Some of you may likely have seen this already, but I thought it was an interesting read. This is significant especially as it's on an HPC industry focused page.

http://www.hpcwire.com/features/ARM-Pro ... 53179.html

Is Genesi purely focused on consumer market, or is there any room for dabbling in HPC (blades, etc)?

Take a look at http://www.genesi-usa.com/products/hdb By that time, this design was really ahead of time, and energy efficiency was not yet that major concern, as it is now. That was the time, when my Intel P4 desktop system alone used more energy, than all the rest power consumers in my household combined...

Times are changing, and the referred article was just one of many, which say, that ARM will arrive to data centers already this year is small quantities and will have a noticeable presence starting next year.

BTW: here is already an early ARM cluster prototype: http://www.linuxfordevices.com/c/a/News ... s-R1801e-/

_________________
CzP
http://czanik.blogs.balabit.com/

We'll dabble with anything if we have a customer that wants to pay for it. CzP is right: the G4 Blade design was a little ahead of it's time. It had a customer (Freescale) for the design, and we completed it. It served as a great example to customers of Freescale who wanted to make a G4/tsi110 design that was compact and high performance.

Freescale killed that project itself by announcing the MPC8641D which promptly went nowhere.

We tried the same with the TetraPower G5 workstation (also good as a server..) and yet again, it served it's purpose as a reference design but it was too expensive to market. There were no customers outside of the original design contract.

Problem right now is the number of customers who want HPC ARM blades is very small, and the projects around are basically Great Experiments.

For HPC what they really need is an ARM processor that has lots of CPU cores (I don't mean crazy SIMD SPE type units), lots of memory bandwidth but isn't dedicating die space to media decoding or video output. The number of cost-effective ARM processors on the market that do that right now are VERY thin on the ground. That STMicro chip in the article is about the only one that I've seen (and some people might consider it a bit slow, but actually it does the job - once you use one of these you might wonder why Freescale has no ARM chips for comms :)

With quad core A9 chips around the corner and maybe by the end of the year the first A15 designs being experimented with, it may get some attention, but right now.. MX51, MX53, MX6... there is a progression of updating products that customers want, before we get to chips that don't exist yet :D

_________________
Matt Sealey