From what you describe, I would also recommend 4670. Good lowest power card. You don't have to play at max resolution.
Or wait for what will come below 5750.

I guess unstable is your MB, not the CPU.
Do you use DDR400? Try at 333, may improve stability.
Good PSU could make the difference, too.
And, I guess, your meter measures VA, not W.

Qfan should be quite good for the cpufan (not so good for chafans).
I would not use 3 chafans. If you don't use a graphic card, probably 1 chafan is too many.

Idle CPU should draw about 8W (1.1V, 1GHz)
Chipset-perhaps 10W
What else? Graphic card?
How do you measure 90W?

kissaragi wrote:...a tingle...

You should not connect the cable with the PC and TV on, you may have killed the card's output (do you have ground in your mains socket?)

A CFL is not resistive. PF about 74%. 59W for a 60W bulb is very close (I measured one 55W). I bought this, cheap, measures real power and PF, claims ±0.5% power accuracy, seems quite good: http://www.nedis.com/datasheets/PRODUCT_DATASHEET_EL-EPM02HQ_ENG.PDF What I wrote earlier about 33W 690G was ...

I saw diff 0.7W idle, 4.2W load (AC). At stock voltage.

Really?
They do not measure that.

You can not "segregate". You can only use a proper meter. Some meters show PF also (better or worse), then you can "calculate".

There is an article (fcenter.ru, remember?), mentioning PM300 as an example of a good meter, working correctly with non-linear loads. I would expect it to, because it is not the cheapest... Too bad if it doesn't.

Multiplying you get VA, not W.
The document is about RMS. You need one on PF.

Incandescent bulb is one thing. Easiest for a meter. "Economic" bulb is another thing. Fluorescent light with a choke - still another.
Edit: I am not saying kill a watt shows wrong numbers. Haven't had one.

Does it measure power? At all?
TrueRMS is one thing, power factor is another. Are you sure your PSU has active PFC (have you opened it)?

(so my AC readings are very accurate, but ...so high, that I'm ashamed to bring them up, about 65W in idle :shock: ) Wrong. You can not measure W. Your real W must be perhaps 42W. If you are lucky, you can get 30W from the wall with a cheap ATX PSU (which may make you re-evaluate the need for a pic...

HTPC? 4670 or 4650 (perhaps an overkill already)

Acetone or similar paint thinner. (White spirit - probably not)

You would need some duct. Plus, the possibility to ruin the heatsink.
Accelero L2 Pro fits and is almost silent.

With the meter, according to which I pay for electricity. Devices, similar to kill a watt, may be very wrong (I had one). Not the first computer I measured. Was surprised myself, but here, in SPCR, I read several reports around 30W. It seems, 740G/760G boards are not lower power than 690G (no reason...

I measured 33W(EDIT:WRONG. The meter is not accurate) idle Biostar TA690G, 4850e, cheap 300W PSU, old 250GB HDD, 4 memory sticks 3GB, two optical drives, floppy, and - HD4670! Should be 25W without 4670 (or not... I'm not sure.) The best low power board?? I mean - what barrier? Broken long ago.

This one drew less: http://www.gigabyte.com.tw/Products/VGA/Products_Overview.aspx?ProductID=2907 "Designed with Top Quality Highly Efficient Components" The cooler is crap, and always on. I did not fight the flicker. I use 1 monitor normally. "The freakin thing" has slower memory (not nice). Consta...

I get similar messages from ASUS Probe, when I launch another monitoring program (Speedfan, Everest). Have nothing to do with the PSU.

I HAVE tested two 4670...

The main thing for low power is lower GPU voltage at idle. If the card can change the GPU voltage, it can be any manufacturer's.
But, with 2 monitors, there is another thing. Flickering of the 2nd one, related to lower frequencies.

tbessie wrote: ..a board that works with memory X and Y...

You don't know that. Do you?

...The power monitor itself appears to work ok.. I would not call that OK. It is OK with resistive load only. I don't understand why such things are even produced. A "real" thing should cost about the same. I bought Unitec EIM-812, the same, about 5W filter only (PSU switched off), measures my comp...

my S754 Venice 3400+ can only be undervolted to 1.35V from 1.40V stock you mean - at stock speed? Any Athlon can be undervolted to 1.1V 1GHz. X2 would draw then about 18W at load (I measured some time ago). A bad chip draws twice what a good one of the same model does. s939 X2 must be really cheap ...

I mean lower than 800.
667, whatever.
Of course, memory must be good at its rated speed. Sometimes it isn't.

The prime suspect is still the memory. You must get the error with other sticks before changing the MB.
You still haven't tried lower memory frequency, have you?
I have two CM2X4096-6400C5, 1.90V ver4.2. Voltage above 1.8V does not improve anything (other brands work better at higher voltage).

ghitz wrote: it needs 8 pin power connector

Did you try with 4 pin?
I have a board with 8 pin and a PSU with 4 pin.
If the CPU is like 140W, then yes, 4 pin may be not enough.

bbzidane wrote:...i can probably live without it ...

Exactly.
Try another decoder. But I think be2300 should be able to play mpeg2 with any decoder.
5050e would be a little better. Buy an Athlon II or Phenom II, if you must and if your MB supports, but the power draw may go up.

780G the best