Got one from ebay, and so far it works great. 😉 Finally USB2 speed on this old boy, it was devastating using USB1.

Not to really necro, but I do have a VIA VT6212 USB 2.0 card working fine with my BE6-II motherboard (440BX). NUSB version 3.6 under 98SE.

The instability that I have is caused by hot plugging/removal of USB 2.0 mass storage devices. They tend to work okay if the drive is inserted prior to booting and is never removed.

There is also a constant CPU penalty that is more severe on older processors.

Managed to buy a Startech 7 Port PCI USB Card (Part Code # PCIUSB7) recently. Yeah, it has the VIA chipset instead of the NEC chipset, but more crucially it had a 9pin USB header, which my motherboard doesn't have.

So now I have 2 working USB2.0 ports on the front of my Multi-Boot Retro PC.

I don't know if my 2 cents about this topic helps, but as I recall, I have used one of the better looking VIA Technologies VT6212L based USB 2.0 PCI card attached to ASUS P5A Rev. 1.04.
I do not use any Windows at all, and I used my own custom i586 instruction set compiled Gentoo Linux based LXDE (Lightweight X11 Desktop Environment) GUI with 640 Mb of RAM.
Probably because the mainboard uses Rev. E ALi Aladdin V chipset that can only cache up to 128 MB, the performance of the DE was at an unusable level.
It works, but it is simply too painful to use.
The microprocessor is Intel Pentium MMX 233 MHz with an XGI Volari V3XT (SiS based graphics engine) AGP graphics card.
I did not really use the VIA USB 2.0 port, but I think USB keyboard / mouse did work out of it.
As for the VIA VT6212(L) based USB 2.0 cards, one I used had the somewhat better looking PCB than the 2 or 3 other ones I picked up in the past.
Still, the PCB appears to be a 2 layer PCB.
98% of the VIA VT6212(L) based USB 2.0 cards I have seen appear to have a 2 layer PCB, but their reference design (there is a PDF design guide of this chip floating around somewhere on the Internet) is a 4 layer PCB.
Probably because VIA chip based products are considered cheap low end products (it does not have to be this way), it appears that most manufacturers choose a 2 layer PCB to cut cost.
Not only that, some of them I have seen don't even have a single poly fuse (that "p" logoed small chip near the USB port) on the PCB at all.

giantenemycat wrote on 2024-07-04, 07:31:

Unearthing this deal I came upon a while ago. For those in the UK, SCAN are selling a PCI USB 2.0 + FireWire card with a 3.5" fr […]
Show full quote

Unearthing this deal I came upon a while ago. For those in the UK, SCAN are selling a PCI USB 2.0 + FireWire card with a 3.5" front bay for £0.49. I bought 5 to get the most out of it cause why not, considering the shipping is almost ten times more than the item itself...

https://www.scan.co.uk/products/scan-2x ... ternal-hub

Even though the pictures show x2 USB, x1 FW and whatever that other little port is on the front bay, mine has x3 USB and x2 FW. Compatible in my GA-BX2000 and SE440BX-2, and Windows 98SE somehow has drivers for them on the disc - not sure how since the OS obviously predates USB2.0.

The attachment ali_xpress.jpg is no longer available

Just wanted to say thanks for the heads up on this, I ordered a couple and they work perfectly on a 440BX and 815EP where my VIA one kept disconnecting and causing problems (probably a faulty card)

Front panel looks and works great, I’ve not done any benchmarking with it in yet though

Does anyone know if it's economical to have someone in the UK forward a couple of these to the USA? I don't have any of these ALI branded chips to test and they're a lot cheaper on than this retail website than eBay.

Kahenraz wrote on 2025-01-07, 11:51:

Does anyone know if it's economical to have someone in the UK forward a couple of these to the USA? I don't have any of these ALI branded chips to test and they're a lot cheaper on than this retail website than eBay.

I too would be curious about this. I have lots of USB 2.0 cards, but... that front panel for it is very neat. I'd take a pile of them if the shipping was cheap enough.

What real world speeds are people getting on that ALi card on 440BX motherboards? I have the same chipset on my PCI USB 2.0 card but I don't get great results...

Measuring with ATTO Disk Benchmark, Read/Write tops out at around 16MB/8MB per second on the ALi card - the on board USB does about 1MB/1MB per second on the same drive

Lostdotfish wrote on 2025-02-12, 19:38:

What real world speeds are people getting on that ALi card on 440BX motherboards? I have the same chipset on my PCI USB 2.0 card but I don't get great results...

