GALAXY GeForce GTX 660 Ti GC 3GB Overclocking Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,634
GALAXY GeForce GTX 660 Ti GC 3GB Overclocking Review - We've evaluated the GALAXY GeForce GTX 660 Ti 3GB video card, now it is time to overclock it to its maximum potential with XtremeTuner Plus and find out how it compares to the GTX 670 and Radeon HD 7950. We will also find out if it is best to concentrate on the GPU clock speed or its 192-bit memory speed to get the best performance gains.
 
Oh man this makes me so happy. I can not wait for mine to come! It's coming tomorrow! I'm going to try to get it to 1250Mhz and 7.6 mem clock just as they did. Gonna wreck a 670 for 100 bucks less! I'm sure the 7950 and 670 will easily make up the difference with OCing but who cares, this card is a beast.
 
Oh man this makes me so happy. I can not wait for mine to come! It's coming tomorrow! I'm going to try to get it to 1250Mhz and 7.6 mem clock just as they did. Gonna wreck a 670 for 100 bucks less! I'm sure the 7950 and 670 will easily make up the difference with OCing but who cares, this card is a beast.

I hope you lots of luck with your new card and hope it is an overclocking demon. :)
 
This should clear doubts for many people considering a 660ti over a 670
Massive OC on the memory. AFAIK most 670 and 680 top at 7ghz or less even with overvolting
 
This should clear doubts for many people considering a 660ti over a 670
Massive OC on the memory. AFAIK most 670 and 680 top at 7ghz or less even with overvolting

I wouldn't generalise based on this mini review. A lot of other reviews on OC edition 660Ti cards aren't being so lucky. Getting a card with excellent OC ability is the exception and not the rule. My advice when buying a new GPU is to get an average OC based on multiple reviews and assume you will be OK if your card gets in the right ballpark. Anything over this average is a bonus.
 
Ah a sucker punch

"GALAXY GTX 660 Ti is a whopping 28% faster than the stock clock Radeon HD 7950"
 
So the 660 really *is* memory bandwidth limited, ...very nicely illustrated here by the [H] team, a stirling job done again. Props for yous guys.
 
I wouldn't generalise based on this mini review. A lot of other reviews on OC edition 660Ti cards aren't being so lucky. Getting a card with excellent OC ability is the exception and not the rule. My advice when buying a new GPU is to get an average OC based on multiple reviews and assume you will be OK if your card gets in the right ballpark. Anything over this average is a bonus.

I didn't meant to say all cards would do the same. I'm surprised such a high memory overclock without mods, even more so on a 3gb card.

From what I've seen, many sites are getting 7+ ghz on the memory, and I think the Zotac card runs "stock" at 7ghz.
 
So the 660 really *is* memory bandwidth limited, ...very nicely illustrated here by the [H] team, a stirling job done again. Props for yous guys.

I wouldn't say that, in the past cards that have been bandwidth limited have seen massive gains from clocking up the memory. This card shows a pretty small 10% gain from 30% more bandwidth.
 
I bought an XFX 8800 GTS due in part to some fancy overclocking that Hardware Canucks managed on the card and it turned out to be anything but. So how can we really know that the card sent to HARDOCP was not cherry picked as well? I liked the showing of both resolutions (1920 x1080/1200 and 2560 x 1600); however, when it counted the most you did not display the (1920 x 1080 bench) on the GPU and memory comparison page. Also, no one in their right mind would run a card that has only 2GB of frame buffer on a $1200 30" panel, that is why the 1920 x 1080 findings on the last page comparison was so crucially needed. And in my reference to tweaking, I mean comparing an overclocked card to a higher level card that is ALSO overclocked. Note: The 670 has a bios out that allows the core locked GPU to go from 1175 Mv to 1212 Mv or if you prefer 1.175v to 1.212v.
http://www.youtube.com/watch?v=fR8656az8sU&feature=plcp
Does the 660 Ti have this as well? These 3Gb 660 Ti's just might end up being the best serious gaming bang for buck cards out in sometime, thanks for the review. SLI for mainstream ? : )
 
So how can we really know that the card sent to HARDOCP was not cherry picked as well?

