Maximum PC

Some details from Anandtech:

http://www.anandtech.com/video/showdoc.aspx?i=3651&p=1

Hopefully Nvidia will price competitively with ATI this time around. It looks to be a beast of a gpu. 512 shader processors looks to be very appealing for folding purposes, should be quite a bit faster then those 295s your running staypuff !

Anandtech's server is being slow... I'll have to try later. But if it really does have 512 shader units... ATi probably should have decked the 5870 out with at least 2000 if they had wanted to be competitive. What I can imagine about these cards is- unnecessarily fast ( ), and ridiculously hot.

hopefully it has some more bling involved... it has only 32 more shaders than a 295... however it is supposed to be a single core so that'd be amazing.....

imagine the gtx395 with 1028 shaders???

that one card would be good for over 35k ppd just itself.... time to build another uber rig if that happens....

Anandtech's still slow.
East coast AM office personnel, back to work!
Here it looks as though there's a few new 300 series parts
supported in the new WHQL .inf file for unreleased nVidia chips.
Hopefully they're not rebranding... again.
Should be an interesting launch.

Got it to load finally. Very interesting ideas, but 90% of them seem to be aimed at GPGPU. Looks like the current pattern of ATi releasing the bang-for-buck cards and Nvidia releasing the big, bad, hot, expensive GPU's will continue. as for folding, the 5800 series finally supports OpenCL. Its only a matter of time before an OpenCL folding client comes out that erases Nvidia's advantage (hopefully, competition is good ).

looks like ill be keeping my GTX285 till this comes out since it looks like it will smack the 5870

yeah it'll smack the 5870. For $600 it'll smack the 5870. Two 5850's in CrossFire are probably going to be THE choice for high-res gamers if things pan out like they did this gen. since it's now got GDDR5, it loses the bandwidth problem that caused many to opt for a 285 instead of a 4850 X2 or two 4850 1GBs, since they're faster in most situations, but in bandwidth-limited applications, they would crap out.

It has gddr5 but on a 384bit level, will that bottleneck it compared to a 512bit interface?

GDDR5 on 384-bit is wider then GDDR3 on a 512-bit interface.

Really? Does GDDR5 essentially double each bit or something? a 768-bit memory bus would absolutely dominate.

yeah. GDDR5 transfers four times per clock cycle, while GDDR3 only two. GDDR5=twice as fast. That's why the 4870 has identical memory bandwidth to the GTX 280. also, narrower buses with faster RAM are cheaper then fatter buses with slower RAM.

LMAO!

http://www.semiaccurate.com/2009/10/01/ ... oards-gtc/
it looks like Fermi might not be as far along as Nvidia says it is. While they supply pictures of an "engineering sample" Fermi, it's clearly just a mock-up with no functionality. Now, while this may be semi-accurate.com, and Charlie is an idiot, you can't argue with many of the points:
The screws holding the slot vents on would stop the card from fitting. That's exactly where a division between the slots would be.
The Fermi backplate's position stops one of the SLi connectors from being used.
The Power connectors don't line up with the sauter on the back of the board.

$600 for a gtx285???? where the fuck are you buying vid cards..... maybe for two of them lol.....

? The GTX 280 started at $550. I'm not talking about the present.

Any news on this topic, I can't wait to see these next gen cards, they are going to be beasts.

yep. Nvidia admitted that the "engineering sample" Fermi was a mock-up after all, and that they didn't actually have a working Fermi GPU at the demo, and were lying about running the demo on Fermi. If it ever comes out, it'll be fast... but it's existence facet isn't looking too hot right now.

Actually, a few people who were at the Nvidia day from B3D have said while the card was a mock and Nvidia admitted to that. They did have demos running on the ES sample of Fermi with a GTX285 system being displayed next to it for comparison sake. They estimayed about a 5-10x performance jump in fermi, aka GT300, over GT200 based on the speed/smoothness of the demo. Which some are speculating to mean about 1.5 to 2x, maybe more real world performance increase over GT200.

Im speculating the 5800 series is having some manufacturing problems and ATi is actually losing a ton of money because of faulty units.

Mind explaining why you're speculating that?