I noticed that there were some USB 3.0 cards that used a PCI interface (including a few that used NEC chips like the NEC uPD720200).

Could Poseidon be upgraded to support this faster transfer rate standard?

> Could Poseidon be upgraded to support this faster transfer rate standard?

I guess it could. It's open source after all ;-)

can the PCI bus handle usb3 theorical speeds ?

> can the PCI bus handle usb3 theorical speeds ?

No, even the fastest PCI standard (533 MB/s at 64-bit 66 MHz) can't reach USB 3.0 maximum bandwidth (625 MB/s).

Even in a pc those cards, which btw just have an additional PCI to PCI-e Bridge on board, may fail depending on the used main board.

When they work, they are faster than USB2.0, but not as fast as USB3.0 could be.

Geit

MorphOS Poseidon isn't open source iirc.

> MorphOS Poseidon isn't open source iirc.

None of the changes in or additions to MorphOS Poseidon that were made after platon42's open source release in mid-2009 are part of the open source release, true. Everything that isn't MorphOS-specific has been included of course, and the bounty description says the following regarding the MorphOS-specific parts:

"Included will be: [...]
* pciusb.device OHCI, UHCI and EHCI combined host controller driver for MorphOS (AS IS!)
[...]
Special notes: [...] pciusb.device is only compilable for MorphOS/PPC and may require access to private MorphOS include files that will not be part of the open source release.
[...]
Excluded will be: [...]
* source code for input.device of MorphOS (for obvious reasons)
* MorphOS specific handling of stuff: query.library and mount.library support"

However, I don't know whether what has been open sourced of Poseidon is sufficient to add USB 3.0 support without having access to MorphOS-internal source code.

I would guess that access to sources for ANY supported USB.device should be sufficent to write an usb3.device.

So, if I follow what Andreas has mentioned, the standard PCI bus has less then a quarter of the bandwidth USB 3.0 can potentially provide?
And even if we were able to find a 64 bit PCI card (which seems unlikely) we'd still have less then half the bandwidth?

How is it that most PC cards are using a 1X PCIe connection?

[ Edited by Jim 01.04.2012 - 15:13 ]

> if I follow what Andreas has mentioned, the standard PCI bus has less
> then a quarter of the bandwidth USB 3.0 can potentially provide?

If "standard" shall mean "32-bit 33 MHz" then yes, 133 MB/s is less than a quarter of 625 MB/s, and only slightly more than a fifth of that.

> if we were able to find a 64 bit PCI card (which seems unlikely) we'd still
> have less then half the bandwidth?

Yes, 266 MB/s of 64-bit 33 MHz PCI as can be found in the G4 PowerMacs is less than half of 625 MB/s.

> How is it that most PC cards are using a 1X PCIe connection?

PCIe 3.0 x1 can transfer 985 MB/s. That's more than enough for USB 3.0's 625 MB/s. In case you refer to PCIe 2.x x1 (500 MB/s) cards here, it's true that they can't offer full USB 3.0 speed and thus would need x2 connection (1000 MB/s) at least.

>PCIe 3.0 x1 can transfer 985 MB/s. That's more than enough for USB 3.0's 625 MB/s. In case you refer to PCIe 2.x x1 (500 MB/s) cards here, it's true that they can't offer full USB 3.0 speed and thus would need x2 connection (1000 MB/s) at least.

So, current PCIe 2.0x slots at 1x can almost manage to support full USB 3.0 bandwidth and come close to what a 64 bit 66MHz PCI slot could do?
Fascinating.