[rescue] looking for SGI system, cheap
Sheldon T. Hall
shel at tandem.artell.net
Fri Jan 27 13:53:50 CST 2006
Quoth Steven M Jones ...
> On Fri, Jan 27, 2006 at 10:17:21AM -0800, Sheldon T. Hall wrote:
> > The primary exception to the "reasonably fast" statement is
> web browsing;
> > all the browsers I've tried are excruciatingly slow to
> load, and have pretty
> > dismal page-rendering speeds, too.
> This reminds me... I have a 250MHz R4400 purple Indigo2, and back
> around 1999-2000 I was running distributed.net rc5-64 clients on a
> lot of machines. Anytime I tried to run it on my I2, or on the
> Origin 200 at work, it yielded abysmal performance numbers. Just
> an order of magnitude off what I would have expected for the class
> of hardware.
> I was just wondering if these two anecdotal accounts add up to an
> explanation in some list members mind. Sure, it could just be
> "poor coding," but ...
I think it is exactly that. Well, not so much "poor coding" as just coding
aimed at the machines most likely to run the program.
I ran Distributed Folding (Blueprint.org/U. of Toronto) on my 4xR10K
Challenge L, and did very well. I started late, but when the experiment
closed, I was still gaining ground on the leaders.
Since I can't find anything better to do now, I'm running the Golumb Ruler
thing from distributed.net, and performance is very, very poor.
Given the realities of the installed base available to run distributed.net,
I'd be astounded if the code were not optimized for Intel processors, with
everything else relegated to the "if it runs at all, swell" classification.
If anyone knows of a medical-research distributed computing project that
runs well on IRIX/MIPS, please let me know.
More information about the rescue