Techwell tw2804 powered card

speedkreature · Post by **speedkreature** » Thu Aug 13, 2009 5:49 pm

I've got a PCI card that's designed to handle up to 16 camera with 240Hz total screen bandwidth. Realistically, I don't need more than 8 to 10.
I have 13 of these cards and they were intended to be used with Windows Server 2003 and the software they come with (see video-insight.com).

Currently, Linux (any of the 2.6 branch) does not even detect the presence of this card on the PCI bus as far as I can tell.

The card has 2 Techwell TW2804 video chips, an Altera Cyclone chip, an Altera Max chip, and a Samsung K4S161622H-UC60 (which I believe is memory). The myriad other chips on this card are too small to clearly identify without a magnifying glass.

Is anyone aware of:
1) How to determine if the kernel/hal sees this card?
2) If there is a patch that is available so this card can be recognized?
3) If a card could be donated, what it would take to get it working?

speedkreature · Post by **speedkreature** » Fri Aug 14, 2009 7:15 pm

Through a process of elimination I was able to determine that the kernel is detecting the presence of a device in the PCI slot occupied by the multi-media card, however it is being incorrectly identified by "Actel Device ca03 (rev 59)"

Some extensive googling revealed that there are other cards affected as well, and ID'd just the same. I was unable to find any definitive specs, or a make and model on any of those cards

I think I'll first report this to my distro in hopes it is further pushed upstream--though I think it really ought to go to the kernel developers (who honestly probably couldn't care less) as both Debian and Fedora based distros seem to be affected. Looks like it's Win2K3 for me in the mean time.

speedkreature · Post by **speedkreature** » Thu Jan 06, 2011 4:19 pm

It's been a while, but I've not given up and I have some new information to post.

1) Altera Cyclone development is alive and well in the Linux community. There are even entire distributions devoted to this chip. I haven't actually been able to find a module or any relevant source, however, for this application of this DSP.
2) Source for the Techwell TW2804 already exists for Linux and it is fairly mature.
See: http://tomoyo.sourceforge.jp/cgi-bin/lx ... s-tw2804.c

3) The card's DSP (Altera Cyclone) is actually manufactured by Actel, then rebranded, if my research is correct.

What does this mean for us:

If anyone here knows how to build and import modules for the Cyclone (in this implementation) and the TW2804, we may very well be able to put to rests the inquiry's on cards powered by this combination of chips.

jdhar · Post by **jdhar** » Thu Jan 06, 2011 5:57 pm

speedkreature wrote:It's been a while, but I've not given up and I have some new information to post.

1) Altera Cyclone development is alive and well in the Linux community. There are even entire distributions devoted to this chip. I haven't actually been able to find a module or any relevant source, however, for this application of this DSP.
2) Source for the Techwell TW2804 already exists for Linux and it is fairly mature.
See: http://tomoyo.sourceforge.jp/cgi-bin/lx ... s-tw2804.c

3) The card's DSP (Altera Cyclone) is actually manufactured by Actel, then rebranded, if my research is correct.

What does this mean for us:

If anyone here knows how to build and import modules for the Cyclone (in this implementation) and the TW2804, we may very well be able to put to rests the inquiry's on cards powered by this combination of chips.

It seems unlikely that it's rebranded as an Actel. They are both FPGA companies, and competing ones at that.

And you can't "build modules for the cyclone". The Cyclone is an FPGA (or Cyclone II, III or IV for that matter), it's not a typical "device" chip that has drivers associated with it (like a Realtek chip for example). It's what they are doing inside the FPGA that is relevant. If it's in a video card, it's most likely doing some transcoding or glue logic, none of which would require drivers. You mentioned there are Techwell chips, so those are what need the appropriate drivers, and that chip is most likely connected to the FPGA for acceleration.

Now what's possible is that the original driver you found was intended for a different piece of HW that used an Actel FPGA. If this is the case, I wouldn't count on it working unless they implemented the exact same logic inside the Altera chips, which seems unlikely also since they may as well have used the Actels if that's the case.

Confused?