ATI AMR details hit rumour level

SOME DETAILS APPEAR TO BE COMING out about ATI's AMR. While they are considered 'rumour' level right now, we put more than 50-50 faith in them. AMR will be different from SLI, but also similar, freakily. A while back, we took a gander at the concepts behind tiling, AFR and split screen on NVidia's and ATI's approaches, theoretically in ATI's case.

What some of these sources are saying is that AMR will have a "master" and a "slave" card. These sources also say that the "master| will be a new board layout for AMR, and there is no word whether it will work without AMR.

The problem here is that OEMs don't like carrying multiple SKUs for the same chip. If you want to have an X700 AMR rig to sell, you need to design an X700 "master" and an X700 "slave", keep both in stock, explain to the mouth-breathers why the two "slaves" don't work the way they thought. Essentially you have to endlessly explain everything to customers. No one likes having to humiliate themselves by doing that. Sources are saying that the OEMs are balking at the idea of making a low volume "master "card, so ATI will probably be the only one to do it.

The other word out of different sources is that, as with SLI, ATI will have a "dongle" to connect the two cards. If true, and without a bridge, PCIe would get hammered into paste with all that traffic. It is good to contemplate the notion that ATI is making a long term solution rather than a short term hack.

Overall, the rumours have some interest. No word on performance, and I would cringe a little at the image quality implications if the VGA to VGA dongle part is true. Also, it may limit the inter-card communications, or at least force that to be passed over PCIe. Shades of 3DFx eh? I guess we will know for sure around Computex time.

The dongles are reported to go from DVI to DVI, or from VGA to VGA, which may limit what cards you can put where, and how it would work out. That said, since most cards worth a damn have both outs, this should not cause many problems. µ