nVidia's Fermi to be called GTX 470 and 480

[thilanliyan]

What I don't understand is...if the 3 series are just going to be rebadges and one number generation down from Fermi, why not just keep current cards the way they are and make Fermi into the 3 series?

 

[Qbah]

What I don't understand is...if the 3 series are just going to be rebadges and one number generation down from Fermi, why not just keep current cards the way they are and make Fermi into the 3 series?

Maybe they managed to redesign the chips for 40nm? And a 256bit+GDDR5 memory setup for lower cost and complexity?

 

[zerocool84]

What I don't understand is...if the 3 series are just going to be rebadges and one number generation down from Fermi, why not just keep current cards the way they are and make Fermi into the 3 series?

Simple marketing

 

[v8envy]

The rumor is fact. Look at http://twitter.com/NVIDIA, second tweet down.

We already have the 310 as a GT210 rebadge for the 3 series. They are selling now. The only remaining question is whether there will be other members of the 3 series and whether or not they'll be 40nm with DX10.1 or just a straight re-sticker job like the GT210->GT310 and GT220->GT315.

 

[SHAQ]

It said end of February? Did I read that correctly? Surely not.

 

[aigomorla]

Gah...

I dont think i can wait around much longer for Nvidia to come around..

And i think there new gpu will FLOP.
Otherwise we would see stuff leaked about it.

You cant hide all the teasers, unless there is nothing to tease with.

 

[RussianSensation]

It said end of February? Did I read that correctly? Surely not.

I think it's end of March 2010:
"The leading supplier of graphics processors has reportedly notified its partners that it would officially launch the GeForce “GF100” graphics chip in March ’10" - http://www.xbitlabs.com/news/video/...raphics_Cards_Finally_Get_Official_Names.html

 

[Lonyo]

They want to catch up with AMD but it would be too obvious if they did it all in one go, so they are skipping to 4xx now, and then next time round the refresh or whatever later this year will be the 5xx cards, while AMD do an HD5xxx refresh (presumably) and then they will hit 6xx a bit after AMD have released their HD6xxx cards. No more number disadvantage.

 

[RussianSensation]

Lonyo, according to Wreckage, the cards are so fast that a single GTX380 will be at least as fast as a 5970. Therefore, this warrants 2 generations number increase

 

[Hyperlite]

lonyo, according to wreckage, the cards are so fast that a single gtx380 will be at least as fast as a 5970. Therefore, this warrants 2 generations number increase

haha

 

[tviceman]

Lonyo, according to Wreckage, the cards are so fast that a single GTX380 will be at least as fast as a 5970. Therefore, this warrants 2 generations number increase

Hahaha, all ridicule aside we still don't know anything yet, but I wouldn't be surprised if there are a few games (namely Stalker Call of Pripyat and/or Alien vs. Predator) where the high end Fermi part competes well with 5970. M guess is performance will be a mixed bag and will be 10-45% faster than a 5870 depending on the game.

We shall see.

 

[ronnn]

I think it's end of March 2010:
"The leading supplier of graphics processors has reportedly notified its partners that it would officially launch the GeForce “GF100” graphics chip in March ’10" - http://www.xbitlabs.com/news/video/...raphics_Cards_Finally_Get_Official_Names.html

Nice link!

"Nvidia plans to release a rather mysterious code-named GF104 chip in Q2 2010. The GF104 graphics processing unit will target the high-end market, according to the media report. "

Wonder what the GF104 is and why high end, unless fermi is mid range.

 

[tviceman]

Nice link!

"Nvidia plans to release a rather mysterious code-named GF104 chip in Q2 2010. The GF104 graphics processing unit will target the high-end market, according to the media report. "

Wonder what the GF104 is and why high end, unless fermi is mid range.

I think they have their information slightly off; I believe the GF104 chip will be the mid-range DX11 parts.

 

[yacoub]

Guys, you're no longer cool unless you have the S series 1U rack with four GPUs inside it.

 

[RussianSensation]

Hahaha, all ridicule aside we still don't know anything yet, but I wouldn't be surprised if there are a few games (namely Stalker Call of Pripyat and/or Alien vs. Predator) where the high end Fermi part competes well with 5970. M guess is performance will be a mixed bag and will be 10-45% faster than a 5870 depending on the game.

We shall see.

I am not sure about STALKER: Call of Pripyat.
http://www.xbitlabs.com/articles/video/display/radeon-hd5970_8.html

2560x1600 16AF
GTX285 = 14.1
GTX295 = 14.1
5970 = 46.9 (3.3x faster than GTX285).

