GF100 (Fermi) previews and discussion

[JAYDEEJOHN]

The breakdown was 2 to 1, or 2 5series/other 1 cypress

 

[poul]

I know that this article is not about Fermi, but it is not out of context. I get the impression from reading that article, that Evergreen was not manufacturable on TSMC's 40nm std. process. But they did not wait around for TSMC to fix it for them, they worked on solutions them self. Fermi is 3.2 billion transistors and 400mm2, significantly denser than Evergreen (Cypress) 2.154 billion transistors and 334mm2. It might have somthing to do with why Evergreen is yielding a bit better than Fermi.

 

[JAYDEEJOHN]

Whatll be bad also is if it doesnt show before metro comes out

 

[Mousemonkey]

I know that this article is not about Fermi, but it is not out of context. I get the impression from reading that article, that Evergreen was not manufacturable on TSMC's 40nm std. process. But they did not wait around for TSMC to fix it for them, they worked on solutions them self. Fermi is 3.2 billion transistors and 400mm2, significantly denser than Evergreen (Cypress) 2.154 billion transistors and 334mm2. It might have somthing to do with why Evergreen is yielding a bit better than Fermi.

I did post that the other day but it seems to have gone unnoticed.
http://www.tomshardware.com/forum/forum2.php?config=tomshardwareus.inc&cat=33&post=281372&page=1&p=1&sondage=0&owntopic=1&trash=&trash_post=&print=0&numreponse=0&quote_only=0&new=0&nojs=0

 

[razercultmember1]

im really hoping the gtx285 or the gtx295 will drop after fermi comes out down to 250 would be very very nice

 

[Mousemonkey]

The breakdown was 2 to 1, or 2 5series/other 1 cypress

Where did you get those numbers? And how many desktop parts does that equate to then?

 

[Amiga500]

Eh, questioning nVidia's engineers seems like folly to me. nVidia is run by some egotistical people, absolutely, but saying "they didn't do their homework" is retarded. I doubt that would be an issue with their experienced engineers.

The engineers can only recommend.


Trust me... we are ignored every f**king day by bean counters, salesmen and lawyers who "know better".


It wouldn't surprise me if Nvidia were not having similar problems.... where basically the tail is wagging the dog.

 

[Mousemonkey]

The engineers can only recommend. After which they are told to do the impossible and be quick about it!


Trust me... we are ignored every f**king day by bean counters, salesmen and lawyers who "know better".


It wouldn't surprise me if Nvidia were not having similar problems.... where basically the tail is wagging the dog.

Fixored! :lol:

 

[JAYDEEJOHN]

Thats from Wavey Davey hisself.

As to the Anand article, I too posted on it, and only a few replied back.
One thing tho, Fermi is closer to 550mm not 400, and its density has caught up not surpassed ATI, which is good, or may be the cause

 

[Griffolion]

Hi everyone, just wanted to get a view of everyones opinions of the dire straits NV is currently in regarding Fermi yields/clocks/TDP etc. From what i've read it aint looking good (unles semiaccurate.com is a ATI fanboy/satire website) for the green camp in terms of getting a DX 11 card.

Oh well, discuss.

 

[EXT64]

Well, I think there is already one of these threads, but anyway...

To me, I don't think it is quite as dire as that made it out to be (but who knows, I have no proof). However, a lot of what he was saying was fact (not the NVidia bashing, but all the TSMC troubles) and really what I took away from that (and the Anandtech article on Evergreen) was that we are lucky to have any new cards at all right now! Hopefully NVidia will get it straightened out. We need them to get DX11 rolling.

 

[Griffolion]

Well, I think there is already one of these threads, but anyway...quotemsg]


Sorry if i repeated a thread, i did take a look for one.

As a NVidia liker and loyal buyer i just would hate to see them being beaten so badly by ATI for much longer than this. Also considering in the long term than if Fermi does turn out to be this massively flawed architecture, ATI will have more or less 1 generation's worth of research on NV and so will easily be able to top revised Fermi architecures with revised Cypress chips etc.

To be honest, if NVidia was to completely take their ego and concern for losing face out of the equation, with the mess they made they might as well just sit in R&D for the next year and perfect an architecture on 28nm, tapeout December til June 2011 and come back with a particular focus for the mainstream.

