tweaktown review GTX660Ti

[exar333]

...lose? Toms Hardware has always given reputable results... I dont need justification for my 7950, i've justified it by outdoing a friend's GTX 670 in all benchies and average frames. The 660Ti is absolute trash for the price of $300. My 7950 was $309.99 and overclocked to outperform 670s and 680s. Even when the 660ti is OCd its still not that much better than it was stock. Think what you wish, i preach truth.

Also, lets be honest, even Nvidia fanboys can admit the 7970 is a fast card... so a $300 midranged nvidia gpu is superior in all benchmarks according to the graphs you just posted. Really...?

Thats really sad that you justify a GPU purchase by 'out doing' a friend's rig. How about justifying the purchase by utilizing it for games or GPGPU stuff (you know, what it was made to do...).

 

[Hypertag]

what coupon code?

http://www.newegg.com/Product/Produc...82E16814130782

 

[toyota]

http://www.newegg.com/Product/Produc...82E16814130782

oh its just that one card and it has the crappy reference cooler. :thumbsdown:

 

[KompuKare]

Can't see any reason at all to pick this over a solid 7950 such as the MSI TF3 unless someone doesn't OC at all and cares about performance/watt.

Without OC, performance/watt is almost identical at 1080P:

and the at 1600P, the 660Ti is behind the 7950:

Of course, overclocked to the max a 7950 is likely to consume more but then it's going to run faster too.

HT4U.NET got the following (wonder are they going to update their review since they missed the power consumption overclocking results):
______________660TI_____7950
Game (Hawx):__147.85W___156.58W
Furmark:_______147.85W___207.82W

Which shows that for their game load, the difference is very little. Strange is that Nvidia's power control seems to be very accurate unless they made a mistake in the Furmark results: the exact same to result to two decimal places.

 

[RussianSensation]

KompuKare, ya they are close at High settings. Put up settings to the max and turn up MSAA and this card falls apart vs. the 7950 and compared to the 670 it's miles apart with Ultra quality settings. It's not wonder NV cut the price $100 vs. the 670. Interesting how Xbitlabs, Tom's, Computerbase all show the MSAA deficiency. This card's performance doesn't seem consistent among reviews which is odd to the least. Makes you wonder who actually used MSAA and who didn't.

This card does well in BF3, that's about it.

Good to see Tom's didn't fall for NV's FXAA/no MSAA marketing wool.

 

[zaydq]

Thats really sad that you justify a GPU purchase by 'out doing' a friend's rig. How about justifying the purchase by utilizing it for games or GPGPU stuff (you know, what it was made to do...).

Its really sad that someone on this forum actually IN PERSON put the two together and conducted a seried of tests to prove a point. I guess its really sad that review sites tend to the exact same thing. I did it to prove a good buy in my 7950 thread. Oh, just FYI, i do use it for what it was intended for .

Sorry TomsHardware, Anandtech, hardware canucks etc etc etc, its really sad you guys do hardware reviews and compare video cards with others to see which ones are best.

 

[zaydq]

KompuKare, ya they are close at High settings. Put up settings to the max and turn up MSAA and this card falls apart vs. the 7950 and compared to the 670 it's miles apart with Ultra quality settings. It's not wonder NV cut the price $100 vs. the 670. Interesting how Xbitlabs, Tom's, Computerbase all show the MSAA deficiency. This card's performance doesn't seem consistent among reviews which is odd to the least. Makes you wonder who actually used MSAA and who didn't.

This card does well in BF3, that's about it.

Good to see Tom's didn't fall for NV's FXAA/no MSAA marketing wool.

Don't use Toms Hardware benchmark graphs, Nvidia fans are so butthurt by it .

The only logical reason Toms has such different results is because, just as you said, they don't fall for FXAA/no msaa marketing ploy and they play it on a level playing field.

 

[blackened23]

It is pretty clever that nvidia sent aftermarket cards to all the websites - so most people are making judgement calls based on overclocked 660 cards.

 

[OILFIELDTRASH]

Why is everyone ignoring the fact that the majority of 660 reviews are benching it alongside the 570 instead of the 580?

