NVIDIA Kepler GPU Speculation thread

[RussianSensation]

Well, you're not any better than anyone from the hordes of "7970 60% better than 580" fanboys.

Stripping off "rambus and 1024 cores" from GTX780 we'll get another 10-20% faster than 7970 for 200$ more, 6 months later. Get real.

I linked rumors posted by 2 websites, these are not specs I made up of thin air. Did I say anything about those being factual?

So you think GTX680 will launch in 6 months for $749 and have 10-20% faster performance from HD7970? Because that's what I am reading from your comment, because I never said any of those things. All I said is that I think Kepler will surpass the 25-30% performance advantage AMD set for now by about 25% (my own estimate, not a fact). You don't have to believe it. This is why it's called "speculation thread", not "facts thread".

 

[Saico]

? I linked rumors posted by 2 websites, not my own made up specs. Did I say anything about those being factual? Did I mention anything about GTX680 being $200 more than HD7970 or launching in 6 months?

All I said is that I think Kepler will surpass the 25-30% performance advantage AMD set for now by about 25% (my own estimate, not a fact).

Do I need to repeat my post?

 

[AnandThenMan]

We don't need another g80/g92 hack job for nvidia to cut it.

Amd built its new gpu from the ground up i would assume nvidia will do the same.

Nvidia doesn't need to do any such thing, and won't. Fermi was already much more advanced than Evergreen. Fermi+enhancements will be more than enough to combat AMD for 2012 at least IMO.

 

[Gloomy]

512-bit bus

How fast is NVidia gonna get that to go?

 

[RussianSensation]

Do I need to repeat my post?

Ya, you need to explain yourself better or probably consider responding in a less derogatory manner. I posted "rumored" specs that you shut down with your 10-20% faster figure and $200 higher price. You pretty much accused me of some fanboism when all I did was post specs, upon which I extrapolated a conservative 25% increase for the highest end GK100 part. If you don't agree with those specifications, or you think Kepler is only going to be 10-20% faster, just say so. Starting your thread with attacking me isn't helping anyone.

 

[blackened23]

Whether Kepler is late is a relative thing. As soon as the 7970 ships in volume, Kepler could be considered late. I have already said, I will be quite surprised if Nvidia has no answer for the next 5-6 months, it's not like Nvidia didn't know what AMD was doing, they were and are fully aware I'm sure.

I don't really understand why Kepler would be months out, Nvidia already did the hard stuff with Fermi, seems to me a tweaked and updated design would not be nearly as challenging.

I dunno. So far there is no prototype chip for the kepler. That would indicate the chances of a new card in 6 months is near zero. It also just recently taped out.

From chatting with someone in the know, it can take a signifigant amount of time from tapeout > prototype, and then the prototype has to be validated. And then its generally 6 months after the validated prototype that cards are released.

 

[VulgarDisplay]

I myself really wish nvidia had cards out right now becuase I'm ready to upgrade from my 5870. I don't plan on upgrading past 1920x1080 any time soon, and I think a 7970 would stay viable at that resolution for a very long time. Hard to judge price/performance till nvidia answers the 7970, but it's tempting to just buy one now and be done upgrading for 2 years.

 

[RussianSensation]

I myself really wish nvidia had cards out right now becuase I'm ready to upgrade from my 5870. I don't plan on upgrading past 1920x1080 any time soon, and I think a 7970 would stay viable at that resolution for a very long time. Hard to judge price/performance till nvidia answers the 7970, but it's tempting to just buy one now and be done upgrading for 2 years.

Get an HD7970 now, and if Kepler is way faster, you can always sell the 7970. At which games does your HD5870 struggle at 1080P though? BF3?

 

[nsavop]

I dunno. So far there is no prototype chip for the kepler. That would indicate the chances of a new card in 6 months is near zero. It also just recently taped out.

From chatting with someone in the know, it can take a signifigant amount of time from tapeout > prototype, and then the prototype has to be validated. And then its generally 6 months after the validated prototype that cards are released.

