GALAXY GTX 660 Ti GC OC vs. OC GTX 670 & HD 7950 @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,603
GALAXY GTX 660 Ti GC OC vs. OC GTX 670 & HD 7950 - We take the GALAXY GeForce GTX 660 Ti GC 3GB video card, overclock it to its maximum potential and compare it to an overclocked GTX 670 and overclocked HD 7950, both of which represent high-end overclocks. Will the GTX 660 Ti crumble, or will it stand up to an overclocked GTX 670 and HD 7950?
 
Yeah, AMD got back into their historical position of providing great value for the dollar with the 7950 price cuts. Not a very good deal at $449 (or whatever it launched at) but at the $320 range it is very nice.

In a way though, it's almost too bad you got such a good overclock on the 7950 - since that seems like a very good result, it kind of skews the comparison of what the average buyer is going to get. It would be nice if there was a way to test "average" overclocks to give a better picture. Kind of a shame it took AMD so long to get its drivers act together also.
 
Yeah, AMD got back into their historical position of providing great value for the dollar with the 7950 price cuts. Not a very good deal at $449 (or whatever it launched at) but at the $320 range it is very nice.

In a way though, it's almost too bad you got such a good overclock on the 7950 - since that seems like a very good result, it kind of skews the comparison of what the average buyer is going to get. It would be nice if there was a way to test "average" overclocks to give a better picture. Kind of a shame it took AMD so long to get its drivers act together also.

Cards like Sapphire HD 7950 Vapor-X, Sapphire HD 7950 950 Mhz edition, MSI Twin Frozr HD 7950 are all using HD 7970 PCB designs. with voltage overclocking 1150 Mhz seems to be achievable quite consistently. still not guaranteed. But thats the case with any other card.
 
Yeah, AMD got back into their historical position of providing great value for the dollar with the 7950 price cuts. Not a very good deal at $449 (or whatever it launched at) but at the $320 range it is very nice.

Agreed, I tried to rationalise the 79x0 release prices comparing to the 3GB GTX 580 prices. In hindsight I should have proclaimed both as extortionate.

In a way though, it's almost too bad you got such a good overclock on the 7950 - since that seems like a very good result, it kind of skews the comparison of what the average buyer is going to get. It would be nice if there was a way to test "average" overclocks to give a better picture. Kind of a shame it took AMD so long to get its drivers act together also.

I think [H] did a great job on choosing the cards for the test. IMHO these results represent the best potential overclock you could expect on each respective card. On average consumers will get lower overclocks than these good samples and the respective performance delta will be similar. As far as drivers go it took a while for Nvidia to fix the Vsync stutter problems. On the whole I would give Nvidia the edge on drivers, but certainly not a big enough lead that it would influence my opinion.

What the conclusion shows is what I have felt for years, a 5%-10% margin of difference between two cards is not enough to notice a difference. In overall terms the difference between 50 and 53 FPS is not going to be noticeable IMHO. Though with an average of ~16% in favour of the 7950, and being a similar price, I still stand by my choice that it is a better option overall. Well as long as you have no aversion to AMD of course :)

We can all agree that the release of the 660Ti has brought an excellent level of competition. Prices are finally looking good for the mid range :)
 
Last edited:
I cannot believe how much things have improved since Riva 128 days! Fantastic 3 articles on the fantastic Galaxy 660 Ti video card. Not a brand I'd ordinarily buy but seems like a safe bet to me.
 
Excellent review. With the 7950 price drops, it makes it much more appealing. The 660Ti would be a gold feather in nvidia's cap if the MSRP was $259 or less.

I still liked the years of the "mid-range" cards costing about $200. $300 is too much for "mid-range" if you ask me.
 
Excellent review. With the 7950 price drops, it makes it much more appealing. The 660Ti would be a gold feather in nvidia's cap if the MSRP was $259 or less.

I still liked the years of the "mid-range" cards costing about $200. $300 is too much for "mid-range" if you ask me.

