TrashPanda wrote on 2022-12-22, 07:50:

...the G200 is even worse being the much older card and neither is suited for anything other than DX6 gaming at low resolution and colour depth.

I also suspected it would be slower, and the only reason I even tried the G200 was because the PC came with it installed when I got it 😆 So, someone thought they were getting an "upgrade" when they installed a dedicated graphics card, but in reality they would have been better off with the integrated graphics.

devius wrote on 2022-12-22, 10:36:

TrashPanda wrote on 2022-12-22, 07:50:

...the G200 is even worse being the much older card and neither is suited for anything other than DX6 gaming at low resolution and colour depth.

I also suspected it would be slower, and the only reason I even tried the G200 was because the PC came with it installed when I got it 😆 So, someone thought they were getting an "upgrade" when they installed a dedicated graphics card, but in reality they would have been better off with the integrated graphics.

Yup . .the i815 wasn't terrible in the day, it wasn't great either but you could game on it if you didn't have anything better.

acl wrote on 2022-12-20, 07:49:

Quadro NVS 280 is a "business" GPU with NV34 chip (Fx5500) but clocked a bit faster. It exists in PCI version and can be found for around 20€. It needs a DMS-59 cable and provides dual VGA or dual DVI output.

Is the 128-bit Quadro NVS 280 PCI faster than a 64-bit Radeon 9250 PCI?

Spark wrote on 2022-12-22, 19:07:

acl wrote on 2022-12-20, 07:49:

Quadro NVS 280 is a "business" GPU with NV34 chip (Fx5500) but clocked a bit faster. It exists in PCI version and can be found for around 20€. It needs a DMS-59 cable and provides dual VGA or dual DVI output.

Is the 128-bit Quadro NVS 280 PCI faster than a 64-bit Radeon 9250 PCI?

I don’t believe so. The 9250, 9200 and 7500 are my top performing pci cards (that I can get to work that is on socket 7)

PCI GeForce/Quadro have bandwidth issues with OpenGL games and games with TnL support in general. Although it won't be much noticeable on most S7 builds. Radeon PCI don't have such issues, but drivers are not good for weak CPUs. Pick your poison.

TrashPanda wrote on 2022-12-20, 23:23:

PCI FX 5500 cards from China are dirt cheap and easy to get on Evilbay, much easier than any PCI variant of the FX5200 Ultra 64 or 128bit. The only issue with the China PCI FX5500 cards is that they can be very picky about which older PCI boards they will post in, this is mostly due to the older boards simply not having a robust enough PCI voltage supply to feed the FX5500.

A PCI 128bit FX5200 Ultra shouldn't have any power issues with older boards but again they are stupidly expensive since the sellers do know that they are a sought after model, so grabbing a dirt cheap 64bit version might be the only option if you are unwilling to pay that premium.

I had another requirement: Fanless operation. This excludes the FX5200 Ultra cards (obviously) and many dirt cheap cards as well. I ended up buying a 128-bit non-ultra FX5200 which arrived today. Some of those cards are designed to work without +3.3V, but not the one I bought 🙁. I will receive powerful +3.3V step-down modules shortly that I ordered after trying to run a 9200 PCI in a 486 board. That 9200 card seems to regulate all voltages down from +3.3V, so I ordered beefy modules. The FX5200 that arrived today has core voltage on a switch mode step-down from +5V, but it has RAM and BIOS at +3.3V from the PCI bus, and I guess the I/O voltage of the NV chip is also at +3.3V. There are level shifter pins on the PCI interface for 5V compatibility, so I have no reason to believe that the I/O voltage of the NV chip is connected to the VI/O supply of the PCI bus.