We take a very hard line with manufacturers when it comes to this. Most of these guys know better than to send us cherry picked non-retail hardware. The Galaxy card we reviewed came to us out of retail stock.

If you go back and look at GPU overclocking results in our reviews you will see that we usually tend to hit "average to high" overclocking compared to what we see reported here in the forums.

All that said, yes, I think we got a "good" sample in the Galaxy card. I think some of the overclocking prowess has to be contributed to the card design as well. This Galaxy card was designed to be a good overclocking and that can be seen in its power layout and PCB design.
 
I'm looking forward to the comparison of all the overclocked cards in an apples to apples form. Going back and looking at the XFX 7950 Black Edition overclocked results makes the 7950 look like a better buy once both are overclocked. Especially given that you can get 7950's now for around $320 or so, which is on par with most of the custom 660's on the market.
 
Ah a sucker punch

"GALAXY GTX 660 Ti is a whopping 28% faster than the stock clock Radeon HD 7950"

A good sample stock 7950 could achieve a 45%-50% overclock with voltage tweak from its stock 800Mhz. This of course does not = a linear increase, so it would work out at around 35%-40% in actual performance.

At max overclocks on good sample cards the 7950 should be around 10%-15% on average faster over a GTX 660Ti. Of course some games will be extreme outliers and it depends on the quality of the individual cards. My own testing of a HD 7950 at 1150 OC put it around 10% faster than a stock GTX 680.

Right now we as consumers have a lot of choice, though prices are still slightly high IMHO. It really comes down to manufacturer preference this generation. For the 1st time in a few generations Nvidia and AMD are actually trading blows at the top end.

Given the same price range.
GTX 680 and HD 7970 GHz edition trade blows.
GTX 670 and HD 7970 trade blows
GTX 660 Ti and stock HD 7950 trade blows.

Normally when the dust settles Nvidia are usually around 15% ahead at the top end. This simply didn't happen this generation when the pricing settled to its current level.
 
Last edited:
So the 660 really *is* memory bandwidth limited, ...very nicely illustrated here by the [H] team, a stirling job done again. Props for yous guys.

I didn't get that at all. A 30% memory increase yields 10% increased performance, which doesn't scream bandwidth limited to me. The same 30% overclock on the core yields a similar 10% performance increase, does that mean it is core limited?
 
I don't think I'll worry too much about OC'ing my two 660s for the near future until there's a game that needs it. These things should last for a while.

Having come from two 560ti's though, I was in la-la land thinking this card would take two 6-pin PCI-E connectors. Glad this review pointed out it required a 6+8, so I could finally bite the bullet on a new power supply.

Really interested to see the comparison of all the cards OC'd though. All we need is something to fuel the 660ti vs 7950 debate even more :rolleyes:
 
Kyle & Company, thanks for the edited note at the end about the all-OC slugout yet to come. That should be a whopper.

Of course, nVidia has a simple solution at hand to any pricing being out of whack versus value when compared to (especially OC'ed) AMD cards: do what AMD did and cut prices. For a shot of perspective, try and argue in favor of the 7950 when comparing launch price to launch price. We'll just see what the picture is like in two months or so.

Or we'll find out in the OC vs. OC article that the vaunted claims of 7950 superiority evaporate when using [H]-style real-world testing instead of canned benchmarks. Wouldn't be the first time.
 
All we need is something to fuel the 660ti vs 7950 debate even more :rolleyes:
I don't even know if there is a truly meaningful debate to be had. There are clearly scenarios where the 7950 is better, and there are clearly scenarios where the 660 Ti is better.

AMD and NVIDIA both have their market segmentation programs going on where they sponsor some game studio or engine developer and come out with better benchmarks. It's hard to "win" over the competition when you don't have that level of access.

