Wednesday, May 12th 2010
Mid-Range NVIDIA DX11 GPUs On Track for Summer Sales
NVIDIA's Fermi architecture seems to be finally coming of age, with the company planning to introduce mid-range GPUs derived from the architecture, which are expected to be energy-efficient and competitive with mid-range lineup of rival ATI. Identified so far by some sections of the media as GF104, the new GPU will form base for a number of new SKUs that will capture key price-points in the mainstream segment.
A new report suggests that NVIDIA might well be at the end of developing this GPU, and will be ready to talk about it to the public as early as in June, at the Computex 2010 event in Taiwan. Select groups of people, particularly board partners, OEM vendors, etc., will also be able to see engineering samples in private. In the following month (July), NVIDIA will be in a position to launch these mainstream SKUs, just about in time for Summer, which is one of the two main shopping seasons in major markets.
Source:
KitGuru
A new report suggests that NVIDIA might well be at the end of developing this GPU, and will be ready to talk about it to the public as early as in June, at the Computex 2010 event in Taiwan. Select groups of people, particularly board partners, OEM vendors, etc., will also be able to see engineering samples in private. In the following month (July), NVIDIA will be in a position to launch these mainstream SKUs, just about in time for Summer, which is one of the two main shopping seasons in major markets.
14 Comments on Mid-Range NVIDIA DX11 GPUs On Track for Summer Sales
maybe it was GT212 but if memory serves the rumored specs seem to match this GF104 basically.
It'll be interesting to see what energy efficiency gains they have made here. TDP, Bang per Buck, and Bang per Watt will be the interesting things for me (in no particular order).
I hope with these lower cards being released we'll start to see price drops w/the Ati 5k series.
To respond to wolf, GF104 has never been that thing tho, it has always been a Fermi derivative, although it's rumored specs have changed as time has passed. It started as a GF100 with more shaders, minus the heavy cache capabilities and minus full 64 bit support. These cut down features have remained, but the number of cores has gone down dramatically, first to 512 SP so that it would be what G92 was to G80 and now rumors alledge it would be a 256 SP chip. I wouldn't take any of these rumors too seriously until something can be confirmed tho.
It would seem to me that the performance-per-watt coefficient is defined with Fermi. There isn't much you can do about it; you might make a few % efficiency gains when underclocking and undervolting, but you wont rewrite the fundamentals.
I watch with interest.
*the process is leaky, the chip rather than fix that problem it makes it worse (Cypress fixed it with redundancy on some areas, expertise adquired with RV740 did pay off). They can fix it "easily" on a different chip (they can screw it again just as easily, I give that to you). The thing is that the chip didn't work as it should, but it's the chip. They had it taped out in summer 2009, not much later than Ati had Cypress, the thing is that first silicon didn't work as pretended and they thought they could fix it on revisions (95% of the times they do). They could have made the fix on GF100 if they had made a full respin (B1) from the start instead of trying so many revisions, that took more time in the long run. So they screwed it, but that's just luck too. Making chips is a gamble in many aspects and sometimes you just have bad luck.