The only reasonable configuration is 3 x 1GB. Otherwise you're talking about an impossible 2GB + 512 + 512 which is nuts.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
GALAXY GeForce GTX 660 Ti GC 3GB Video Card Review
- Thread starter FrgMstr
- Start date
The only reasonable configuration is 3 x 1GB. Otherwise you're talking about an impossible 2GB + 512 + 512 which is nuts.
Yea I was thinking as much but a shorter PCB length would be nice.
Thanks man, this is exactly what I was looking for. I am hoping it would somehow reach $250 shipped and I'm game. Right now Newegg got it for $300 even, which is still tempting.checkout Bit-tech's review of the Zotac 660 Ti.
Zotac GeForce GTX 660 Ti 2GB amp! Editon
http://www.bit-tech.net/hardware/2012/08/16/nvidia-geforce-gtx-660-ti-2gb-review/3
The performance is really good and the bang for the buck is actually pretty good but unless it gets down to $250 some people will probably go for a 560 Ti.
.
http://www.newegg.com/Product/Product.aspx?Item=N82E16814500268&Tpk
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
r.e. AA, Max Payne - We used 4X MSAA in the apples to apples, BF3 - We used 4X MSAA in the apples to apples and found one instance of 4X MSAA playable at 1080p, Batman - Found an instance of playable MSAA at 1080p, Skyrim - Found instances where we were able to use MSAA across the board, Witcher 2 Doesn't support any type of MSAA only it's built in Shader Based AA.
So out of all our games, we used MSAA in every game capable of supporting it, Witcher 2 doesn't support MSAA. Therefore, your statement "Also why did not test any forms of AA, other then FXAA" is entirely incorrect.
Keep in mind also we use the best ratio of AA (any form) plus in-game quality settings plus performance. Often, FXAA comes off as the best choice for AA because it reduces aliasing on everything, has near 4X MSAA quality, and no performance hit. So, in games that support FXAA it makes sense to use it and raise a resolution or raise an in-game quality setting to get the best experience rather than using MSAA and having to turn down graphics options. We aim for the BEST gameplay experience in IQ and Performance.
r.e. 660 Ti vs. non-overclocked 7950. We tested against a standard stock clocked GTX 670 and 7950 so that gamer's out there who have these cards have a baseline to compare the new 660 Ti to. We compare by price, and by price, standard 7950's are closer in price to this video card than overclocked ones are, plain and simple. There are lots of 7950's and GTX 670's out there of custom cards with custom clock speeds, we cannot cover all of them in a review. If we use one overclocked card someone will complain it was clocked too high, or something will complain it was clocked too low, there is a big range out there. Since we can't evaluate the whole range, we'll use a stock card as a baseline so you can see how it would compare to a non-modified card. If you are familiar with overclocking on these cards, then you pretty much already know how much of a percentage performance increase you get by overclocking them, so all you have to do is take a calculator and you can smartly extrapolate how much faster our 7950 or GTX 670 results would be if overclocked, and make that comparison in your head.
One other thing to keep in mind is that on launch evaluations we are introducing the GPU for the first time, and for that first time it makes sense to compare it to stock cards to get a baseline and see where it is going to fall in performance. As we do more evaluations of retail custom cards we'll use overclocked cards in those comparisons if the price and comparison makes sense. The launch evaluation is not the first and last only evaluation of cards using this GPU, per usual we will evaluate many new 660 Ti's and make appropriate comparisons that make sense for the evaluation, we will use overclocked cards in them if it makes sense based on price and performance.
We wish we could cover every card under the sun, but 6 cards took literally as long as we had to evaluate this card, our evaluation method is hands down the slowest testing method out there, it takes time to evaluate performance on each card as we do. We don't just run timedemo's and go get a sandwich, if we did that, sure, we could have 20 cards for you to compare by, but by evaluating real-world gameplay hands-on, takes time.
Last edited:
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
Not in my experience:
http://hardocp.com/article/2012/08/16/galaxy_geforce_gtx_660_ti_gc_3gb_video_card_review/3 - 4X MSAA at 1080p in MP3 at the bottom.
http://hardocp.com/article/2012/08/16/galaxy_geforce_gtx_660_ti_gc_3gb_video_card_review/5 - 4X MSAA at 1080p in BF3 at the bottom.
http://hardocp.com/article/2012/08/16/galaxy_geforce_gtx_660_ti_gc_3gb_video_card_review/6 - 4X MSAA at 1080p in BF3 on the 2GB model at the top.
http://hardocp.com/article/2012/08/16/galaxy_geforce_gtx_660_ti_gc_3gb_video_card_review/9 - 8X MSAA across the board even at 2560x1600
http://hardocp.com/article/2012/08/16/galaxy_geforce_gtx_660_ti_gc_3gb_video_card_review/10 - 8X MSAA on the 2GB model at 1080p
No breakdowns vs. 7950 or 670.
