GTX 660Ti Reviews

[chimaxi83]

Talking about the bundled Borderlands 2.

PhysX in action:
http://www.youtube.com/watch?v=ZIu23UjKGOk&feature=player_embedded

I like PhysX (good video showcasing it), but not enough to only buy Nvidia, or keep me from going back to AMD (which I did). I'll just throw my old, dusty 460 in and keep the effects.

 

[RussianSensation]

Pulling an "80%" number out of your ass, just like most every other statistic, is meaningless.

You can't be serious? JHH himself stated 80% of NV users don't tweak GPU settings in games/NV control panel during GTX690 launch feed. That's the whole reason he said his firm was working on GeForce Experience It's not a # I pulled out of thin air.

Not even in the same league.

Soon GTX660Ti will be faster than a 680...hehhe.

670 is on average 15-16% faster:

and
http://www.computerbase.de/artikel/grafikkarten/2012/test-nvidia-geforce-gtx-660-ti/7/

 

[Phynaz]

Does Nvidia have anything like AMD's edge-detect AA?

I was really hoping power consumption would be lower - that's my number one decision point as I'm held back by a 450w PSU (that I'm not going to replace) and I just don't like the heat.

Finding it difficult to jump on this card at this price. Although considering how long I keep my cards (a long time) the extra $50 may not make much difference.

Blah, even $20 cheaper and the decision would have been easier.

 

[RussianSensation]

What brand PSU? How many amps are on the 12V rail? What cpu do you have/Oced?

450W PSU should be fine with 660Ti/7950. 660Ti is efficient, using < 150W at peak, with average coming closer to 134-136W.

 

[toyota]

Does Nvidia have anything like AMD's edge-detect AA?

I was really hoping power consumption would be lower - that's my number one decision point as I'm held back by a 450w PSU (that I'm not going to replace).

Finding it difficult to jump on this card at this price. Although considering how long I keep my cards (a long time) the extra $50 may not make much difference.

Blah, even $20 cheaper and the decision would have been easier.

you have no specs in your sig but that psu may be plenty. list your specs and what the exact power supply is that you have.

 

[f1sherman]

@RS we all know that underdog doesn't need to screw-up big time to find themselves in trouble, especially when market-leader executes well.

Tahiti is more overweight than Kepler, but it's very lean compared to NVIDIA total-GPGPU attempt i.e. Fermie

I am actually amazed that AMD did pack all their overhead in only 350mm2.

 

[blastingcap]

Harvested chips. Yawn.

 

[blastingcap]

GTX460 was a $199 card and 2 years ago NV considered it a Gamer's Sweet Spot:

Now GTX660Ti is NV's sweet spot and it's $299. AMD thinks the sweet spot is $275 7870 -- $330 7950. That's pretty expensive for mid-range cards from the last 2-3 generations.

It looks like the GPU makers are passing on the costs on to us this round. It seems both AMD and Nvidia now position graphics cards within the $275-330 price-range as the sweet spot for gamers. Previously, the sweet-spot for gamers was something between $149 and $249 ($199/229 for the 460 768/1GB and $179/$239 for 6850/6870). It remains to be seen whether gamers will actually start to buy more expensive solutions to play the latest titles or will wait for price-cuts or introduction of new products. To me it looks like both AMD and NV are trying to push the mid-range GPU pricing upmarket.

http://www.extremetech.com/computin...y-with-tsmc-claims-22nm-essentially-worthless

Look at these charts. The picture is clear. Gone are the days of big jumps in price/perf with every new node, at least until a disruptive technology comes around. The move to larger wafer diameters is nice, but it's a once-in-a-long-time kind of deal. After that we'll be stuck on that wafer size for a long time. Can't count on wafer sizes to help price/perf consistently.

 

[Phynaz]

What brand PSU? How many amps are on the 12V rail? What cpu do you have/Oced?

450W PSU should be fine with 660Ti/7950. 660Ti is efficient, using < 150W at peak, with average coming closer to 134-136W.

you have no specs in your sig but that psu may be plenty. list your specs and what the exact power supply is that you have.

Thanks guys.
PS is whatever HP was using in their Elite series at the time. I've never checked the OEM.

Currently driving a stock i7-870, a stock 5770, and a couple of spinners. If it helps, the 260GTX was an option on this PC, which I believe was in the 145w range?

My plan for mid-September is to get a U2412M and something to drive it. I wanted to try a CUDA card this time around to see how it works with the video apps.

