lack of 64-bit
MS is going to support ARM starting windows 8. I have a sense that it's also going to be for 64-bit.
The most encouraging part of this news article is a chip company demonstrating a new architecture using FPGA and not wasting huge sum on getting a test vehicle done for architecture eval.
Hope good sense prevails and this continues....
AMCC has come up very fast with the working FPGA prototype of 64bit ARM V8 Series of Processors. When can the silicon will release it will be in high demand if the manufacturer announces the support for the operating system.
My Mom the Radio Star Max MaxfieldPost a comment I've said it before and I'll say it again -- it's a funny old world when you come to think about it. Last Friday lunchtime, for example, I received an email from Tim Levell, the editor for ...
A Book For All Reasons Bernard Cole3 comments Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...