That's not true they have working silicon in house, also mentioned is Kepler was taped out in September.

It's looking like a Q2 launch maybe Q1 for GK104.

 

[Saico]

Ya, you need to explain yourself better or probably consider responding in a less derogatory manner. I posted "rumored" specs that you shut down with your 10-20% faster figure and $200 higher price. You pretty much accused me of some fanboism when all I did was post specs, upon which I extrapolated a conservative 25% increase for the highest end GK100 part. If you don't agree with those specifications, or you think Kepler is only going to be 10-20% faster, just say so. Starting your thread with attacking me isn't helping anyone.

I'm just really surprised by your attitude. Yesterday you were bashing and moking 7970 for not being as good as the speculations suggested it to be. And now you're posting your own speculations and want people to believe in them. :whiste:

You're posting the unconfirmed rumours that were floating around for a few months already. Creating a new redundant thread with old rumours the day new tech launches. Err I smell biasiness. ()

That's not true they have working silicon in house, also mentioned is Kepler was taped out in September.

It's looking like a Q2 launch maybe Q1 for GK104.

And they didn't confirm it's a Kepler. GK100 to be precise. Thye have some card in the works to counter 7970 but everyone is saying the GK100 is not anywhere close to prototype.

 

[blackened23]

That's not true they have working silicon in house, also mentioned is Kepler was taped out in September.

It's looking like a Q2 launch maybe Q1 for GK104.

28nm silicon isn't a kepler prototype, also, that same article says fiscal Q2 which puts it at the 2nd half of 2012. My understanding is they're doing a die shrink 28nm fermi which will be released around the May-June timeframe.

edit: yep.

"Before Kepler arrives, the GPU maker is expected to release a series of die shrinks of Fermi built using TSMC's 28nm fabrication process. "

The performance keplers are supposed to be released Q3.

http://www.eetimes.com/electronics-news/4230979/Nvidia-Kepler-GPUs-to-trail-AMD-s-next-generation

The higher end GK104 GPU with 384-bit memory interface and GDDR5 memory looks set to launch in the early second half of 2012, while the dual-GPU version of that card, the GK110 will come to market sometime in the third quarter of next year. This pushes Nvidia’s highest end GK112 card –sporting a 512-bit memory interface-- out to late 2012 or even early 2013.

Nvidia would not confirm the roadmap, but denied any delay to its Kepler launch schedule noting, “our transition to 28nm is going better than 40nm, and yields are better than our original plan.” Current generation Fermi cards were notoriously delayed when they launched in the first half of 2010, owing to a complex design and large die size.

 

[RussianSensation]

I'm just really surprised by your attitude. Yesterday you were bashing and moking 7970 for not being as good as the speculations suggested it to be.

You are confused. There are 2 distinct things that you somehow clumped together as one:

1) I am unhappy with stock HD7970 for real world performance gain, not because it didn't meet rumored performance. For example, if the rumored numbers were + 80%, and HD7970 beat GTX580 by 50% on average, I would be happy.

2) The rumoured specs came true, but the the published preview numbers for "reviewer's guide" that apoppin talked about showed 40-60% increase over GTX580. I thought it was poor form to provide Reviewer's guide #s far higher than real world performance. I understand that AMD would want to position their card in the best light but going from 40-60% estimate to just 25-30% is too large of a discrepancy for me not to comment. Luckily, apoppin addressed this in his preview and noted that AMD's estimates were not related to real world average of 25-30% and only showed best case scenarios (i.e., outliers).

And now you're posting your own speculations and want people to believe in them. :whiste:

Huh? The specs I provided here is not my speculation. It's information provided by TechPowerup and 3DCenter.org. But clearly you have not read my post. I never presented those #s as facts, especially not something I made up in my sleep. I am simply providing information I found as a rumor. Please focus on reading comprehension. Don't accuse me of making up information and presenting it as fact when I clearly said "rumor mill" in my post.

You're posting the unconfirmed rumours that were floating around for a few months already.