Measuring with ATTO Disk Benchmark, Read/Write tops out at around 16MB/8MB per second on the ALi card - the on board USB does about 1MB/1MB per second on the same drive

M5271
ALi P1394 / USB2.0 / MS / SD Controller
https://theretroweb.com/chip/documentation/m5 … f6527985707.pdf
^ This looks like a chip you'd normally find on the motherboard itself...

16 MB/s -> 104 Mb/s
8 MB/s -> 64 Mb/s

Those rates of data transfer actually sound decent for that era of hardware and USB 2.0, so if there is a limiting factor it could be your CPU, some quirk of a particular motherboard. or entirely a software problem.

Also, the ALi M5271 claims to support PCI Rev. 2.2 and 66 MHz operation, while the Intel 82443BX Intel® 440BX AGPset: 82443BX Host Bridge/Controller) only supports PCI Rev 2.1 and 33 MHz operation.

The M5271 is an interesting chip. It would be even better if there was a USB/Firewire/SD combo card. There weren't any consumer desktop motherboards I'm aware of that used 66MHz PCI, but the chip can handle it in case you need it for some workstation application.

bracecomputerlab wrote on 2024-12-28, 18:16:

As for the VIA VT6212(L) based USB 2.0 cards, one I used had the somewhat better looking PCB than the 2 or 3 other ones I picked […]
Show full quote

As for the VIA VT6212(L) based USB 2.0 cards, one I used had the somewhat better looking PCB than the 2 or 3 other ones I picked up in the past.
Still, the PCB appears to be a 2 layer PCB.
98% of the VIA VT6212(L) based USB 2.0 cards I have seen appear to have a 2 layer PCB, but their reference design (there is a PDF design guide of this chip floating around somewhere on the Internet) is a 4 layer PCB.
Probably because VIA chip based products are considered cheap low end products (it does not have to be this way), it appears that most manufacturers choose a 2 layer PCB to cut cost.
Not only that, some of them I have seen don't even have a single poly fuse (that "p" logoed small chip near the USB port) on the PCB at all.

I thought the card I have was 4-layer, but I see enough stuff (including power) routed on the outside that it might be 2-layer. It's hard to tell with black solder mask (it even blocks light around vias, so I can't see if there are internal layers by shining light through it), but it sure does show flux residue (I didn't add flux - that's just the flux core in the solder). I see that I will need to clean the edge connector if I ever find a computer to put it in. Everything I have with a fast enough CPU already has USB 2.0. This is the card I tested earlier and found that it severely affects performance with only a mouse connected.

This one isn't as cheap as some of the other cards shown in this thread. The regulator you can see is for the 2.5V core supply, and the really cheap cards simply use a silicon diode (1N400x series) off the 3.3V supply to make something around 2.6V instead. This card doesn't have a full-length PCI edge connector, so only half of the 5V pins are used. It has footprints for polyfuses, but it came from the factory with ferrite beads, which won't provide any protection at all. I replaced them with actual polyfuses, but because of the shorter edge connector, I used 750mA instead of the more common 1.1A. 1.1A would probably be more stable with 2.5" external hard drives, but one of those old Y cables to get power from two ports would work as well.

I don't remember if I had my usual rant on 5-port VIA cards in this thread, but I just saw one in the latest Epictronics video (where the VP6 was finally fully working). The VT6212 only supports four ports, and those 5-port cards never have an onboard hub chip. That means that the internal port is connected directly in parallel with one of the external ports, and nobody tells you which port is unusable (you have to follow the traces yourself). Also, the stub resulting from the traces leading to the now-dead port will cause problems at high data rates. The NEC chips actually do support five ports.

Lostdotfish wrote on 2025-02-12, 19:38:

What real world speeds are people getting on that ALi card on 440BX motherboards? I have the same chipset on my PCI USB 2.0 card but I don't get great results...

Measuring with ATTO Disk Benchmark, Read/Write tops out at around 16MB/8MB per second on the ALi card - the on board USB does about 1MB/1MB per second on the same drive

In my P3 440BX machine I notice a big performance hit in games like Quake 3 and UT when I put a USB 2.0 VIA based card in it. Do you experience anything like that with the ALI based card?

I don't think that this has anything to do with any of the various USB 2.0 branded chips on Windows 9x, but rather now Microsoft wrote the USB driver stack. If performance is tanking, then there is probably some kind of busy loop that the system spends time in, probably at the kernel level. My best guess is that there is some kind of active polling being done, and this is what is gobbling up CPU cycles. There aren't enough wait states in the loop to yield back to the rest of the system, or perhaps it's just poorly threaded. I don't think that the performance problem is "solvable" on Windows 9x, other than using a beefier CPU to try and compensate for the performance loss.

This hypothesis can be tested by swapping USB 2.0 cards in and out of Windows 2000, just to see if there is a performance hit there as well.