NVIDIA reveals Fermi's successor: Kepler at 28nm in 2011, Maxwell in 2013

[SolMiester]

Do you honestly think that Nvidia has no refresh planned for the gf100 or nothing to counter the 58xx series refresh or 68xxx?

I just dont see them releasing in May 2110 and have nothin more till the second half of 2011.
I would think the gtx 480 refresh was planned well before the gtx480 was released and just because the oringinal gf100 line was delayed, it does not mean the refresh will be also.
If there is no refresh planned mabe Nvidia knows something we don't? Mabe the 6000 series won't increase performance as much as the leaked benchmarks (which I still think mostly were bogus) make us believe?

So far I havent seen any official performance numbers, heat, noise, and especially price on the 6000 series. I've seen the 6700 is supposed to release at the end of OCT and some faster cards to follow in Dec, followed by the low end in 2011. Even this is not a official release date, just broad speculation on ABOUT when they COULD be released.

I think Nvidia is keeping there mouth shut for the refresh. Mabe because they dont have crap to offer but they could be just waiting for AMD to release and once again trump there performance "like they almost allways do"
I'll believe it when I see it, is my attitude. 😉

My thoughts also!

 

[Kenmitch]

Where have you been? TMSC had process issues on that node. One of the reasons why AMD was supply constrained nearly the entire time Nvidia wasnt in the DX11 market.

I'm not defending the guy but I think he's just saying AMD didn't seem to have too much of a problem getting the 5xxx series out on time. So one would think that yes TMSC had some issues with the 40nm node but it would seem like nvidia had more issues with the production of Fermi than they let on. Now if this was TMSC's fault or nvidia had to re-spin a couple of times to out perform the 5xxx series only nvidia knows the answer. I'm leaning towards a couple of re-spins more than pointing the finger at TMSC and blaming them for making Fermi so late.

 

[MrK6]

Do you honestly think that Nvidia has no refresh planned for the gf100 or nothing to counter the 58xx series refresh or 68xxx?

Nope, why would you think that? I said performance would stagnate because it seems that NVIDIA might not introduce anything that will break boundaries within that time frame. Anything's possible, but it seems like they've dug themselves into a hole with the GF100.

I just dont see them releasing in May 2110 and have nothin more till the second half of 2011.
I would think the gtx 480 refresh was planned well before the gtx480 was released and just because the oringinal gf100 line was delayed, it does not mean the refresh will be also.

That would be difficult to call, depending on what exactly their issues were and how much redesigning they called for.

If there is no refresh planned mabe Nvidia knows something we don't? Mabe the 6000 series won't increase performance as much as the leaked benchmarks (which I still think mostly were bogus) make us believe?
So far I havent seen any official performance numbers, heat, noise, and especially price on the 6000 series. I've seen the 6700 is supposed to release at the end of OCT and some faster cards to follow in Dec, followed by the low end in 2011. Even this is not a official release date, just broad speculation on ABOUT when they COULD be released.

Anything's possible, but I highly doubt it. Besides, general speculation and leaks for the previous two launches have been pretty accurate as far as I'm concerned. I'm sure there's a little extra PR push behind it, which may make the cards seem better than they truly are, but I fully expect the Barts XT part to hang with the 5850 and 5870 performance levels (maybe more towards the 5870 in newer games). That doesn't mean the final product will meet my expectations though.

Since when does a lack of confirmation regarding unannounced products confirm there ISN'T new products being worked on for the current node?

It doesn't, how did you come to that conclusion?

 

[Grooveriding]

So basically this amounts to the first of many 'announcements'

This is the same as that first 'announcement' we got when the 5XXX series was just about to release.

Much like there was an announcement of a date that a release date would be announced, then on that day an announcement of another day where a release date would be announced.

Then on that day, a release date was announced. Then on that release date we got a paper launch and three weeks later you could actually buy.

So they have nothing new until 28nm. Nothing from nvidia for another nine months, just refreshes of their 4XX series.

I hope they deliver some serious performance when they release that 28nm card late next year and deliver it at the same time AMD releases their 7 series on 28nm.

It's pretty disappointing seeing them fall into a routine of press releases but no actual hardware for two year stretches.

