I don't think we're talking about how many pipelines the R520 HAS... more like how many will be working.
I'll join in the speculation party -
I think R520 has 32 pipes, and they've been re-spinning it over and over to try and get a decent number of chips off the line that have at least 16 of those 32 working as advertised. Ever since ATI first started hyping the R520, the rumored specs have been getting worse and worse. Sounds to me a lot like marketing writing checks that engineering and manufacturing can't cash.
"Our next GPU will score 35,000 on 3Dmark05!!!"
*whispers to engineer* "Get to work, we need a GPU that scores 35k."
I don't think ATI deserves any respect for not paper launching. The only reason they aren't is because Nvidia had cards on shelves on launch day, and ATI would look like schmucks if they did another paper launch.
I think ATI tried to get really fancy with the SM3.0 implementation, probably adding their own little twists (SM 3.12c...only on ATI! hurray!), and all they managed to do was balloon the transistor counts out of control and create a design that they couldn't build. We'll get something, and it will probably be competitive...but I don't think it will be anything like what ATI wanted to put out...which doesn't give me the best feeling.
Time will tell.
I'll join in the speculation party -
I think R520 has 32 pipes, and they've been re-spinning it over and over to try and get a decent number of chips off the line that have at least 16 of those 32 working as advertised. Ever since ATI first started hyping the R520, the rumored specs have been getting worse and worse. Sounds to me a lot like marketing writing checks that engineering and manufacturing can't cash.
"Our next GPU will score 35,000 on 3Dmark05!!!"
*whispers to engineer* "Get to work, we need a GPU that scores 35k."
I don't think ATI deserves any respect for not paper launching. The only reason they aren't is because Nvidia had cards on shelves on launch day, and ATI would look like schmucks if they did another paper launch.
I think ATI tried to get really fancy with the SM3.0 implementation, probably adding their own little twists (SM 3.12c...only on ATI! hurray!), and all they managed to do was balloon the transistor counts out of control and create a design that they couldn't build. We'll get something, and it will probably be competitive...but I don't think it will be anything like what ATI wanted to put out...which doesn't give me the best feeling.
Time will tell.