| ultima said: Wait a minute, these double cards work pretty much the same way as SLI? As in the memory pool isn't shared, but rather mirrored? Very lame if true... |
Yep, they do!
Plus, the Double-GPU cards generally have lower clockspeeds and in extreme cases PCI-E contention.
It all essentially boils down to advertising.
Back in the day when 3dfx had a couple of VSA100 chips on a single card, they would advertise them as 64Mb/128Mb etc'.
That is, each GPU on the card had that much memory all to itself, no sharing.
Fast forward to today and AMD and nVidia will advertise a high-end dual-GPU card with 3-6Gb when each GPU actually only gets half of that.
However, there is one exception to that rule, some compute scenarios will use the entire memory pool. :)
It's almost as bad as Hard Drive manufacturers stating 1Gb as 1,000 Megabytes, when in fact it's 1024 Megabytes, so you always get less than the advertised capacity in windows. :P
www.youtube.com/@Pemalite
