NVIDIA GF100 & Friends speculation

[Silus]

Can we at least know if the Fermi based GeForce products, are inline with some of the speculation around here ?

 

[Archaeolept]

it has a compelling 40% lead in HeatMark 2010?

given that speculations in this thread have ranged all over the map, the reality of it must lie somewhere inline with some of that. otherwise, anyone who really knows anything is bound by nda or professional courtesy.

oh, I too am dieing to know some hard facts... but that won't really happen until independent reviewers are given actual cards.

 

[trinibwoy]

their pre-2010 launch.

Oh I thought you were referring to something that actually happened.

their putative 2009 launch date... or you must know that is what he meant, and are making the counter claim that they did not know that they wouldn't be able to deliver. That seems unlikely, that they had no idea at all how far behind they were. Maybe they hoped to squeeze in some cards before the new year, though.

I think there are enough claims about Fermi floating around already Whether Nvidia knew or not isn't particularly interesting though as the result is the same.

 

[Archaeolept]

since I can't sleep I'll post the latest rumours - Fudzilla is claiming 20% chip yields currently, and that the gf100 is 20% faster than the 5870.

Chip yields seem to be the topic of discussion lately as the much larger Nvidia Fermi chip is struggling with yield issues as our sources suggest that the actual yields are as low as 20% and are not improving quickly according to sources close to the fabs.[...]
Those in the known[sic] are claiming that Fermi despite the yield and thermal issues is only about 20% faster than Cypress

http://www.fudzilla.com/content/view/17205/1/

fud is a bit more reputable than a poster on an internet gaming forum, but not by too much

 

[eastmen]

NVIDIA GF100 pre-release speculation

20% faster seems alot more reasonable than some of the other claims though

 

[rpg.314]

since I can't sleep I'll post the latest rumours - Fudzilla is claiming 20% chip yields currently, and that the gf100 is 20% faster than the 5870.

http://www.fudzilla.com/content/view/17205/1/

fud is a bit more reputable than a poster on an internet gaming forum, but not by too much

If indeed it is just 20% faster, the GTX 360 should end up quite close to 5870, which will be bad as a ~550mm2 chip is going up against a 334mm2 chip.

NV will prolly try to stem the losses by restricting the supply to consumer markets. It's a shame that AMD has no real traction in the professional and HPC markets so far despite having more efficient gpu's for ~1.5 yrs now.

 

[Davros]

and that the gf100 is 20% faster than the 5870.

And the 5870 is roughly (I said roughly, dont bite my head off) 20% faster than a gtx285
so fermi is 40% faster than a gtx285
thats bloody terrible the new gen is only 40% faster than the last, bloody terrible..

 

[annihilator]

And the 5870 is roughly (I said roughly, dont bite my head off) 20% faster than a gtx285
so fermi is 40% faster than a gtx285
thats bloody terrible the new gen is only 40% faster than the last, bloody terrible..

That makes is 45%, not 40.

Did you also think it was bloody terrible when 5870 turned out to be approximately 35-40% faster than the last generation?

 

[Rys]

NVIDIA GF100 pre-release speculation

I've started a new thread to keep the GF100 product pre-release noise away from the architecture discussion. So, have at it, and please try and add some value. I'll slowly move the non-architecture posts from the other thread over here as time goes by, so the arch one can be archived as something useful.

Latest rumour mill stuff had the first GeForce product at 20% higher than 5870. Rumour courtesy of Fudzilla.

 

[eastmen]

That makes is 45%, not 40.

Did you also think it was bloody terrible when 5870 turned out to be approximately 35-40% faster than the last generation?

IS the 5870 only 35-40% faster than the 4870 and 4890 ?


Anyway both of these companys have added alot more to these chips than there was on dx 10 chips. hopefully this will amount to better visual quality and speed when developers start using it.

 

[DegustatoR]

thats bloody terrible the new gen is only 40% faster than the last, bloody terrible..

Yeah, almost like Cypress to RV770.

 

[Psycho]

Rys shouldn't have moved that piece of misinfomation...
Cypress has more like 70% higher framerates(*) than rv770, and ~40%, not 20% over gtx285 - http://www.computerbase.de/artikel/...009/18/#abschnitt_performancerating_qualitaet
So 20% over 5870 would actually be a slightly higher generational gain (over gtx280) than cypress over rv770.

(*) Ofcourse the card (and especially chip) is faster than that, but system limitations et al. in current games...

 

[AlexV]

IS the 5870 only 35-40% faster than the 4870 and 4890 ?

Depends on what you're doing and how you're doing it, mostly. I'd not expect things to be hugely different with Fermi (in that the distance from GT2xx stuff will be variable).

 

[Babel-17]

Rys shouldn't have moved that piece of misinfomation...
Cypress has more like 70% higher framerates(*) than rv770, and ~40%, not 20% over gtx285 - http://www.computerbase.de/artikel/...009/18/#abschnitt_performancerating_qualitaet
So 20% over 5870 would actually be a slightly higher generational gain (over gtx280) than cypress over rv770.

(*) Ofcourse the card (and especially chip) is faster than that, but system limitations et al. in current games...

Something weird with the 4890 numbers, going from 4xAA/8xA to 8xAA/16xA. Performance increases?

4870 numbers, while not being listed for 2560x1600 aa and aniso performance, also look suspicious.

But I can't read German so maybe this was addressed?

 

[Psycho]

They're average scores for all the tests (except batman), but some (at least stalker and world in conflict) don't have 8xAA mode/scores, so I guess that's why (ie. different set of games for 4x and 8x).

 

[Ailuros]

edit: brainfart

 

[Lonbjerg]

At 2:30 in that video, why does the sled fly off without the jet attached exactly the same way it did with the jet?

My guess would be that they only simulate the physics properties of the single components and not the whole inner workings of a rocket slegde...but that is a guess I don't have the demo myself.

 

[Sontin]

A small piece of GF100 information from ChrisRay:

However, I can say this at this point as I expect alot of leaks in the nearby future. Fermi dramatically improves on of icecold's biggest complaints about Gt200 and really all cards are having this problem. I expect he'll be happy to see it addressed GF100's tessellation engine is impressive. Its certainly not software and there certainly is dedicated tessellation hardware beyond just the HULL/Shader routines. Its been an incredibly long couple months of me being quiet. But thats at an end really shortly.

http://www.rage3d.com/board/showpost.php?p=1336134574&postcount=1573

 

[Mize]

Well then, let the leaks begin Chris!
[Mize starts checking for PMs]

 

[mapel110]

A small piece of GF100 information from ChrisRay:


http://www.rage3d.com/board/showpost.php?p=1336134574&postcount=1573

So Monday is THE DAY, I heard?!