As the FX5200 only has RAM as significant power sink on +3.3V, I hope to get around with 3-4A, and a mini step-down module from ebay can do the trick. I will report back after I tested whether the card works in a 486 system at all. If the card works with injected +3.3V, I will order some mini converters and mount one of them on the card and remove the +3.3V connections to the PCI bus. I expect to have one of the most powerful fanless (non-multi-slot) PCI GPUs then. On the other hand, that exercise might be completely pointless, as the seller of that card pointed out that the GeForce 5200 is highly dependent on CPU clock and even a moderate P3 system is bottlenecked by the CPU.

mkarcher wrote on 2022-12-22, 20:55:

TrashPanda wrote on 2022-12-20, 23:23:

PCI FX 5500 cards from China are dirt cheap and easy to get on Evilbay, much easier than any PCI variant of the FX5200 Ultra 64 or 128bit. The only issue with the China PCI FX5500 cards is that they can be very picky about which older PCI boards they will post in, this is mostly due to the older boards simply not having a robust enough PCI voltage supply to feed the FX5500.

A PCI 128bit FX5200 Ultra shouldn't have any power issues with older boards but again they are stupidly expensive since the sellers do know that they are a sought after model, so grabbing a dirt cheap 64bit version might be the only option if you are unwilling to pay that premium.

I had another requirement: Fanless operation. This excludes the FX5200 Ultra cards (obviously) and many dirt cheap cards as well. I ended up buying a 128-bit non-ultra FX5200 which arrived today. Some of those cards are designed to work without +5V, but not the one I bought 🙁. I will receive powerful +3.3V step-down modules shortly that I ordered after trying to run a 9200 PCI in a 486 board. That 9200 card seems to regulate all voltages down from +3.3V, so I ordered beefy modules. The FX5200 that arrived today has core voltage on a switch mode step-down from +5V, but it has RAM and BIOS at +3.3V from the PCI bus, and I guess the I/O voltage of the NV chip is also at +3.3V. There are level shifter pins on the PCI interface for 5V compatibility, so I have no reason to believe that the I/O voltage of the NV chip is connected to the VI/O supply of the PCI bus.

As the FX5200 only has RAM as significant power sink on +3.3V, I hope to get around with 3-4A, and a mini step-down module from ebay can do the trick. I will report back after I tested whether the card works in a 486 system at all. If the card works with injected +3.3V, I will order some mini converters and mount one of them on the card and remove the +3.3V connections to the PCI bus. I expect to have one of the most powerful fanless (non-multi-slot) PCI GPUs then. On the other hand, that exercise might be completely pointless, as the seller of that card pointed out that the GeForce 5200 is highly dependent on CPU clock and even a moderate P3 system is bottlenecked by the CPU.

Wait, you want to use a 5200 with a 486? I really doubt that's going to work, even if you can get it to POST. I would think that the drivers have to require MMX and possibly other instructions. And even if it does allow you to install them on a 486, the drivers + Windows 98 alone will likely consume so many resources that any games will be unplayable.

You could get a Riva 128 and it would still be severely CPU bottlenecked by the 486... but at least it might work.

Also, for what it's worth, I don't think a PCI 5200 Ultra exists, and by far the vast majority of 5200 non-ultra cards are 64bit. If you did find a 128bit 5200 PCI it would definitely be a decent card to hang onto for a late-90s early-2000s PC, so be careful not to fry it trying to make it work in the 486 board.

Ozzuneoj wrote on 2022-12-22, 21:37:

Wait, you want to use a 5200 with a 486? I really doubt that's going to work, even if you can get it to POST. I would think that the drivers have to require MMX and possibly other instructions. And even if it does allow you to install them on a 486, the drivers + Windows 98 alone will likely consume so many resources that any games will be unplayable.

Yeah, I fully understand that. I'm experimenting on a "too much of everything" themed 486 system together with a friend, and this is just about probing the limits of what's technically possible. The real thing to put into a 486 computer for gaming is a Voodoo 1 or something like that.

Ozzuneoj wrote on 2022-12-22, 21:37:

If you did find a 128bit 5200 PCI it would definitely be a decent card to hang onto for a late-90s early-2000s PC, so be careful not to fry it trying to make it work in the 486 board.

The 128bit 5200PCI loses in 3Dmark agains both 64bit 5200AGP card that I also have at hand, which was unexpected. Most late-90s to early 2000s PCs do have an AGP slot, so they will work better with the 5200AGP. Nevertheless, I am indeed going to be careful when I try to rework that card into a 5V-only compatible card.

