Gas Guzzler

[zandor]

Ok, I've managed to overload a 400W PSU. It seemed to run ok, but an alarm kept going off every time I started the machine up telling me to go to the power section in my bios.
For now I've yanked the GF2MX200 board I was going to use to run triplehead w/ independant resolutions, so I'm either in spanned or dualhead, which is ok for now.
At any rate, now I need a PSU rated for more than 28 amps on the 3.3V line (that's the one that was/is running low). It's at 3.2 now & no alarms are going off, but it drops to bit over 3.1V if I stick that GF2 in.

Just in case anyone's curious, here's the rest of the specs:
Dual PIII-1000 CUs
Asus CUV4X-D
512MB Micron 7E registered ECC SDRAM
Matrox Parhelia 128 retail
3com 905b ethernet
3 lvd scsi hdds (18.2@10k, 9.1@10k, 18.2@7.2k)
Plextor 40max CD & Plextor 8/20 CD-R
Symbios based (Tekram) SCSI
Sound Blaster Live 5.1
4 12V case fans (along w/ the chip coolers & PSU fan)
IBM Model M keyboard
Logitech Trackman FX
21" (20" V) Hitachi CM823F
19" (18" V) Hitachi CM715
20" (18" V) Sun (Sony) GDM-20E20

Oh, and the Parhelia seems to work fine. At least it runs the reef demo... I'll have to go try some games now. I've also already found a bug & a workaround. It seems the mode list gets messed up sometimes if you boot in triplehead. Switching to single head & back (no reboot needed) seems to take care of the problem.

[Agent31]

I was about to recommend the <a href="http://www.topower.com.tw/home/atx12v_520p4.htm" target="_blank">520Watter</a> I just bought, but I looked at the specs and saw that it too is only rated to 28amps on the 3.3V rail ...

[Wombat]

The cheap-o solution is to get a second supply and throw the HDs on it

[zandor]

I'm not sure that would help.
It's the 3.3V line that's overloaded, and the drives only draw on the 5V & 12V wires.
I guess I'll have to do a bit of shopping. I'm not in a real big hurry. Until they release Linux drivers for the Parhelia I'm going to be doing any real work (programming) on my Sparc anyway.
Any comments on Enermax? They've got a few units with 38 amps @3.3V output.

[Greebe]

You can bridge two ATX PS's together which is what I think Wombat was trying to say.

Typically the 3.3v supply rail is generated from the 5v supply side. It can handle only so much of one or the other ie greater 5vdc load less available to the 3.3vdc rail and vise versa.

[Agent31]

my flatmate has an oldish Enermax in his Athlon 1.2 box, he's not really taxing it though so I can't say how good it is under stress...

[Tomasz]

Zandor,
The Antec TruePower power supplies receive my wholehearted recommendation. The 480 provides 30 amps on 3.3 and the 550 provides 32 amps:

http://www.antec-inc.com/p_truehome.html

Personally, I have used several of the 330 and 430 models without any problems, and as a positive side-effect they are probably the quietest commercial power supplies I have ever owned.

[Tomasz]

Oh, and by the way, these power supplies contain dedicated 3.3V, 5V, and 12V output circuits (hence the name "truepower") so you do not run into trouble on the 3.3V circuit if the 5V circuit is overloaded. This might solve your problem...

Did I mention I really like these power supplies and use them exclusively for my systems?
I sound like a sales rep or something
Anyways, good luck with your choice.

[VJ]

Are your harddrives spinning up at the same time (esp. the 10K drives) ?

I have mine set to spin up upon SCSI-start signal (is a jumper setting), as 10K drives tend to drain the powersupply when spinning up (they consume far less power once there are up and running). Now, when I turn my system on, no drives are spinning; when the SCSI-controller scans the bus, it first gives a signal to my first drive (id 0: Quantum Atlas 10K), waits for it to spin, and next gives a signal to my second drive (id 1: IBM Ultrastar 36 LZW).

Most LVD-drives can either spin up:
1. immediatly (mostly this is default)
2. upon start command
3. after a delay (calculated based on id number)

Just check the jumpers in the manual.


Jörg

edit: I'm running this on a 250 W (not sure though, it could as well be a 350 W)
Pentium II 450
384 MB SDRAM
Matrox Millenium G200
Adaptec 2940UW
-id0: Quantum Atlas 10K (internal)
-id1: IBM Ultrastar 36 LZX (internal, also 10K)
-id4: Plextor 4220 (internal)
Adaptec 2906
-id2: Agfa Snapscan 1236s
-id5: Iomege Zip 250 (external)
Toshiba 1612 DVD (on IDE primary master, internal)
3com 3c900b combo (ethernet)
4 80mm casefans (in a LianLi PC70)

(sorry, I just read your 3.3 V is giving problems...)

[LvR]

Typically the 3.3v supply rail is generated from the 5v supply side

I doubt that. All of the supplies I have seen use a dedicated 3.3V winding in addition to the 5V and 12V windings on the main switching transformer. I have never seen a second switchmode regulator to do the conversion from 5V to 3.3V being employed.

A linear regulator is also out - its simply too lossy to be employed in this situation (1.7V to be dropped at around 30Amp = 50Watt+ on a 300Watt supply) and on top of that 1.7V headroom is extremely marginal to constuct a high current/high power linear regulator from.

I would also advise against "bridging" ANY regulators together in order to improve capacity - its a guaranteed method of eventually generating copious amounts of smoke from one or both of the supplies even if the output voltages are exactly the same.

[KvHagedorn]

I have about as much junk as you do (3-10k scsi drives, dual P3s, CD, CDRW, Lots of Fans, etc.) I have a PC Power and cooling TurboCool 450 that seems to have done the trick for me. They now have upgraded this model to 475 watts. Check their site:

http://www.pcpowercooling.com/