Sounds about right. Voodoo 3 is also below 15W.

Thanks.... These kind of numbers are so hard to come by regarding any old GFX card. (Cough.. nVidia... Cough)

Sometime way back I compared idle power consumption of a bunch of video cards I had. I was measuring total consumption at the wall and comparing it with different cards plugged in.
One of the cards I tried was a Matrox G200 AGP. I don't know how comparable that is, but I don't have a 400.
However, I didn't test under load, I was only testing idle at a 2D desktop.

At a 2D desktop, the G200 was nearly the same (0.3W less from the wall) as an STB Velocity 128 Riva128 AGP. It was roughly 1.5W higher than the various ATI 3D Rage Pro variants I tested, and 1.1W higher than an ATI Rage 128 Ultra 16MB 64bit AGP.

Maybe that helps to corroborate something. However, under load they could diverge, and it wasn't even a G400, it was a G200.

shamino wrote:

Sometime way back I compared idle power consumption of a bunch of video cards I had. I was measuring total consumption at the w […]
Show full quote

Sometime way back I compared idle power consumption of a bunch of video cards I had. I was measuring total consumption at the wall and comparing it with different cards plugged in.
One of the cards I tried was a Matrox G200 AGP. I don't know how comparable that is, but I don't have a 400.
However, I didn't test under load, I was only testing idle at a 2D desktop.

At a 2D desktop, the G200 was nearly the same (0.3W less from the wall) as an STB Velocity 128 Riva128 AGP. It was roughly 1.5W higher than the various ATI 3D Rage Pro variants I tested, and 1.1W higher than an ATI Rage 128 Ultra 16MB 64bit AGP.

Maybe that helps to corroborate something. However, under load they could diverge, and it wasn't even a G400, it was a G200.

This would be the 'easiest' way to figure out these numbers (test them at the wall under load) and obviously you'd never get accurate numbers (as one will be testing power draw for en entire system) but one could get numbers like the ones you mentioned "Card 1 eats 5W more than card 2, but card 3 uses 3W more etc".

But the thing is that this seems kinda senseless as these cards have comparatively very low power draw to begin with.

I just did a similar method of estimating power consumption. I first measured power at the wall with onboard video (Intel Extreme i845). Total idle system power was 51W with standard VGA driver, and 54W once the Win98 driver was installed. I used the Unreal flyby to estimate max power. With Intel graphics, the system maxed out at 103W. So I used these as my baseline.

First add-in card was GeForce FX 5200 AGP. It idled at 58W and maxed out at 106W.

Second add-in card was Radeon 9800 Pro. It requires additional power from a molex connector. It idled at 96W and maxed out at 158W.

I'm not sure how to estimate the add-in card power, but I can tell the FX 5200 barely uses more power than onboard video while the 9800 Pro uses significantly more power.

clueless1 wrote:

I just did a similar method of estimating power consumption. I first measured power at the wall with onboard video (Intel Extre […]
Show full quote

I just did a similar method of estimating power consumption. I first measured power at the wall with onboard video (Intel Extreme i845). Total idle system power was 51W with standard VGA driver, and 54W once the Win98 driver was installed. I used the Unreal flyby to estimate max power. With Intel graphics, the system maxed out at 103W. So I used these as my baseline.

First add-in card was GeForce FX 5200 AGP. It idled at 58W and maxed out at 106W.

Second add-in card was Radeon 9800 Pro. It requires additional power from a molex connector. It idled at 96W and maxed out at 158W.

I'm not sure how to estimate the add-in card power, but I can tell the FX 5200 barely uses more power than onboard video while the 9800 Pro uses significantly more power.

What you could do is test a couple (more low-end or older) cards which actually have good measured max power consumption (don't just use 1, use 3 or 4 or something so you can see if your measurements are somewhat realistic) and then test the cards you're actually interested in.

The thing is that some GPUs might actually use more power with better performing CPUs and this might skew your measurement results.

Perhaps you could compare the FX5200 and the R9800 Pro with the cards you want to measure?

Tetrium wrote:

What you could do is test a couple (more low-end or older) cards which actually have good measured max power consumption (don't just use 1, use 3 or 4 or something so you can see if your measurements are somewhat realistic) and then test the cards you're actually interested in.

The thing is that some CPUs might actually use more power with better performing CPUs and this might skew your measurement results.

Perhaps you could compare the FX5200 and the R9800 Pro with the cards you want to measure?

Nice thought. Did you mean "...some GPUs might actually use more power with better performing CPUs..."

I have noticed that on a slower system (P3 933) some graphics cards seem to use disproportionately less power than in more powerful systems. I still have not had the courage to test the 9800 Pro in my Dell Dimension 4100 (the P3 933) because it has a 200W PSU with a hard-to-find proprietary connector. If I fry that baby then I lose my main Win98 retro rig. 😵 Still, part of me thinks, "the 9800 Pro will use less power. That PSU will have more than enough overhead to power it". I just can't bring myself to try, though.

clueless1 wrote:

Tetrium wrote:

What you could do is test a couple (more low-end or older) cards which actually have good measured max power consumption (don't just use 1, use 3 or 4 or something so you can see if your measurements are somewhat realistic) and then test the cards you're actually interested in.

