GeForce 6200.

The Serpent Rider wrote on 2022-04-23, 10:04:

GeForce 6200.

It's an interesting option, they are not much common to find but I'll look for one of these. On the ATi side after the various Radeon 9250 or similar did they release some newer cards before the PCI-EX bridged few ones?

X800 cards were the last generation with native AGP support. And X300 and X600 were just rehashed Radeon 9600 with PCIe support. Nvidia had to make new low-end AGP chip, because FX series chips were unmarketable trash by that point.

afaik NV40 was native AGP and released after the R420, and the AGP 6200 was a new GPU released a full year after, so I think Geforce 6 being the last AGP gen would be more accurate?

Geforce 7300GT was available in an AGP version, though almost certainly thanks to a bridge chip of some kind .

I clearly misunderstood the question, please ignore.

386sx, I don't know what do you mean by a GPU but Volari Z7 might be one of the latest PCI chips made.

SPBHM wrote:

so I think Geforce 6 being the last AGP gen

Only NV40 and NV44 have native AGP. And it's obvious that only NV44 had PCI versions.

What about matrox? 🤔

There is the pci-x matrox. Is that of use to you?

Not agp? Idk but seems like may be useful for your situation. I think they work in 32-bit slots?

PCI-X usually can work in regular PCI, yes.

386SX wrote on 2022-04-23, 12:11:

The Serpent Rider wrote on 2022-04-23, 10:04:

GeForce 6200.

It's an interesting option, they are not much common to find but I'll look for one of these. On the ATi side after the various Radeon 9250 or similar did they release some newer cards before the PCI-EX bridged few ones?

6200 PCi are pretty easy to find NOS, there is a few sellers on eBay selling them, pretty sure one of the sellers actually has a large stock of NOS 6200 PCI cards and sells them for a reasonable price for a PCI card.

Just a note it seems like 256mb 6200 cards appear to be compatible with older drivers than the later gen 512mb+ chips. Also really early cards can be unlocked to a Geforce 6600 apparently.
Edit: A2 chips are unlockable and A4 chips are hardware locked.

NV44 can't be unlocked, because there's nothing to unlock.

Putas wrote on 2022-04-23, 19:51:

386sx, I don't know what do you mean by a GPU but Volari Z7 might be one of the latest PCI chips made.

I didn't remember that 2D only card interesting indeed even if a bit of OpenGL or Direct3D acceleration would have been nice at least for compatibility reasons with apps (like a web browser that'd still need OpenGL 2.x support).
But thanks I'll read about it meawhile I'm reading it demand 1,4 watts 😉, it sure has an interesting nostalgic look without any heatsink on it. I wonder if it was specifically designed for PCI or multiple bus.

ODwilly wrote on 2022-04-24, 06:39:

Just a note it seems like 256mb 6200 cards appear to be compatible with older drivers than the later gen 512mb+ chips. Also really early cards can be unlocked to a Geforce 6600 apparently.
Edit: A2 chips are unlockable and A4 chips are hardware locked.

I don't remember well the 6200 early GPU version if they were higher end GPUs like the early Radeon 9500 Pro situation was, but in any case I'd have problem on a 30W limited PCI bus to unlock anything that surpass that limit and looking at the various 6200 heatsinks I suppose they might be already close to 25W or similar power demand. 😉

But it's an interesting choice. Of these old card on a modern o.s. environment I tested the FX 5200 256MB 128bit PCI and it wasn't bad (well.. it's a PCI FX 5200... not many expectations anyway), it has WDDM1.0 drivers (only 32bit version worked the 64bit beta didn't in modern W8.1 o.s.) and the open linux driver with OpenGL 1.5 support. Not bad but the card is quite old and demand more power than the mainboard and cpu itself but at least had a native AGP chip that I suppose had less problems talking to a PCI bus instead of the PCI-EX GPUs with their complex IC bridges that results in a more complex communication task. I suppose the less things get translated on the already bridged PCI mainboard bus the better might be (as read in the past also old PCI bus were bridged but I suppose the chipset supported PCI in a faster more native way than latest mainboards maybe using it as a retro options and there's nothing to tweak, drivers are standard in any o.s. and the only thing is the PCI release time in the bios).

So the idea was to test a sort of low power still somehow installable PCI card that has less possible overhead while mantaining some feature that would be usable accelerations, in the GUI, video.. talking about basic low end features of course. To get as far as possible in the past I'm waiting for these Rage XL PCI to start from it as a first real PCI compatible native not-bridged chip to see if this improve the latency introduced in more modern solutions.

Sphere478 wrote on 2022-04-24, 03:57:

What about matrox? 🤔

There is the pci-x matrox. Is that of use to you?

Not agp? Idk but seems like may be useful for your situation. I think they work in 32-bit slots?

I never read a lot about PCI-X, would they demand more watts to work than the PCI standard specifications?

Good indicator of wattage is the heat managment going on. Big beefy heatsink, lots of watts. Little passive one, not so many.

386SX wrote on 2022-04-24, 07:50:

Sphere478 wrote on 2022-04-24, 03:57:

What about matrox? 🤔

There is the pci-x matrox. Is that of use to you?

Not agp? Idk but seems like may be useful for your situation. I think they work in 32-bit slots?

I never read a lot about PCI-X, would they demand more watts to work than the PCI standard specifications?

I’m pretty sure that the additional pins of 64-bit PCI are only I/O signalling voltage and address lines, so no additional power draw.

I’ve run my PCI-X Matrox Parhelia on regular PCI (in a Pentium 4 Shuttle XPC) a few times with no ill effects.

Interesting thanks. Meanwhile I'm waiting in the next weeks the Rage XL PCI not cause I've many expectations from it but I'm curious to see the differences using a supported (I think linux should use a working DRM module for mach64 cards) native PCI card and then trying the other cards that used bridge chips.
The onboard GPU instead had a story of missing driver options that's why I'm interested in the single PCI slot option (which ironically reading the bios specifically intended it for external vga while having a PCI-EX x4 and x1 lines that were not connected to a PCI-EX bus on the PCB). Lately adding a second module of DDR3 ram I found that this Atom SoC can support even 8GB DDR3@1066 rams which is quite impressive for such cpu and I think the usage of the second ram slot did boost a bit the performances compared to a single 4GB module while it still has a single channel memory bus.

I tested the Rage Pro PCI 4MB SGRAM into this Intel NM10/CH7 board and it's interesting in linux that even when the correct MACH64 driver is loaded and has quite lot of lines in the logs (detecting everything, chip clock, even the chip model, the chip foundry,..) and seems to accelarate what it can more or less, the final result GUI speed is much slower than the still cpu demanding GMA500 driver that seems to use modeset only in the Xorg log file. I think this is what I'll get with the Rage XL PCI card once it will arrive. Even if it seems detected and quite with many internal information of such old video card even a light GUI like LXDE can't get benefit from the Rage Pro video chip. I suppose these GUIs are much more demanding than these cards features even if not much different from any Win 9x/ME/XP GUI from a design point of view.
EDIT: reading around it looks like there's not really a modern DRM/DRI acceleration but from what I've read in the logs there's a XAA acceleration detected but performances are quite slow. OpenGL support were discontinued I think but there might be some third party DRM driver for mach64 chips to test. I hoped the official mach64 module would have solved this but it might need some Xorg.conf modding I don't know to enable a faster GUI.