I find it a little sad that we have this segmentation, but it is what it is. For the games I care about the most, and/or the engines they use, NVIDIA is at this time providing more performance for the dollar. I can just as easily make a strong case for AMD.

I do enjoy some balls-out OC slugging, but if people make their purchase decisions based on them, well, I find that questionable.
 
is an overclocked 7950 28% faster than a stock one? I don't even know :(

At the end of the review:

Editor's Note: For those of you that are interested in OCed 660 vs. OCed 670 vs. OCed7950, keep you panties on, we are working on it currently. We will also be taking a more focused look at 660 vs. 670 memory bandwidth limitations while using AA after that.
 
It's interesting how cannondale06 is being awfully quiet after their loud raucous the other day when the initial review was released.
 
Of course, nVidia has a simple solution at hand to any pricing being out of whack versus value when compared to (especially OC'ed) AMD cards: do what AMD did and cut prices.

This is nothing new, all companies drop their prices depending upon market demands.

For a shot of perspective, try and argue in favor of the 7950 when comparing launch price to launch price. We'll just see what the picture is like in two months or so.

How is this relevant? You can't compare the price of a GTX 660Ti to the price of a HD 7950 from 6 months ago and say the GTX 660Ti is cheaper. Right now the GTX 660Ti and HD 7950 are similar prices, that is all that maters when comparing video cards.

Or we'll find out in the OC vs. OC article that the vaunted claims of 7950 superiority evaporate when using [H]-style real-world testing instead of canned benchmarks. Wouldn't be the first time.

This would be nice to find out. This mini review shows that a good sample Galaxy 3GB 660Ti can overclock to on average around 10% higher than a GTX 670. My own testing showed a good sample 7950 clocking to on average around 10% faster than stock GTX 680 speeds.
 
How is this relevant? You can't compare the price of a GTX 660Ti to the price of a HD 7950 from 6 months ago and say the GTX 660Ti is cheaper. Right now the GTX 660Ti and HD 7950 are similar prices, that is all that maters when comparing video cards.

As I said, to provide some perspective. Is it relevant to pimp the 7950 when, stock to stock, it trades blows with the 660ti for a slightly higher price, even after 6 months of aggressive price-cutting? Is it not interesting, worth pondering, that the 660Ti is so competitive even when still at full launch retail? What would the value picture be if nVidia cuts the price by 10% once the launch rush is over? What if it goes on sale for Christmas? I doubt after such massive cuts already that AMD can take the 7950 lower. nVidia has margin galore to play with.

Speaking as someone who no longer pays $300-plus for video cards, such considerations are relevant to me, and I can wait to see where the prices go. I'm even curious to see what a 650/650Ti might offer, especially since I game at 1600x1200, which is even lower pixel count than 1920x1080.

Also, for me, an AMD product would have to be an absolute, unequivocal slam-dunk on price, performance, and features to even consider buying one. I have had too many unsatisfactory experiences with them. If that bothers you, I refer you to the heading at the top of the thread. This is the nVidia forum, and I post here rather than in the AMD forum for a reason.
 
As I said, to provide some perspective. Is it relevant to pimp the 7950 when, stock to stock, it trades blows with the 660ti for a slightly higher price, even after 6 months of aggressive price-cutting? Is it not interesting, worth pondering, that the 660Ti is so competitive even when still at full launch retail? What would the value picture be if nVidia cuts the price by 10% once the launch rush is over? What if it goes on sale for Christmas? I doubt after such massive cuts already that AMD can take the 7950 lower. nVidia has margin galore to play with.

No, it isn't interesting. The price of a card on release has no bearing on its current price/performance ratio. The only metric that matters is that right now the GTX 660 Ti from this review is priced similarly to non-reference HD 7950 OC editions.

Speaking as someone who no longer pays $300-plus for video cards, such considerations are relevant to me, and I can wait to see where the prices go. I'm even curious to see what a 650/650Ti might offer, especially since I game at 1600x1200, which is even lower pixel count than 1920x1080.

