Nvidia Kepler (6XX-series) discussion

[Lowlie]

Die area ~340mm^2
705MHz core clock, 1411 MHz shader clock
32 ROPs
128 TMUs
2 GB memory, 256-bit bandwidth

We're looking at an equivalent to the HD 7900 series more or less. Obvously this isn't a 'flagship' worthy card in the traditional sense, but it's all Nvidia will have for half a year. I fully expect GK110 to take the outright performance crown for a single GPU video card in November or so, but I also expect AMD to come out with a Tahiti XT respin at higher clocks and maybe additional CUs at a lower price.

Those are wrong clocks btw ^^
I might have underestimated the power of this card. All i can say is everything linked to this thread so far are wrong and only pits and pieces are correct.

[DeltrusDisc]

Those are wrong clocks btw ^^
I might have underestimated the power of this card. All i can say is everything linked to this thread so far are wrong and only pits and pieces are correct.

Care to link us to a credible source then?

[Lowlie]

NDA is a bitch

All he could say was that those clocks are wrong but in what direction.. I guess we need to figure out ourselves. But i doubt its less than 705 ^^

[kidsafe]

Yeah wasn't trying to imply those specs as a fact, though I believe the 32 ROPs is correct. The 705MHz rumor comes from VR-Zone, which also says that dynamic core clocks go from 300 to 950 MHz.

[Drunkenvalley]

NDA is a bitch

All he could say was that those clocks are wrong but in what direction.. I guess we need to figure out ourselves. But i doubt its less than 705 ^^

Is "he" part of the NDA too, or would you tell us/link to who it is?

[Lowlie]

---------- Post added 2012-03-11 at 05:46 PM ----------

[/COLOR]

Yeah wasn't trying to imply those specs as a fact, though I believe the 32 ROPs is correct. The 705MHz rumor comes from VR-Zone, which also says that dynamic core clocks go from 300 to 950 MHz.

I know mate, 80% of the stuff that leaks out seems to be pure guesses so far.. The 20% facts are the obvious ones, how much memory and memory bandwidth.

Funny thing is that its the same guys over and over who leak these false numbers. In a hope of site hits i guess? But everyday they change their story and fix the "fact" benchmarks they are providing. It was so much better few years ago, when a leak actually gave a hint to at what direction to go. But now, you cant have any trust on them whatsoever just cause of few people who decide to post this fake crap.

---------- Post added 2012-03-11 at 05:55 PM ----------

I hope you managed to read what i posted earlier since i somehow removed it...

But ill say it again that

He is working on the local site that is focused on computer hardware and i dont wanna reveal anything more so i dont get him into trouble cause of me. But you can find him around San Francisco if you wanna try to squeeze more info out of him

---------- Post added 2012-03-11 at 08:04 PM ----------

Also worth mentioning, it does not have the hotclocks from fermi based cards. The clocks are wrong, i guess they are alot higher like 950 something other than that i do not know yet. Since finding out most of the rumours are fake and bullshit i have bursted into a waiting mode myself and im actually exited about this release. I hope it lives up to the expectations.

[Drunkenvalley]

I hope the NDA is over soon. I want some damned benchmarks already.

That said, the GK110 was previously rumored to be 40-45% faster than the HD 7970. Considering this is the GK104 apparently being released with the new cards, instead, I'm curious if it's going to be a tighter race.

[dicertification]

http://videocardz.com/30951/nvidia-g...ilable-for-549

If the pricing is right I imagine we'll see performance within a few percent of the 7970. If it was to smash the 7970 I'd expect a higher price, but you never know. Also a *lower TDP than the 7970 which is surprising giving Nvidia's history of pushing their hardware. Perhaps the headroom isn't there on the process currently?

edit: Just dawned on me. They may have a 680ti in the works, and the current 680 was not meant to be the highest kepler. The lower TDP just throws me off so much. Perhaps different methods of measurement?

[noteworthynerd]

I hope the NDA is over soon. I want some damned benchmarks already.

Ugh, me too. Like I mentioned elsewhere, I'm not expecting anything amazing, but it would be nice to see NVIDIA get back in the game, they've been slipping lately. =P

[Drunkenvalley]

Honestly, if nVidia's cards aren't more than a few % ahead then I'm more than a little buzzkilled. >.>

[Lowlie]

It will be tighter race than 40-45% but not the 3-6% i was speculating before. Now im sure it will be faster but not by that much. 3-12% sounds more realistic to me. Also, dont start staying wake all nights in hopes of NDA to stop just yet o.-

http://muropaketti.com/toimitus-ja-b...ssa-nvidian-ke

Gotta say i like nvidia style though. AMD gave away belt puckle with their bulldozer test chip, thats gotta be a reference to something which i hope you will understand, since i dunno how to explain it in english

[dicertification]

Giving away cookies....