 

[bryanW1995]

Not sure....But on the other hand wasn't there a rumor that AMD was in discussions with Intel about a possible joint venture in the GPU market....After all isn't that how a rumor gets started 🙂

jhh is strongly against using global foundries b/c they are the "enemy". however, with tsmc's continued issues they would be wise to at least explore additional options. hmmm, I wonder if intel has any spare fab space...

 

[Genx87]

I'm not defending the guy but I think he's just saying AMD didn't seem to have too much of a problem getting the 5xxx series out on time. So one would think that yes TMSC had some issues with the 40nm node but it would seem like nvidia had more issues with the production of Fermi than they let on. Now if this was TMSC's fault or nvidia had to re-spin a couple of times to out perform the 5xxx series only nvidia knows the answer. I'm leaning towards a couple of re-spins more than pointing the finger at TMSC and blaming them for making Fermi so late.

They were out on time. But they were supply limited due to the issues and they had to yank out functionality in the 5000 series to get it done. Nvidia had more at stake with their chip(Tesla, Quadro lines, and gaming) and didnt have that flexibility. That said, the 40nm node issue was a fiasco for both companies.

 

[Daedalus685]

They were out on time. But they were supply limited due to the issues and they had to yank out functionality in the 5000 series to get it done. Nvidia had more at stake with their chip(Tesla, Quadro lines, and gaming) and didnt have that flexibility. That said, the 40nm node issue was a fiasco for both companies.

Question is though... was 40nm a fiasco or is TSMC a fiasco.. Even worse, if the sort of issues we see start to become commonplace the smaller we get both companies are going to have to have a rather large change of strategy to succeed.

 

[happy medium]

Nope, why would you think that?

Easy one, because when do they not have a refresh or higher clocked faster card.? , Both AMd and Nvidia do it.

might stagnate

seems that NVIDIA

seems like they've dug themselves into a hole

would be difficult to call,

Anything's possible, but I highly doubt it

general speculation

but I fully expect the Barts XT part to hang with the 5850 and 5870 performance levels (maybe more towards the 5870 in newer games).

Thats what this thread is all about, just like I said in an earlier post its all guessing, and wishfull thinking.

Dont get me wrong the speculation/spin is quite entertaining. 🙂

 

[notty22]

So basically this amounts to the first of many 'announcements'

This is the same as that first 'announcement' we got when the 5XXX series was just about to release.

Much like there was an announcement of a date that a release date would be announced, then on that day an announcement of another day where a release date would be announced.

Then on that day, a release date was announced. Then on that release date we got a paper launch and three weeks later you could actually buy.

So they have nothing new until 28nm. Nothing from nvidia for another nine months, just refreshes of their 4XX series.

I hope they deliver some serious performance when they release that 28nm card late next year and deliver it at the same time AMD releases their 7 series on 28nm.

It's pretty disappointing seeing them fall into a routine of press releases but no actual hardware for two year stretches.

Its their dime this week, GPU Technology Conference
http://www.nvidia.com/object/gpu_technology_conference.html

Isn't this what every company does ? AMD was doing it during Intel's developer forum. Its to let know , they are still alive.

 

[happy medium]

So they have nothing new until 28nm.

Is this a question? Or do you know something we don't?

Nothing from nvidia for another nine months, just refreshes of their 4XX series.

another question? or once again do you know something we dont.

Edit: like said before ,I dont know or claim to know crap, but while I think speculating is fun, writing it like its a fact is just plain FUD.

 

[Kenmitch]

I just dont see them releasing in May 2010 and have nothing more till the second half of 2011. I would think the gtx 480 refresh was planned well before the gtx480 was released and just because the oringinal gf100 line was delayed, it does not mean the refresh will be also.

If there is no refresh planned maybe Nvidia knows something we don't? Maybe the 6000 series won't increase performance as much as the leaked benchmarks (which I still think mostly were bogus) make us believe?

Take a deep breathe and relax 🙂

I doubt all the blame with Fermi's delay can be placed in TSMC's lap. I'm sure some of the delay was caused by the need to re-spin/tweak the design. One could say that AMD didn't seem to have much of a problem with TSMC's 40nm node other than supply. As far as Fermi's replacement goes it's in the old PR slide shown in the photo.

Now does nvidia have something to counter the 6xxx series launch? We'll have to wait and see I guess....I'm kinda surprised they haven't released a dual gpu card yet!