This thread is called "speculation thread". If you were aware of the 768-1024 CUDA cores for months, then you should have shared those with us, because I sure wasn't aware of those. I guess you have some contacts in the industry?

Creating a new redundant thread with old rumours the day new tech launches. Err I smell biasiness. ()

I missed the part where I was the original creator of this thread....:hmm:

And they didn't confirm it's a Kepler. GK100 to be precise.

GK104 and GK100 are rumoured to be "Kepler" parts. Not sure what you mean. NV never comments on rumours. So of course they wouldn't confirm that these were Kepler products.....If you have information that GK104 and GK100 are not Kepler but are Fermi parts, please link your source.

Thye have some card in the works to counter 7970 but everyone is saying the GK100 is not anywhere close to prototype.

Who is "everyone"? Do you have some connections in the videocard industry? I don't know this "everyone" crowd that has specific knowledge on Kepler. If you found additional information, share it from a reputable source.

All I know is that I read 2 websites with Kepler information, saw this thread and added information which may or may not be true. In your case, you keep derailing this thread by personally attacking me and adding no valuable information on the Kepler videocards at all.

 

[Saico]

All I know is that I read 2 websites with Kepler information, saw this thread and added valuable advice. In your case, you keep derailing this thread by personally attacking me and adding no valuable information on the Kepler videocards.

All I know, is that some Nvidia fanboys are really upset with AMD having a nice lead with their new GPU's and try really hard to spoil this launch by lowering real-world perfomance numbers for it and posting false rumours about Nvidia's future offerings. I'm not pointing my finger at anyone. No. Just FYI.

And personally for you RS - I suggest learning to use forum search to find all threads marked by KEPLER to read all old rumours you've been reposting here. The closest one is not even a page away.

Ah, yeah learn to read other people posts too because you keep refering to TPU and 3D as "reputable sources" while they were wrong all the way about 7970.

 

[boxleitnerb]

Some guy in the 3DC forums has been told, that Nvidia will quite likely drop hot clocks for the shaders. He has very good connections, so this could be believable.

That said, the number of shaders could increase manyfold over GF100/110, but have significantly lower clocks (like 1-1.2 GHz as opposed to 1.5-1.6 for GF110).

 

[Silverforce11]

The only Kepler that has a remote chance of launching within the next few months is gk104. But with NV confirming its due for late Q2 however, its chance of launching soon is next to nil.

gk104, 768 cuda cores, a node shrink gtx580. On specs it has potential for 50% gains, but we all know that it never scales perfectly with increased resources. Assume good gains of 80%, it should be ~30-35% faster than gtx580. This is the only part NV has to compete with 7970, because gk100/110 is even further away.

I had said in the other post, if NV needed a respin to fix gk104 delaying it to late Q2, and its just a "simple" die shrink of gtx580.. do you really think they won't need respins for gk100/110, a chip with 1024 cores and new kepler features bloating the die size well beyond gtx580's 520mm2, all on a new node at TSMC. Doesn't take much logic to figure kepler high-end isn't going to be here anytime soon.

If they drop the hotclock design, its even more harder to predict performance. ATM, gtx580's 512 cuda cores run at what, 1.7ghz? If gk104's 768 cores run at 1ghz, it would be slower. It needs to run at ~1.2ghz to reach parity, and >1.3ghz to claim slightly faster than gtx580.. not impossible, but initial rumors had gk104 as "slightly faster than gtx580".

 

[boxleitnerb]

Who is to say that Kepler has not taped out yet? The big one, that is. Just because no one knows, doesn't mean it cannot be true. Think of G80, that was a complete mystery until very very close to launch.

They cannot go much above 520mm2 from a business standpoint and a technological one. Max at TMSC is about 600mm2 and you never want to ride so close to the limit of what is feasible.

Everything you say could be true...but it could also be completely wrong.

