---- I conquered the task of adding the rest of the lower Geforce 6xx cards (605, 610, 620, 630, 640 OEM, and GT 645) - most of which are rebadges. Only the 630 OEM is a cut-down GK107 version of Kepler, with a bit over half the power of GT 640 DDR3 version.

Where do you guys think the 660ti will be on the list?

I think it will be on par with 580

Just a little faster than the HD 7870

Yeah, if it's gonna have a castrated 192-bit bus, then it'd be almost impossible to beat GTX 580 - the GK104 is already severely bottlenecked with 256-bit bus. GTX 670 actually has the same memory clock as GTX 680, unlike all of the previous generations' lesser brethren before it (e.g., GTX 570 vs GTX 580, GTX 260 vs GTX 280, etc..), just so that it could blow HD 7950 out of the water as much as possible (which is what Nvidia successfully did)!

I'm all hoping for that GTX 660 Ti will still have a 256-bit bus. If 192-bit, it'd barely be able to beat HD 7870 - as long as its core is not castrated or downclocked much more than GTX 670.

-----
Anyways, notice that all of the 6xx numbers (605, 610, 620 630, etc..) are already used up - mostly as renamed Fermi products? I would guess that Nvidia is going to move ahead to the 7xx nomenclature very soon for GK11x chips, not long after releasing GTX 660. I could be wrong if Nvidia sticks to the 6xx numbering for the rest of the year, but heck, the 650 and 660 series are the only available numbers not yet filled in.

It makes me wonder if Nvidia is just going to clock GK107 sky-high with 5500Mhz+ GDDR5 memory and then call it GTX 650 (still unlikely to beat HD 7770), right before moving onto the 7xx series. The only Kepler products for the 6xx series would be GK104 and GK107 chips, right? No GK106?

Kepler already *is* a midrange part by default. There's only so much they anticipated cutting it down, or so it seems. Their yields are good. They might as well keep on pounding AMD while they have the lead.

Update: Added GTX 660 Ti, rated at 185 VP, and HD 7950 Boost Edition, rated at 210 VP!!!

Hardware.fr doesn't have a review of it yet, but amazingly Xbitlabs started to do day-0 reviews! Techreport has it up as well, along with pretty much the rest of dozens of review sites. Anandtech explains how Nvidia could have resolved the asymmetrical memory issue of GTX 660 Ti, in how it's somewhat different from the asymmetrical configuration on GTX 550 Ti.

If Kepler is midrange, wth is Tahiti?
-Less than midrange?


If you were running Nvidia, would you not take full advantage of AMD's screw ups?


And it was very difficult for me to get this Day-Zero evaluation up. i got no sleep the last couple of nights (good thing my Internet went out earlier this week or i would be in even worse shape)

The absence of GK100/110 doesn't make GK104 a high end design, it is a high end product in the current market however. Tahiti is AMD's best effort at a new design, it's their high end chip, which just happens to compete with Nvidia mid-range design.

Not the first time we've seen AMD's best effort compete against Nvidia's mid-range chip.

Yeah, the GK104 (GTX 680/670/660Ti) cards are amazing in their generational leap of per-watt and per-mm^2 efficiency - especially for the flagship single-GPU offering!

Looking at the past, GTX 460 was hardly any more efficient than GTX 285, eating maybe 85% as much power while offering similar performance. GTX 480 was not any more efficient than GTX 295, which was just as fast in some games (without factoring in any subtle forms of microstuttering that BtW wouldn't be noticeable at over 100fps on a 60Hz screen anyways). The 40nm round was a kinda disappointing round for Nvidia - HD 5970 blew GTX 480 out of the water while remaining in a similar power envelope.

Now, we're seeing GTX 680 being more than twice as fast as GTX 460, while hardly consuming any more power. The new chip is also smaller, at 294 mm^2 vs GTX 460's size of 332 mm^2. All of this at the same time, in a single generation leap, while becoming the flagship single GPU beating HD 7970 in not only performance, but in power consumption, temperature, and noise output, was something we have not seen since the 6800 Ultra and X800XT times. The 6800 Ultra ate more power, though, while X800XT performed around twice as fast as the previous-gen flagship, ate only a few more watts, and was hardly any hotter or louder. The 7800GTX-512 did manage to amaze maybe even more than the 6800 Ultra did, but it did not come out at the beginning of a new node generation - the original 7800GTX vs 6800 Ultra was maybe as impressive as HD 5870 vs HD 4890.