Rialto on it's own works just fine, but not with a radiator dumping heat in all directions on top of it.

I removed the card from my Pentium 3, it certainly have lots of buggy things with old games support, nevertheless more modern XP style games runs fine. Nvidia 6600 GT is still much better solution for compability sake, both 98 3D games and early XP runs just fine. I only having problem with early DOS and 3D games, but for that i will built another less powerfull system, this is just too hard task to make it all work in one configuration for my personal needs.
I`m sure HD 2600 would work amazing on dual core computer.

That's one issue with this card - even if/when it works - old games do not work because of compatibility, new games do not work because it is going to be too slow.

In dual core system in most cases pci-e would be available and much newer, faster cards would be beneficial. Yes, dual core with AGP is possible, but only if done deliberately for some reason.

I don't see any practical use in DX10 hardware on AGP system, even if it's Radeon 3850.

x264 video hardware acceleration + perf/power metric ?

x264 was nice to have back in the day no doubt, but nowadays... what's the point in trying to play it on vintage hardware?

Also at least from what i've seen (it would be nice to see actual measurements) idle power consumption is quite high on this cards, so performance/power would not necessarily be all that great with low load....

agent_x007 wrote on 2025-07-10, 22:33:

x264 video hardware acceleration + perf/power metric ?

1) Practically pointless after almost 20 years.
2) I don't see much use for that outside of benchmarking. Granted, you can still use some of Radeon X1xxx AGP cards which apparently are software monitored/moddable by ATItool in various ways (voltage, timings, etc).

agent_x007 wrote on 2025-07-06, 20:02:

Yes, I think we are done here.

OK

Archer57 wrote on 2025-07-09, 02:53:

That's one issue with this card - even if/when it works - old games do not work because of compatibility, new games do not work because it is going to be too slow.

I agree completely, but 2600 family (hi, agent_x007) had very different designs.

I held such a Sapphire Radeon HD 2600 XT GDDR4 in my hands, but it was already dead, unfortunately.
Looks and feels very cool.
One can only wish to have this in one's collection at working state.

Studiostriver wrote on 2025-07-06, 19:08:

Here is mine ATI HD 2600 MSI 256mb DDR2 agp

It's like the series "Xena: Warrior Princess" - about old things in a new way.
In those years, heat pipes were not widely used.
Cool thing.

shevalier wrote on 2025-07-11, 14:25:

I agree completely, but 2600 family (hi, agent_x007) had very different designs. https://ua.gecid.com/data/video/200708100000-21 […]
Show full quote

I agree completely, but 2600 family (hi, agent_x007) had very different designs.

I held such a Sapphire Radeon HD 2600 XT GDDR4 in my hands, but it was already dead, unfortunately.
Looks and feels very cool.
One can only wish to have this in one's collection at working state.

Yeah, that's a very cool card, especially with how short-lived GDDR4 was.

Probably somewhere very close to the end of single slot coolers for high end cards too, though i am definitely not a fan of this style of coolers and the way they are built from practical point of view. They suffer from typical issues modern laptops have - heatsink gets completely clogged with dust really fast. And since the screws holding plastic cover have to be unscrewed from the underside of the cooler - whole cooling system has to be disassembled to clean. Regularly, once a twice a year depending on amount of dust around. Very annoying maintenance, and perhaps a reason for dead cards.

By the way - are this affected by bumpgate-related issues, or are they already fixed?

Archer57 wrote on 2025-07-11, 15:03:

Very annoying maintenance, and perhaps a reason for dead cards.

offtopic
Now they've finally started using air filters in computer cases.
But the turbine itself is noisy.
But the rtx5090 design with a gap between the brackets - that looks completely cool.

Archer57 wrote on 2025-07-11, 15:03:

By the way - are this affected by bumpgate-related issues, or are they already fixed?

Presumably not affected, because ATi used different substrate packaging. Obviously still died from dust clogging, crappy case ventilation, etc. Better substrate is not a silver bullet.