As far as the hypothetical performance of the upcoming 6xxx series from AMD it's anybodies guess....But I'd think AMD wouldn't waiste the R&D time and money if it wasn't gonna pay off for them.

 

[happy medium]

Now does nvidia have something to counter the 6xxx series launch? We'll have to wait and see I guess....I'm kinda surprised they haven't released a dual gpu card yet!

I missed the update to the article:
Quote:
"Update: Does this roadmap mean we should expect new GPUs from NVIDIA every two years? Jen-Hsun Huang wouldn't commit to that, but said that there will be "mid-life kicker" product launches in-between each new NVIDIA chip to keep the wheels turning."

That spells refresh to me. So Nvidia's president said this, and you exspect nothing to counter the 6000 series?

 

[SolMiester]

Take a deep breathe and relax 🙂

I doubt all the blame with Fermi's delay can be placed in TSMC's lap. I'm sure some of the delay was caused by the need to re-spin/tweak the design. One could say that AMD didn't seem to have much of a problem with TSMC's 40nm node other than supply. As far as Fermi's replacement goes it's in the old PR slide shown in the photo.

Now does nvidia have something to counter the 6xxx series launch? We'll have to wait and see I guess....I'm kinda surprised they haven't released a dual gpu card yet!

As far as the hypothetical performance of the upcoming 6xxx series from AMD it's anybodies guess....But I'd think AMD wouldn't waiste the R&D time and money if it wasn't gonna pay off for them.

LOL, how condescending!.....the slides show new architecture GPU's, not mid-life kickers...however I too am waiting for the dual GPU

 

[Kenmitch]

I missed the update to the article:
Quote:
"Update: Does this roadmap mean we should expect new GPUs from NVIDIA every two years? Jen-Hsun Huang wouldn't commit to that, but said that there will be "mid-life kicker" product launches in-between each new NVIDIA chip to keep the wheels turning."

That spells refresh to me. So Nvidia's president said this, and you exspect nothing to counter the 6000 series?

I wouldn't call it much of an article. Kinda wondering where it came from? If you look at the date for Fermi (2009) it makes it look like an old PR slide of nvidias roadmap. As far as the mid-life kicker products ???

Anybody find anything related to this from another source? I did a quick look at nvidias site and didn't see anything mentioning it.

 

[SlowSpyder]

I missed the update to the article:
Quote:
"Update: Does this roadmap mean we should expect new GPUs from NVIDIA every two years? Jen-Hsun Huang wouldn't commit to that, but said that there will be "mid-life kicker" product launches in-between each new NVIDIA chip to keep the wheels turning."

That spells refresh to me. So Nvidia's president said this, and you exspect nothing to counter the 6000 series?

How do we know what that means though? In this case it could be a 512SP part with a 750MHz core. It could be a respin that uses 40 watts less. It could be a fully functional GF104. It could be a dual GF104 part. It could be a number of things. I wouldn't say that that it means that a GF100 refresh to fight off the Radeon 6000 is on it's way.

After the 2900XT and 3870 AMD didn't really try and do a whole lot in a battle it couldn't win. They competed where they could, at prices that made sense, than came out with a great bunch of parts in the 48xx series. Nvidia might do the same until 28nm. They might not. But I'd hardly say that this means that they'll have an answer for the Radeon 6000 in any sort of refresh.

 

[RussianSensation]

If Maxwell will be what they say it will be, I think we'll have enough power to do just about anything including ray tracing.

That part has me interested. But Crysis 3 will take it down 🙂 JHH told at the press conference that the Maxwell chip is 40 times faster than Fermi and that it will get up to 10 to twelve times faster than Kepler (in dual precision in GFLOPS per Watt). But that's not necessarily tied to absolute gaming performance. Sure performance/watt is critical, but so are geometric performance increases.

 

[happy medium]

How do we know what that means though? In this case it could be a 512SP part with a 750MHz core. It could be a respin that uses 40 watts less. It could be a fully functional GF104. It could be a dual GF104 part. It could be a number of things. I wouldn't say that that it means that a GF100 refresh to fight off the Radeon 6000 is on it's way.

