AMD Radeon Technology Group, Senior VP and General Manager, Mike Rayfield to Resign

[Basard]

4870, 5870, 6970 and 7970 were pretty solid GPUs TBH

Youre right, I was drunk lol. I had a 5870 for a while--the super fancy 2GB. It was a hand-me-down though.

 

[Totally]

"Mike Rayfield joins Intel" announcement in a couple weeks, then.

This immediately crossed my mind after reading "AMD RTG exec resigns"

 

[bonehead123]

sing it now ya'll...

"Everybody.... WANG Chung tonite"

 

[gamerman]

R.I.P. amd...
rats and for sure leadersengineer ppl leave sinking ship...

bye amd, welcome Intel for GPU world.

 

[Dimi]

Welcome to Number Wang!

 

[Jism]

Keller left AMD too once the design was complete. All it took was tape-out and mass production. Nothing new here.

 

[RichF]

No, 290x was the beginning of the end, showing a GCN that was already pushed beyond healthy limits and far into inefficiency. When you have stock cards doing 90 C+ at normal loads, your shit isn't viable for a future iteration. And we know what came after that.

Note the power consumption of the RX580. It is rapidly moving the same direction again.

Temperature is related to die size and node characteristics, not just internal architecture. Put even the most efficient architecture on a low power process with a dense design library, then bump up the clocks on that process to obtain high performance, and watch performance-per-watt plummet.

Large die sizes also typically come at a temperature cost. There is also the issue of using higher leakage transistors to get more performance, something Nvidia reportedly did with the GTX 480. Use too many of those and you get a product like the GF100, so hot it can't even be sold in fully-enabled form.

As I understand it, though, die size can also be deceptive, if it is expanded in order to put more dummy material around to keep the hot spot problem down. There is also the issue of library density. Higher density libraries are good for keeping die size down but typically scale more poorly when it comes to clocks.

AMD also simply put on simple blower coolers that weren't adequate for cooling the product, as Nvidia did with the 480. It's also my understanding that the wider bus designs typically have lower performance-per-watt (not counting HBM). The 512-bit bus wasn't free performance, although AMD said it was able to reduce improve the efficiency of a 512-bit bus dramatically versus its earlier 384-bit design.

With Hawaii AMD still isn’t entering the big die race that defines NVIDIA’s flagship GPUs, but AMD is going larger than ever before. At 438mm2 Hawaii is AMD’s biggest GPU yet, and despite AMD’s improvements in area efficiency Hawaii is still 73mm2 (20%) larger than Tahiti. The fact that AMD is able to improve their gaming performance by 30% over Tahiti means that this is a very good tradeoff to make, it just means that AMD is treading new ground in doing so.

Similarly, at 6.2 billion transistors Hawaii is AMD’s largest GPU yet by transistor count, outpacing the 4.31B Tahiti by 1.89B transistors, an increase of 44%.

Comparison between Hawaii and Tahiti

 

[oxidized]

Keller left AMD too once the design was complete. All it took was tape-out and mass production. Nothing new here.

???????????????????????

What that's supposed to mean?