I'm sorry but that's just lame. As lame or lamer than the belt buckle thing.

[Drunkenvalley]

Giving away cookies....

I... guess they think the fans are... cute...?

[Lowlie]

I hope there is a cookie on the retail box, or im gonna skip this series and never look nvidias products again.

[Rennadrel]

http://videocardz.com/30951/nvidia-g...ilable-for-549

If the pricing is right I imagine we'll see performance within a few percent of the 7970. If it was to smash the 7970 I'd expect a higher price, but you never know. Also a *lower TDP than the 7970 which is surprising giving Nvidia's history of pushing their hardware. Perhaps the headroom isn't there on the process currently?

edit: Just dawned on me. They may have a 680ti in the works, and the current 680 was not meant to be the highest kepler. The lower TDP just throws me off so much. Perhaps different methods of measurement?

Considering Nvidia is always more expensive then AMD, whether their products are drastically more powerful or not, their prices will be higher. Besides, AMD could just pull the rug out from under Nvidia's feet and price drop all their current models.

[kidsafe]

Considering Nvidia is always more expensive then AMD, whether their products are drastically more powerful or not, their prices will be higher. Besides, AMD could just pull the rug out from under Nvidia's feet and price drop all their current models.

AMD has no reason to drop prices if the Nvidia card debuts at $550. I have no reason to believe it will outperform an HD 7970 when it has a smaller die and only a 2GB/256-bit framebuffer. If anything the GTX 680 should be the one that's cheaper.

[Synthaxx]

AMD has no reason to drop prices if the Nvidia card debuts at $550. I have no reason to believe it will outperform an HD 7970 when it has a smaller die and only a 2GB/256-bit framebuffer. If anything the GTX 680 should be the one that's cheaper.

If these specs are even final. Everything is still up in the air. Until NVIDIA openly comes out with specs, nothing is a given.

Also, 2GB VRAM is still a massive amount even these days. There's certainly only a handful of games that will use that much. Again, IF those specs are real, they've changed their design from having a smaller amount of smart cores (compared to AMD having lots of dumb cores).

For the purpose of gaming, particle effects should see a nice improvement (traditionally not where NVIDIA excels), but shader effects may see a dip in performance (traditionally a strong point for NVIDIA). Texture performance should increase if they've actually increased the number of TU's.

Overall performance is still going to be limited by the CPU though, except in graphical-only data representations (which typically aren't done in a 3D world in the same manner as a game) and games where there's very little AI or logic running in the background, which is limited almost exclusively to Indie titles and benchmarks, neither of which were good measures of actual gaming performance.

[tetrisGOAT]

edit: Just dawned on me. They may have a 680ti in the works, and the current 680 was not meant to be the highest kepler. The lower TDP just throws me off so much. Perhaps different methods of measurement?

What I'm confused about, is what card is even 'supposed' to be released on March 23rd. GTX 670 Ti or GTX 680? I hear very conflicting things about both, and their names just randomly slapped on whichever fits best atm.

I've only seen the leaked pictures on the GTX 670 Ti.
But yes. I also believe nVidia has more powerful things, since I read somewhere that the highest end cards would be released toward the end of the year.
As well as AMD potentially having their new series out by then as well.

[Drunkenvalley]

Don't recall who or when (and too lazy to dig now), but someone proposed that nVidia is going to release cards based on the GK104 first as their 600 series and save the GK110 for the 700 series, meanwhile AMD are also probably throwing something into the mix.

[DeltrusDisc]

Considering Nvidia is always more expensive then AMD, whether their products are drastically more powerful or not, their prices will be higher. Besides, AMD could just pull the rug out from under Nvidia's feet and price drop all their current models.

Waaaait a minute. Where exactly has AMD been always cheaper than NVidia? I'd say they're about even, the only grounds we really have to even ponder this idea is the 580, but that's a special card that AMD simply never made competition for, the 6970 and 570 are about evenly matched, same with the 560 Ti and 6950.

Then you have the 590 and 6990, the 6990 was slightly better, but they were still about evenly matched, as I recall, or I think the 6990 may have been more expensive.

Hell searching 6950 on newegg and 560 Ti, seems like AMD's card is a little more pricey.