Ahoy!

I have restructured the reference values page. It now has more data and is far better to consume than that old markdown table. It now contains the exact voltage, current and power measurements and multiple results per card.
I usually test:

• Windows desktop
• Aquamark 3
• 3DMark 2001SE
• 3DMark2003

For DX9 I now run the S.T.A.L.K.E.R. Call of Prypjat Benchmark as well.

You can find the new reference page here: https://tehsiggi.github.io/agp-power-monitor/

I had to re-measure all cards to gather complete data.
So far in that list:

• Gigabyte Radeon 9200
• HIS Radeon 9550 256MB
• PowerColor Radeon 9200 256MB
• Sapphire Radeon 9100
• HIS Radeon 9600 256MB
• Sapphire Radeon 9600 Pro 256MB DDR2
• ATI Radeon 9800 Pro 128MB (Hynix)
• Hercules 3D Prophet 9700 Pro 128MB

Many many more to come. I have a good flow now to measure cards so that goes relatively fast.

I just stumbled upon your thread after being furious about how wrong many AGP graphics cards' TDPs are on pages like Wikipedia, techpowerup and many more.
I was basically just looking for an energy efficient Geforce (because these are more DOS compatible than ATI cards) that at least beats the whole 3dfx lineup to use in a Palladine LCDPC case (which is essentially a poorly ventilated mATX case and LCD screen put together to look like a CRT, with only a single 60x25 exhaust fan).
And the TDP of, say, a Radeon 9500 Pro, is being reported as anything from 12W (some czech site) to 50W (Wikipedia), so none of that data was helpful to me at all.

I own about 200 AGP graphics cards and a decent benchtable, and would love to contribute. Sadly, I cannot build an AGP Power Monitor myself due to a physical disability preventing me from soldering, but maybe I could buy one off you, or lend you cards for measuring.

Here's the bigger part of my collection.

Hey Manfred! Glad to see that there are people who share my unhappiness regarding real world power consumption data of AGP cards.

If you like, i could just send you one unit and you can perform measurements yourself! I'm not doing this for money, and since I have some spare parts around, that shouldn't be too big of deal.

I suppose that'd be easier than you and me sending a dozen of GPUs around to measure them.

However, if you're more comfortable with that, I'm also glad to go that way!

That would be amazing! I'm already benchmarking cards from the first Geforce and Radeon generations until the end of the AGP era on my Phenom II system anyway.

What do you need from me? Also please let me at least cover shipping.

Danger Manfred wrote on 2025-06-30, 20:53:

That would be amazing! I'm already benchmarking cards from the first Geforce and Radeon generations until the end of the AGP era on my Phenom II system anyway.

What do you need from me? Also please let me at least cover shipping.

I envy you for that Phenom platform! 😁

But then we'll know how we go. Give me a couple of days to prepare everything. I'll check in detail today which parts I might be missing and order stuff (suits me well, needed things anyway). Once that's done and I have a running unit, I will let you know!

Cheers

Added a couple more cards: 9200s, 9000s, 9500 and a Geforce 6200. More to come.

https://tehsiggi.github.io/agp-power-monitor/

tehsiggi wrote on 2025-07-01, 12:19:

Added a couple more cards: 9200s, 9000s, 9500 and a Geforce 6200. More to come.

https://tehsiggi.github.io/agp-power-monitor/

These breakdowns per rail are fantastic, thank you for sharing.

What's the GeForce 6200XT 128MB ? 😉

Thermalwrong wrote on 2025-07-03, 22:42:

tehsiggi wrote on 2025-07-01, 12:19:

Added a couple more cards: 9200s, 9000s, 9500 and a Geforce 6200. More to come.

https://tehsiggi.github.io/agp-power-monitor/

These breakdowns per rail are fantastic, thank you for sharing.

What's the GeForce 6200XT 128MB ? 😉

A typo, will change later. It's been the 6800XT 😁

I fixed the typo.

I also added a GeForce 4Ti 4200 as well and was surprised how large the idle consumption of that card is. There's only a small difference between idle and load. Since it's using the reference design and works fine, I assume that's normal. So whenever you want to build something, that does not burn a lot of energy in idle, the Ti4200 is not the card for you.

tehsiggi wrote on 2025-07-04, 11:22:

I fixed the typo.

I also added a GeForce 4Ti 4200 as well and was surprised how large the idle consumption of that card is. There's only a small difference between idle and load. Since it's using the reference design and works fine, I assume that's normal. So whenever you want to build something, that does not burn a lot of energy in idle, the Ti4200 is not the card for you.

Thanks 😀

Regarding the Ti4200 that is a little surprising though that does explain why my PNY Ti4200 64MB card has so much discolouration around the regulator area if it can't go into an idle power mode.

It makes sense too, the 2d and 3d clock system that nvidia used didn't get introduced until the Geforce FX series & later. Probably idle power is bad on the GF2 & GF3, but GF4 would be the worst since it's mostly an expanded and overclocked geforce 3 core.