AIDA64 has a good CPU+GPU stress test.

Some time ago I had a P3 system:

ASUS P3B-F
P3-933 Coppermine (fanless)
4*128Mb DIMM PC133
AGP Radeon 8500
PCI Ethernet
PCI USB 2.0
ISA Roland LAPC-I
DVD Plextor PX-130A
FDD Epson SD-800
HDD Hitachi 80Gb

It ate 60-70W in Windows 98, and 90-100W in W98 DOS box. Up to 113W though under 3DMark 2000. Measured by Zalman ZM-MFC3 panel.

440BX Tekram P6BX-An
Slocket Tualatin 1400S @ 1.6 with 92mm fan
3x 256MB PC133
Geforce 6200 with 80mm fan
1x SSD
1x HDD
1x DVD
PCI Promise Controller
PCI Ethernet
PCI USB 2.0
ISA AWE64 Gold
2x 120mm case fans

The max draw from the wall ever seen from the above system is 95W AC. At ~75% efficiency from the ancient PSU, that means ~72W DC tops. So as weird as it sounds your 155W PSU is way overkill for the application 😉 .

However if that PSU is long in the tooth the problem might be that the capacitors inside are aged, tired, and passing a lot of nocive ripple into your system.

Please tell us the exact PSU model and year. Pics of the PSU inside would be great for assesment.

EDIT - Could this be the PSU in question?:

My approach for stress testing a power supply is much more crude - I just hang a bunch of 12V halogen lamps off the voltage rails and measure the voltage drop. 😎

d1stortion,
Thanks for the AIDA64 suggestion. I ran the stress test for about 30 minutes and nothing bad happened. I'm going to assume this system is OK.

bristlehog,
Thanks for the comparison. I assume that my Voodoo 3 3000 eats about as much power as your Radeon 8500. In that case, my system probably consumes a similar amount of power as yours.

TELVM,
The PSU is the same IBM FRU number, but a different model. The label on mine reads:
Astec Model aa21480
IBM Part 00n7684
IBM FRU 00n7685
Date Code 0128

So this PSU was made in the middle of 2001. I didn't want to disassemble it for fear of breaking something, but I checked it out with a flashlight as best I could. I couldn't detect any blown or bulging caps.

This PSU works differently from others I've seen, in that its fan doesn't blow air out of the case. It pulls air into the case. Some of the NetVista models had a duct that would direct this airflow onto the CPU's heatsink for cooling. My model had a heatsink/fan combo and no air duct, so this seems a bit counter-productive. If I ever do pull the PSU to conduct a proper inspection, I'm going to turn the fan around so that it ejects hot air out of the case.

I'd really like to get something a bit beefier for this system in the 200+ watt range, but it seems like I'm just being paranoid. Even if this PSU did fail, there are still plenty of NOS replacements available on the web and eBay.

133MHz,
One of these days, I'm going to buy a Kill-A-Watt tool to see just how much power my electronics are consuming.

KT7AGuy wrote:

... Astec Model aa21480 ...

Roger, found pics of its guts at a Polish forum.

Surprisingly well built for such a tiny and venerable PSU! 😎 Quite decent input filtering, thick heatsinks, and good Nichicon jap caps of generous size (2200uF) filtering the output. Good stuff!

I'd stick with it, and just recap if and when the neccessity arose. There's plenty of decent PSU in there for a PIII system.

KT7AGuy wrote:

... If I ever do pull the PSU to conduct a proper inspection, I'm going to turn the fan around so that it ejects hot air out of the case ...

Notice the reason those caps are still alive after 12 years is the good cooling they've been receiving with fresh air directly from the exterior (as opposed to eating hot air from the inside warmed by the CPU). Heat murders electrolytic caps.

TELVM,

Thanks for the info. You've given me confidence in this old PSU. I'm going to leave it alone and won't tamper with the fan. If I ever detect any problems, I'll just consider recapping it.

Your research is much appreciated. Thanks again!

My pleasure!

That's why today's PSUs are at the bottom, to feed them colder air from the outside:

As opposed to heated air from the inside:

Your PSU was designed to last, and lasted it has! 😎

I'm sorry to resurrect such an old thread, but I have an update.

I ran my system with 3DMark 99, 2000, and 2001 with a Belkin Conserve Insight power monitor attached.

At idle in DOS, the machine consumes about 75 watts.
At idle in Win98SE, the machine consumes about 55 watts.
Running 3DMark 99, it averaged about 79 watts, with the occasional peak to 83 watts.
Running 3DMark 2000, it averaged about 80 watts, with the occasional peak to 86 watts.
Running 3DMark 2001, it averaged about 83 watts, with the occasional peak to 87 watts.

