A motherboard based on an Intel B75, Q75, Q77, or H77 chipset (Ivy Bridge processor support). Here's one I use (Gigabyte GA-P75-D3):

https://www.gigabyte.com/Motherboard/GA-P75-D3-rev-10#ov

BTW, I've used PCI video cards with both Native PCI and bridges (using a PCI to PCIEx1 adapter) on the same board, and I haven't noticed any difference. Same with network cards. Sound cards tend to have a lot of trouble with these bridges. I suspect some other precision data collection cards might also suffer from the added latency.

fosterwj03 wrote on 2021-08-20, 14:27:

A motherboard based on an Intel B75, Q75, Q77, or H77 chipset (Ivy Bridge processor support). Here's one I use (Gigabyte GA-P75-D3):

https://www.gigabyte.com/Motherboard/GA-P75-D3-rev-10#ov

BTW, I've used PCI video cards with both Native PCI and bridges (using a PCI to PCIEx1 adapter) on the same board, and I haven't noticed any difference. Same with network cards. Sound cards tend to have a lot of trouble with these bridges. I suspect some other precision data collection cards might also suffer from the added latency.

Thanks! So did you try that adapter into a modern PCI not-native bus in a mainboard that already has its own internal bridge right? That's the config I'm testing now, basically two bridges are working for the opposite tasks just cause the mainboard manufacturer decided this smart choice of having only a single fake PCI bus when the chipset NM10 already had the PCI-EX line to use at least in a x1 configuration.

No. The fastest PCI card I have is a Matrox G200. I think it uses native PCI internally on the card/chip (hard to tell from info available online). The G200 works the same on both Native PCI slots and PCI slots using a PCIE bridge.

I also have a Q67 motherboard with only 2 Native PCI slots and 2 PCIE slots. I use the PCI-to-PCIEx1 adapter with this board when I need a third legacy PCI slot. Like I said, I see the same performance with my video cards and network cards with this board on both the native slots and the bridge adapter.

You might see some additional latency penalty on a "modern" PCI card with a lot on onboard memory, but I suspect the PCI bus bandwidth will create the biggest bottleneck on the system. I doubt you'll detect much of a difference in Windows 7/8/10 video benchmarks.

Most powerful one I could think of is X79 (and X79 doesn't use a bridge chip).

It would be interesting to test something like a GT 430 PCI (Native) vs GT 430 PCI (Bridged) vs GT 430 (PCIEx16) on the same motherboard. I would expect the PCIEx16 to win hands down.

fosterwj03 wrote on 2021-08-20, 16:10:

It would be interesting to test something like a GT 430 PCI (Native) vs GT 430 PCI (Bridged) vs GT 430 (PCIEx16) on the same motherboard. I would expect the PCIEx16 to win hands down.

I have PCIe x1 and PCI GT430 cards. Haven't tested th m against each other though. The PCIe card should definitely win in certain circumstances.

I don't know if it's worth it for the kind of testing you are doing but I wouldn't normally bother about whether an ivybridge or sandybridge system ran a PCI graphics card well, because intel HD graphics are a massive step up from their previous onboard efforts. i.e. a low end card is not a night and day improvement like it was over previous gens. I wouldn't put the chipsets you get on PCI cards into the system on PCIe cards and expect an improvement. So in the case of those systems it's only worth upgrading to something that was more midrange. Not sure if you gain any backwards compatibility with using older PCI graphics, I know older (early XP era) stuff I tried on my laptop with HD4000 was fine.

You could use one of these with the PCI GT 430 for a PCI Native slot vs Bridge comparison:

https://www.amazon.co.uk/KALEA-INFORMATIQUE-A … e/dp/B016MUQW6M

I have one of these (purchased from Amazon US for $15). It uses the same Asmedia bridge chip found on motherboards that don't have native PCI from the chipset.

What I'm testing now are a GT610 PCI and the GT610 PCI-Ex version but the mainbord chipset I'm testing the first in, has only a single old PCI bus that most probably use a mainboard PCI-EX to PCI bridge not as a single external IC but I suppose inside the southbridge NM10. On the vga obviously there's a dedicated PI7C9X bridge that does the opposite from PCI bus to PCIe x1 GPU.
I'm quite sure of this cause the PCI list shows the 82801 PCI bridge of the mainboard and the mainboard did have a desoldered (by factory) mini-PCIex connector.
I did test the GT610 PCIe into a G41 Socket 775 system and the native PCIe connection even if used at maximun 50% still gave a 30% to 50% boost mostly (even if the cpu was much faster indeed). I'm not sure anyway if I can test the PCI GT610 into the G41 chipset having two PCI bus that I didn't check if native or bridged.

