NVIDIA GeForce GTX 400 series official today

Robert Hallock 12 Apr, 2010 at 12:30pm ET Article published in Tech

The day is here, NVIDIA fans! NVIDIA’s newest GPU architecture is officially on the prowl in retail and lookin’ mean.

Hitting the streets in two flavors, the GeForce GTX 480 and the GeForce GTX 470, NVIDIA’s new GPUs are based on an all-new architecture known as Fermi. Designed from the ground up with a mind towards GPU computing, the green team also tacked on enough gaming hardware to give their newest kit a 5-15% lead over competing chips from AMD.

“The top-of-the line in a new family of enthusiast-class GPUs, the GeForce GTX 480 was designed from the ground up to deliver the industry’s most potent tessellation performance, which is the key component of Microsoft’s DirectX 11 development platform for PC games,” NVIDIA said in a press release. “Tessellation allows game developers to take advantage of the GeForce GTX 480 GPU’s ability to increase the geometric complexity of models and characters to deliver far more realistic and visually compelling gaming environments.”

“The GeForce GTX 480 is joined by the GeForce GTX 470 as the first products in NVIDIA’s Fermi line of consumer products…. The remainder of the GeForce 400-series lineup will be announced in the coming months, filling out additional performance and price segments.”

Unique features for the new duo include NVIDIA 3D Vision, which enables stereoscopic 3D support on systems with 120Hz LCDs; 3D Vision Surround, which brings 3D stereoscopy to multiple displays; PhysX hardware physics support and CUDA, for outstanding programmability and a wide array of excellent developer tools and support channels.

The GTX 480 and GTX 470 have a street price of $499 and $349, respectively, but stocks are limited and the cards are sold out across the web.

	GeForce GTX 480	GeForce GTX 470
Codename	GF100	GF100
Process Node	40nm	40nm
Die Size	529mm²	529mm²
Transistors	3.2 billion	3.2 billion
Core Clock/Shader Clock	700/1401MHz	607/1215MHz
Memory	1536MB GDDR5	1280MB GDDR5
Memory Clock	924MHz (3696MT/s)	837MHz (3348MT/s)
Bus Width	384-bit	320-bit
Memory Bandwidth	177.4GB/s	133.9GB/s
Shaders	480	448
Texture Mapping Units	60	56
Render Output Units	48	40
Thermal Design Power (TDP)	250W	215W
Power Consumption	480W (Approx.)	400W (Approx.)
PCIe Pins	6-pin & 6+2-pin	2×6-pin

Comments

12 Apr 2010 ~ 12:40pm TurboPenguin Mine is coming tomorrow w00t!
12 Apr 2010 ~ 12:53pm Sledgehammer70 Not entirely sure today is the direct launch day. Nvidia said week of the 12th... either or Newegg and other retailers have been selling cards since last week. They seem to sell out within 10 minutes of posting them online.

I guess people want their space heaters.
12 Apr 2010 ~ 1:10pm mirage The performance is much higher where it matters. I have seen and handled many hot cards before. And, this is the card I want.
12 Apr 2010 ~ 1:21pm TurboPenguin Tomorrow, I'll post some number from my home machine.
24 Apr 2010 ~ 6:09pm zingwoosh3 ati fanboys have been crying all month. They got their panties in a twist because their cards are slower. So like little 14 year olds they run around and spew exaggerated lies about it being warmer and using more electricity. Making it sound so much worse than it is. And the electric comes out to be about $8 more a year assuming they have no lives and play 4 hours a day, EVERY day all year. So unless you have a tiny case with almost no fans, it wouldn't be a problem. And then again, if you have a lousy case like that it would be a problem for the upper ati anyway. And why do ati fanboys get SO worried over a card that isn't theirs? Because they want to make themselves feel better and believe they have the best card. It's called being selfish and stupid. Also the 470 is probably the best deal in many ways. Obviously the 480 works thermally, so the 470 would be even cooler, cost less etc. And it has roughly the same performance of a 5870, but costs less money. So it's a perfectly fine card option for anyone WITHOUT a card who needs one. Of course ati fans with an ati card in their case will never admit that another option is suitable and they should just be ignored. I spent days reading specs everywhere to find out that 95% of them are dishonest as they are highly biased since they own their cards already. I'm looking at this from the point of view of having nothing and picking one. My point being there isn't a thing wrong with the nvidia. I'm getting a 470 when more become available. Plus I like knowing that the tessellation hardware will improve a few games here and there.
24 Apr 2010 ~ 6:17pm zingwoosh3 I meant that I spent days reading forums everywhere (along with specs), to find out that 95% of ati fans are dishonest when it comes to giving nvidia any credit for anything. Their egos just won't let it happen.
24 Apr 2010 ~ 7:46pm Cliff_Forster

zingwoosh3 wrote:

ati fanboys have been crying all month. They got their panties in a twist because their cards are slower. So like little 14 year olds they run around and spew exaggerated lies about it being warmer and using more electricity. Making it sound so much worse than it is. And the electric comes out to be about $8 more a year assuming they have no lives and play 4 hours a day, EVERY day all year. So unless you have a tiny case with almost no fans, it wouldn't be a problem. And then again, if you have a lousy case like that it would be a problem for the upper ati anyway. And why do ati fanboys get SO worried over a card that isn't theirs? Because they want to make themselves feel better and believe they have the best card. It's called being selfish and stupid. Also the 470 is probably the best deal in many ways. Obviously the 480 works thermally, so the 470 would be even cooler, cost less etc. And it has roughly the same performance of a 5870, but costs less money. So it's a perfectly fine card option for anyone WITHOUT a card who needs one. Of course ati fans with an ati card in their case will never admit that another option is suitable and they should just be ignored. I spent days reading specs everywhere to find out that 95% of them are dishonest as they are highly biased since they own their cards already. I'm looking at this from the point of view of having nothing and picking one. My point being there isn't a thing wrong with the nvidia. I'm getting a 470 when more become available. Plus I like knowing that the tessellation hardware will improve a few games here and there.

