AMD "Tonga" Silicon Features 384-bit Wide Memory Interface

[Red_Machine]

*yawn* My 580 has a 384-bit wide memory bus. It's four years old.

 

[nunyabuisness]

is it just me. or is the 285 showing some serious performance issues? im looking at those benchmarks you posted. its 50% slower then a 280! and in australia I have to pay $120 bucks more for a 285 vs a 280!! AMD is seriously almost done as a company. no innovation for years. still using 32nm 2d transistors. still using far too much power!
AMd has to release something new QUICK or it faces extinction

 

[GSG-9]

is it just me. or is the 285 showing some serious performance issues? im looking at those benchmarks you posted. its 50% slower then a 280! and in australia I have to pay $120 bucks more for a 285 vs a 280!! AMD is seriously almost done as a company. no innovation for years. still using 32nm 2d transistors. still using far too much power!
AMd has to release something new QUICK or it faces extinction

Don't be tooo hard on them.

I am using an Nvidia 690 that I bought for half price ($550) a year ago. It's been out for 2 years.
Beats anything anyones come out with. Buying a peak card after the next gen comes out is pretty awesome.

 

[THE_EGG]

*yawn* My 580 has a 384-bit wide memory bus. It's four years old.

My gtx260 sp216 has a 448-bit bus....I fail to see the point of your post. From what I've tried out with gaming performance, the width of the memory bus doesn't mean all that much in the real world (though there were many other variables that contributed too) and if having a smaller width means lower cost without sacrificing (much) performance then I'm all for it.

I'm actually excited about this. It should mean that Tonga is very adaptable for different performance segments which in turn should mean low price points.

 

[Casecutter]

Man you all miss'd that article is saying "Fiji" GPU as a 380X in February as a TSCM 20nm part, so what does that mean for this a full-Tonga?

IF AMD can get 20nm part the best's the 970/980 in 5 mo's I might say it's not that dire... I mean it took Nvidia 6 mo's to find a "part" on an current fab that performed better than a 7870 and then was terrible perf/watt while more often $30-40 more.