 

[ShintaiDK]

I always love the "select the reviews with the worst numbers" and then proclaim it was the only thing worth reading

It never fails.

 

[Olikan]

It is pretty clever that nvidia sent aftermarket cards to all the websites - so most people are making judgement calls based on overclocked 660 cards.

...yeah, it's remembering dragon age 2 reviews

 

[zaydq]

I always love the "select the reviews with the worst numbers" and then proclaim it was the only thing worth reading

It never fails.

how else do you get such low numbers if all the other sites dont? They're doing something right that brings out the cards weakness...

 

[RussianSensation]

I always love the "select the reviews with the worst numbers" and then proclaim it was the only thing worth reading

It never fails.

That's funny that you say that. The minute someone here links reviews with MSAA you call them "select reviews" but people spending $300 on a card won't use MSAA or consider overclocking? Pretty much on average most reviews agree that this card trades blows with a 7950 at stock speeds, but once you use MSAA, up the resolution or consider overclocking, it's slower.

I think it's fair to provide information that can benefit people. For example, at high resolutions 660Ti has no power left. Even at 1.3ghz overclock, it can't even beat a stock 7970 or a GTX670. That's not cherry-picking, that's informing people who want this card for 2560x1440/1600 that it's not great. HD7950 debuting at $450 was also lackluster if you remember.

Source

The funny part is that it's the Europoean reviews that are most exposing the weakness of the 660Ti with MSAA and you call all of them "select reviews". This card is good for the $300 price, but it isn't a GTX670/7950 killer.

Why is everyone ignoring the fact that the majority of 660 reviews are benching it alongside the 570 instead of the 580?

Probably because it seems like there was barely any progress in the last 20 months on the NV side. GTX580 is still fast because it has strong tessellation performance, has 192GB/sec bandwidth and plenty of ROPs. This makes 660Ti look poor in certain cases. Granted the 660Ti is way cooler, consumes less power and costs $300. So in that case it is an improvement over the 580, but considering it's been 20 months, this generation is not setting the world on fire.

 

[exar333]

how else do you get such low numbers if all the other sites dont? They're doing something right that brings out the cards weakness...

I have to agree here, with one point to add.

Users who do not plan to crank-up MSAA settings will likely see this as a very good card. That would apply to most people who are purchasing this for absolutely cutting-edge games that otherwise would just not cut-it at all with acceptable framerates with high MSAA enabled. Another group would be those who plan to play on large multiplayer maps (like BF3)where performance can really take a hit with a lot of people in the mix.

I definitely don't think its a 'bad' card, but compared to the new priced 7950, its not as good. Again, drop the price $30-50 and its a different story. A little less performance for a decent amount less $$$ (660ti) or more performance for more $$$ (7950).

 

[ShintaiDK]

how else do you get such low numbers if all the other sites dont? They're doing something right that brings out the cards weakness...

Ye that must be it. It cant be the other way around, it must not be the other way around!

 

[ShintaiDK]

That's funny that you say that. The minute someone here links reviews with MSAA you call them "select reviews" but people spending $300 on a card won't use MSAA or consider overclocking? Pretty much on average most reviews agree that this card trades blows with a 7950 at stock speeds, but once you use MSAA, up the resolution or consider overclocking, it's slower.

Sure sure mr cherry picker. Yet your 300$ card buyers obviously all play on 27-30" screens or tripple monitor setups while overclocking.

Not a fun one eh?

 

[Arzachel]

Sure sure mr cherry picker. Yet your 300$ card buyers obviously all play on 27-30" screens or tripple monitor setups while overclocking.

*image*

Not a fun one eh?

Now jump on a larger mp map and watch the 660ti numbers take a nose dive. The card has 1.5gb ram, with the other 512mb slapped on just for looks. I wouldn't be shocked that it's actually capped at 1.5gb in the drivers.

 

[toyota]

Now jump on a larger mp map and watch the 660ti numbers take a nose dive. The card essentially has 1.5gb ram.