EDIT: spelling fix

mkarcher wrote on 2022-12-22, 22:13:

Yeah, I fully understand that. I'm experimenting on a "too much of everything" themed 486 system together with a friend, and this is just about probing the limits of what's technically possible. The real thing to put into a 486 computer for gaming is a Voodoo 1 or something like that.

I figured it was something like that. I guess we'll find out if it's possible at least?

I will say though, even a Voodoo 1 would be severely bottlenecked by a 486 in any games that support 3D acceleration, and a Riva 128 is much faster than a Voodoo 1... TNT is much faster than that... TNT 2 much faster than that... Geforce... Geforce 2... Geforce 3... Geforce 4... FX. So yeah, the 5200 is probably about 10x overkill for a 486. 😁

mkarcher wrote on 2022-12-22, 22:13:

The 128bit 5200PCI looses in 3Dmark agains both 64bit 5200AGP card that I also have at hand, which was unexpected. Most late-90s to early 2000s PCs do have an AGP slot, so they will work better with the 5200AGP. Nevertheless, I am indeed going to be careful when I try to rework that card into a 5V-only compatible card.

I like the faster PCI cards for putting into prebuilt systems that have integrated graphics and no AGP slot. Like a Pentium II\III system. They are nice for that kind of thing. 😀

Ozzuneoj wrote on 2022-12-22, 22:34:

I figured it was something like that. I guess we'll find out if it's possible at least?

Well, currently we are at a Matrox G450 PCI in my 486/160. And it took mkarcher a few hours to make the BIOS from the HOT-433 working with it (due to the need to modify the PCI brigde initialisation, since the G450 has a PCI to AGP bridge)

I fear that the 5200 or the NVS 280 is going to be a fruitless combination with a 486 I would suggest trying a Radion 7000 or a 7500 or gf 2/3

But even that may be too recent may have to go even older. The pci version is going to probably keep the card from posting. Let alone the drivers. Suggest asking someone to test the combo (card pcb and mobo apples to apples) before buying something for this attempt. There are issues that keep late gfx cards from working in early pci mobos/processors correctly.

PCI video card with decent 3D ability and define cheap? No? That's correct, does not exist since they are very desirable.

Cheers,

TrashPanda wrote on 2022-12-19, 00:17:

A PCI 5200 or 5500 would be fine, just don't expect amazing DX9 gaming as FX series cards are well known for having poor DX9 performance, but for anything DX7/8 based they will be great little PCI cards.

Note apparently some PCI FX 5200s are rebranded MX 4000s with a slight memory boost (33MHz, mine reads "NV19PL" on the top in big letters). I didn't realize this until a few days ago, which means somehow the person I bought the FX 5200 from managed to mislabel a listing as itself.
Which... is completely no bueno considering EVGA branded it as a DirectX 9 card making no reference to it being a GF2-derivative chip. It was as far as I can tell a literal scam... but that tends to be par for the course for graphics cards.
But yeah, either form of PCI 5200 would probably be more than enough for a Pentium 133.

The Serpent Rider wrote on 2022-11-10, 19:57:

Rage Pro and derivatives are super cheap, but very shitty for 3D in general. ATi were very adamant to implement double-buffered Vsync in Direct3D at all times, without performance to back it up, even in period correct games. So enjoy constant dips form 60 to 30 fps or from 30 to 20/15 fps, due to how double buffering works. It's worse when you have slow CPU.

They were not good for 2D DOS gaming either.

i think the forced vsync in D3D stems from some kind of early microsoft recommendation, because nvidia and 3dfx do it too by default, and would usually only expose the option to force it off after registry tweaks. rage 128 allows to force it on or off at least, but needs a reboot and lacks a "software controlled" option, so even that kind of sucks in comparison.

what i find really strange is how ati went back and thought it's a good idea to force vsync even in VESA DOS games that already didn't have tearing, thanks to page flipping...

