[HH] Hubris Springs Eternal

[Kurt L Keville]

I think if you have a job that can be solved by an FPGA + Xeon, there is probably a way to do it FPGA only. More importantly, they don't discuss the other 6 or 7 variables taken into consideration with the architecture decisions in a Datacenter. Certainly ARMv8 will be way cheaper, non? Facebook is a big company and there is no way they would sign on to this. What percentage of a Big Customer's processing is even floating point? I have to figure not much... now that is how you make a case against ARMv8... :)

________________________________
From: Federico Lucifredi [flucifredi at acm.org]
Sent: Tuesday, July 01, 2014 1:55 PM
To: Kurt L Keville; hardwarehacking at blu.org
Subject: Re: [HH] Hubris Springs Eternal

Looks more like adefensive move against ARM.  For big (BIIIG) customers like Microsoft, Google, or a financial shop, that kind of manic optimization may make sense (and those same customers would have the pull and money to order a custom SoC with an ARM core if they wanted).

Cute technically... but unless you are one of the few budding skynets, designing the FPGA is the costly hurdle. And if you do a sloppy job, what kind of optimization is it? ;-)

Best -F

On 06/26/2014 06:45 PM, Kurt L Keville wrote:
Just so I'm clear here... Intel wants us to give up searching for a solution to our DataCenter problems and come back to the farm... and all it will cost us is an FPGA... if I wanted to program FPGAs, I don't need a Xeon to go along with it!

http://gigaom.com/2014/06/18/intel-will-offer-a-customizable-chip-to-keep-data-center-clients-happy/



_______________________________________________
Hardwarehacking mailing list
Hardwarehacking at blu.org<mailto:Hardwarehacking at blu.org>
http://lists.blu.org/mailman/listinfo/hardwarehacking




--
_________________________________________
-- "'Problem' is a bleak word for challenge" - Richard Fish
(Federico L. Lucifredi) - flucifredi at acm.org - GnuPG 0x4A73884C
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.blu.org/pipermail/hardwarehacking/attachments/20140701/768743f9/attachment.html>