It's been four years since AMD launched RDNA, the successor to the venerable GCN graphics architecture. We take a look through the tech and numbers to see...
I mean… AMD spun off their fabs in 2008/2009, which eventually became glofo. To say glofo 32nm was bad in 2011 is really to say AMD 32nm was bad in 2011 because much of the 32nm research and development would have been locked in from when AMD was still running the fabs. I imagine they’d have been far worse, even, without the capital injections from Mubadala.
Bulldozer was a poor architecture choice even if it had a great node to work off. High clocks and poor IPC was not a design for the future, plus the shared FPU meant that for floating point it was like you had half the cores, they had bet all that work would move to the GPU way too early and CPU performance needs obviously never went away.
Underrated factor of Bulldozer’s downfall was GlobalFoundries 32nm node being a major disappointment. TSMC is a far cry ftom GF.
Didn’t help the APUs on 28nm.
I mean… AMD spun off their fabs in 2008/2009, which eventually became glofo. To say glofo 32nm was bad in 2011 is really to say AMD 32nm was bad in 2011 because much of the 32nm research and development would have been locked in from when AMD was still running the fabs. I imagine they’d have been far worse, even, without the capital injections from Mubadala.
Bulldozer was a poor architecture choice even if it had a great node to work off. High clocks and poor IPC was not a design for the future, plus the shared FPU meant that for floating point it was like you had half the cores, they had bet all that work would move to the GPU way too early and CPU performance needs obviously never went away.