NVIDIA will be launching sometime in calendar Q1 2012 (fiscal Q1 2013) the GeForce 600 series. Our sources point to a January launch, though that's probably a soft launch. Expect NVIDIA to ship product sometime in the Feb.-March window, barring a surprise.

http://www.dailytech.com/article.aspx?newsid=23578

 

[Cookie Monster]

The only Kepler that has a remote chance of launching within the next few months is gk104. But with NV confirming its due for late Q2 however, its chance of launching soon is next to nil.

Link? Where is this Q2 originating from? I dont remember nVIDIA claiming anything other than they were on track if not abit ahead of their own schedule due to supposedly having no problems with the 28nm process.

 

[Saico]

Link? Where is this Q2 originating from? I dont remember nVIDIA claiming anything other than they were on track if not abit ahead of their own schedule due to supposedly having no problems with the 28nm process.

Hey, don't pull a RussianSensation on us. The link is 5 posts above your own.

 

[Silverforce11]

Link? Where is this Q2 originating from? I dont remember nVIDIA claiming anything other than they were on track if not abit ahead of their own schedule due to supposedly having no problems with the 28nm process.

Quote: "We talked to NVIDIA today and they said they plan on releasing their next generation processor in the first half of 2012 and our sources off the record said it would be at the end of that time frame. That means AMD might have the lead for the next six month! Kuduos to AMD for being on top!"

http://www.legitreviews.com/article/1805/16/

Quote: "Nvidia Corp.’s first generation Kepler GPU should emerge in March 2012, according to reports. The GK107, first of the Nvidia Kepler series, a budget card aimed at a mainstream audience,"

Quote: "The first Kepler cards, successors to Nvidia’s Fermi architecture and based on TSMC’s 28nm process, were initially expected to emerge in late 2011 but production has been pushed back to early Q2 of 2012. First silicon is apparently still on track for this year’s end."

Quote: "Despite saying the firm was on-schedule for Kepler, however, an Nvidia spokesman seemingly confirmed the delayed timeframe set out by 4Gamer, saying that the first Kepler GPUs would only be available in OEM systems and channel desktop boards at “some point” during the first half of 2012. “We have no further details to disclose at this time,” he added."

http://www.eetimes.com/electronics-news/4230979/Nvidia-Kepler-GPUs-to-trail-AMD-s-next-generation

They may launch the gk107 soon, a budget card. This is what probably taped out in September to test the waters on TSMC's 28nm,

 

[RussianSensation]

All I know, is that some Nvidia fanboys are really upset with AMD having a nice lead with their new GPU's and try really hard to spoil this launch by lowering real-world perfomance numbers for it and posting false rumours about Nvidia's future offerings. I'm not pointing my finger at anyone. No. Just FYI.

I have owned Radeon 8500, Radeon 9700, HD4890 and HD6950. You should stop thinking there is some crazy conspiracy theory happening about why some people are disappointed about the HD7970. Certainly, some people are happy with HD7970. People who wanted to buy a $500 GTX580 just got a much better card in HD7970. Lower power consumption, 25-30% more performance, better overclocking, more VRAM, win for them all around. But just because someone isn't happy with paying $550 for 40% more performance, doesn't make them a fanboy.

I read a lot of reviews on the card and for me (not for you, or for Joe Smith), me, its performance is unsatisfactory, not in stock form, not on current drivers, not for the price. I am supposed to sing praises for this? :whiste: That's hilarious.

I wanted HD7970 to be MUCH faster not because I was going to buy the HD7970, but because that gave me an idea of how much of an upgrade HD7950 would be. And now, I probably have to skip this entire AMD generation since the performance increase for the highest card wasn't enough, which means I have no chance to get any card for $250 in the near future to get a 30-40% performance increase over my 6970....

You think that's fanboism? If you think getting 30% more performance for $550 from GTX580 is amazing, that's your opinion and no one is stopping you from buying the card.

I think you are strongly mistaken if you think I wouldn't rip this card if it was a GTX680 and provided only a 25-30% increase over a GTX580, because I would.