The thing is that some CPUs might actually use more power with better performing CPUs and this might skew your measurement results.

Perhaps you could compare the FX5200 and the R9800 Pro with the cards you want to measure?

Nice thought. Did you mean "...some GPUs might actually use more power with better performing CPUs..."

I have noticed that on a slower system (P3 933) some graphics cards seem to use disproportionately less power than in more powerful systems. I still have not had the courage to test the 9800 Pro in my Dell Dimension 4100 (the P3 933) because it has a 200W PSU with a hard-to-find proprietary connector. If I fry that baby then I lose my main Win98 retro rig. 😵 Still, part of me thinks, "the 9800 Pro will use less power. That PSU will have more than enough overhead to power it". I just can't bring myself to try, though.

Ehh, you're correct 😊

You can download datasheet from mine page and try to count it from min/max voltage in electrical specification chapter. But not sure how accurate it will be (its chip only or whole card?) and i dont know how exactly do it. Maybe someone with proper education will help with that.... http://vgamuseum.info/index.php/cards/item/21 … ium-g400-sh-max

vlask wrote:

You can download datasheet from mine page and try to count it from min/max voltage in electrical specification chapter. But not sure how accurate it will be (its chip only or whole card?) and i dont know how exactly do it. Maybe someone with proper education will help with that.... http://vgamuseum.info/index.php/cards/item/21 … ium-g400-sh-max

I had a quick look, but it's really long and I should be going to bed (almost 2am here). The electrical specifications seem to be more about the pins and not about the board design as a whole (it's very detailed).

Matrox made both the GPUs and the cards themselves, is that correct?
But at any rate, I think both values can be of use, in theory at least. The GPU only number is better when you want to know which cooling solution will suffice. The entire board's power consumption is more useful for instance if you want to 'calculate' whether a certain PSU will be enough for your graphics card.
But the reason I doubt that this matters much is because of all the parts on a usual graphics card, the GPU is the vast primary source of power consumption, the rest of the graphics card (like the RAM chips) use a lot less power.

A simple way to 'calculate' how much power consumption your graphics card has, could be to see how good it's cooling solution is (usually bigger heatsink means more heat...usually that is) and if the component is supposed to have extra cooling nearby or not (like a case fan or a PSU fan). I don't know how much use it would be if you'd know the maximum operating temperature of any particular GPU

Voltage info starts from page A-10 (in chapter Appendix A: Technical Information - then choose electrical specification). I heard that power consumption can be counted by multiplying Volts and ampers. But theres too much various voltages, im not engineer and in the end this could show only power taken by chip only and not whole card. Sadly all official documents about G400 cards have no info about power consumptions (sometimes could be found in user manuals).

The datasheet does not state supply currents, so no luck.

Tetrium wrote:

vlask wrote:

Matrox made both the GPUs and the cards themselves, is that correct?

In case of G400 there was also bios-less Gigabyte card for their own motherboards.

clueless1 wrote:

Nice thought. Did you mean "...some GPUs might actually use more power with better performing CPUs..."

Yes. - Even different GPUs on the same CPU.
They will be different because different drivers utilize the CPU to different degrees.
-
So these different watts readings for different GPUs are not really indicative of (entirely due to) differences in the power the cards take from the slot.
IOW the power used by the CPU will be different for different GPUs.
(So differences determined by measuring total system power are misleading because it's the sum of CPU and GPU differences, not just GPU.)
.
.
If your goal is to find how much power the card draws from the slot.....
To do that properly (conclusive meaningful results) you would need to fabricate a test jig that can measure the power to the cards directly.
If you do that you can also determine the driver CPU loading differences by looking at the total watts changes compared to the card watts changes.
.

This is an "artist's impression" of how one could measure power usage ;p

adalbert wrote:

This is an "artist's impression" of how one could measure power usage ;p

Yes - That's the idea. [Although I don't think the military cares how much power their video cards use.]

A more refined method would be to take the measurements with a data-logging card which could output voltage and amps to EXCEL or something.
.

Split core Hall transducers would be the way to go. No dis assembly required.

adalbert wrote:

This is an "artist's impression" of how one could measure power usage ;p

Wouldn't it be possible to rig an AGP slot (and perhaps also PCI slot for PCI cards) which keep their data connectors connected to the rest of the board, but have all connectors that supply power to the graphics card be routed via some power measurement device (including molex's etc). This would also give the most natural results.

Another way 'could' be to have the card power on via some rigged AGP/PCI slot which isn't connected to a real computer, but has it's GPU put to work very hard and measure the power usage that way (this solution seems more complicated though, mostly because of drivers?).

edit: I wasn't looking right (I should go to bed), it's basically what you did in your artist concept. But I'd still suggest you create both AGP and PCI ones and also include the molex's and other power connectors 😀

Or you could try logging data in that way

(i don't know why I'm making such stupid images, I usually don't do this)
(btw i'm MA in communication in media and graphic design)

adalbert wrote:

Or you could try logging data in that way

(i don't know why I'm making such stupid images, I usually don't do this)
(btw i'm MA in communication in media and graphic design)

I think they are fun! You should do it more often 😁

adalbert wrote:

(btw i'm MA in communication in media and graphic design)

Btw, you still need an avatar 😜