The Serpent Rider wrote on 2025-07-11, 21:45:

Presumably not affected, because ATi used different substrate packaging. Obviously still died from dust clogging, crappy case ventilation, etc. Better substrate is not a silver bullet.

Well, ATI definitely made some chips defective in the same way, xbox360 GPU being the most well known example. Though that is 90nm IIRC, which coincides with radeon X1nnn, not HD2nnn. What's interesting is at which point they stopped doing that. I'd guess HD2nnn probably should not be affected, but i am definitely not sure about that...

Hello good people. Somehow today I have felt less risk-averse than usual, and did install the card) I feel I have to contribute back to your friendly community by writing a report)

The card did not burn in the AGP 4x slot of the TUSL-2C. The Everest even showed it as 4x@4x (the previous owner had the same Everest showing the same card as 8x@8x). I suspect the startup beeps were somewhat different. The card's cooler started at bootup, then stopped, then started again. This was indeed scary) Especially because the HD2600 does not give a welcome onboarding message before the main BIOS, like the FX5200 does, so you have to endure some dark screen until the main BIOS shows up. Did not notice much noise compared to FX5200 though. Maybe this was because I was not using the external power source.

I did not glue the heatsink to the Rialto (because I knew the previous owner did not bother to cool the Rialto, and it was supposed to be a quick test initially, just to see if the card is even working at all). The power supply used was the 300W Powerman. The drivers used were Omega 3.8.442 - and compared to FX5200's those were somewhat less easy to find. I expected the proper Catalyst will just pop up in Google, but that was not the case, had to dig a little deeper than expected, and dug this Omega set out. It asked if I wanted an AGP or PCI-E version installed. I think that was pretty cool, as I was mucho afraid of installing the wrong PCI-E drivers and thus killing the card or the system.

Concerning how much of an overkill was using this card on obviously NOT suitable Tualatin-class system, I have tested these games:
1)Need For Speed Underground - no difference to FX5200. I remember that switching (involuntarily) from Mendocino to Tualatin gave me a lot of extra oomph here. Maybe this one relies more heavily on the processor.
2)Little Big Adventure in DosBox looked more smooth, and I have noticed the game was using more of the screen. This is an XP system now, so no pure DOS games could be tested.
3)Max Payne 1 gave me a lot of new higher resolutions, and was quite playable at those. Something I did not have on FX5200.
4)Tomb Raider Legend and Anniversary looked significantly better. The Shaders 3.0 support, of course. Not much performance boost at higher resolutions though. Just smoother picture at lower res. Maybe Radeon's 512Mbs of DDR2 give some extra textures compared to 128Mbs DDR that the FX5200 has. There is a striking texture quality contrast between 1Gb and 512 MBs in those two games, it's a fact I know at first hand.
5)Tomb Raider Underworld launched) This was that the only game the FX5200 was unable to launch, because of the required Shaders 3.0. Felt exactly like playing the Tomb Raider 2 on a P-75 with CirrusLogic 5430: sorry, moms and pops have to work pretty hard to give you even this much money, kiddo, so you'll have to do with what you have here)
6)Max Payne 3 was not even supposed to be there, tried it just to make sure the disks are not enough, you need Steam to play anyways.

The downside: much higher temperature. Just in plain Windows, running nothing, The Everest gave me 97F (36C) for the processor, 88F (31C) for the motherboard, and whopping 120F (49C) for the Radeon, fan at 20%. The baby is on fire indeed. When I took the card out after all, it felt pretty heavy Uranium warm, and I mean not only the heatsink, but the whole board, too. Something I'm not used to with FX5200. The Rialto was lukewarm though. The SensorsView did not give me the Radeon's temperature, but the rest of the temperatures corresponded to Everest.

Thanks to your kind advice and comments I knew I should not be experimenting for too long. There was always a lingering thought that all these nice vignettes might be just enabled by some poor struggling badly overheated cap) Because I knew where exactly to look, I could clearly see the card is not sustainable in this system, running at almost double temperature than the rest of the components. After all it was a shameless fun day, made me think of that one time when I had the borrowed Voodoo 1 for a couple of days back in the 90s))