> what kind of UHCI controller chip exactly pegasos2 have ?

The Peg2's UHCI controller is not a dedicated chip but is inside the Via VT8231 southbridge chip.

> The Peg2's UHCI controller is not a dedicated chip but is inside the Via VT8231 southbridge chip.

Checked in OF via the "cd /pci/usb" and then ".properties", saying not VT8231, but VT83C572.

I can check it at home.
But Pegasos2 have Via VT8231 southbridge with 4port (2x2) USB 1.1
and Via VT83C572 is PCI to USB controller with (probably) 2port USB1.1
Cannot be VT83C572 part or subset of VT8231?

Now I see, it can be caused by identical DeviceID (3038h) of VT83C572 and USB controller of VT8231.

>> ".properties", saying not VT8231, but VT83C572

> Cannot be VT83C572 part or subset of VT8231?

Yes, I think that the VT83C572 is the UHCI controller contained in the VT8231 southbridge chip. The VT83C572 is available as both a dedicated chip, a controller inside UHCI/EHCI USB2 chips (like VT6202) and a controller inside multi-function chips (like VT8231 southbridge).

Quote:
Even unicorns occasionally appears...
Here on ebay is exactly such adapter.

adapter: AGP to 66MHz PCI or PCI-X Adapter
author: einstein186
description:
This adapter allows the use of 66 MHz PCI or PCI-X cards in a 3.3V or Universal AGP.
As AGP is only an extension to the 66 MHz PCI Revision 2.1 specification, it inlcudes almost all the funtionality of a 66MHz 3.3V PCI slot. But as AGP uses a different connector it is by default not possible to install high performance 66 MHz PCI or PCI-X adapters in the AGP. We have developed this adapter to allow the use of high speed interface cards like 66 MHz PCI or PCI-X Gigabit Ethernet and SATA adapters in headless systems. The adapter uses a 6 Layer PCB design according to the AGP specification. With the theoretical throughput of 266 MB/s this adapter allows unprecedented I/O bandwidth on AGP based systems.
PCI cards relying on the LOCK signal, which is missing from the AGP, might not work properly with this adapter.
You are buying directly from the creator of this product. Please get in touch, if you have any questions.

So, we can experiment wit Pegasos 2 and PCI(e) cards...

@sailor
That sounds interesting, what graphics cards would you like to test?
I suppose for Matrox Parhelia we have no drivers...

[ Editiert durch polluks 05.04.2023 - 12:51 ]

@polluks:
Matrox Parhelia is nice card, but most probably not more powerful than Radeon 9800 PRO, which is now in my Pegasos 2. Not count the drivers .

But there are PCI HD cards - one of them (HD4350) works for me in 33 MHz PCI slot.
- when bigfoot finishes new gfx drivers, these cards could be faster than 9800 PRO.

Largest profit can be from testing PCIe cards. Like X1950 or HD Northern Islands. Maybe they will not works because of Smart firmware initialization behind the PCI-PCIe bridge - we will see.
And second what could work very nice here is SiI SATA1 or Promise SATA2 cards. Promise SATA2 I tested some years before ( see first post ) and speed was limited by slow PCI 33 MHz slot.




[ Edited by sailor 05.04.2023 - 15:18 ]

> there are PCI HD cards - one of them (HD4350) works for me in 33 MHz
> PCI slot. [...] Largest profit can be from testing PCIe cards. Like X1950
> or HD Northern Islands. Maybe they will not works because of Smart firmware
> initialization behind the PCI-PCIe bridge - we will see.

Wouldn't this be the same as with your PCI HD4350 where SmartFirmware doesn't work but MorphOS does? After all, it shouldn't matter whether the bridge is soldered on the graphics card itself or on an adapter the graphics card is installed to.

I doubt that an AGP to PCI to PCIe adapter will get you anywhere with a Pegasos2 speedwise. AFAIR, the Pegasos2 only gives you a PCI 66MHz slot with an AGP connector, that means that non of the additional AGP DMA transfers really work which are important to get any of the more recent (AGP) cards up to speed - if graphics acceleration works at all without proper DMA, that is ...
Even if you can get any newer card to work somehow, don't expect that you reach any better performance from it which makes the whole setup kind of pointless IMHO.

Quote:

And this is exactly thing what should be tested.
Look here - you have to translate from Czech language. Sam440ep-flex is now much better computer with PCIe R9 270X.

Sam440ep-flex is similar case like Pegasos2 - AMCC440ep has only PCI controller ( this time with PCI connector ), exactly as Marvell Discovery 2 in Pegasos 2 ( that one with AGP connector ). Both controllers has 66 MHz.
Moreover - Sam shares 66 MHz PCI bus with SATA driver and bridge with slow devices. Pegasos 2 has 66 MHz bus dedicated for gfx, with own Hostbridge.

If you take a look on article above, you see that bus speed ( write to VRAM ) with PCI-PCIe bridge is much slower ( latency from pcie reverse bridge ) than classic PCI/AGP cards. Despite this disadvantage, pcie cards are better.

Such thing cannot be exactly theoretically calculated - they should be tested in real conditions. And of course, it also much depends on drivers - PCIe cards are much more powerfull, but drivers should utilize them.
Thus I am looking forward to bigfoot's great work.




[ Edited by sailor 06.04.2023 - 08:36 ]

Quote:

It will be great. if yes. I agree with you that it should be this way.

But we will see - PCI HD4350 has incorporated PCI-PCIe bridge 8111 ( or 8112 ) on card. It works.
If I tested these times another real PCIe card ( I forgot the type) with external bridge 8111, it not worked.

So chance there is, we will see.


[ Edited by sailor 06.04.2023 - 09:21 ]

>>> here is succesfull 2GB with rev 2B1 2GB on Pegasos 2 rev 2B1
>>> [...] it looks that it works with linux.
>>> But this is the only Pegasos 2B1 revision with 2GB I know about.

>> It's even the only pre-2B5 board with reliably working 2 GiB *I* know about ;-)

> sometimes miracle happens, Prober is lucky man.

Plot twist:
https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=13597&forum=11&start=9