I'm with you on that mid-range price lament! I'm still running an HD5850, great mid-range card for its day, that was $249 at release...I wish I could still get that bang for the buck.
 
While checking NewEgg for the price of the XFX Black just now, I saw that the standard XFX 7950 is 289.99 after MIR. Holy Toledo!

Time for NV to do some price cuts, unlike the 5XX series.
 
Great reviews! Thanks for running all of the cards through the paces in an apples-to-apples comparison.
 
IMHO these results represent the best potential overclock you could expect on each respective card.
This is just false. There are many many overclocking threads on hardware forums, both for the 7850 and 7950 and their actual average with vcore mod is ~1.2ghz, their typical max is at around 1.3ghz, very similar to kepler which seem to max around 1.3ghz boost as well.

In fact, almost every single review with these cards even at stock voltage could hit the CCC max 1050mhz.

So in essence, [H] just put a max 660ti & 670 vs a below average 7950 OC, and the radeon still won.

Not to mention 4x MSAA only at 1080p (why not also bench 8x MSAA?), and no MSAA at 1600p... why?

Heck, there's even a big thread in this forum: http://hardforum.com/showthread.php?t=1679776
1.15ghz and 1.2ghz are the norm for 7950.
 
Is that drop repeatable in Batman at stock clocks? Such an excessive hit in framerate makes me wonder if crc checks are kicking in due to the memory oc.

This is just false. There are many many overclocking threads on hardware forums, both for the 7850 and 7950 and their actual average with vcore mod is ~1.2ghz, their typical max is at around 1.3ghz, very similar to kepler which seem to max around 1.3ghz boost as well.

In fact, almost every single review with these cards even at stock voltage could hit the CCC max 1050mhz.

So in essence, [H] just put a max 660ti & 670 vs a below average 7950 OC, and the radeon still won.


1.15ghz and 1.2ghz are the norm for 7950.


Thats bs, 7950 needs good cooling for 1200mhz and that is something that most cards don't offer. I would say that 1100mhz is more representative of an average 7950 oc. These cards suck a lot of juice at the 1.2v that you need for 1200mhz and that translates to a lot of heat. I couldn't hit 1200mhz before I water cooled my cards and that was with the MSI TF3. Even the better custom coolers are going to get loud at 1200mhz and they do suffer a bit in crossfire. 1200mhz is not an oc that I would expect when buying a 7950. I also wouldn't expect 1300mhz out of a gtx670 but it isn't that far off from what most cards will hit.
 
Last edited:
This is just false. There are many many overclocking threads on hardware forums, both for the 7850 and 7950 and their actual average with vcore mod is ~1.2ghz, their typical max is at around 1.3ghz, very similar to kepler which seem to max around 1.3ghz boost as well.

In fact, almost every single review with these cards even at stock voltage could hit the CCC max 1050mhz.

So in essence, [H] just put a max 660ti & 670 vs a below average 7950 OC, and the radeon still won.

Not to mention 4x MSAA only at 1080p (why not also bench 8x MSAA?), and no MSAA at 1600p... why?

Heck, there's even a big thread in this forum: http://hardforum.com/showthread.php?t=1679776
1.15ghz and 1.2ghz are the norm for 7950.

Out of all the 7950 reviews we have done, the XFX card got the highest overclock, therefore it made sense to use it. We haven't had any other 7950 that can overclock higher. I'd consider 1195 from the stock frequency of 800 on a 7950 a GREAT overclock. That is a 49% GPU overclock, which is pretty darned impressive. We did another 7950 review and it only hit 1050, and 1195 is 145MHz higher than 1050, so it is significant.

I've seen more 1.3GHz overclocks on 7970's than I've seen on 7950's, sorry, we didn't have a 7950 that can reach that. I've personally not seen a 1.3GHz overclock on a 7950.

I would not consider the XFX 7950's overclock a below average overclock at all, I would consider it an above average overclock. Average seems to be around 1050-1100 for 7950 overclocks. 1200 is on the high-end.

