NVIDIA Kepler speculation thread

[trinibwoy]

And I don't necessarily agree to your examples in 1 & 2 either. Fermi had 1 or more hw bugs, but it wouldn't had been manufacturable in late 2009 due to crappy yields and in extension too high costs anyway and it's not like R600 had a joyride with 80nm@TSMC which NV for some weird reason avoided completely.

Well my conclusions were based on how well the follow ups fared. Fermi was eventually vindicated, R600 was still slow in RV670 guise.

It's my understanding that their original plan for Kepler was for quite some time to go with the performance part first and later on to go for the top dog.

If that really was their plan I hope they aimed higher this time else they won't have anything to face Tahiti. That's also doubtful because Titan was supposed to get daddy Kepler by the end of 2012 - I assume for that to happen the chips have to be ready well before then.

One message that might be of some value is "earlier than expected". So when was GK104 really expected after all, since no next generation 28nm GPU made it on shelves in 2011?

Summerish?

 

[Alexko]

A trickle of 7970's isn't a threat.

I'm really not sure that's fair. Newegg went out of stock for about 24 hours (maybe a bit more, I'm not sure) but they have two SKUs right now, both of which are only restricted to 10 per customer, so it might be a bit on the tight side, but supply seems about adequate.

 

[axehandle]

NV's competitor missed Christmas too, and now nobody has any money left to spend on expensive graphics cards.

I know I'm not everybody but I'm also not nobody and I have cash in hand and I'm ready to upgrade now. My current graphics card (HD5850) is faulty and I randomly get BSOD when running video files so it's not just upgrade fever..I kind of need a new card now. I was displeased with my AMD card so I want to go Nvidia but I don't want to buy a Fermi if Kepler is around the corner. Hence why I am so personally disappointed that Kepler was a no-show at CES.

 

[trinibwoy]

I'm really not sure that's fair. Newegg went out of stock for about 24 hours (maybe a bit more, I'm not sure) but they have two SKUs right now, both of which are only restricted to 10 per customer, so it might be a bit on the tight side, but supply seems about adequate.

Perhaps but honestly even a flood of $500+ cards isn't a threat on its own. Like I said, a timing advantage of a few months only plays out if your competition's response is weaksauce.

I know I'm not everybody but I'm also not nobody and I have cash in hand and I'm ready to upgrade now. My current graphics card (HD5850) is faulty and I randomly get BSOD when running video files so it's not just upgrade fever..I kind of need a new card now. I was displeased with my AMD card so I want to go Nvidia but I don't want to buy a Fermi if Kepler is around the corner. Hence why I am so personally disappointed that Kepler was a no-show at CES.

Then you will never be happy unless both companies launch competing parts at the same time. There's no guarantee of that so just buy what you need when you need it. Buyer's remorse is a bitch but so is life.

 

[Sinistar]

Kyle Bennett over at [H] says he is hearing June for Kepler

 

[Bob]

R600 had a joyride with 80nm@TSMC which NV for some weird reason avoided completely.

All of G8x (where x != 0) was on an 80nm process at TSMC. Linky

 

[axehandle]

http://www.fudzilla.com/component/k2/item/25503-jensen-on-kepler

After the keynote, we managed to ask Jen Hsun Huang, Nvidia CEO, a single question. We asked, where is Kepler? The answer we got was that we have to be "patient" about it.

And apparently Nvidia's partners are getting a little antsy.
http://www.fudzilla.com/home/item/25431-nvidia-partners-concerned-about-amds-28nm-lead

Meanwhile AMD’s 28nm lineup looks very good and despite silent promises made by Nvidia to its partners that “Kepler is much better than that“, they are expressing serious concerns about AMD’s lead, or rather they how to go compete against AMD until Kepler launches. Of course, at this point nobody knows what to expect from Kepler in terms of performance and all we have to go on are Nvidia’s own rosy predictions.

 

[seahawk]

Silence on NV´s side is usually a better sign, than the times when NV went around creating FUAD.

 

[neliz]

Silence on NV´s side is usually a better sign, than the times when NV went around creating FUAD.

remove emotion from the equation and you get more of the Cypress <-> Fermi situation.

One party has low-end and mainstream 28nm parts out a couple of months before the other (sales) and has the comfort of a dual-GPU spoiler product (marketing).

 

[DegustatoR]

There will be some spoilers in the upcoming products, that's for sure.