If it beats 5970 in this game, Fermi is the product of the year hands down!

But in a game like Wolfenstein or Far Cry 2, which perform better on NV hardware, I can definately see Fermi tearing it up.

 

[tviceman]

Guys, you're no longer cool unless you have the S series 1U rack with four GPUs inside it.

That's all you got? I prefer the S series HD-1U++ rack with triple SLI support in a quad crossfire setup.

 

[Ben90]

I think nvidia is in deep trouble.

They first said the Fermi has 512 shaders. Soo well stick to that until otherwise notified.

The ATI 5890 has 1500 shaders. That is 3 times more shader. What does nvidia have to say about this. I think nvidia blew it!

Leave a comment on newegg please

 

[Dribble]

I think they have their information slightly off; I believe the GF104 chip will be the mid-range DX11 parts.

GF104 chip is allegedly the X2 part (i.e fermi equivalent of GTX 295).

 

[Borealis7]

Leave a comment on newegg please

Double Burn!

for me, this changes nothing. the name has no meaning.

perhaps that rumor that nVidia was working on a "second generation fermi" was founded on the basis that the first Fermi was expected to be called 3XX, so 4XX would be the 2nd generation?

 

[plopke]

^_^ Most of you guys are all drooling about this news. But I assume that nvdia will not be releasing mid-ranged at the launch? Whenever that launch might be, with mid-range i mean at the price point of 5750-5770 and who knows what price the 5830 will be.

Anyway I have to build a new pc very soonish , looking around I assume these card will not be here until march or april. And i guess the midrange will even be later?

Since I can not wait that long , I wonder if i should go for a cheap card(5750 or lower) now , but on the other hand nvdia hasn't proofen one thing yet. If I read stuff like http://www.nvidia.com/content/PDF/f...DIA_Fermi_Compute_Architecture_Whitepaper.pdf , it just feels like bit of deja vu (PS3-cell , intel larabree,...) , promissing everything to please everybody but at what cost.(price,heat,power,difficulty to work with,...).

Time will tell I guess but the longest I can wait for a new card is the 5830 (if that is 5feb)

 

[PingviN]

Fermi is named 400 so that Nvidia can give G92 another go as the "new" 300-series.

It's obvious.

 

[tviceman]

Fermi is named 400 so that Nvidia can give G92 another go as the "new" 300-series.

It's obvious.

You might not be far off. If nvidia can shrink the gt200 down to 40nm, add in dx11 support, increase clock speeds, reduce the bus down to 256 bits and use ddr5 vram, it would probably be a good part for release.

 

[Munky]

You might not be far off. If nvidia can shrink the gt200 down to 40nm, add in dx11 support, increase clock speeds, reduce the bus down to 256 bits and use ddr5 vram, it would probably be a good part for release.

I don't think adding DX11 is as trivial as you make it sound. If they do shrink and rebadge the gt200, I can at best see new DX10.1 capability and maybe a few tweaks, not to mention a new memory controller on a narrower bus.

And how is that "good?" In that thought, if AMD could shrink the rv790 to 40mn and give it a healthy clock boost, it would also give the 5850 a good run for its money, and leave the 5770 in the dust, but I wouldn't consider it a good card compared to all the new features and improvements a 5850 has.

 

[lifeblood]

You might not be far off. If nvidia can shrink the gt200 down to 40nm, add in dx11 support, increase clock speeds, reduce the bus down to 256 bits and use ddr5 vram, it would probably be a good part for release.

It is possible to add in DX10.1, but not DX11. It's to big of a change. Most of nvidia's engineering team should be working on Fermi and it's followon. Spending the amount of time to add DX11 is too expensive and takes away from other things.

A 40nm DX10.1 GT200 with improved clock speeds using gddr5 would be an excellent midrange card until they can get a Fermi derivative into the midrange.

 

[v8envy]

Upon further reflection I am starting to think DX11 may be doable on the GT200 parts. Unlike ATI's dedicated silicon, tesselation on the Fermi is done entirely in shaders. So, it may be feasable to take the same approach with the GT200 shrink.

But without the architectural improvements of Fermi I'd imagine this "software" implementation might perform very poorly if backported to this generation's hardware.

Anyhoo, I wouldn't spend > $100 on any new nv product without proper DX11 support. And I may not be alone there. Which means there's probably room for a G92 relabel of the GT250, but not much higher than that in the 3 series lineup.