NVIDIA GF100 & Friends speculation

That would have totally depended on what Nvidia would have priced Fermi at. Considering the state of 40 nm at the time was extremely bad yields for Fermi would have been even worse than now when producing the same A3 spin. I don't think they could have priced a 480 low enough to force 5870 to readjust price at that time.

Now fast forward to today and I could certainly see 5870 being lower in cost as both cards would have benefitted simultaneously with fixed 40 nm production thus allowing Nvidia to attempt to be more price competitive.

Regards,
SB

 

yes, the fan may be smaller, but the power is significantly higher. its a ~2A fan. The cost of a fan isn't in the plastic blades, fyi.

Also, the fermi heatsink is a far more advanced heatsink than the others shown. First its an exposed heatpipe contact cooler, second it has 4 heatpipes distributing thermals to the fins vs the others which have 3.

 

Well when you find those cheap 50gig+ discs let me know...

Problem with BR is that it is expensive and it has no real drivers. If it had come out 5 years earlier and didn't have HD-DVD to compete with it might have been able to establish itself.

 

HD-DVD is dead and buried. remember how many years it took DVD to replace VHS? the same thing will happen with BR. it will take a few more years for BR drives to become common in PCs. DVD will not last forever.

 

Back on topic now, please? &more pretty pictures!

 

Has anyone ran the DX11 benchmark between 5870 and GTX470 on Metro 2033 ( from STEAM ) ?

 

Hey is it to much to ask for a 5850 at $200 bucks. Unemployment only goes so far !!!!

 

Its a TWIMTP title. I have little doubt that it will be faster on nvidia. That is why I don't buy those titles until they are cheap.

 

After reading Anand's Evergreen development story, I doubt it was a matter of "speccing up" in 2008, but rather "speccing down". Execution tells you if it was the right move or not

They could probably sense even somewhere in mid 09' that NV will not have it easy if ever to go into production within the year. But at that spot it's way too late to change anything fundamental.

That and the fact that AMD "serves" the high end/enthusiast segments since RV670 with dual chip SKUs. Flipping back to Anand's article they changed plans to go for a high end single core chip somewhere in early 2008.

Yields at TSMC are IMHO still nowhere near "perfect" and supply constraints are a headache too. When TSMC kicks in Fab14 aside of Fab12 in ~Q3 then they estimate themselves a potential of 160000 wafers/quarter compared to the current 80000.

With that in mind I'm not willing to "forecast" anything until I see GF100 launching and how availability exactly looks like.

 

Erhm?
Cysis Warhead has reduced I.Q. over Crysis, what are you talking about?

 

One of the GTX 480 OC Editions consumes 305W
http://translate.google.com.tr/tran...etimi_305_Watt_olabilir-18648.htm&sl=tr&tl=en

 

Same here. I yanked my drives over a year ago and now I share an external drive between my desktop and laptop. So far I've only used it for Win7 and the free copy of FC2 I got with my card. Steam all the way. But I guess we are limited to the capacity of retail media and it's too bad.

OTH, is that a factory overclocked 480 pulling > 300w? How did they swing that?

 

Great! Can't wait to see the performance of such a beast!

 

I don't believe OEMs buy factory overclocked cards, other than the likes of Alienware/Voodoo/Falcon NW/etc. Unless you're building a high-volume PC for the masses, who cares about PCI-e certification?

 

Breaking the 300 watt barrier shouldn't matter for those cards' intended audience, but it does help play into the zomg Fermi is teh powerhungry beast bad juju.

 

Ehm.. if a slightly OC'ed part does indeed break the 300W barrier.. doesn't that mean that it is a power-hungry device?

As far as I know, real OC'ed cards are not showing up for a good while.

 

well, of course, but it's not really a line that I would think nvidia would be happy with pushing...

Just think that it's a bit of a bad idea to mention these things yet... unless they really need something to point forward to, ie. in order to say "sure, we're not beating the 5870 by much, but just wait till next month!1"

 

If with 50% more transistors and 50% more power they can't beat the competition by a tangible margin then they aren't going to get there with little tweaks. If the 470 is as close to the 5870 as those early numbers indicate then I see no reason for them to be pushing the limits of the PCIe spec on the 480 just for a few % points. Why not just overclock the 470 instead?

 

The 470 may just be close to the 5870 because of frame buffer limitations.

Of course if those benchmarks of the 10.3s vs 9.12 are correct the 5870 might get a whole lot faster just in time for the 470/480 launch