Even if I assume a max spike of 90 watts, I'm still 65 watts below the maximum rated for my PSU.

Thanks again to everybody who replied previously!

I'm now tempted to try running a GF3 Ti500 in this system to see how it performs and how much power it consumes.
If I do it, I'll update this thread again later.

The thing to really be concerned with is how much power is drawn from each of the separate rails under worst case conditions. If none of the rails are ever being overloaded, then you're good.
The total wattage of all rails combined isn't normally an issue.

If that motherboard reports voltages, then you can judge from that. Preferably get the voltages to appear in a utility like Speedfan so you can see what they're doing while the system is under a load.
If it doesn't report voltages, then you could measure the rails with a multimeter while the system is under stress.
But really, you can probably trust that if it's been working for this long and hasn't had any problems, it's probably doing fine.

I was rolling and I couldn't resist. I uninstalled the Voodoo 3 and popped in my GF3 Ti500. I used drivers v45.23 for this test.

Immediately, I could tell the GF3 sucks more power. Using the GF3, the system idled at 68 watts in Win98SE. This was an increase of 13 watts over the Voodoo 3. Still, it was less than I expected.

Running 3DMark 99 was not as straightforward as it was with the V3. The first two tests consumed about 96-98 watts. The rest of the tests averaged about 74 watts. The most interesting thing to note was the dramatic improvement the GF3 provided over the V3. Everything about it was better. The framerate was much higher, the visuals were crisper, the colors were deeper, etc. The GF3 was just far superior to the V3 in every way.

3DMark 2000 was just as dramatic an improvement for the GF3 as 3DMark 99 was. Overall, the tests averaged about 100 watts with the occasional peak at 106 watts. The visuals and framerates were just stunning compared to the V3.

3DMark 2001 continued the trend of displaying a dramatic improvement over the V3. The tests averaged about 102 watts with an occasional peak of 109 watts. The GF3 was easily 3X-4X more powerful than a V3, even with the CPU bottleneck of a 1ghz Coppermine and PC133 RAM. It was also still within the realm of being a feasible card to run with this 155W PSU.

If this were my only system... If this system had at least one more free PCI slot... If I had not already built a very nice image for this box... Well, then I would certainly drop the V3 for the GF3 and a pair of V2s in SLI mode. As it is, I think I'll keep this rig as it is: a spare legacy Win98/GLIDE gaming machine that can be beaten, abused, and flogged without guilt.

If I can find a Voodoo 4500 or 5500 AGP for really cheap, I'll consider rebuilding a new image for it. Otherwise, it will continue to run Red Baron 3D nicely like it always has.

Thanks again to everybody who responded to me! 😀

Shamino: I wish I knew how to measure the voltage on each rail. If I could, I would and I would reply to let you know what my findings were with this system. Sadly, I don't know how. One of these days, I'm going to take an intro class for electronics and learn some of that black voodoo magic. In the meantime, rest-assured that any quality PSU over 200W for a Voodoo system is overkill. I have proved it.

KT7AGuy wrote:

I wish I knew how to measure the voltage on each rail. If I could, I would and I would reply to let you know what my findings were with this system. Sadly, I don't know how. One of these days, I'm going to take an intro class for electronics and learn some of that black voodoo magic.

+5V and +12V are easy, you can get them from the molex hard drive power connectors. With the meter in the Volts mode, you just put the negative (black) probe against the case of the power supply (this is your ground), and the positive (red) probe into one of the unused drive connectors. The red wire on those connectors is +5V, the yellow is +12V.
I'm not trying to push you into it or anything, but if you're curious then that's the easiest way to do it.
Anytime you do something like this you do have to be careful not to short anything together with one of the probes. I like to wrap them with tape so only the tip is exposed. But measuring at the drive connectors is pretty safe.
3.3v would be more tricky, since it's not at those connectors.

I have some of this same power supplies lying around, you can have one if you pay shipping.
They are also from IBM thincentre's P4 systems with brand HI-PRO ACBEL or DELTA electrics. same builds format etc.
And total power is 185w. as seen in my attachment.

how do i test a psu (just to see if it starts) without connecting it to a mobo

paperclip into green wire and a black wire on the atx connector.

tnx i bougt a 550w coolermaster for 3 euros it was totally dust free which either means its never been used or it was dead when bought from new