Not necessarily.

The Serpent Rider wrote on 2023-08-11, 14:48:

Not necessarily.

How should I know?

Install any sound card. They require DMA to function properly, except FM synth.

Oh ok, already done that. A Vibra16 VX was ok.

Thoughts on using such an adapter in a g3 mac? Is there a card that could take advantage of it?

So it's some kind of IRQ issue. I've tried to install Radeon X series cards and all of them also refused to work even in normal PCI slots too. Apparently, because old Radeon cards are identified as two separate entities for each display output. And primary entity always has Code 12 error.
In other words, video card and PCI-PCIe bridge actively try to sabotage each other with resource sharing and TR-DLS is too dumb to figure it out (or specifically HP Phoenix BIOS).

The Serpent Rider wrote on 2023-08-12, 00:38:

So it's some kind of IRQ issue. I've tried to install Radeon X series cards and all of them also refused to work even in normal PCI slots too. Apparently, because old Radeon cards are identified as two separate entities for each display output. And primary entity always has Code 12 error.
In other words, video card and PCI-PCIe bridge actively try to sabotage each other with resource sharing and TR-DLS is too dumb to figure it out (or specifically HP Phoenix BIOS).

I think you are right.
That is why I feel it depends from the bios rather than the motherboard or chipset itself.
That said, with the 370DLE I encountered the exact same problem with bridged pci cards (Matrox p690 and GeForce 610) that use a different bridge chip, Pericom.
So I think is a combination of bios and chip used on the bridge.

That being said, now I also have some doubts that PEX8112 is working in 66Mhz mode. Even 6800GS can score 8k+ points in 3DMark on normal PCI.

The Serpent Rider wrote on 2023-08-12, 09:49:

That being said, now I also have some doubts that PEX8112 is working in 66Mhz mode. Even 6800GS can score 8k+ points in 3DMark on normal PCI.

There is a selector on the S370DLE that allows 66Mhz on pci64 slots.
I obtain better results (close to double) when enabling it.
I do not think a 6800gs is that much inferior to hd5570 in dx8.
And 3dMark does not saturate the bus much probably.

Mamba wrote on 2023-08-12, 11:10:

There is a selector on the S370DLE that allows 66Mhz on pci64 slots. I obtain better results (close to double) when enabling it. […]
Show full quote

The Serpent Rider wrote on 2023-08-12, 09:49:

That being said, now I also have some doubts that PEX8112 is working in 66Mhz mode. Even 6800GS can score 8k+ points in 3DMark on normal PCI.

There is a selector on the S370DLE that allows 66Mhz on pci64 slots.
I obtain better results (close to double) when enabling it.
I do not think a 6800gs is that much inferior to hd5570 in dx8.
And 3dMark does not saturate the bus much probably.

What test does saturate it better? I definitely did not see 2x performance gain on my Matrox from 33 to 66..... not even close. Could be the card is so weak to begin with.

Sphere478 wrote on 2023-08-11, 17:35:

Thoughts on using such an adapter in a g3 mac? Is there a card that could take advantage of it?

There's a few PCIe cards for the later G5 powermacs like the 6600LE and 7800GT but not much else.

Grem Five wrote on 2023-08-12, 13:28:

Mamba wrote on 2023-08-12, 11:10:

There is a selector on the S370DLE that allows 66Mhz on pci64 slots. I obtain better results (close to double) when enabling it. […]
Show full quote

The Serpent Rider wrote on 2023-08-12, 09:49:

That being said, now I also have some doubts that PEX8112 is working in 66Mhz mode. Even 6800GS can score 8k+ points in 3DMark on normal PCI.

There is a selector on the S370DLE that allows 66Mhz on pci64 slots.
I obtain better results (close to double) when enabling it.
I do not think a 6800gs is that much inferior to hd5570 in dx8.
And 3dMark does not saturate the bus much probably.

What test does saturate it better? I definitely did not see 2x performance gain on my Matrox from 33 to 66..... not even close. Could be the card is so weak to begin with.

Cannot answer to your question as my statement was a guess also.
I can say you are not alone on your boat, you can see others in the same situation with Matrox on this forum.

Mamba wrote on 2023-08-12, 14:39:

Cannot answer to your question as my statement was a guess also.
I can say you are not alone on your boat, you can see others in the same situation with Matrox on this forum.

Why earlier in the thread I asked your settings in the software you used. One of my pet peeves in the retro scene is we all want to know if our hardware is working up to spec but everyone decides to F with the settings they use when they test and normally dont say what those settings are in their posts. Makes it incredibly hard to get an idea how your and their particular hardware is working comparatively. Not a rant towards you as you told me what you were using but in general, from other threads I have seen my card does better than most but many times I guessing on the settings they are using.

Stock settings in the software to me personally is the best way to test. Personally I'm interested in one of those adapters you have to test other cards but I got my P3tdle with dual 1.4s and The pci64 Matrox for cheaper that what you can get a single 1.4 Tualatin processor on ebay for so I'm in no rush to add the cost of an adapter to test right now but hopefully in the future.

Grem Five wrote on 2023-08-12, 17:04:

Mamba wrote on 2023-08-12, 14:39:

Cannot answer to your question as my statement was a guess also.
I can say you are not alone on your boat, you can see others in the same situation with Matrox on this forum.

