660 Ti SLI, very high money efficiency?

Hardware Hub GPU
1

Ive been looking through a lot of benchmarks recently, and the 660 Ti is popping up more and more often near the top of the list, but now there seems to be a lot of tests done with two 660 Ti's running in SLI, and the performance per dollar seems to be absurd.

With one of the closest competitors being the 680 coming in around 500 dollars,http://www.newegg.com/Product/Product.aspx?Item=N82E16814127693,  whereas 660 ti's in SLI are around 600, http://www.newegg.com/Product/Product.aspx?Item=N82E16814127696 (you might want to add another 50-70 for the more expensive PSU, but thats not what im interested in at the moment.

FPS per dollar (from http://www.xbitlabs.com/articles/graphics/display/geforce-gtx-660-ti-sli_4.html)
MSI Power Edition 2x660 TI SLI:


BF3: AF16x, AA4x, $4.95/fps

Sleepng Dogs:AF16x,AA4x, $8.95/fps

Borderlands 2: AF16x, AAOff, $5.88/fps

Metro 2033: AF16x, AA4x, $10.53/fps

Asus Gtx 680:

BF3:AF16x, AA4x, $6.09/fps

Sleeping Dogs:AF16x,AA4x, $11.90/fps

Borderlands 2: AF16x, AAOff, $6.57/fps

Metro 2033: AF16x, AA4x, $13.04/fps

All of these numbers were derived from tests done @ 1920x1080, there seems to be a less pronounced diference @ 2560x1080, but i wanted to show what most of us would be playing at.

My verdict is that if you have the extra 100 Dollars to spend, and your Mobo has 2x 3.0x16 PCI express slots, and your PSU is rated above 750-850 watts, then 2x660Ti's in SLI is a definite winner. (2x660Ti's in SLI showed a 19.3% higher fps over that of the single 680)

Note: (The performance shown with 3 660Ti's in SLI is comparable to a single 690, but that's 3-way SLI compared to a Dual-GPU card.)

2

The 7870XT is a better bang for buck, not sure about crossfire scaling though. The 660ti is the best nvidia option by far. 

3

Yeah I've read about a lot of people crossfiring those, the reason i put the work into this post was so i could show how efficient 660 TI's in SLI are, I have heard of nobody using them in SLI, and thought it was odd due to how efficient it is.

4

im gonna have to disagree im not a fan of the 660ti. only cause the performance difference between the 660 and the ti is not that great, and the price leap is a bit of a joke tbh. i personally think the 670 and 660 is the best buys from nvidia. but nvidia cards do scale well in sli when game's are optimized for it. as shown even in linus tech tips recent titan review where the 660ti actually beat the titan in a few cases. 

5

660 ti is well worth it over the 660, although I have the 660 ti I would still recommend a 7870 xt or a 7950.

6

its really not, the bus bandwidth strangles the extra cuda cores it has and in gamming bench marks it either only performs marginaly better or in some cases exactly the same. it not worth the extra mone. here's just one bench vid to show this. i can dig up more if you want.

https://www.youtube.com/watch?v=iLY5DAH9xbk

7

Crossfire and SLI are still not adopted by enough game manufacturers to make it worth it. Very few games will utalize 2 cards and even some will perform worse with 2. Just buy whatever the best single GPU you can get for the money is and it'll be far more worth it.

8

660ti in sli pretty much kicks ass of any single card in most instances. If I had the choice between a 680 or two 660ti's, I'd choose the latter.

9

Well I don't think your factoring in how much you can overclock the 660 ti compared to the 660. Also I would like to add this video which shows that the only game that the 192-bit bus bandwidth "strangles" the 660 ti is Metro 2033. Every other game is at least a 9-10fps increase and thats at stock. From what I could tell the 2 cards he benchmarked were $60 apart and for the 5fps price increase for the 670 which is $90 more than a 660 ti I don't see how you think the 670 is priced right and the 660 ti is not.

http://www.youtube.com/watch?v=HaXVrCFvL94

10

+1 I would never go for that 5-10% fps increase in "some" of the games optimized for SLI. On top of the higher power consumption and higher temperatures.

11

so you are willing to pay $60 more for 9-10fps in more at stock in some games? cause that does not exactly say im a bang for buck card vs the 660 also with assasins creed 3 they both peformed the same, and in some games like dirt 3 your only getting 6 extra frames. and actually in every game the cuda is being strangled by the bus why you might ask? cause it's a trimmed down 670 the only difference is the bus between the 2 cards, and the 670 still pulled ahead everytime. why cause it had more bandwidth to use with the exact same number of cores as the ti. nvidia should have dissabled some of extra cores on the ti cause they are not being used, the keplar arcutechiture has it's limits ya know.

also in linuses bench the lack of performance gap between the 660ti and 670 was due to cpu limitation which linus pointed out several times. and no you cant overclock the ti that much more than the 660, whats your coreclock? shaderclock and memory clock at oc? cause i bet you cant even pass the 1300mhz coreclock on that card.

so yeah i would rather pay the extra 30$ for the 670 over the ti to get full use out of everything the card has to offer unlike the ti. also why the 670 at all. cause the performance gap between 670 and 680 are lot like the 660 and 660ti performance gap other than the fact the 680 will actually outperform the 670 everytime not just half the time. for the ti and 680 they have pretty bad diminishing returns when it comes to price to performance compared to the 660 and 670.

also take a look at the 7950 and the tahiti le specs bact to back, cause you will notice they reduced bus bandwidth and the core count on the tahiti, cosidering the tahiti le is a trimmed down 7950 that means the tahiti came about the the same way 660ti did. in otherwords nvidia fucked up, with the tahiti le you get what you pay for, cause you not paying extra for something you cant use.