The test (just for curiosity, nowdays vga are so expensive I find more interesting the old low end cards) was to see how much I could boost a very weak low power mini-itx system using a very modern low power vga but the mini-itx board has only the "probably fake" PCI so two bridges are working here the vga one and the mobo one.

I guess you could hack a PCIe 1x extender into the PCIe lines the bridge chip uses on the motherboard, cutting off PCI but giving you "pure" PCIe for a graphics card.

Is it accurate to say that Intel used PCI to ISA bridges for years starting with Triton?

If that's accurate, then why all the hooplah with PCIe to PCI bridges?

BitWrangler wrote on 2021-08-20, 16:31:

I guess you could hack a PCIe 1x extender into the PCIe lines the bridge chip uses on the motherboard, cutting off PCI but giving you "pure" PCIe for a graphics card.

Yes that would be the "fastest" connection but unfortunately the mini-pciex connector was factory desoldered as an option for more expensive mini-itx models that has that installed. The internal iGPU GMA probably was PCI express.

mockingbird wrote on 2021-08-20, 16:33:

Is it accurate to say that Intel used PCI to ISA bridges for years starting with Triton?

If that's accurate, then why all the hooplah with PCIe to PCI bridges?

I am not expert but I suppose the convertion from PCIe to PCI might be more complex considering what I read from feature list of these IC bridges. PCI and PCIe seems to work quite differently and need more features for such high performances applications.

mockingbird wrote on 2021-08-20, 16:33:

Is it accurate to say that Intel used PCI to ISA bridges for years starting with Triton?

If that's accurate, then why all the hooplah with PCIe to PCI bridges?

I would say that PCI-to-PCIE bridges work well in most applications if you still need legacy peripherals depending on the motherboard manufacturer's implementation. I've seen good and bad.

Some peripherals suffer more from the added latency than others (sound cards sound crackly in both playback and recording), and precision instruments don't work well.

If you absolutely need a PCI card for your application (either for retro uses or real work), I think it's worth the trade-off to have the capability to use legacy peripherals.

mockingbird wrote on 2021-08-20, 16:33:

Is it accurate to say that Intel used PCI to ISA bridges for years starting with Triton?

Yes. Intel started using this with the PIIX south bridge on the 430FX chipset.

The last official support for PCI to ISA bridges was with PIIX4 south bridge.

mockingbird wrote on 2021-08-20, 16:33:

If that's accurate, then why all the hooplah with PCIe to PCI bridges?

I think that the use of bridges was for providing legacy support due to the shrinking relevance of the PCI bus over the time for mainstream users.

Removing the actual PCI bus from later chipsets was for cost reasons since it's cheaper to manufacture then without this.

Does anyone know why the Sparkle GeForce 8500 GT PCI video card does not work with the Gigabyte GA-X58-UD5 motherboard? The system simply does not start. It just power cycles endlessly.

I have two of these Sparkle cards, both the same. Both work in other boards, but not this one.

By contrast, the following PCI cards work without problem in this board: PNY GeForce 8400 GS, Jaton GeForce 9500 GT, Zotac GeForce GT 610.

Errius wrote on 2021-08-20, 17:01:

Does anyone know why the Sparkle GeForce 8500 GT PCI video card does not work with the Gigabyte GA-X58-UD5 motherboard? The system simply does not start. It just power cycles endlessly.

I have two of these Sparkle cards, both the same. Both work in other boards, but not this one.

By contrast, the following PCI cards work without problem in this board: PNY GeForce 8400 GS, Jaton GeForce 9500 GT, Zotac GeForce GT 610.

It might be a resource conflict (probably memory areas). Have you tried it in different slots?

There is only one PCI slot, hehe.

(The older model, GA-EX58-UD5 did have two, but I no longer own one.)

ETA: I see from my 3DMark records that the Sparkle card did indeed work with the GA-EX58-UD5. The problem is with the newer GA-X58-UD5 rev. 2.0.

ETA2: The card I tested with the EX58 was a SF-PC85GT256U2. I no longer own this card. The two cards I am testing with the X58 are SF-PC85GTT3T256U2. I don't know if this makes a difference.

ETA3: That second number is not a typo. Picture added.