Again, I must stress that the only metric that maters is what price a particular card will cost you right now. It doesn't matter if it was $1000 dollars six months ago, or if it will be $20 at Xmas. All that matters is its price/performance ratio right now.

Also, for me, an AMD product would have to be an absolute, unequivocal slam-dunk on price, performance, and features to even consider buying one. I have had too many unsatisfactory experiences with them. If that bothers you, I refer you to the heading at the top of the thread. This is the nVidia forum, and I post here rather than in the AMD forum for a reason.

Good for you, you don't like AMD, you are at least honest in your bias. I care not one jot about Nvidia or AMD, I purchase whatever one has the best price/performance ratio and feature set. That entitles me to post on either board on this forum, please don't try the utterly childish, "this is a Nvidia forum, please leave unless you love Nvidia" BS. You have been a member here for 8.1 years so you must be more mature than that.

There are many people out there who rely on unbiased information to ensure they get the best deal. That's why reviews here at [H[ show performance of cards from both AMD and Nvidia in a particular price bracket. Just like you many can't afford, or simply won't pay over a certain price for their GPUs. Many also only purchase a GPU every 3 years or so and unlike you they are not tied to a particular brand. To them a 10%-15% performance difference means a lot considering how long they plan to use their card.
 
Right now the GTX 660Ti and HD 7950 are similar prices, that is all that maters when comparing video cards.

If you don't mind buying used you can pick up a 7950 in the B/S/T forum for good price/performance. With the 660Ti just being released it will be hard to find a discounted one on the B/S/T forum.

I added the galaxy 660ti 3gb to my newegg cart and after shipping/taxes it was around $370. Was real close to ordering it but slept on it. The next day for $100 less I bought a 7950 off the forum.
 
No, it isn't interesting. The price of a card on release has no bearing on its current price/performance ratio. The only metric that matters is that right now the GTX 660 Ti from this review is priced similarly to non-reference HD 7950 OC editions.

It is interesting in a "houses used to cost $25,000" kind of way, but it certainly shouldn't influence anyone's opinion of the cards now, or influence the purchase decision in any way other than "they've already cut the price three times so I don't know how much more they can lower it in the next few months" thinking. I think it does illustrate just how much AMD got away with overpricing these cards on launch due to Nvidia's sub-optimal release schedule. Which is good for AMD, I guess.
 
So you guys and gals want to see the GALAXY 660 Ti GC overclocked put up against overclocked 7950 and GTX 670, so I'm going to use the ASUS GTX 670 DirectCU II TOP video card for the overclocked 670, and the Sapphire HD 7950 OC, those cards we got the highest overclocks on. I have a feeling I know what's going to happen, but maybe I'll be surprised :p
 
Last edited:
i really dont get the point why people flame about "well if i only overclocked this higher priced card it would eat that lower priced card for dessert" first of all if your buying something like a gtx660 ti you wouldnt even be looking at a 670 since its obviously out of your budget unless you have the ability to stretch yourself out a little bit like some of us do... but for most people getting to stock level performance with cards that are priced $100-200 more is exactly why they choose the 660ti...

second, the 7950 can be viewed as a direct comparison because of the price drop amd implemented today... while i dont think that it is better than the 660ti because of its power consumption and heat output when overclocked at 1150+mhz, which is required to match an overclocked 660ti's performance, its also $30-40 more than the 660ti as of right now (amazon prices)

so maybe next time when reviews like these pops up we can take the results with a grain of salt knowing that the higher end cards were comparing it to are better performers when overclocked but that isnt the whole point of this review...
 
So you guys and gals want to see the GALAXY 660 Ti GC overclocked put up against overclocked 7950 and GTX 670, so I'm going to use the ASUS GTX 670 DirectCU II TOP video card for the overclocked 670, and the Sapphire HD 7950 OC, those cards we got the highest overclocks on. I have a feeling I know what's going to happen, but maybe I'll be surprised :p


I bet the 670 wins!!! Just IMO right now until proven wrong. :D
 
It's interesting how cannondale06 is being awfully quiet after their loud raucous the other day when the initial review was released.