Zing, full disclosure... I'm probably the ultimate AMD fanboy...

That said, to be fair, did you see the marketing leading up to Fermi? Seriously, this card was supposed to crush the 5870, but it beats it by 10%, maybe a little more on an optimized title. Beyond that its supposed to cure cancer. I'm not making this up, it was in the marketing while they were killing time re spinning the chips.

Listen, here are the facts as I see it....

Nvidia is late to the DX11 party because Fermi was much harder to manufacture than they would have liked it to be.

Nvidia used a FUD marketing campaign to make users feel like if they did not wait and see how theirs was going to shake out, that they would be missing the ultimate gaming experience, when the fact is, its a marginally faster card, and it should at least be that, it took they were seven months behind AMD. Point being, Nvidia did not deliver what they promised you.

AMD's cards do run cooler and are more energy efficient. If that does not matter to you, okay, fine, but its a fact and a huge part of chip design is in that efficiency. You can dispute its importance if you like, but it does not change the facts. If an AMD "fanboy" wants to say, hey our cards and more efficient and run cooler... well, its not like we are making it up.

Here is how I see it. Nvidia promised to change the world with Fermi, and all we got was a card that is a little faster and runs really hot. You can pretend that your not a tiny bit disappointed, but seriously, who would be the fanboy then?
25 Apr 2010 ~ 6:40am Snarkasm Aaaaand Cliff just can't stay away.
25 Apr 2010 ~ 12:29pm clifford_cooley zingwoosh3 did say ati fanboys had been crying all month. Did you expect him to lay down and take it or respond?

Personally I'm an ati fanboy myself. However in the event of starting to fold, I am now considering nVidia as an option. I'm not going to sit and wait three years for a GPU3 client with better ATI support.
25 Apr 2010 ~ 1:10pm Cliff_Forster

clifford_cooley wrote:

zingwoosh3 did say ati fanboys had been crying all month. Did you expect him to lay down and take it or respond?

Personally I'm an ati fanboy myself. However in the event of starting to fold, I am now considering nVidia as an option. I'm not going to sit and wait three years for a GPU3 client with better ATI support.

Hey, and my response is reasonable. It states the facts as I think most folks understand them.

Clifford, about improved GPU3 clients, and I know this is going to be unpopular, but I am going to throw it out there...

One of these days I am going to write an extremely unpopular editorial about why folding at home is no longer about helping people, its become about E-Peen and clever marketing.

We could do more to help Alzheimer patients by making a direct monitory contribution to a charity factoring the money we would save from video cards and power spent. The folding at home project is a grossly inefficient way to help anyone.

I'm not saying the GPU should not be used for research and helping people, I'm just saying, this kind of distributed computing project is grossly inefficient, and people should be aware of that. Its doing more damage by overdrawing its fair share of resources both in the form of power, and bandwidth then its helping anyone. I know why it works, ultimately, they have made charity a game. Hey, we get points! But I don't know, there is just something that does not sit well with me about that. I understand the human part of why it works, its a really clever piece of social engineering, but at the end of the day, wouldn't it just be more effective for everyone that is involved in the folding at home project to turn off their machines and send Stanford a nice donation to build a far more efficient localized super computer? I realize this is probably unrealistic, but that does not mean the question does not have some merit.

Its like robbing Peter to pay Paul. Lets burn more carbon and congest the internet in hopes that one day something will come of it. Now, like I said, I get the social engineering side of it, and why it gets people excited to participate. Thats not a bad thing, but I just can't resist being the nut job in the room that says, hey, would turning your computer off when your not using it and using the money saved as direct monetary contributions to important charities a better way to help people?

Nvidia marketing folding at home performance as a differentiated feature for computer enthusiasts, I get it, I don't blame them, I just wonder if anyone stops to look at the larger picture. Are they doing it to help people, or are they doing it to oversell us on a graphics product?
25 Apr 2010 ~ 1:56pm clifford_cooley For the most part I agree with you.

My question is not why nVidia has a better GPU client. My question is why ATI does not have a comparable client? I have decided to contribute to the folding at home cause whether it helps or not. I may not have a super computer but I will fold what I can. And if nVidia is better for it then that is what I will use. I do wonder if the guys at AMD see things the same way and choose not to support Folding at Home.

I do realize its not AMD's place but they could step in for a good cause even if its not as efficient as a localized super computer would be. After all doing it this way they don't have to worry about the funds to support the servers as they run year after year.
25 Apr 2010 ~ 1:59pm clifford_cooley When are the cut down versions of the GeForce GTX 400 series expected (As in the 450 or 460)?
25 Apr 2010 ~ 3:06pm Snarkasm NV's client is better because the GPU client was built on CUDA. ATI didn't have a comparable framework, and they had to port it.

Now, Cliff, claiming F@H over-uses bandwidth is laughable - the work units are miniscule, they take hardly any bandwidth at all. And while it may have overgrown its original intention, F@H was a way to utilize idle cycles on machines that would otherwise be on all the time anyway. I despise shutting down my computer. I may as well do something with it.

I daresay Stanford gets more use out of spare cycles than they could out of $20 a year, but maybe that's my fuzzy math.

I like how something favors NV and Cliff cuts it down. First tessellation, then Folding... what's next?
25 Apr 2010 ~ 6:35pm Cliff_Forster

Snarkasm wrote:

I like how something favors NV and Cliff cuts it down. First tessellation, then Folding... what's next?

Have I gotten to PhysX today?