Brackle
Old Timer
- Joined
- Jun 19, 2003
- Messages
- 8,568
I find it interesting how your results differ then other websites on some of the apples to apple benchmarks.
http://www.amazon.com/gp/product/B00GJ7YP04/ref=ox_sc_act_title_3?ie=UTF8&psc=1&smid=A10RFI7E4FIYNW
http://www.amazon.com/gp/product/B003SP4QAE/ref=ox_sc_act_title_4?ie=UTF8&psc=1&smid=ATVPDKIKX0DER
http://www.amazon.com/gp/product/B001NEK2GE/ref=ox_sc_act_title_6?ie=UTF8&psc=1&smid=ATVPDKIKX0DER
http://www.amazon.com/gp/product/B00GPRW9HY/ref=ox_sc_act_title_7?ie=UTF8&psc=1&smid=A3MS1I85ZQ44M2
http://www.amazon.com/gp/product/B008IUV50S/ref=ox_sc_act_title_8?ie=UTF8&psc=1&smid=A119K31BOSN2QF
http://www.amazon.com/gp/product/B00074H6VA/ref=ox_sc_act_title_9?ie=UTF8&psc=1&smid=ATVPDKIKX0DER
http://www.amazon.com/gp/product/B003SP4QAE/ref=ox_sc_act_title_4?ie=UTF8&psc=1&smid=ATVPDKIKX0DER
http://www.amazon.com/gp/product/B001NEK2GE/ref=ox_sc_act_title_6?ie=UTF8&psc=1&smid=ATVPDKIKX0DER
http://www.amazon.com/gp/product/B00GPRW9HY/ref=ox_sc_act_title_7?ie=UTF8&psc=1&smid=A3MS1I85ZQ44M2
http://www.amazon.com/gp/product/B008IUV50S/ref=ox_sc_act_title_8?ie=UTF8&psc=1&smid=A119K31BOSN2QF
http://www.amazon.com/gp/product/B00074H6VA/ref=ox_sc_act_title_9?ie=UTF8&psc=1&smid=ATVPDKIKX0DER
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
Lol. Sorry about that; was in a rush to leave in case they sold my card or something. Thanks for the answer. But now, I'm also curious about the one to this question...:
Which I figured as well, but now that I think about it, it's not unreasonable to think there might be a catch on nVidia's end on how the memory controller behaves -- information we aren't privy to, as stated in the Anandtech review.
If it's just the most simplest scenario (1GB x 3) then I apologize and will promptly shut up.
EDIT: I have the card in my hands!
Just ordered two to upgrade from my two 560ti. It's a pretty good performance bump for only ~$300 or less out of my pocket after I sell my two 560s, and one of the borderlands coupons.
Close to the performance of one 690 with SLI 660s for a few hundred less? Yes please!
Close to the performance of one 690 with SLI 660s for a few hundred less? Yes please!
GoldenTiger
Fully [H]
- Joined
- Dec 2, 2004
- Messages
- 29,729
Tech report, guru3d, hwc, etc. All say differently. I trust places like those and hardocp over simply people picking one review from a small Site to believe wholesale. While it is certainly constrained a bit by memory bandwidth, it is a superb performer still and the price is insane. Once oc'd it gains more than the 670 does and catches up to OC 670 easily, often hitting 680 stock levels in guru3d review for example.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
you could have bought a single 7970 ghz for 250 bucks cheaper and gotten the same playable performance with more than 4x AA at 1920x1080. and if you are at 2560x1600 then you really wasted your money.
Commander Suzdal
2[H]4U
- Joined
- Jul 29, 2004
- Messages
- 2,134
My thinking behind posing the question was, sure they would put 1 GB on each channel. But NVidia has apparently applied some secret sauce to one of the channels to make 512MBx512MBx1GB work as well as possible. If Galaxy can't undo or retune those tweaks, does it mess with optimum performance? If they can retune the controller, does that make a noticeable difference in memory performance? Just curious about what's going on under the hood.
GoldenTiger
Fully [H]
- Joined
- Dec 2, 2004
- Messages
- 29,729
Go check guru 3d review of the msi oc. It runs within a couple percent both ways of a gtx 680 even at 2560x1600.
Does he mean this GPU as in the 660 ti? Or does he mean THIS GPU as in the Galaxy GC 3GB? (which sounds good to me, considering that's the one I just bought =D)
MavericK
Zero Cool
- Joined
- Sep 2, 2004
- Messages
- 31,910
Please include 3 vs 2GB version in the follow up, great review other then that
+1. It's not really clear if the 3 GB makes any real difference on its own.
But the MSi one only has 2 gigs correct? I want to know how much that extra gig of RAM helps. Because I am assuming that it should remedy the problems most people call out on the 2 gig versions. Those being, the 660Ti's lackluster AA and high res texture performance. Well at least that is what people were saying before the card was released, not it seems even the 2 gig versions can handle AA fine.
Please include 3 vs 2GB version in the follow up, great review other then that
This please. Also, Psyside from Overclock.net forums?
lol, it is I, Joemomma360
GoldenTiger
Fully [H]
- Joined
- Dec 2, 2004
- Messages
- 29,729
It's nothing to do with capacity, but rather bandwidth.
