VOGONS


Reply 21 of 28, by vvbee

User metadata
Rank Oldbie
Rank
Oldbie

A compatibility matrix should be compiled. Professorprofessorson says he has a massive library of games so that's a good starting point. Ideally though have someone who hasn't yet committed to a side.

Reply 22 of 28, by BeginnerGuy

User metadata
Rank Oldbie
Rank
Oldbie

Well, per my original post I still want a card that can do duplicate monitors for my capture card and it looks like my choices remain either geforce 5 series or getting a dvi splitter or similar.

Still havent selected a card but im browsing ebay. Perhaps you guys can suggest which display driver version I should go with, and I'll test these matters myself. That said my categories will really just be smooth or crippled, i don't care about 75fps vs 100fps for example.

I'm totally happy with the performance of the geforce 4 ti i have in there, and i figure my cpu even holds back that. It's only flaw is it cant do duplicate monitors with its dual output.

Sup. I like computers. Are you a computer?

Reply 23 of 28, by appiah4

User metadata
Rank l33t++
Rank
l33t++

Radeon X800 AGP and Matrox G550 can drive dual dvi monitors as well..

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 24 of 28, by BeginnerGuy

User metadata
Rank Oldbie
Rank
Oldbie

Hmm I think the x800 is a bit much. Matrox could be doable but i think that will kill some of the games I have installed like return to castle Wolfenstein.

That said though, I'm running an SE440BX-2 which I believe is limited to 2x AGP. I believe the geforce 4 ti im running is rated 8x so I dont see any issues with backwards compatibility, short of shooting the bandwidth down bigtime.

I bought an abit board with 4x agp and 133 bus but alas it doesn't work and I havent had time to go over it yet.

Given all that I'm thinking about just grabbing any cheap entry level geforce 5 series and calling it a day.

Sup. I like computers. Are you a computer?

Reply 25 of 28, by cxm717

User metadata
Rank Member
Rank
Member

I would just give it a try, the low end FX cards look to be cheap. I can't see even an FX5200 being slower than a Geforce4 in your setup. I would just watch out for cards with gimped memory performance (like 128 vs 64 bit memory bus), that seems common with lower end cards like the FX5200 or say Radeon 9200. There might also be some 9550/x300/x1050 or x700 Radeons that have dual DVI also. Or even a Matrox Parhelia... 🤣. I got my Parhelia for $20 a bit ago

Reply 26 of 28, by swaaye

User metadata
Rank l33t++
Rank
l33t++

Something that comes to mind with 440BX is I saw freezing if I used drivers newer than 56.64. You should really just use 43.45 or 45.23 with FX cards though if you're running DirectX 8 and older games.

Reply 27 of 28, by shamino

User metadata
Rank l33t
Rank
l33t
ProfessorProfessorson wrote:

Thats a sweet article. However, the problem with their testing is that they are using it on a more modern specification for ATX and AGP using 8x slot, on a Athlon 64 board that relies heavily on 12 Volt to power everything on the motherboard, knocking it down when needed, so you cant totally trust whether or not your 3.3 and 5v leading to certain components is even coming off the psu +5/+3 rails, or if its being knocked down from the +12 on the motherboard. This was a time when AGP standards dropped dropped from relying heavy on putting power towards the 3.3V pins in the slot and opted to swap over to doing 1.5V and to 0.8V. So +12v dependence is heavy there and the system in that test was built with that in mind, with a power supply built to that ATX requirement.

So while that chart will apply greatly for something from that generation, it wont help so much for older Socket A, 370, and Slot boards that dont act that way since it wont be as cut and dry for them, which is why you end up with a lot of modern supplies getting bent out of shape with some FX and 6800 cards paired with old boards.

Their charts measured power consumption at the card, not at the PSU. That's exactly what they should do, and why they're so useful compared to other measurements that usually get taken.
Power drawn from the AGP slot was measured by isolating those pins so they were no longer connected at the slot. The corresponding circuits on the video card were then connected to an alternate supply which could be measured.
https://web.archive.org/web/20041010012745/ht … owercons_3.html
The charts they've posted thus represent the power drawn by the card at each of the available connections (3.3v AGP, 5v AGP, 12v AGP, 5v molex, 12v molex). The accuracy might not be very high, but what they are measuring looks valid to me.

I agree that it would be a mistake for someone to assume that 3.3V at the AGP slot is always coming from the 3.3v rail on the PSU, because as you mentioned there are some motherboards that use the +12V rail to provide onboard 3.3v. It's a minority, but there are a decent number of them.
I haven't noticed any boards that do this for 5v, but it's certainly possible that some are out there.

What is needed is someone willing to take a wide selection of older boards, and also a wider selection of the FX cards other then just a few, because manufactures didnt always stick with stock designs.

I don't know whether the cards they tested are nVidia reference designs, but most cards are.
If the power drawn by the card is known, then I don't think testing with different motherboards is necessary. If you're trying to determine how the load of the video card will translate upstream at the PSU, then you just need to probe the motherboard a bit and confirm where the AGP slot's power rails are being fed from.

That chart they did kind is kind of scary when they are having one card, the weaker FX 5900 ultra, pull much heavier on the 12volt rail then the 5950 one does.

I don't know that generation very well so I might be wrong, but I think the FX5900U came out earlier. It looks it must be one of nVidia's first +12V heavy cards. They may have decided it was causing problems, prompting them to shift the load around when they designed the FX5950U.
Either that, or just 2 different designers that did things differently.

