year new

ah. it has been a while, yes, since my last post. a direct result of the work/blog equation -- more work equals less blog and vice versa. not that there isn't a ton of work remaining to be done, just that my machine is rendering a rather complicated composite which a client wants to check up on in an hour or less. so, to expedite the rendering, i'm working elsewhere -- i'm doing the blog bit on a g5 this time, as opposed to my normal windows workstation.

so, what's been happening of late?

random memories below:

fires in the sky

went to watch the fireworks show put on by the makati city government at the ayala triangle with roomie and friends. first time ever for me to watch such a thing live. quite a show. the ringing thumps of the cannons as the fireworks shells streaked into the night, their trajectories barely visible as trails of dim red sparks; the echoing crack of the shell explosions as the pyrotechnics bloomed overhead; the visible, reflexive cringe of the crowd as the expanding sparks looked close enough to touch -- or burn.

one detail. there were a bunch of rather dimmer (by comparison) circular yellow fireworks that 'graced' the sky during lulls in the whole show. it was only later that it occurred to us that it was a neat little conceit on the makati government's part. seems that they had gone so far as to have likely engaged the fireworks specialists to have their logo writ in yellow sparks across the canvas of new year sky... ...their money to burn.

still and all, a nice event. although given the relatively tight confines of the ayala triangle (we were on paseo de roxas next to the new stock exchange tower), we really had to look almost straight up to see the show. hmm. no, lying on the street may not be a good idea either, with people moving around looking up all the time.

amazing tech

by way of backgrounding, at the very start of this company, when the matter of hardware choices were being bruited about, i had the occasion to mention to the powers-that-be that one option for render power at relatively low cost would be the "hammer" line of processors from advanced micro devices: opteron. Unfortunately, given the speed at which this enterprise was put together, the only choice given was a dual intel xeon solution from the company's standard hardware supplier. hence we have soldiered on with that original configuration, now numbering in the mid-20's.

last few weeks, i got word that a brand-name supplier had delivered a dual opteron rackmount server for testing. hmm. ought to be interesting indeed.

and it was. out of the gate, the dual opteron had posted a 13% performance disadvantage versus the dual xeons. what's so interesting about that, you ask. well, it isn't bad at all given that the dual opterons were at a disadvantage of a total of 41% in the gigahertz game too, so the result is quite impressive. then the memory was upped to parity with the xeons (at the start, the dual opteron was delivered with half the ram) -- and the render benchmark improved to a 9% performance disadvantage. i asked for the top of the line opteron in the dual processor line, just to see what it would do versus our top of the line (then) xeons. supplier responded with chips just one level shy of their top offering. results: an amazing 6% advantage over the dual xeons (and still with a 31% GHz shortfall. two more tests, one a kernel change, one a missing software install, and the dual opteron now outperformed the dual xeon to the tune of 11%.

hooray for the underdog.

now we'll just have to see if the numbers are compelling enough for the powers-that-be to switch to the opterons...