 

[seahawk]

remove emotion from the equation and you get more of the Cypress <-> Fermi situation.

One party has low-end and mainstream 28nm parts out a couple of months before the other (sales) and has the comfort of a dual-GPU spoiler product (marketing).

I am quite confident, that the difference in availbility will decrease with the smaller chips.

 

[Alexko]

Perhaps but honestly even a flood of $500+ cards isn't a threat on its own. Like I said, a timing advantage of a few months only plays out if your competition's response is weaksauce.

Agreed, $500+ cards are never going to see much volume, no matter how fast they are—unless they start giving blowjobs or something.

But Pitcairn is scheduled for February, isn't it? And I think Cape Verde isn't far behind… It all hinges on just how many wafers TSMC can supply, at what price, and of course on yields, but a 3~6-month advantage could be a very big deal.

 

[trinibwoy]

Agreed, $500+ cards are never going to see much volume, no matter how fast they are&mdash;unless they start giving blowjobs or something.

Newegg is OOS again. If you consider that launch day availability should be much higher than incremental shipments it's clear that there weren't very many chips to go around. Let's see how long it takes for everyone who wants a 7970 to get one.

Also, the 7950 will be cheaper and therefore in higher demand. Insufficient availability to meet that demand is the only real plausible reason for holding back the launch.

But Pitcairn is scheduled for February, isn't it? And I think Cape Verde isn't far behind&hellip; It all hinges on just how many wafers TSMC can supply, at what price, and of course on yields, but a 3~6-month advantage could be a very big deal.

Yeah, there's a lot of pent up demand for a fast $150 card. The upper end of the midrange ~$250 will have to contend with discounted 560's and 6970's though as usual.

The bolded part surprised.. who? We all knew there was GF100b in the pipes since the launch of GF100

From TPU: "If two weeks ago somebody told us that today NVIDIA would be hard-launching a new high-end graphics processor under a new product family (the GeForce GTX 500 series), we'd have laughed out loud."

 

[Alexko]

Newegg is OOS again. If you consider that launch day availability should be much higher than incremental shipments it's clear that there weren't very many chips to go around. Let's see how long it takes for everyone who wants a 7970 to get one.

Also, the 7950 will be cheaper and therefore in higher demand. Insufficient availability to meet that demand is the only real plausible reason for holding back the launch.

But that could also be because yields are actually pretty good and they don't want to sell Tahiti Pros when they could be selling Tahiti XTs instead. Either way, I'm sure it well get clearer within a few weeks.

 

[Kaotik]

From TPU: "If two weeks ago somebody told us that today NVIDIA would be hard-launching a new high-end graphics processor under a new product family (the GeForce GTX 500 series), we'd have laughed out loud."

Hardlaunch at said point could have been surprise, but the fact that GF100b was in the pipes definately wasn't

 

[trinibwoy]

Hardlaunch at said point could have been surprise, but the fact that GF100b was in the pipes definately wasn't

Agreed. Though I think the point was that the name, configuration, launch date and performance were not leaked even though there was a lot of availability on launch day. So it is possible to keep things under wraps. Granted it's probably easier to hide a tweaked chip than a new architecture on a new process.

 

[DegustatoR]

Granted it's probably easier to hide a tweaked chip than a new architecture on a new process.

Actually, that's quite the opposite. A tweaked chip is harder to hide and it's much easier to estimate it's performance and features since they're basically known from the previous version of that chip.

As for June and other stuff people are hearing somewhere:

The reason behind the January 31st launch date might have something to do with the Nvidia's plan to launch its 28nm Kepler based GK104 GPU sometime around that date and AMD simply wants something to coincide with Nvidia's launch, or ruin its day.

 

[no-X]

Actually, that's quite the opposite. A tweaked chip is harder to hide and it's much easier to estimate it's performance and features since they're basically known from the previous version of that chip.

I can't agree. GF110 was just a new revision of a released and well known product. TSMC workers had several Fermi revisions in hands, so it was just another one for them, nothing special. That's the most likely reason why anybody didn't expect it earlier.

 

[rpg.314]

He is soo... credible....

 

[Dave Baumann]

Perhaps but honestly even a flood of $500+ cards isn't a threat on its own. Like I said, a timing advantage of a few months only plays out if your competition's response is weaksauce.

Don't underestimate the importance of momentum, especially in a channel evironment where you have partners with engineering resources.