 

[EXT64]

I'm not overly concerned, in the long run at any rate. Sometimes it takes a complete failure to bring out the best. And I just realized (sorry) you did not repeat a thread, this topic was brought up in the Fermi sticky at the top of this forum (not in its own thread, which it probably deserved).

 

[microterf]

I am an ATI fan, but I really would love to see the fermi response here real soon. When both companies are rolling, we (the consumers) win with lower prices and fast cards

 

[shadow187]

So..if ATI's refresh comes shortly after nVid's FERMI release, will the cycle of ATI's next-gen cards beating nVidia to the punch continue?

 

[edyn]

I wonder if I'll be able to get a Fermi card and just leave in my 9800GTX in my computer as a PhysX card, or if that is even worth the hassle seeing how there are so few PhysX games.

I have an Antec 900.

9800 eats way too much power and is some seriously overkill solution for a physX dedicated card.

I currently use some old 9500 as physX slave and its more than enough for all games I have played, and alooooot less power-hungry than a 9800 .

 

[EXT64]

Not sure. NVidia's next gen (after fermi) may not be delayed at all and thus could come out before or after ATI's.

 

[EXT64]

I can't remember for sure, but I thought the optimal spot was about 9600GT-9800GT for dedicated PhysX cards.

 

[sabot00]

Batman: AA, Dark Void, and probably others want a 9800GTX/GTX+/GTS 250 for a dedicated PhysX card.

 

[reprotected]

I think Fermi will be one of the best advancements (mentioned by NVIDIA). I hope the jump will be about or better than the jump from GeForce 7 to 8. Anyone remember that time? 7950GX2 (Dual Cards) < 8800GTS (Single Card), 7900GT (High End) = 8600 GT (Medium-Low End), etc.

I'm just wishing that the GTX 480 won't be two graphics cards in one, and it will beat the HD5970 with flying colours.

 

[AMW1011]

I don't think that we can hope for nearly that much with Fermi, reprotected, but I suppose anything is possible.

My take? The GTX 480 will be 10% slower than the 5970. The GTX 470 will be in between it and the 5870 while the GTX 460 is in between the 5870 and the 5850. All decent market positions if priced correctly.

 

[edyn]

Batman: AA, Dark Void, and probably others want a 9800GTX/GTX+/GTS 250 for a dedicated PhysX card.

Possible, have not tried these games yet. Having already used a friend's 9800 i would still not recommend it because its really wasting so much power.
The 9500 and 9600gso have both being doing fine under crysis and modern warfare2 for physx tho. There also alot cheaper.

This being said, having an endless budget i would never say no to a 250 as physX slave but thats not the case 🙁...

Having to deal with a limited budget and having to watch over my power bill, I can't stress enough that these 2 cards really are doing "good" for a much lower price and power consumption.

 

[Griffolion]

I don't think that we can hope for nearly that much with Fermi, reprotected, but I suppose anything is possible.

My take? The GTX 480 will be 10% slower than the 5970. The GTX 470 will be in between it and the 5870 while the GTX 460 is in between the 5870 and the 5850. All decent market positions if priced correctly.

Actually its predicted that even if the bad manufacturing process dictates they have to fuse off some shader units to stabilise voltage issues (about 12.5% performance decrease predicted) the GTX 480 will still be roughly 10% faster than the 5870. But in response to reprotected, its nice to be optimistic but things aren't have haven't looked that rosey in a while so its best to make some serious considerations as to what your next card is gonna be.


Oh and YAY i started an interesting thread lol .

 

[MagicPants]

I expect Fermi will be faster but not by that much. The 480 will be somewhere between a 5870 and a 5970. What nvidia really needs to capitalize on is how flexible the Fermi architecture is. PhysX is a good step but it's not enough.

I'm hoping nvidia will be able to do something with the quality of rendering that the ATI 5000 series can't.

Speaking of which I just got 3d vision today, when it works well it's breathtaking but only a few games do it right. It really makes you hate any kind of HUD whatsoever.

 

[JAYDEEJOHN]

Remember tho, when using the lessor card, often the clocks are run up a bit to work out the lessor amounts of shaders etc.
But, with the rumors of Fermi having high power usage, this will be harder to achieve, sort of a rock n hard place scenario