NVIDIA's GTX 400 Series Gets A Launch Date: March 26th
by Ryan Smith on February 22, 2010 12:00 AM EST- Posted in
- Ryan's Ramblings
On the heels of a rather unusual (and poorly received) announcement this morning that they'd be showing off the GTX 400 series at PAX East this year, NVIDIA has made a second and much more to-the-point announcement today.
And at this point that's all we know. Specifications, performance, pricing, launch quantities, etc remain to be seen. Perhaps more interesting is that this is on a Friday. We can't immediately recall a Friday GPU launch, even for a refresh part. Like everything else, the whether this has any significance remains to be seen.
Facebook
Twitter
RSS
112 Comments
View All Comments
Parhel - Tuesday, February 23, 2010 - link
Where did you get the idea that the GTX470 will be on par with the HD5870? If the GTX480 is only expected to beat the 5870 by 5%, the GTX470 should come in just a bit faster than a HD5850 . . .This doesn't bode well for NVidia considering that ATI could most likely get a HD5890 out the door at 1Ghz, possible with 2GB memory and six display ports as well.
samspqr - Tuesday, February 23, 2010 - link
I did some math, see here:http://www.semiaccurate.com/forums/showthread.php?...">http://www.semiaccurate.com/forums/showthread.php?...
(the magic number is 0.61: that's the power I used to convert raw power into real world performance, and it seems to work quite well)
ATI only needs a 925MHz card to match GTX480; at 1GHz they would have the same advantage over GTX480 as that one will have over 5870
fshaharyar - Tuesday, February 23, 2010 - link
Hey guys you are forgetting the main point here there first spin off came with only 448sp's that means going for the brute part that is 480 they might just launch the 470 early as due to the lower clock scale due to half hot clock being around 600mhz which restrict until the yeilds are better.so it is an option that 470 might be around 5% to 8% faster than 5850 but ATI has the scale tilted with price cuts and at that time their yeilds will be very good with overclocks going near 7-800 mhz.
which will narrow the gap between 470.
Now the ball rolls into Nvidia's die, sorry Side.
blyndy - Wednesday, February 24, 2010 - link
I read the semiaccurate links and the main point isn't that Fermi is having teething problems or how Nvidia should position it against ATI until yields improve, those problems are moot.The main points, according to semiaccurate are:
-Fermi is a complete balls-up in terms of manufacturing
-It is super hot, and power hungry and with no headroom for either (result: Fermi certified cases)
-yields are <10% and <10,000 chips will be made
-Nvidia will sell each chip for $400+ US, but at a loss! (the 5870 chip is sold for $80)
-Fermi was far-from-finished general-purpose compute research project that was dusted off after their first GPU design hit some problem, and rushed into production
-It is a general-purpose compute architecture with GPU functions tacked on
-Just to FIX the current design would take 6+ months
-Even if yields were equivalent to the 5870 it would still be economically uncompetitive against the 5870
-Fermi 2 is a fantasy at this point, definitely not on TSMCs 40nm, and TSMCs next process, 28nm, is also a fantasy
-Nvidia management got way too complacent on R&D spending, engineer numbers and process research
-Fermi is the 'Laughabee'
tterremmotto - Wednesday, February 24, 2010 - link
People should stop giving undue validity to the technical opinions of people, with little professional experience or proper formal education on a field as complex as computer architecture and fabrication technologies.As a person working in the belly of the beast, it is laughable the level of disconnect between the realities of the field, and the opinions expressed by some of these so-called "experts."
Anandtech and ars being the few glaring exceptions among the litany of mainstream gadget-oriented sites which filled with writers which have absolutely no effing idea of what they are talking about. I also blame tech companies for cater to these people and allow marketing to run rampant...
Targon - Friday, February 26, 2010 - link
So, how do you explain the delays in getting this GPU out the door and no engineering samples making their way into the hands of review sites? This alone implies that Fermi is in trouble, and may easily miss an April, May, or even June release date.It normally takes three or more months from when review sites get their samples to when it is actually released. With NDA, the PUBLIC may not get the information, but review sites have the inside scoop months in advance. If Anandtech and the other sites just don't have them yet, that isn't a good sign for Fermi.
tterremmotto - Monday, March 8, 2010 - link
Because there is a big difference between being 1 Quarter late to a psychological deadline. And the non-sense that the chip is broken and unfixable that chap over in the UK, who knows jack squat about the realities of the industry, is quacking about.Yes, AMD beat NVIDIA to market in the DX11 parts. But the only ones making up the deadline for Fermi are NVIDIA themselves, not web bloggers who decide when a product should have been released.
Targon - Friday, February 26, 2010 - link
So, how do you explain the delays in getting this GPU out the door and no engineering samples making their way into the hands of review sites? This alone implies that Fermi is in trouble, and may easily miss an April, May, or even June release date.It normally takes three or more months from when review sites get their samples to when it is actually released. With NDA, the PUBLIC may not get the information, but review sites have the inside scoop months in advance. If Anandtech and the other sites just don't have them yet, that isn't a good sign for Fermi.
Galid - Thursday, February 25, 2010 - link
You know what, it's logical. Think about it. 40nm has problem about yeilds and still has. The bigger your chip is the bigger the problem of yeilds occur in your batch. ATI had problems with that and it's still not as good as they'd like.Imagine now ATI has problems with a chip ALOT smaller than GF100 is supposed to be. It's more than double in size, the problem with yeilds with double size doesn't means half less liveable chips, it's more like a quarter. And those still can't be called no defects fully enabled chips cauz of heat and wattage problems(40nm problems).
Plus the architecture, they got big groups of simds, they can't disable one on 512 cores, they come in BIG blocks of 32. ATI did the right thing by making lots of smaller groups, why they got 1600 cores instead of 512(supposed for fermi).
boe - Tuesday, February 23, 2010 - link
It seems odd they don't have any out for advance review and benchmarking if they are coming soon. My guess is either the cards aren't as good as nVidia claimed they would be, they don't have drivers that work reliably or they have such a manufacturing issue they can't produce many and they won't have more than a few hundred available for the release.I'm not an ATI fanboy - I still can't ever find the silenced 5870's in stock since I started searching in November. So ATI clearly can't get enough of these out the door either.