This please. Also, Psyside from Overclock.net forums?
Hi lol!
Brent_Justice
Moderator
- Joined
- Apr 17, 2000
- Messages
- 17,755
It has certainly happened before, and it won't be the last time. All I can tell you is that we play the games exactly as gamers do, there are no tricks it's fraps recording us play the game over a period of time, we used the latest drivers and latest game patches, and our real-world results are what you'd experience if you came over to my house sat down in front of my computer and played the games on these cards.
DooKey
[H]F Junkie
- Joined
- Apr 25, 2001
- Messages
- 13,576
Can't ask for more than that, Brent. Thanks for the time you put into this!
Ur_Mom
Fully [H]
- Joined
- May 15, 2006
- Messages
- 20,691
Nope, not at all. 2600K overclocked, 16 GB RAM, SSD.... Video card is the slowest and the most needed upgrade.
You said 30" monitor in your previous post ... just get a 680 or 7970. 2560 is metric ton of pixels to push.
Brent,
This is the main reason I like your reviews. Most synthetic benchmarks I just skip over. Thanks for the effort.
Some Tri-SLI numbers here. Unfortunately, they're completely useless because the driver versions are not consistent. Older cards (e.g. the 680) are presented with figures from benchmarks with older drivers. This is, of course, where [H] wins hands-down because [H] present a consistent platform in their reviews.
- Joined
- May 18, 1997
- Messages
- 55,651
And as I have said MANY TIMES, we get different results when we evaluate real world gameplay vs. benchmarks, that is why we spend the resources doing real world gameplay. If I could do a bunch of benchmarks and get the same data, I would save the money and do it that way.
So yes, I agree with you, we find it interesting how sites get differing results. That said, we stand behind our results 100% when it comes to giving you our opinions on GAMING on the cards. Benchmark monkeys are better off getting their data elsewhere.
Viper87227
Fully [H]
- Joined
- Jun 2, 2004
- Messages
- 18,017
This is exactly how it should be. As usual, kudos to you guys for going out of your way to tell people what sort of gameplay experience they should expect instead of just throwing up page after page of raw numbers and calling it good.
Kyle let me ask you this. Would you go with the 660ti or the 7950 for the future. I know the 660ti is a great performer but the specs have me on edge a bit for newer upcoming games such as crysis 3. I game at 1920x1080 and dont plan on stepping up to anything over this res.
I like the 660ti as I said & I really want to get into Physx for borderlands 2 since the Tech Demo for BL2 was amazing but I feel it might be gimped a bit in specs for higher games like Crysis 3.
I am purchasing a card tonight Im just getting my research complete before I make it final
I like the 660ti as I said & I really want to get into Physx for borderlands 2 since the Tech Demo for BL2 was amazing but I feel it might be gimped a bit in specs for higher games like Crysis 3.
I am purchasing a card tonight Im just getting my research complete before I make it final
Worry less about the specs, and more about the performance. The card slots in just about the right place for its price point, so I don't think you can say it is "gimped". Of course it is less capable than a 680 on the spec front, but it is also $200 cheaper. What do you expect?
That being said, I'd wait to see overclocked performance numbers before you made up your mind - my guess is the 7950 will pull ahead at higher resolutions, but we'll have to see.
Right thats understandable. I more then likely will grab the 660ti or a 670 only because I really want to get into Physx. That Borderlands 2 tech demo for physx was amazing and it made a comparing it to a non physx card just made it look that more awful
loafer87gt
Limp Gawd
- Joined
- Feb 2, 2005
- Messages
- 474
How so? What is garbage about this performance?
I agree, as it beat the 580 on every chart on that page. That's not garbage, lol.
cannondale06
[H]F Junkie
- Joined
- Nov 27, 2007
- Messages
- 16,180
look at the 2560x1600 graphs for yourself. its $600 in video cards basically getting matched by a $450 single 7970 ghz card. gtx660ti SLI is only 6% faster with 4x AA and only 3% faster with 8x AA. why spend 150 bucks more for that minuscule difference and put up with dual gpu issues when the cheaper single card will give you the same playable results?
the guy was asking about 2560x1600 results so SEE ABOVE.
Last edited:
Commander Suzdal
2[H]4U
- Joined
- Jul 29, 2004
- Messages
- 2,134
I think it's sad that there's such a hangover from the pre-release spin doctors that people still can't seem to accept the results that are right there before their eyes. Look at the charts. The claim that the 660Ti would be gimped by bandwidth is patently false. On EVERY game, its ability to enable playable AA is equal to the 7950 and the 670, at least on the Galaxy card. It remains to be seen whether that is the result of the factory overclock or the extra RAM, but the point remains.
Ah sorry, I don't do sli/cfx at all. Yah looks like 660Ti runs out of steam in an SLI setup for 2560.
To be expected, I think it was the 8800GT 320 that everyone was in an uproar over, then the memory limits came very relevant.