DonCatoli wrote on 2022-12-24, 00:22:

The Serpent Rider wrote on 2022-11-10, 19:57:

Rage Pro and derivatives are super cheap, but very shitty for 3D in general. ATi were very adamant to implement double-buffered Vsync in Direct3D at all times, without performance to back it up, even in period correct games. So enjoy constant dips form 60 to 30 fps or from 30 to 20/15 fps, due to how double buffering works. It's worse when you have slow CPU.

They were not good for 2D DOS gaming either.

Which 2D DOS games looks (or works) bad especially ? I'm just curious. Right now I have Rage IIC PCI in my P133 rig and can check it.
Also can check same games with Rage Pro AGP in P3/BX.

zyga64 wrote on 2022-12-29, 07:41:

DonCatoli wrote on 2022-12-24, 00:22:

The Serpent Rider wrote on 2022-11-10, 19:57:

Rage Pro and derivatives are super cheap, but very shitty for 3D in general. ATi were very adamant to implement double-buffered Vsync in Direct3D at all times, without performance to back it up, even in period correct games. So enjoy constant dips form 60 to 30 fps or from 30 to 20/15 fps, due to how double buffering works. It's worse when you have slow CPU.

They were not good for 2D DOS gaming either.

Which 2D DOS games looks (or works) bad especially ? I'm just curious. Right now I have Rage IIC PCI in my P133 rig and can check it.
Also can check same games with Rage Pro AGP in P3/BX.

I remember my Rage IIC was usless for Keen Dreams aka the stuttering screen scrolling.
Look over the ATI card list in the DOS compatibilty chart @ https://gona.mactar.hu/DOS_TESTS/

Disruptor wrote on 2022-12-18, 00:47:

Geri wrote on 2022-11-20, 00:01:

If thats not an option, go for the matrox G200. But avoid the g400 and above.

What's the problem with the G400 and G450 ?

The G400 / G450 runs slower with very weak CPU than the G200. About 30-40% slower when paired with a Socket7 Cyrix in my computer.

DonCatoli wrote on 2022-12-30, 23:09:

zyga64 wrote on 2022-12-29, 07:41:

DonCatoli wrote on 2022-12-24, 00:22:

They were not good for 2D DOS gaming either.

Which 2D DOS games looks (or works) bad especially ? I'm just curious. Right now I have Rage IIC PCI in my P133 rig and can check it.
Also can check same games with Rage Pro AGP in P3/BX.

I remember my Rage IIC was usless for Keen Dreams aka the stuttering screen scrolling.
Look over the ATI card list in the DOS compatibilty chart @ https://gona.mactar.hu/DOS_TESTS/

i'd say that chart is way overblown as far as i'm concern, i've tried many ati video cards in dos, aside from lacking VESA vbe 3.x ; they all work fine in msdos. the commander keen games don't dictate how "compatible" a video card is.

And besides commander keen is known to have that screen tearing scrolling bug for many video cards and has a patch to fix that issue.

Stainlesscat wrote on 2023-01-03, 11:01:

i'd say that chart is way overblown as far as i'm concern, i've tried many ati video cards in dos, aside from lacking VESA vbe 3.x ; they all work fine in msdos. the commander keen games don't dictate how "compatible" a video card is.

And besides commander keen is known to have that screen tearing scrolling bug for many video cards and has a patch to fix that issue.

I am inclined to agree with you and I've said it here before. People frequently treat the number of green vs red/orange boxes in a row as some kind of benchmark of compatibility, but in that sense it has turned into a "synthetic" benchmark for a lot of us because we have absolutely no intention of ever trying to run most of those games, and if we do we're probably not going to force them to run in the modes/situations where they do have problems. Like, who runs Mario Shareware on their retro PC, really? 😀

There are games that have quirks that are not listed on the chart as well. I remember playing a relatively obscure old game (Alien Logic) and getting a lot of weird shimmering pixel artifacts on S3 PCI cards, and yet the seemingly less compatible Matrox Millennium ran the game perfectly. This game isn't on the chart.

So yeah, I think it's a useful source of information but I think it has biased a lot of people for/against certain hardware without them having personal experience with it. I'm sure I'm guilty of using it this way as well at some point. It'd be best if we stopped using it to tell people which hardware is worthless due to supposedly awful incompatibilities though.