I think you guys are worrying about the odd memory configuration too much and are making inaccurate assumptions. I have that same odd memory setup with a 192bit and 1gb of vram and my performance does not tank when I go over 768mb of vram. also if I oc my 288sp card to 950/1900 to match the 810/1620 336sp regular gtx560 card I can basically match it in every benchmark.

 

[Silverforce11]

Singleplayer BF3 bench = irrelevant, never has been relevant, period.

This card is great for 1080p and 4x MSAA, which is good enough for the $300 market segment.

It's weakness is 8x MSAA and higher res, but so what, these ppl who buy mid-range cards aren't going to be using such settings. It will sell very well and if demand is high, probably cannibalize NV's sales of the 670/680, remember, same gk104 selling for $300 << gk104 selling for $400/500.

Ppl who buy 7850 or 7950s do so because they are overclockers, these NV release isn't likely to draw them away, because OC 7950 for ~$300 is STILL the best perf/$, and by a large margin.

 

[Arzachel]

I think you guys are worrying about the odd memory configuration too much and are making inaccurate assumptions. I have that same odd memory setup with a 192bit and 1gb of vram and my performance does not tank when I go over 768mb of vram. also if I oc my 288sp card to 950/1900 to match the 810/1620 336sp regular gtx560 card I can basically match it in every benchmark.

Well I was going off the statement that the memory bus is limited to 64 bits once it goes over 1.5gb, cause that would absolutely tank performance. But the first part still stands, larger mp maps would stress the bottleneck far more than singleplayer benchmarks.

 

[RussianSensation]

[/IMG]

Not a fun one eh?

How did you miss Post #455 where I picked 1920x1080 games? I linked SKYRIM because it was to counter the notion that this card doesn't take a hit with AA or the 24 ROP/192-bit bus doesn't negatively affect it. It does affect it just not in all games. Furthermore, someone said that 660Ti SLI setup would work great for 2560x1440 and again those benchmarks I was linking for that purpose.

I already linked Computerbase (average), Xbitlabs (average), Tom's Hardware (average) and you dismissed all those despite me/other people linking 1080P benchmarks.

Computerbase has 7950 and 660Ti neck and neck at stock speeds at 1080P. I didn't say that's not true. The difference is I don't ignore that HD7950 can overclock from 800mhz to 1.1ghz and match a GTX680. On that account, GTX660Ti is going to lose. If you don't overclock, well it doesn't matter for you then.

 

[mundosold]

I really don't get the appeal of this card. Seems like the new 7870@$350 like it was a few months ago--too expensive for a midrange card, performance too low to be high-end. Almost no OC headroom, severely gimped when running what you'd expect a $300+ card to run (high real AA at 1080p).

Really I think 7850 wins at the $230 range, 7870 at the $270-300 range, 7950 at the $330 range, and it is getting even debatable about the 670 at the $400 range when 7970's are available at around $370-400 these days. And I am not an ATI fanboy in the slightest (I was quite pissed at them with the old CrossfireX 5770 setup thanks to the awful drivers, though their drivers are tolerable for single cards. Was also annoyed how overpriced Pitcairn was when it launched).

ATI is still sitting pretty and I wouldn't be surprised to see retailers all increase the price of the 7870 back to $300 AR.

 

[zaydq]

Sure sure mr cherry picker. Yet your 300$ card buyers obviously all play on 27-30" screens or tripple monitor setups while overclocking.

Not a fun one eh?

Lets totally use AIB 660tis to prove points and reference 7970s/7950s running old drivers shall we?

Lol

I also use my PC on a 42" 1080p television. My $300 card is running my huge monitor with no issues. So your point?


Also, did anyone notice Tom's is the only one to use ref designs all across their benchmark charts. They didn't use AIB 660s against ref boards for other model cards. HMMMM?

 

[ShintaiDK]

Lets totally use AIB 660tis to prove points and reference 7970s/7950s running old drivers shall we?

Lol

I also use my PC on a 42" 1080p television. My $300 card is running my huge monitor with no issues. So your point?


Also, did anyone notice Tom's is the only one to use ref designs all across their benchmark charts. They didn't use AIB 660s against ref boards for other model cards. HMMMM?

Look again, reference design is there.

 

[zaydq]

Look again, reference design is there.