hmm, wouldn't it have been better not to build such a HUGE chip for once. I know they think it will work, but it just means they will have more 340-379 cards (whatever number they will give it) and less of the performance cards of 380+.
I hope they made a smart choice while designing the chip, that even if parts fails, they still can make a good chip from it. But more smaller chips would have been better if you ask me.
That's bad news... So far ATI hasn't been able to fully cash in on their DX11 lead due to supply troubles, but just as these are getting fixed Nvidia announces another delay. I hear champagne corks flying in the ATI headquarters.
Have they taken to much hay on their pitchfork? Every statement they release boasts about the sheer scale of Fermi, yet it's always accompanied by another delay. It's windy at the top...
ATI's rumored to have a 5890 in the works, so however powerful Fermi is, ATI's is gonna have a response by either releasing a cheaper 5890, or taking back the performance crown with the 5890. Nvidia is really shooting themselves in the foot here. What are they gonna do when ATI lowers the price the the 5850 and 5870 by a $100 making it a way better bang for your bang then Fermi. And by march ATI is gonna have all of its low and mid range cards out, which are the true money makers for a GPU company.
nVidia's new cards will be very fast and a bit overpriced.
ATI will fight back against nVidia's flagship by releasing the 5890.
Depending on the performance of the nVidia cards ATI may pull ahead or they may still be behind which will for ATI to release thier new flagship at a lower price. ATI may still try to attain the top spot thus causing them to release a higher performing card. 5900 series maybe? Only time will tell but I like to see that nVidia is trying to innovate and I am excited to see the benchmarks for these new "Fermi" cards.
I think nVidia was surprised by the current ATI performance and had to delay in order to bring them up to par. I think their constant mention of Cuda technology and virtually nothing on the "pixel pushing" supports this assertion. Maybe they are on to something... but it's not going to be significant in this generation of cards. Most software doesn't even need the computing power a core i7 provides. Why do we need parallel co-processing via GPU? nVidea made a monster chip that will most likely under perform where it matters, and cost a fortune.
[citation][nom]SneakySnake[/nom]ATI's rumored to have a 5890 in the works, so however powerful Fermi is, ATI's is gonna have a response by either releasing a cheaper 5890, or taking back the performance crown with the 5890. Nvidia is really shooting themselves in the foot here. What are they gonna do when ATI lowers the price the the 5850 and 5870 by a $100 making it a way better bang for your bang then Fermi. And by march ATI is gonna have all of its low and mid range cards out, which are the true money makers for a GPU company.Sigh, i think Nvidia is falling behind[/citation]
yeah , the first point of improvement would be to increase the memory width by a bit in spite of gddr5 ... 256 bits gddr5 is quite good , but look at how the 5770 underperforms compared to 4890 having all the same theoretical specs for everything else except the memory width .
but regarding the article , that image (from a video) is not fermi based work ... i thought everyone knew this by now including you , marcus ... if there's a doubt i will find that link to the forum where this is mentioned by the creator .. i wonder if he knows about all these references to his work being passed as fermi based .
Meh, I can wait. I rather have a stable GPU than a non-stable one. It's going to kick ass. NVIDIA has never let me down, and I am all about performance. For those of you who do not understand the architecture of the chip, it's the most complex chip NVIDIA has ever made... so calm down and be patient.