12

He said it twice so bit of exaggeration? and yet in Crysis 2 where he says it is cpu limited at around "90" fps which the 660 does not even reach also in Skyrim with this cpu limitation the 660 ti still manages to get an extra 12 fps.

No the difference in cost between a 660 ti and a 670 is not a mere $30 it is $90 compared to the 660 to 660 ti which is $60 and doubles the fps increase.

Yes you can get the 660 ti PE to 1300mhz but most games will crash with it so 1299mhz is really its limit.

I realise that the AMD cards at the moment are a way better deal which is why I recommended a 7870 xt or 7950.

13

several times means more than once, so no it's not an exaguration and in all the benchs you can tell the 670 is being held back by the cpu, numbers dont lie, if the oc was higher you would see more than double the frame increase from the ti too the 670 not just 50% and at no point was the ti card being held back by the cpu, which can be told by the performance gap between the ti and 670 cause if they both were being held back by the cpu you would see near identical fps them like you did with the ti and 660.

also in the uk the msi ti power edition is normally £280 and a refrence msi 670 is £290 oh and a twin frozr 670 is £300 so for a an extra £10 i could have a refrence card that would beat the ti oc'ed at stock let alone oc'd at all. the most expensive 660 is £200 and the cheapest ti £220 which means there is a bigger price gap. all that aside you have to pay an extra $90 for a 670 from the 660ti? wtf man. shop around or something. cause the last i checked everything was more expenseive over here. so yeah a 670 over a ti anyday of the week.

also at the price point of the ti where your starting to get diminishing returns on the price to performance ratio hence why the 660 is the sweet spot. atleast at the price point 670 you actually recover some price to performance and then with 680 it just goes out the window, with price to performance. which is why the 670 is the high-end sweet spot and the 660 mid range sweet spot for nvidia not the ti

14

Your trying to say the i7 3930k (overclocked to 4GHz) is "holding back" the GPU? No.

So what if the price is better in the UK? In New Zealand the 660 ti PE is $500 and the 670 reference is $630. 

Msi 660 ($239) http://www.newegg.com/Product/Product.aspx?Item=N82E16814127699

Msi 660 ti PE ($299) http://www.newegg.com/Product/Product.aspx?Item=N82E16814127697

Evga reference 670 ($375) http://www.newegg.com/Product/Product.aspx?Item=N82E16814130782

Msi 670 PE ($389) http://www.newegg.com/Product/Product.aspx?Item=N82E16814127685

 

15

there is no trying about it, it does. numbers dont lie. most games dont even use more than 4 threads iv seen it plenty of times otherwise why would people oc higher just for e penis i think not, like i said shop around, i quoted aria pc and overclockers and they weren't even discounted i wasnt even trying and i could look at other retailers i know about and beat those prices once again, there's plenty of retailers. and you just quoting one retailer say's it all really. try harder in future, and you might actually get value for what your buying

16

So surely if the numbers don't lie that means everybody will get pretty much the same results as Linus meaning that the 660 ti will beat the 660 by 9-10+ fps while the 670 will beat the 660 ti by 5fps? Your logic makes no sense.

As for saving money here you can save your $25 on the 660 http://pcpartpicker.com/part/http://pcpartpicker.com/part/msi-video-card-n660tf2gd5oc

And I'll save my $20 http://pcpartpicker.com/part/msi-video-card-n660tipe2gd5oc

17

my logic make's perfect sense, you are simply overlooking things i have pointed out, on purpose i suspect. even tiny tom logan from oc3d benches his gpu's at 4.6ghz on a 3930k. and the results are exetremely different the 670's performance goes through the roof compared to linuses. and he has a consistent track record, hes even in the 6ghz club top ten at 7th for ocing on the 775 socket. he know's what hes doing. and he knows 4ghz is not enough. so i think i will take his test result's for what the 670 can actually do, over linuses which only proves that a 4ghz oc bottles the 670, heck when you put them back to back it proves the the 670 is being bottlekneck at 4ghz.

18

So Logan's review video for the 670 is being bottlenecked by his stock i7 3820 since it has less performance than a 3930k? Yet when he overclocks the card he finds there is a significant boost in the fps and that my friend is not a bottleneck.

http://www.youtube.com/watch?v=PUBJzNGK4Pg

19

oh come off it already. even linus was getting better frames with a non oc'd 670 in most cases. cause the cpu was clocked higher, and what part of 4 THREADS do you not get. you need to learn how to read into the results better and actually draw realistic conclusions. you keep picking isolated instances where the card works well in gpu limited scnarios and in the world of science we call those anomalys. if you could actually factor in all the variables properly to make a valid point i could take you seriously. this debate is over go back to the drawingboard as im done with your incoherence.

21

Do you mean 7930?