Why earlier in the thread I asked your settings in the software you used. One of my pet peeves in the retro scene is we all want to know if our hardware is working up to spec but everyone decides to F with the settings they use when they test and normally dont say what those settings are in their posts. Makes it incredibly hard to get an idea how your and their particular hardware is working comparatively. Not a rant towards you as you told me what you were using but in general, from other threads I have seen my card does better than most but many times I guessing on the settings they are using.

Stock settings in the software to me personally is the best way to test. Personally I'm interested in one of those adapters you have to test other cards but I got my P3tdle with dual 1.4s and The pci64 Matrox for cheaper that what you can get a single 1.4 Tualatin processor on ebay for so I'm in no rush to add the cost of an adapter to test right now but hopefully in the future.

I hear you.
I always run everything on stock and expect to see results equally obtained, otherwise we are comparing apple to tomatoes.
From what I can say, 3dMark01 scaled in a strange way when near 10k points, meaning power needed to reach 10k from 8k feels more than to reach 6K from 4K.
Feels like logarithmic.
That is why I think your score is accurate.

OR we think our retro top tier gpu are stronger than what they are in reality (or the lesser ones are better, take your pic).

At the moment you can get a 8112 for around 15€.

Ok, I've acquired Supermicro 370DL3 and can confirm that it scales with PCI 66Mhz. I've got 100 fps in 3DMark2001SE Nature test and that's with 1Ghz Coppermine CPU.

BTW does these boards require any BIOS modding to support Tualatins on top of adapters?

====================================

Results with 1Ghz Coppermine and GeForce 6800GS on Supermicro DL3 (CL2 timings):

Quake 3 Demo v1.09 / High settings preset (800x600x32) Demo001 - 130.6 fps
Quake 3 Demo v1.09/ High settings preset (800x600x32) Demo002 - 124.2 fps
3DMark 2001 SE (stock settings) - 7637 points

Detailed report:
Game 1 - Car Chase - Low Detail - 86.9 fps
Game 1 - Car Chase - High Detail - 24.8 fps
Game 2 - Dragothic - Low Detail - 126.2 fps
Game 2 - Dragothic - High Detail - 66.7 fps
Game 3 - Lobby - Low Detail - 85.8 fps
Game 3 - Lobby - High Detail - 37.5 fps
Game 4 - Nature - 103.3 fps

As you can see, even 1Ghz CPU can be on par or even outmatch 1.4GHz CPU with slower non-dedicated PCI bus in many tests. Plus Supermicro allows to mod memory timings to improve performance even further.

No,
Just put the latest bios.

The Serpent Rider wrote on 2023-08-14, 18:15:

Ok, I've acquired Supermicro 370DL3 and can confirm that it scales with PCI 66Mhz. I've got 100 fps in 3DMark2001SE Nature test […]
Show full quote

Ok, I've acquired Supermicro 370DL3 and can confirm that it scales with PCI 66Mhz. I've got 100 fps in 3DMark2001SE Nature test and that's with 1Ghz Coppermine CPU.

BTW does these boards require any BIOS modding to support Tualatins on top of adapters?

====================================

Results with 1Ghz Coppermine and GeForce 6800GS on Supermicro DL3 (CL2 timings):

Quake 3 Demo v1.09 / High settings preset (800x600x32) Demo001 - 130.6 fps
Quake 3 Demo v1.09/ High settings preset (800x600x32) Demo002 - 124.2 fps
3DMark 2001 SE (stock settings) - 7637 points

Detailed report:
Game 1 - Car Chase - Low Detail - 86.9 fps
Game 1 - Car Chase - High Detail - 24.8 fps
Game 2 - Dragothic - Low Detail - 126.2 fps
Game 2 - Dragothic - High Detail - 66.7 fps
Game 3 - Lobby - Low Detail - 85.8 fps
Game 3 - Lobby - High Detail - 37.5 fps
Game 4 - Nature - 103.3 fps

As you can see, even 1Ghz CPU can be on par or even outmatch 1.4GHz CPU with slower non-dedicated PCI bus in many tests. Plus Supermicro allows to mod memory timings to improve performance even further.

I've also checked Radeon X550 and this motherboard has no weird issues with two "virtual" GPUs chilling in one slot. Quite a fancy setup for Windows 98/early Windows 2k era.

EDIT: 8061 points on Radeon X800GTO!

Finally people showing some PCI @66Mhz ownage, nice work

I hope I can get some benches with a E7505 PCI-X 66-133Mhz soon

I also have the PCI-X to 4x PCIe adapter somewhere here.

I can dig out my ASUS X58 workstations board, which has PCI-X 133 Mhz, but no idea if PEX8112 can handle such abuse.

The Serpent Rider wrote on 2023-08-14, 22:51:

I can dig out my ASUS X58 workstations board, which has PCI-X 133 Mhz, but no idea if PEX8112 can handle such abuse.

the product brief pdf says no.

The PEX 8112 has a single parallel bus segment supporting the PCI v.3.0 protocol, and a 32-bit wide parallel data path runni […]
Show full quote

The PEX 8112 has a single parallel
bus segment supporting the PCI
v.3.0 protocol, and a 32-bit wide
parallel data path running up
to 66MHz.