Moore's Law puts supercomputers out to pasture because power – not just the cost of electricity, but the availability of juice – is the biggest constraint at the big supercomputing centers. And sometimes the lack of budget helps lock the gate, and HPC cloud computing butchers the cow. That's the case with the massive 28-rack …
Being a New Mexico native myself I can tell you wasting money on pie on the sky projects is hardly something New Mexico can afford. As one of the poorest states in the nation this article show much of the reason for it (poor leadership in general, as well as having reservations with 3rd world conditions also doesn't help). I no longer live in New Mexico because of the combination of a lack of good paying jobs and the fact the place is a dump in general (beautiful wilderness but the cities yuck). Still it does tend to have a lot of supercomputers in general due to having several national labs that mostly serve as government welfare for the state these days.
Hope you made it at least as far as Austin, True, the state tends to more or less suck, but Austin is solid gold.
Chopping it up for metal or spares makes sense but chopping it up and giving bits of it to individual universities would just burden them with a load of old scrap. I read it as being old enough to being pre memory controller and CPU being on the same die so these "xeons" would be 5400 series that needed southbridges? Do you get the cooling system and plumbing?
Worse it's 5300 series.
"Oh, and asdf, red or green?"
I'm not asdf, and I'm a Californian, but that's an "and" statement, not an "or" question. IMO, of course. 2000 starts in the hot-house, 200 of which, in 25 varieties, are already 15" tall and showing buds ...
Unusable "super computers" that are under 25 years old should be scrapped. Over 25 years old? Restore 'em & put 'em in a museum ... the more we forget history & all that ;-)
Foggy Computing is a substitute for HPC -- Seriously? So where on the TOP500 list do these hazy computers rank currently?
The "hazy computers" are at a not so hazy 102 in the top 500 http://www.top500.org/site/50321
Location, location, location
The thing to do would be to move it somewhere with hydroelectricity. Then the extra power usage would have no carbon footprint, as well as being less of a cost issue.
Re: Location, location, location
Also, install it somewhere that has a cold climate and use the waste heat to warm the building it's housed in. If we had high speed data backbones, massive computers could be installed in the northern cities and used to heat large buildings.
Re: Location, location, location
That would be somewhere like northern Scandinavia then. Oodles of hydro and damned cold most of the time.
Presumably why Google are planning two data centres up there, one in Hamnina, Finland and one across the Baltic in Luleå, Sweden.
...the first "but, will it run Crysis?"
Re: In before...
It may. If you put GPUs in it.
According to spec one of the possible blades going into Altix ICE has 16 x PCIe for "networking". You cannot plug a FAT GPU into it, but a "thin" GPU (workstation class) should fit nicely.
So it depends which blades have they got. If they got the oldest ones (which have the slot) they can upgrade it at a very low cost to a very decent machine.
Its amazing how cheap second hand, high-end computing equipment is on the fleabay, but when it makes as much noise, kicks out the same amount of heat, and costs almost as much to run as a kettle, it's difficult to get that passed by the home office for personal use.
Re: Home Use
yarp. i can agree there. I have a Dell T7400 Workstation with dual 3ghz quad cores and 64gig of ram. was a goodbye present from a previous company i worked for.
Complete overkill for most home use but flies though anything and I use it for the odd bit of CAD and messing around with VMware (shame the free one only supports 32gig) But the 1kw PSU kick out a right noise and a bit of a rattle. Which means I'm loathe to use it unless i have to.
Maybe we can sell it
To SETI, they are always after computing power.
FInd somewhere with plentiful solar energy and have the world's first renewables powered supercomputer.
Come to think of it, $20M isn't a lot.
Perhaps keep it intact and use it for climate change research?
There's loads of good ideas for its use in charitable environments, but even for free hardware, its the $2m a year running costs that blow it.
You got anywhere in mind for the place with plenty of solar energy? cos i'd love to the the solar panel rig that will support 861kw kit 24/7. The battery cost let alone the PV panels or the space needed etc...
What? No silicon heaven?
But where do all the calculators go?
Re: What? No silicon heaven?
The place near where all the biros go to.
Re: What? No silicon heaven?
Does Zaphod run a secondhand computer business too?
@AC13:01 (was: Re: What? No silicon heaven?)
That would be my wife's purse, then. She's a notorious biro thief ...
(Note to my fellow Yanks: A "biro" is to "ball point pen" what "Xerox" is to "Copier".)
Move it to Barrow, Alaska? I'm told that they have a lot of oil up there to spin the generators with, and they could certainly appreciate the waste heat from a system like that.
Heck, maybe Barrow, Alaska could become the new supercomputing center of the world?
P.S. I'll get my coat, it's the one with the ice cubes in the pockets.
As someone who runs an SIG ICE 8200 machine I can say that it is a very reliable workhorse.
Muticore isn't the answer to everything, and somethign with dual core CPUs can be pretty efficient.
Dual plane Infiniband on these things also.
OK, if you were buying new you would go for something with multicore and faster IB, but if you were getting these racks for free you're getting a good deal. Assumign though that you are getting maintenance on them too.
As someone above questions, are they getting the cooling gear too - power installations, water cooling etc. outlast a generation of kit at least, so that's a good question.
Sell it to
BTEDan as the "Enterprise" main computer.
Ought to be fine, the massive parallelism and redundancy means that it should run correctly in space if there are enough failsafes.