And personally for you RS - I suggest learning to use forum search to find all threads marked by KEPLER to read all old rumours you've been reposting here. The closest one is not even a page away.

I thought Kepler might have 768 SPs, but no way did I think it would have 1024 SPs.

Ah, yeah learn to read other people posts too because you keep refering to TPU and 3D as "reputable sources" while they were wrong all the way about 7970.

Somehow you missed the entire point if what a rumor actually means. Looks like you have no intrerest in contributing to any new information on Kepler, but instead have some axe to grind regarding my comments on HD7970's performance and "people all over the interent making up rumors on Kepler and downplaying HD7970's performance".

If you want to discuss Kepler, feel free to contribute.

 

[Rifter]

Creating a new redundant thread with old rumours the day new tech launches. Err I smell biasiness. ()

Russian did not start this thread. Reading english is obviously not your strong point. i suggest you learn how to do that before taking part in the discussion, and the attacks dont help your case either.

 

[Saico]

Russian did not start this thread. Reading english is obviously not your strong point. i suggest you learn how to do that before taking part in the discussion, and the attacks dont help your case either.

And I didn't say it was him. Obviously you can't read yourself.

 

[RussianSensation]

A couple things I wanted to add.

1) Apparently, it was already known on August 4, 2011 that NV would not have any commercially available Kepler products for sale in 2011. They said so themselves. See this article. Based on that we should probably forget about any Kepler parts in January.

Since this thread is about anything and everything we can find and consolidate on Kepler, here is more interesting bit.

2) This particular article shows information contrary to what I provided earlier from TPU and 3DCenter, even with different code-names.

"... it will take Nvidia almost a year to fully roll-out the Kepler family, according to a newly published information.

In a bid to avoid a situation with massive delay of the new family, Nvidia will start with introduction of relatively simplistic products, code-named GK107 and GK106, according to information published by 4Gamer.net web-site. Although the GK107 (128-bit memory) will support DirectX 11.1, unlike the GK106 (256-bit memory bus) it will not feature PCI Express 3.0. The more powerful GK104 will feature PCIe 3.0 and 384-bit bus, whereas the GK110 is projected to carry two of such chips. Both GK104 and GK110 will be available later than the less advanced parts. The most advanced Kepler-family chip will be code-named GK112. The product is projected to feature 512-bit memory bus. The flagship single chip solution will be the last in the Kepler 1.0 family and will be presumably released towards the end of 2012."


TPU linked information which shows that Kepler's high-end single-GPU card has a code-name GK100, with no specified launch date.

Xbitlabs states that Kepler's high-end single chip card has a code-name GK112 and will launch only in Q4 2012.

 

[AtenRa]

My estimation is,

GK100 will be launched first in Q1 2012.

Specs

1024 Cuda Cores
512-bit memory
64 Rops
1024kb L2 memory
128 TMUs
~6bil tranistors
500-550mm2 die size

Performance estimation,

With current Fermi architecture at the same clocks it could be up to 80-100% faster, i will say 50-60% on average depending on the game, resolution and settings. Compute and Tessellation performance will get the biggest boost in performance, Anti Aliasing will get a nice performance boost as well. High resolution will get much better scaling due to 128 TMUs and 64 ROPs.

 

[Saico]

My estimation is,

GK100 will be launched first in Q1 2012.

Specs

1024 Cuda Cores
512-bit memory
64 Rops
1024kb L2 memory
128 TMUs
~6bil tranistors
500-550mm2 die size

Performance estimation,

With current Fermi architecture at the same clocks it could be up to 80-100% faster, i will say 50-60% on average depending on the game, resolution and settings. Compute and Tessellation performance will get the biggest boost in performance, Anti Aliasing will get a nice performance boost as well. High resolution will get much better scaling due to 128 TMUs and 64 ROPs.

Alright. I'll save this one. It's meaningless to be here anymore. See you in april.

It won't be very wise of me to start laughing at this specs now, because if this turns out to be true I'll make a fool of myself. But but but I'll make sure to bring this up if you are wrong.