Nothing big, just some people were shown a private demo of Barcelona, all they talk about is AMD's future plans, I want to know if it was fast, baby!
Here:
http://blogs.zdnet.com/Berlind/?p=364
At least some reliable sources reporting this news. Tis looking that tad bit brighter for AMD. It would be better if we had some photos of this demo though Not ones of Task Manager benchmarks either
Talk is cheap... If they are so convinced that they have such good product, release engineering samples of both products to the reviewers WITHOUT NDA's... In other words, walk the walk.
The teraflop is all about the R600 it just happened to be running on a Barcelona system. This isn't terribly impressive as a singe 8800 does 500 gigaflops perhaps a bit more.
The teraflop is all about the R600 it just happened to be running on a Barcelona system. This isn't terribly impressive as a singe 8800 does 500 gigaflops perhaps a bit more.
Anyone got a link to show how much an 8800 does???
The teraflop is all about the R600 it just happened to be running on a Barcelona system. This isn't terribly impressive as a singe 8800 does 500 gigaflops perhaps a bit more.
Anyone got a link to show how much an 8800 does???According to Nvidia, around 550Gflops which would make it and the R600 very similar.
What a bummer. Considdering the R600's time to market, I would have expected it to blow the 8800 out of the water. It won't be bad but it wont be as wonderful as what I had hoped for. (just my thoughts anyway).
I might be wrong but the pcb seems somewhat smaller than the one from gf8800 gtx. Maybe the same size that gf8800gts.
Therefore, the cooler is freaking ridiculously huge. If graphic cards keep on going like this, they'll pass my toaster in kw\h, and maybe reach my washing machine in kw\h count.
What a bummer. Considdering the R600's time to market, I would have expected it to blow the 8800 out of the water. It won't be bad but it wont be as wonderful as what I had hoped for. (just my thoughts anyway).
I'd imagine the R600 should be slightly faster than the 8800GTX though, and worth buying if you're already in the market for a new card. Also seeing as no Direct X10 titles have hit yet, I wouldn't worry too much as I'm sure there are still plenty of people who have no made the jump, giving ATI plenty of business as well.
One question : How good at scaling is sli? 550 + 550 = 1.1 terraflop
In a best case senario you can get literally 2x the performance, but generally this isn't the case with games. For general purpose code, it's really difficult to say as it's never been used before. I'm also sure that when AMD said 1 teraflop, they were talking about theoretical performance.
MIPS and MFLOPS are just a marketing gimmick, with no meaning for real world performance, and this stands on Chapter 1 in any decent book on computer architectures.
For example, an architecture could do 1 ADD per clock, but only one MUL every 5 clocks, and another architecture is able to do one MADD (combined multiply add) per clock.
If they have the same frequency and same number of ALUs, both architectures would have the same MFLOPS score, yet the second architecture would be way faster in real world situations.
MIPS and MFLOPS are just a marketing gimmick, with no meaning for real world performance, and this stands on Chapter 1 in any decent book on computer architectures.
For example, an architecture could do 1 ADD per clock, but only one MUL every 5 clocks, and another architecture is able to do one MADD (combined multiply add) per clock.
If they have the same frequency and same number of ALUs, both architectures would have the same MFLOPS score, yet the second architecture would be way faster in real world situations.
Note the word "theoretical"; I never once said real world performance.
Wow, the last half of 2007 could be pretty big for AMD.
Yes, the 8800GTX SLI is already capable of over a Teraflop with CUDA compiled programming on any fast compatible motherboard and a C2D.
It does not have to be a Quad core, just like the AMD system would not have to be a Barcelona core for the test if the video subsystem was the one doing the work.
Those numbers are interesting, but I think whether or not AMD can hold off Penryn will be dependant on whether these figures represent a performance lead per clock over Clovertown, or absolute performance lead over Clovertown. If Barcel
Not ones of Task Manager benchmarks either 