 

[AnandThenMan]

http://www.extremetech.com/computin...y-with-tsmc-claims-22nm-essentially-worthless

Nvidia is always complaining about TSMC I would not put too much stock into them doing it again. Meanwhile AMD will probably have 22nm graphics products first if the long held pattern continues.

But on the Nvidia customer loyalty discussion, I see it ALL the time. People either have no clue who AMD is, or consider them low end or unreliable, they really have no idea what AMD is all about. Can't blame Nvidia for that, AMD has gawd awful marketing, in some cases the marketing they do is worse than none at all.

 

[zaydq]

^ true story.

 

[RussianSensation]

@RS we all know that underdog doesn't need to screw-up big time to find themselves in trouble, especially when market-leader executes well.

Tahiti is more overweight than Kepler, but it's very lean compared to NVIDIA total-GPGPU attempt i.e. Fermie

I am actually amazed that AMD did pack all their overhead in only 350mm2.

I said it before and I'll say it again, I think Fermi was underrated. Take Fermi GTX580, apply 28nm shrink to it and it'll be a wonderful card. Obviously I have no way to prove it but Fermi as an architecture was excellent. It didn't have many weaknesses other than TMUs at high resolutions. It's main liability to me was that NV went all on GPGPU + gaming in on 40nm node.

If you think about it Kepler really is just Fermi enhanced. NV didn't do much to revise Fermi with Kepler. They chopped off shader clock, gutted GPGPU/dynammic scheduler and applied 28nm shrink to control leakage and thus power consumption. That 1.4-1.6Ghz shader clock on Fermi was a liability (as Pentium 4 taught us that high frequency is the enemy of power consumption). Then all they did is compensate for the clock speeds with many shader cores, upped the TMUs because gutting GPGPU provided additional space for more functional GPU units. Kepler isn't some revolutionary beast, but just a rebalanced Fermi architecture into a streamlined gaming card and it happens to benefit from 28nm node shrink.

Global Foundries sites, "28nm transistors offer up to 60% higher performance than 40nm at comparable leakage with up to 50% lower energy per switch and 50% lower static power." ~ Source

I can't see TSMC being much worse. Can't discount the massive benefit a full node shrink provides. NV set themselves up nicely since with 294mm^2 die and experience building 500mm^2+ chips, they have ample room to add more functional units and performance. AMD has a monumental task with HD8000 series. They have already ballooned to 365mm^2. There are some theories of how AMD may improve performance with HD8000 series (Use Google Chrome to translate) but from what I read they all point to even large die size. I am not an engineer but it sounds very difficult to contain HD8000's power consumption, while increasing die size and adding even more functional units. Maybe AMD is counting that 28nm node will mature and they'll be able to have a better stepping on which to build HD8000 series.

What's impressive is how NV made a 180* turnaround in just 2 years since Fermi. In 1 generation, they are now leading in performance/watt, performance/mm^2 of die space for gaming. Now it is AMD who somehow has to figure out a way to build a chip that works in compute and gaming. If NV decides to drop 500mm^2 gaming Kepler, AMD will have no shot unless they drop GPGPU. The question is will NV drop 2880-3072 SP GK110?

http://guru3d.com/news/gk110based-surfaces--has-2880-shader-processors/

GK 110 -- "Big Kepler" specs based on K20:

7.1 Billion Transistors
2880/3072 Shader processors :wub:
15/16 SMX (shader engines with 192 Shader-ALUs)
384-Bit wide GDDR5 memory bus
6-Pin & 8-Pin power connectors for Tesla K20-cards
3x higher double precision

 

[railven]

Props to nVidia for delivering yet another solid product. The $300 crowd should be happy now.

The BL2 bundle should be icing on the cake!

 

[f1sherman]

@RussianSensation

If I remember correctly ideal direct Fermi shrink would give something like 580 old cuda cores. But I read once that there is no such thing as direct shrink

So it's certainly not obvious what it would end up like.
Other than, while saving a die area - hotclock was indeed a TDP liability.

 

[RussianSensation]

TechReport has a couple nice summary charts:

GTX660Ti Zotac AMP! looks much smaller than 670. Looks like an awesome card for a very small case. The circuit board is under 7" long, and the card measures 7.5" to the pointed tip of its cooling shroud. That's impressive.