Reply 28 of 28, by ProfessorProfessorson

User metadata
Rank Member
Rank
Member

Their shunt method works ok. You can use this method to test your AGP slot too, to make sure its putting out adequate power by soldering to the AGP pins on the back side, but you will have to use AGP cards not sipping power from a molex. Something like a TI4600 would be ideal because it pushed the power draw on AGP slots pretty high. This shunt method was one of the methods used to confirm what was going on with the Asus P3B-F AGP slot with Geforce and Voodoo 3 cards, because people had so many issues with that board. And its a method they should have used in their test also on a few boards to get a broader range or results.

Ive found that a lot of companies were taking Nvidia reference boards back then and altering the choice in components used. Sometimes to a small degree with caps values, sometimes more major with swapping out things like Os Con caps in favor of tantalum capacitors of a different value and routing them slightly differently. While that may, or may not come into play with power draw, the really strange thing is that anyone would bother to change the reference design to begin with, unless it was on something like the Gainward Golden Samples to get higher clocks. I mean if you're sticking to stock clocks, no need to screw around with a tried and proven component choice.

The thing with the FX5900U, I dont know the deal on that but what I do know is my Gainward Golden Sample is a FX5900, flashed to a 5950U, and I basically got the same issues with it as the 6800 with new supplies. Less shutdowns though, more whine from the supplies and games closing out. Its convinced me to buy a couple of those PowerMan 350 watt psu as new old stock a few years back. Those supplies are pretty solid, but eventually I am going to have to recap them because they were made around 2005-07 range.

vvbee wrote:

A compatibility matrix should be compiled. Professorprofessorson says he has a massive library of games so that's a good starting point. Ideally though have someone who hasn't yet committed to a side.

As much as I would love to do this, I would not have a way to capture video for any youtube presentations or anything. I dont keep any modern capture card equipment on hand. What I have is a older Philips based PCI capture card. It does work with TMPGEnc and VirtualDub if I remember correctly, but isnt stable in Windows 7 and wont work in Windows 10, and only does Svideo, RF, and Composite capture anyway. Also, I would need to pick up a FX 5600 or a non-gimped 128-bit FX 5200 and record it with like my HD webcam. My current Pentium 3 setup is in a Antec Titan case with a EVGA 500watt.

I already have everything ziptied down because I have a cathode lighting setup in there and had to get it cleaned up and tied down, and I dont have any spare large molex coming from the supply to power any grafx cards at this point. I had no intention of ever using my 6800 and 5950 cards again so I consider this build fairly permanent. The only spare molex I have in there is the diskette drive one, in case I decide to pop a Radeon 9700 back in. Even if I did have a spare large molex available, I would run into the prior power issues mentioned anyway in that build with that psu. Really this is the kind of test you would want to do on a later Nforce 2 or Socket 754 board with a modern supply in Windows 98 or ME if you desire testing the beefier FX cards. That way you can push the card hard and get better results. If you are only concerned with compatibility, all you would need is a 5200 or 5600.

If someone wants to start testing, just start picking titles that were not leaning towards Glide so much so that you know they have solid DX or OpenGL modes. I mean yeah you can run Dethkarz, but opting to run that game in DX 7, instead of a Voodoo in Glide will present slowdown during heavily lighted areas of track, but it is stable at least. But yeah in general just start trying games like Hundred Swords, House of the Dead 2, Blood 2, Shogo, Heavy Metal Fakk2, Quake 2, Half Life, Dune 2000, Bang Gunship, Midtown Madness, Carmageddon 2, Sega Rally 2, Need for Speed 3 and High Stakes, Unreal Gold and UT, Hydro Thunder, NOLF, Resident Evil 2 and 3, Hitman, Rollcage 1 and 2, MK4, etc. Thats the kind of stuff I prefer to play on my 98 box. If you opt to test Shogo or Blood 2, make sure vsync is turned off on the card. That game has an issue with vsync on Nvidia cards.

Also, I am not committed to any side. If I was Id be using an FX currently. Im using a TI 4200 instead. I actually bounce back and forth between cards because I like to tinker some. I have a few boxes worth of cards dating back to the Permedia 2 and early Rage and TNT Vanta flavors, on up. Even have a Real3D Starfighter and Kyro II here. What I am currently lacking is a FX 5200 or 5600, and also a Radeon 7500 and 8500, as I have not had any Radeons from that generation in about 10 years or so.

My experience with the FX line has only been with the gimped 5200 and a PNY 5600 that was sold as a 256mb card but was actually 128mb that had to be returned to Circuit City, and the FX 5900 cards. I dont remember the 5200 and 5600 fondly due to said issues, but they did run games fine without crapping out due to bugs when I used them. Funny enough I used the 5200 to beat The Matrix on 800x600. Looking back at it now, that game really sucked. But yeah I never had any problems with any 5900 cards outside of my power issues, and the 5900 was my card of choice for a solid 2 years there for my Windows 98 builds, when I was still running Athlon XP stuff.

I would be using one right now to mess around with off and on, but the Asrock 754 board I have is woefully unstable so I cant use my FX or 6800 anymore in any reliable way, and I dont have a single Socket A board with good caps now. Im just not willing to recap the stuff right now. I am already behind with recapping other parts like my Slot A board, and older game systems I have. I still have two NeoGeo MVS boards waiting to be recapped because the TK caps are not up to snuff anymore, along with a couple spare 32X boards, a Sega Genesis Model 2, Snes mini, spare Famicom, etc.

Spending money on any more 754 or Nforce2 gear for something I might use for 1 to 2 hrs every 6 months just to prove a point isnt a option. I need to throw money towards stuff I would want to use more. If someone is willing to send me a 5200 or 5600, or when I eventually buy one I will be more then happy to test stuff again with that specific line and test specific games on request that I either own or can get ahold of if I dont using a older version of Fraps. Till then though its like a back burner sort of thing. Honestly the OP should stick with his GF 4. I cant recall if anyone has even mentioned this, but his cpu is a major bottleneck as is anyway. Swapping to a faster card wont help in most of his games.