I am curious to hear from him as well as I held off my order of a 660Ti after his comments as to how crippled the architecture was at 2560 x 1440.
 
so maybe next time when reviews like these pops up we can take the results with a grain of salt knowing that the higher end cards were comparing it to are better performers when overclocked but that isnt the whole point of this review...

I disagree on this point. I find comparison to the higher end cards very informative. It lets potential buyers know how close they get to top end preformance for a lower price. Or vice-versa, it lets people who paid premium prices for their GPUs how much they could have saved. :)
 
Wow [H] must have gotten a golden chip/card. Anand wasn't able to exceed 7Ghz on the memory on either of their 3 cards and nor was Toms or most of the other major sites I read. I wonder if it has anything to do with the odd memory configuration on the Galaxy?
 
Wow [H] must have gotten a golden chip/card. Anand wasn't able to exceed 7Ghz on the memory on either of their 3 cards and nor was Toms or most of the other major sites I read. I wonder if it has anything to do with the odd memory configuration on the Galaxy?

Some other sites have had 660Ti cards that went to 1300+ core but never as high on the VRAM. Others had cards that woul barely overclock at all. Anandtech had 3 OC edition cards that would only reach a max of 5% overclock. This worked out around 15% faster than a stock GTC 660Ti. As always it must be remembered that purchasing a card based on overclock results from one site is idiocy IMHO. Read as many OC results form as many reviews as possible (even from crap sites) and average the results to get a figure to aim for. This is true no matter what card you are looking at.

This mini review shows the Galaxy 3GB OC GTX 660Ti on average 25% faster than a stock HD 7950 over these 5 games. This average is only 17% faster when we remove the outlier that is Batman. The Batman result demonstrates very well how one game that clearly favours one brand over the other can really skew results. It is the same when we see Dirt Showdown reviews showing a HD 7870 beating a GTX 680 by a large margin. All of these extreme cases should be excluded IMHO. A similar good sample HD 7950 will overclock 40%-50% giving a real performance increase of around 30%-40%. If we exclude the obvious outliers on both sides then overall the HD 7950 should be faster given these basic facts.
 
I am curious to hear from him as well as I held off my order of a 660Ti after his comments as to how crippled the architecture was at 2560 x 1440.

I'll let you know how one card runs on a 2560x1600 compared to my old gtx480. Wasn't really upgrading for this monitor, I'm putting this pc on my 1080P 60" tv which will be sli'd down the road...
 
So you guys and gals want to see the GALAXY 660 Ti GC overclocked put up against overclocked 7950 and GTX 670, so I'm going to use the ASUS GTX 670 DirectCU II TOP video card for the overclocked 670, and the Sapphire HD 7950 OC, those cards we got the highest overclocks on. I have a feeling I know what's going to happen, but maybe I'll be surprised :p



That should be a good comparison. I was glad to see this comparison is coming.
My MSI 670s are very gently OC'd and in game hit 1245-1265 MHz gpu speeds.

This 660 Ti looks to be a pretty good overclocking card.....I bet it's very close to the 670....except for that mild memory limitation.
 
Regarding the insanely high memory overclock:

It is likely that some kind of error correction kicks in at these speeds, isn't it? I have never ever seen a memory OC that high (in absolute terms and in percentage). 10% more fps for 28% more bandwidth is meh.

Btw, could [H] include info about software revisions in their reviews? Drivers and games, that is. How do we know that all benchmarks were done with the latest drivers and that some values were not taken from older reviews with older drivers?
 
Btw, could [H] include info about software revisions in their reviews? Drivers and games, that is. How do we know that all benchmarks were done with the latest drivers and that some values were not taken from older reviews with older drivers?

Driver revisions used for the cards is documented in the system setup page of each review, and the game revision is documented in the maximum playable settings chart for each game within the review.
 
Back
Top