[TT] NVIDIA GEFORCE GTX 660 Ti 2GB Reference Video Card Review - Page 15

Sums it up nicely.

My thoughts exactly, Why aren't the rest of you seeing the obvious flaws of this card?? I'm not bashing it but there's far to much green eye'd stench going around in this thread

Nvidia didn't fail with their "GK110," it was probably cancelled for launch back in 2010, the GK104 was supposed to be Nvidia's top GPU for the GT 600 lineup.
I hope you've realized the GF104 could have competed with the HD 5870 back in 2010 (if clocked equally high,) but Nvidia had to release the GF100 for their Tesla line. This time around, Nvidia have taken their time releasing their Tesla-cards and didn't need to release a huge gaming/rendering/compute-monster hybrid to compete with AMDs HD 7970.

Nope indeed, nVidia didn't fail GK110, they failed GK100.

Aside from your speculations, there's tangible evidence all over the web (GTX 670Ti, nVidia admitting interconnect problems, stripping FP64 ALUs) that nVidia didn't plan to lead this generation with GK104. GK100 was dropped early in the fab stage because of poor interconnect design (high clock skew and power dissipation). The first "big kepler" we'll see is the mid-life kicker GK110.

Also, no, GF104 could most definately not compete with Cypress in 2010. Not with Fermi's broken memory controller, high leakage and clock skew. GF114 could somehow tread into Cypress and Cayman territory with high enough clocks, but that's because of the tweaked Polymorph Engines (comparable with tesselators applying z-culling), somewhat fixed memory controllers and limiting leakage by utilising high impedance transistors in latency tolerent logic blocks. All of this allowing higher clock targets, and a higher IPC (best of both worlds really).

That's also the reason why nVidia is able to counter Tahiti this gen with their second tier GPU. They did well in their execution this time around. But they needed GK100 and need GK110 rather sooner than later. No company designs a product, then decides to keep it paperware because of the lack of competition only to shorten the lifespan of the product. That's bad R&D management. You should understand that nVidia also has Maxwell and the next architecture in their pipeline and the competition isn't exactly idling around...
Edited by TheBlademaster01 - 8/2/12 at 6:03am

Enjoying this thread.

So let's say for giggles we gave some credibility to the TT results, basically we're looking at performance in the general vicinity of the 7950, give or take?

I would say below 7950 but I'll wait for a review from a site that doesn't bench ancient titles -___-

As long as you dont play with AA, and if you dont whats the point? Im not putting a lot of stock into their review though.

Wowwww... so I was going to get a 670 ftw for my new rig i'm putting together... but now i'm thinking i should just get two 660tis.... if one performs like a 560 ti sli (which i have now), i should double my performance in sli.

Yes I know it won't be exactly double.

I wouldn't go for 660ti SLI because of bad AA performance.

You would or would not get 660ti over one 670?

Also, let's just say that the 660ti comes out around the 15th like so many people are saying. How long does it usually take non reference models to come out?
Edited by Trelga - 8/2/12 at 9:15pm