New Fermi mid-life kicker replacement "for next year" - Jen-Hsun Huang

Printable View

Show 100 post(s) from this thread on one page

09-30-2010, 05:54 AM
SKYMTL

Quote:

Originally Posted by K404

Well if Huang is saying next year... with his endless optimism.....card might be ready for next Christmas?

Remember, due to the delays there was a major shake up within NVIDIA's corporate structure and design direction. JHH stated several times at the GTC and subsequent interviews that mistakes were made and the reasons behind said mistakes were found and fixed.

Judging from all the new faces the media has seen in certain technical briefings, etc it is quite obvious the shake up is well under way. Even the general tone of PR phone calls has changed quite a bit and this really has been welcome.

The only problem for the time being is the fact that NVIDIA is now saddled with the GF100 architecture for the lucrative Christmas shopping season. If I had to peer into my crystal ball, it is obvious that we will see NVIDIA lean heavily on the GF104 over the next few months. Whether that means new SKUs being released or not is anyone's guess but even without new products, the GTX 460 cards are perfectly placed to face the new realities of the GPU market.

IMO, the biggest step forward was NVIDIA stating that they had hit a power consumption and TDP wall which will lead to an increase in perf per watt rather than the brute force approach of the GF100 series.

Whether or not all these changes are going to do a lick of good has yet to be seen. However, I don't think we should assume that past failures will be repeated en masse.
09-30-2010, 06:16 AM
JohnJohn

Quote:

Originally Posted by SKYMTL

Whether or not all these changes are going to do a lick of good has yet to be seen. However, I don't think we should assume that past failures will be repeated en masse.

True, usually these kind of screw ups are positive in the sense that they clean and shake all the parasites from a company, amd went through this with r600, now years later, they are more competitive and focused than ever, at least in graphics. In cpus they are doing better, the difference is that here the competitor is mighty intel
09-30-2010, 06:31 AM
Teemax

Quote:

Originally Posted by Svnth

Not really. It wasn't like they were knowingly lying.

http://images.bit-tech.net/news_imag...red/fermi2.jpg
Selective memory is convenient, isn't it?
09-30-2010, 06:46 AM
Teemax

Quote:

Originally Posted by SKYMTL

Remember, due to the delays there was a major shake up within NVIDIA's corporate structure and design direction. JHH stated several times at the GTC and subsequent interviews that mistakes were made and the reasons behind said mistakes were found and fixed.

You probably gave JHH too much benefit of the doubt.

It took ATI quite a while to recover from the HD 2900 XT fiasco. While I believe that JHH is very competent as a leader, turning a big complacent company around is no easy task.

If I recall correctly, in their latest roadmap, JHH still claimed that Fermi was "released" in 2009. So it's not a stretch to say that NVIDIA still has some problems with truthfulness.
09-30-2010, 06:48 AM
SKYMTL

Quote:

Originally Posted by Teemax

Selective memory is convenient, isn't it?

The reference was made to release timelines. Not woodscrewgate.
09-30-2010, 06:54 AM
SKYMTL

Quote:

Originally Posted by Teemax

It took ATI quite a while to recover from the HD 2900 XT fiasco. While I believe that JHH is very competent as a leader, turning a big complacent company around is no easy task.

IMO, they turned things around pretty darn quick. From what I recall R600 was released at the beginning of summer 2007 and the RV670 was released in November or December 2007. 6 months isn't all that much. ;)
09-30-2010, 12:32 PM
revenant

coming to a gaming PC near you soon..er.. sometime in 2011. the GTX480 MLK edition.. I like it. :)
09-30-2010, 12:39 PM
Pantsu

http://i.hardware.fi/storage/picture...ainaminute.png

Most companies have actually adopted Valve time by now. It'll be released soon enough.
09-30-2010, 12:57 PM
saaya

Quote:

Originally Posted by SKYMTL

And "next year" could be only 3 months away.
Remember, CES is coming up. :D
And don't forget that NVIDIA's answer to the 6000 series for the time being could be price cuts or the release of new GF104 / GF106 based products. There are many ways to "answer" a challenge.

could be price cuts? thats what they have been doing for a few weeks already trying to saturate the market sell as much gf104 as they can... and a full blown gf104 wouldnt make a big difference, thats a less than 10% performance boost...
a full blown gf100 would make more sense, thats where a 10% perf boost matters... but it looks like they still dont have enough yields to even do a marketing stunt with a full blown gf100...

