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

There have been rumors about this on some sites. One site said "there is a big chance that you'll see this new Fermi, let’s call it Fermi 1.5 even in late 2010."

Well, lets squash this right now with words from Jen-Hsun Huang himself at GTC. ( http://www.youtube.com/watch?v=rK7CVrZez0Y&hd=1#t=4m ) Jen-Hsun Huang was asked about new product release schedules. His answer:

"The basic [product release schedule] cadence has to follow process advancement. Unless there is a major new node in process, it's hard for us to come out with enough transistors to fundamentally change the architecture in some meaningful way. Going forward I think what you're going to find is a new processor node about every two years. In between we will likely have a "mid-life kicker" if you will. Something that enhances the micro-architecture features. You are going to see this coming in the near future for next year, the mid-life kicker for Fermi."

this must be giving tsmc nightmares already hahaha :D
if this is fermi 1.5 then what is gf10x? fermi 1.4?

We already have a thread for that bamtan :shrug:

http://www.xtremesystems.org/forums/...d.php?t=259492

Quote:

---

Originally Posted by Olivon

We already have a thread for that bamtan :shrug:

http://www.xtremesystems.org/forums/...d.php?t=259492

---

this is news, not speculation. nvidia will have no answer for amd 6000 series this year.

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.

New GF11X, full GF104 and full GF106 = GTX 5XX series ?

Anyone taking this seriously after the stunt nVidia pulled with "here's the puppy" Fermi is naive to put it mildly. The only response nVidia can have this year is another space heater and we've had enough of those.

Quote:

---

Originally Posted by SimBy

Anyone taking this seriously after the stunt nVidia pulled with "here's the puppy" Fermi is naive to put it mildly. The only response nVidia can have this year is another space heater and we've had enough of those.

---

first, that's a misquote.

secondly, why would you be gullible to believe what he said? the card is a different issue entirely and it was reasonable to be skeptical. there is no benefit to lying about this as opposed to demoing a graphics card half a year before launch.

Quote:

---

Originally Posted by Chumbucket843

secondly, why would you be gullible to believe what he said? the card is a different issue entirely and it was reasonable to be skeptical. there is no benefit to lying about this as opposed to demoing a graphics card half a year before launch.

---

What he's saying is that Nvidia has already pulled that stunt before where they announced the announcement of the paper launch of a card that ends up being made out of wood and paint for the first 3 months of "public" existence.

The goal of this announcement was to tell people who were thinking to buy amd to wait a little longer and possibly get a better card with Nvidia. I mean, come on, just as the amd 6xxx specs are leaked, we get an announcement about fermi 1.5... nice timing!

Quote:

---

Originally Posted by antiacid

The goal of this announcement was to tell people who were thinking to buy amd to wait a little longer and possibly get a better card with Nvidia. I mean, come on, just as the amd 6xxx specs are leaked, we get an announcement about fermi 1.5... nice timing!

---

And that's what we call marketing.

Quote:

---

Originally Posted by SKYMTL

And that's what we call marketing.

---

feels just like august to december of 2009, any AMD gpu news is followed by an Nvidia announcement with some kind of positive spin

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

576 sp gf110 is what I'm thinking. :yepp: Some claim the kicker would come already this year, but that sounds rather soon.

I wonder where that guy in the hood with the green screen is, looks like he could be called back into business to tell us 'it's coming' any time now :p:


Seriously though, I think this is good news. Hopefully some kind of process shrink for lower heat, less power consumption, and a few performance tweaks. I'd be quite happy with that, particularly in the mobile area.

Quote:

---

Originally Posted by Serra

I wonder where that guy in the hood with the green screen is, looks like he could be called back into business to tell us 'it's coming' any time now :p:

---

http://img94.imageshack.us/img94/9667/66972171.jpg

:rofl::rofl:

JHH isn't the most trustworthy individual. I hope he's telling the truth, but my default assumption is that any forward looking statements he makes are half truths at best.

in the world of business and marketing half truths are often used (omitting information) (e.g. plank of wood representing fermi) in order to portray the message they desire.

Mid-life kicker can represent from a visual demo, to some specs on an upcoming product, the message is so ambiguous and diffuse that I would take it with a grain of salt (possibly a giant one)

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.

---

I really appreciate you being the voice of reason. And see you @ CES. :D

Quote:

---

Originally Posted by Russian

I really appreciate you being the voice of reason. And see you @ CES. :D

---

More like "the voice I want to hear", no?

Other posts are quite reasonable to me. After all, they were citing historical facts of the past Fermi announcements, not speculations ;)

Quote:

---

Originally Posted by Teemax

More like "the voice I want to hear", no?

Other posts are quite reasonable to me. After all, they were citing historical facts of the past Fermi announcements, not speculations ;)

---

Not really. It wasn't like they were knowingly lying. There were constant challenges in the new architecture... obviously they had their problems. A kicker product is not going to have those same problems...

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

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 :)

50% enlargement out of GF 104 design is within reason, but doubling up or further enhancement will need 28 nm process node. Nonetheless, i don't think this "kicker" will be on sale this year, late Q1 2011 at the earliest IMHO, and end Q2 2011 most likely. It won't bring the undisputed GPU performance king title, but atleast will be quite respectable than watching GF 100 got butt raped by Cayman XT.

that whole kicker sounds like a marketing gimmick to me, a kicker doesn't necessarily mean hardware upgrade, mid life kicker could even be only for mobile market... like I said before, that message is so diffuse I would not make much of it until some factual info was leaked out... or if gigantic ballbag fell on top of my foreheard...

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.

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

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?

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.

Quote:

---

Originally Posted by Teemax

Selective memory is convenient, isn't it?

---

The reference was made to release timelines. Not woodscrewgate.

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. ;)

coming to a gaming PC near you soon..er.. sometime in 2011. the GTX480 MLK edition.. I like it. :)

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

Most companies have actually adopted Valve time by now. It'll be released soon enough.

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...

lol i like your sig saaya

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.

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.