 

[Ajay]

"The best case scenario is always going to be that the entire 192bit bus is in use by interleaving a memory operation across all 3 controllers, giving the card 144GB/sec of memory bandwidth (192bit * 6GHz / 8). But that can only be done at up to 1.5GB of memory; the final 512MB of memory is attached to a single memory controller. This invokes the worst case scenario, where only 1 64-bit memory controller is in use and thereby reducing memory bandwidth to a much more modest 48GB/sec. If we had to pick something, on a pure performance-per-dollar basis the 7950 looks good both now and in the future; in particular we suspect it’s going to weather newer games better than the GTX 660 Ti and its relatively narrow memory bus." ~ AT's review.

So this is a really a 1.5GB card, not a 2GB card.

Fiddle sticks! I thought 2GB was going to have some drawbacks. A non-overclocked version w/only 1.5GB @ $249 would have been a better option, IMO.

I hope NV allows AIB makers to release 3GB models. I'd be much happier with the longer term prospects of such a model, especially since I'd like to upgrade my monitor next year.

 

[notty22]

Fiddle sticks! I thought 2GB was going to have some drawbacks. A non-overclocked version w/only 1.5GB @ $249 would have been a better option, IMO.

I hope NV allows AIB makers to release 3GB models. I'd be much happier with the longer term prospects of such a model, especially since I'd like to upgrade my monitor next year.

There are 3GB models, Hocp tested one.

We also experienced the GALAXY GTX 660 Ti GC video card providing for the most part better performance than a Radeon HD 7950. While the Radeon HD 7950 can currently be found for around $319 with MIR, which is about $20 cheaper than the GALAXY GTX 660 Ti GC, it doesn't seem worth it just to save $20. The GALAXY GTX 660 Ti GC is a much better value than the Radeo HD 7950 at $339. In all our gameplay the GALAXY GTX 660 Ti GC provided a better gaming experience than the Radeon HD 7950.

 

[RussianSensation]

They'll have 3GB models, but they'll be $339.99 such as this galaxy version. At that point you may want to wait for a $350 deal on the 670.

Notty22, that conclusion is odd to say the least. They tested a $340 + $7 shipping 660Ti 3GB vs. an 800mhz 7950 and concluded that a factory pre-overclocked 660Ti is faster? For $350, you can get an HD7950 with 950mhz clocks or for $370 an HD7970. I am pretty sure a $340 3GB GTX660Ti doesn't look so hot compared to either of those. That's not even taking into account that a $317 MSI TF3 will clock to 1.1ghz+ and has 3GB of VRAM standard.

800mhz 7950 is slow

925mhz 7950 is fast

1025mhz 7950 is very fast

 

[cmdrdredd]

That is why HardOCP is a joke IMO.

 

[Attic]

That is why HardOCP is a joke IMO.

I generally like the take from those guys, but sometimes they are way off. This is one of those times.

 

[RussianSensation]

That is why HardOCP is a joke IMO.

Last time HardOCP 'cremated' Fermi saying it's super hot and loud. They supposedly were hitting > 90*C at 100% fan speed on them at stock speeds on an open test bench. I got home, fired up my reference 470, cranked voltage to 1.087, 760mhz in 5 min of Crysis runs. Didn't hit above 81*C max at 760mhz at 70% fan speed, that's before changing the TIM and high-flow bracket installed. Ever since then I don't trust their reviews half the time. I got my 470s right around June 2010, so at that point the node wasn't even mature yet. So it's not like I got some cherry-picked Fermis.

 

[f1sherman]

Really?

I trust every single one of them out there.
S|A and kitguru included

 

[f1sherman]

BTW... I am disappoint.

Leaving out the mother of all reviews

 

[RussianSensation]

BTW... I am disappoint.

Leaving out the mother of all reviews

Wow that's a great summary. But check out what happened when they compiled all the averages:

1920x1080 4AA
660Ti = 100%
7950 = 101.5%

1920x1080 8AA
660Ti = 100%
7950 = 108.5%

2560x1600 4AA
660Ti = 100%
7950 = 108.6%

2560x1600 8AA
660Ti = 100%
7950 = 115%

So 660Ti is overall slower than a 7950.

Oh oh, the MSI TF3 $317 "Overclocker's Dream" is looking better than ever. Our forum members are hitting 1200mhz on this bad boy. That's a 50% overclock on top of a stock 7950 / 660Ti's speed! 3GB of VRAM comes standard.

 

[notty22]

BTW... I am disappoint.

Leaving out the mother of all reviews