Quote:

Originally Posted by jaredpace

A revamped 768sp "Double gf104" at ~ 500mm2 would be my guess

thats a contradiction in itself though, double gf104 at 500mm2? how??? thats only possible on 28nm and then it would be less than 500mm2 i think? :D

Quote:

Originally Posted by Tha Last Meal

GF110 will come with 40nm I think . and I think that GF110 will come like GF104 . less Temperature and less Power Consumption and huge Performance :)

right, in other words it will be faster AND cooler AND consume less power... and i assume it will cost less too? will its exhaust smell like roses maybe? :D

come on man, thats wishful thinking and unrealistic...
it will either be much faster and consume the same, most likely, or will be slightly faster but run cooler and consume less, less likely...

Quote:

Originally Posted by SKYMTL

Remember, due to the delays there was a major shake up within NVIDIA's corporate structure and design direction. JHH stated several times at the GTC and subsequent interviews that mistakes were made and the reasons behind said mistakes were found and fixed.

really? like what? how exactly was the management problem fixed? major management shake up? first time i hear about that... :confused:
so did anybody get canned? or they restructured? how?

the real problem is more that of the strategy...
massive chip PLUS new process PLUS new architecture... all at the same time... and they havent even talked about that so im not sure they changed their strategy...

all we know is that their next chip is supposed to me more than twice as fast as fermi, at least in dpfp, and that makes it a massive chip... on 28nm... so depending on when it comes out, they might be going for the very same combination of risky events like they did with fermi and all the management changes and whipping peoples backs is not going to change anything...

Quote:

Originally Posted by SKYMTL

Judging from all the new faces the media has seen in certain technical briefings, etc it is quite obvious the shake up is well under way. Even the general tone of PR phone calls has changed quite a bit and this really has been welcome.

so there is more/new technical pr and better communication with the press... thats good! but that sounds 100% like good old nvidia... if a product fails, use more marketing... im not saying there is no management shake up, but you cant conclude anything from pr changing...

Quote:

Originally Posted by SKYMTL

The only problem for the time being is the fact that NVIDIA is now saddled with the GF100 architecture for the lucrative Christmas shopping season. If I had to peer into my crystal ball, it is obvious that we will see NVIDIA lean heavily on the GF104 over the next few months. Whether that means new SKUs being released or not is anyone's guess but even without new products, the GTX 460 cards are perfectly placed to face the new realities of the GPU market.

right now? yes! when 6k comes out... we will see...

Quote:

Originally Posted by SKYMTL

IMO, the biggest step forward was NVIDIA stating that they had hit a power consumption and TDP wall which will lead to an increase in perf per watt rather than the brute force approach of the GF100 series.

thats like saying, their product sucked, and they know thats not good, so im sure their next product will suck less... :D
if its going to take 300W to beat the 6k series, do you honestly think nvidia will be humble and nice and go for a slightly slower 250W part? honestly? :D

to sum things up... i think the smart thing to do is to have a massive chip on the old node before you start tinkering with the new one.
cause when a new node arrives its usually not ready to hold the same amount of logic as the old mature node at comparable yields at least
thats how nvidia used to do things... then they got under pressure at some point and jumped on the new node asap, and it worked out "ok"... so they continued with that... they stumbled a bit with every new node after that, until they fell with gf100... the smart thing to do now, imo, would be to go back to the way they did things before... so a massive gf10x on 40nm out by the time 28nm arrives...
so far everything points to that...
09-30-2010, 01:18 PM
Baron_Davis

lol i like your sig saaya
09-30-2010, 01:55 PM
ToxAvenger

Quote:

Originally Posted by SKYMTL

IMO, they turned things around pretty darn quick. From what I recall R600 was released at the beginning of summer 2007 and the RV670 was released in November or December 2007. 6 months isn't all that much. ;)

RV670 was not really "turning things around" it merely made R600 a reasonable product, both from a cost and a power consumption perspective. It still did not challenge Nvidia in any serious way.

Nvidia needs to change their design philosophy (like AMD did) and make a balanced part, not a balls to the walls who cares how big it is GPU.
09-30-2010, 02:14 PM
Chumbucket843

Quote:

Originally Posted by ToxAvenger

RV670 was not really "turning things around" it merely made R600 a reasonable product, both from a cost and a power consumption perspective. It still did not challenge Nvidia in any serious way.

Nvidia needs to change their design philosophy (like AMD did) and make a balanced part, not a balls to the walls who cares how big it is GPU.

they decreased die size by 2.5x for pretty much the same performance. they also decreased power by a larger factor. in fact some dont consider the 3870 high end. i would consider this turning the corner after r520 and r600 which were late and hot. since then, they have had 3 good gpu's in a row. nvidia had a similar streak after nv30, peaking at g80 or g92 depending on what you consider success.

i'm afraid that design goals are not as simple as that. achieving them is even more complex. there are many things that you can do wrong.

Show 100 post(s) from this thread on one page