After the 2900XT and 3870 AMD didn't really try and do a whole lot in a battle it couldn't win. They competed where they could, at prices that made sense, than came out with a great bunch of parts in the 48xx series. Nvidia might do the same until 28nm. They might not. But I'd hardly say that this means that they'll have an answer for the Radeon 6000 in any sort of refresh.

good points all around 🙂

 

[happy medium]

As far as the mid-life kicker products ???

Anybody find anything related to this from another source?

I see quotations around mid life kicker which leads me to believe it came from the president of Nvidia himself.
To me thats better then a review/news sites rumor mill.

 

[notty22]

I wouldn't call it much of an article. Kinda wondering where it came from? If you look at the date for Fermi (2009) it makes it look like an old PR slide of nvidias roadmap. As far as the mid-life kicker products ???

Anybody find anything related to this from another source? I did a quick look at nvidias site and didn't see anything mentioning it.

They are in the middle of their GPU Technology Conference, I think that the CEO spoke to this during a speech. http://blogs.nvidia.com/ntersect/20...(The+NVIDIA+Blog+-+GPU+Technology+Conference)

 

[SlowSpyder]

That part has me interested. But Crysis 3 will take it down 🙂 JHH told at the press conference that the Maxwell chip is 40 times faster than Fermi and that it will get up to 10 to twelve times faster than Kepler (in dual precision in GFLOPS per Watt). But that's not necessarily tied to absolute gaming performance. Sure performance/watt is critical, but so are geometric performance increases.

I was going to post something similar before your edit. 🙂 This is DP GFLOP performance per watt. They could technically not make a part that is any faster, in fact it could be slower, and be true to the graph... not that I think Nvidia would do that. Maybe they expect a certain power saving feature that will be implemented on whatever process Maxwell will be built on? Maybe Nvidia has an architecture shakeup in mind. Who knows.

 

[Kenmitch]

I see quotations around mid life kicker which leads me to believe it came from the president of Nvidia himself.
To me thats better then a review/news sites rumor mill.

When I talked to my source at AMD he said " Were gonna kick nvidia to the curb just like Apple did " so does that make you feel better about the 6xxx series hypothetical performance now 🙂

What I meant is that if it was a press-release as others claim then why isn't it all over the web....We all know how the Big Green Machines PR department works.

Thread-crap/thread-derail. Please keep it on topic. Nvidia is the topic, your post here is inflammatory and designed to generate flame-responses. (which is exactly what has transpired in this thread following your post here) Not acceptable.

Moderator Idontcare

 

[tviceman]

It doesn't, how did you come to that conclusion?

I came to that conclusion because you said it yourself.

I didn't see a roll eyes emoticon, so I'll ask for confirmation: you haven't heard of any new high-end parts for 40nm? If so, that's a shame, as we're going to have a long dry spell of performance until 28nm arrives.

 

[Janooo]

I missed the update to the article:
Quote:
"Update: Does this roadmap mean we should expect new GPUs from NVIDIA every two years? Jen-Hsun Huang wouldn't commit to that, but said that there will be "mid-life kicker" product launches in-between each new NVIDIA chip to keep the wheels turning."

That spells refresh to me. So Nvidia's president said this, and you exspect nothing to counter the 6000 series?

... and you would believe him? How about the fake Fermi board?

 

[HurleyBird]

JHH isn't exactly the most trustworthy executive.

 

[blastingcap]

... and you would believe him? How about the fake Fermi board?

NV has a history of grandiose promises that they sometimes keep and sometimes don't. They talked up Fermi in '09 but it got delayed over and over again and eventually wound up being only 10-15% faster than their competitor's card, at much higher temps, noise, and power draw... and missed its target clocks. GTX280 also missed its target.

Also, as others noted, the chart is for performance/watt, not performance. How could Kepler NOT have better performance/watt than Fermi when it's on 28nm? Cypress is already way above Fermi in perf/watt and it's still on 40nm. And NI (HD6xxx) will probably be higher still.

Last year JHH said that he doesn't like to talk about products that don't exist yet.

Today, a month before the launch of HD6xxx and probably just a couple of weeks before the media starts getting review cards for HD6xxx, he talks and talks about NV GPUs several months to a few years away, in what could be interpreted by some as an attempt to steal AMD's thunder.

Dirty.

But effective, judging by NV's stock price movement this morning.