Definitely - +1

I won't be able to do anything any time soon, but it would be good to have the info.

Its 8' of pvc to 20' of copper, straight down. The water pushes down through a 1/2" pvc tube and slowly rises against the copper. This makes the loop about 60' and can drop the water to about 50F under no load. The air running by the radiators add the load as the room temp increases, but the water doesn’t seem to go up much. I even turned the heater on to get the room to 100F. Turned it on full blast and the water never got above 60F. It's not fast like an air conditioner, but it gets the job done if you turn it on before the day heats up for a 900 sq.ft lower flat in my neck of the woods.

The reason why I bring this up really is the best crunching is not all about speed. For those of us that crunch 24/7, we run into the same problems the IT department do. Like TCO, or more so in our case, cost/flop ratios. Overclocking may get you speed, but at what costs?

My costs have to stay low as I don’t have a ton of money, but I do power 4 machines that replace the 8 I was running. In doing so, have dropped my cost to flop ratio by greater than a factor of 4, by upgrading to newer tech machines and high efficiency power supplies.

Just moving from single to dual core machines improves things greatly. Not just to improve output here, but to no have to have all the other electronics to run the second processor.

I run big on CPU heatsink metal to keep the need for the CPU fan to run on much more than low, dropping cooling costs.

What I am really saying though is there are many roads that make up the city of best crunching. But this thread is definitely the highway.

I just read through this whole thread, some good stuff in here ... but I have to mention one thing.

P4s with Hyperthreading ... quite frankly for crunching it's a disaster. Turn it off or limit your client to 1 "core" only. The Northwood chips in particular just don't have enough cache to run two units simultaneously. To put it a different way, it's faster to run one unit and then another one than to run two at the same time on those chips. The new i7 Intel chips, however, are a whole different ball game. They do have the cache and the horsepower to run with HTT enabled, effectively getting an 80% improvement in output.

In simple terms:
Hyperthreading on a Pentium 4 = BAD, BAD, VERY BAD.
Hyperthreading on an i7 = GOOD.

I used to think like you and you can certainly find some old posts of mine where I said it.
Until I have finally decided to do some more serious testing on my old P4HT 2.4 GHz to have real figures to discuss.
And to my own surprise I was obliged to recognize that the global work throughput was about 25-30 % higher with two simultaneous tasks versus a single one.
There are drawbacks from a statistical viewpoint because the benchmark is far too high and the claimed credits are overestimated, but if you run Rice WUs as mine is doing since my test the crediting algorithm of this project is reducing the granted credits as they should.
Also the CPU time is counted full on both threads therefore that single processor is recording two days of runtime per day.
But it is doing 25-30 % more real work per day, that's it.

I am also aware that some particular P4HTs are producing errors when running two tasks simultaneously (that may also depend on the selected projects) and in this case it is obvious that Boinc on these failing machines should be set for running one task only (do not disable HT itself as far as possible, it helps even when running a single task). But when it runs fine and clean why lose 25 % of crunching power?

Cheers. Jean.

Not to make you change your choice, simply to state the facts: when I have run these tests it was with HCC WUs because they are the more consistent as far as duration is concerned. With other projects you can never know if a WU runs faster because it ran better or because the work to do was smaller.
More details on these tests are there comparing the quad core 2.4ghz to the p4 3.2ghz - and the i7

After the tests I have finally chosen Rice for normal work because I prefer when WUs are not lasting too long, not because HCC was causing problems.

Cheers. Jean.