Yes, we used 4X MSAA at 1080p, Skyrim BTW was at 8X MSAA at 1080p. Why not bench at 8X MSAA in every game? Let's first consider the price point of this card, 1600p and 8X MSAA is not what this card is meant for, you want that you get a GTX 670 or better, consider who this card is priced for. For your information, I am going to work on an article that covers high levels of AA to find out where the 660 Ti's wall is, but that is irrelevant to the overclocking comparisons, which do stand on their own.
 
Last edited:
Thats bs, 7950 needs good cooling for 1200mhz and that is something that most cards don't offer. I would say that 1100mhz is more representative of an average 7950 oc.

Are you smoking something? There's the MSI TF, Sapphire Flex or OC versions, Gigabyte windforce, Asus, Powercolor PCS+, HIS IceQ.. they all are great cooler custom cards. Price?

http://www.newegg.com/Product/Produ...-1&isNodeId=1&Description=radeon+7950&x=0&y=0

Cheaper than the Galaxy version in this review.

You have a thread on this forum with 7950 owners, go check it out. <1.2ghz on the core is weak, to suggest its the best is flat wrong.
 
IMHO these results represent the best potential overclock you could expect on each respective card.

Out of all the 7950 reviews we have done, the XFX card got the highest overclock, therefore it made sense to use it. We haven't had any other 7950 that can overclock higher. I'd consider 1195 from the stock frequency of 800 on a 7950 a GREAT overclock. That is a 49% GPU overclock, which is pretty darned impressive. We did another 7950 review and it only hit 1050, and 1195 is 145MHz higher than 1050, so it is significant.

I've seen more 1.3GHz overclocks on 7970's than I've seen on 7950's, sorry, we didn't have a 7950 that can reach that. I've personally not seen a 1.3GHz overclock on a 7950.

I would not consider the XFX 7950's overclock a below average overclock at all, I would consider it an above average overclock. Average seems to be around 1050-1100 for 7950 overclocks. 1200 is on the extreme high-end.

Yes, we used 4X MSAA at 1080p, Skyrim BTW was at 8X MSAA at 1080p. Why not bench at 8X MSAA in every game? Let's first consider the price point of this card, 1600p and 8X MSAA is not what this card is meant for, you want that you get a GTX 670 or better, consider who this card is priced for. For your information, I am going to work on an article that covers high levels of AA to find out where the 660 Ti's wall is, but that is irrelevant to the overclocking comparisons, which do stand on their own.

While i don't doubt its the OC results you obtained, your samples are few, just a few cards at most. While there's a massive thread at overclockers.net and other enthusiast OC forums which does show both the 7850 and 7950 average out at ~1.2ghz. I simply feel its not accurate at all to suggest <1.2ghz is the "best", when you put it vs. 1.3ghz kepler, which we do know is its max potential.

1600p could simply be with 4x MSAA, if you feel it lacks the grunt for 8x MSAA. Having no AA for an enthusiast site/review is... ?

In most of the games you tested, the OC cards could handle 8x MSAA at 1080p, easily.
 
I own two MSI TF 7950s, neither one did 1200mhz on the stock cooler. I've seen quite a few other people on the forums say pretty much the same thing. These cards draw a lot of juice when you pump the core voltage, why are you surprised that they require really good cooling?

1200mhz on a 7950 is by no means a guarantee like you seem to think.

Only when I slapped on a couple of waterblocks could I do 1200mhz+.
 
I can hit 1.2ghz on a stock MSI Twin frozn III heatsink. Paid $309 for the sob too.

This is great news. This shows I was right about the 7950 > 660ti easily. Specially with the price drops.

I mean you can get a MSI Twin frozn III cheaper then the 3gb galaxy they are selling. AND the MSI Twin frozn III comes on a 7970 PCB as well
 
I own two MSI TF 7950s, neither one did 1200mhz on the stock cooler. I've seen quite a few other people on the forums say pretty much the same thing. These cards draw a lot of juice when you pump the core voltage, why are you surprised that they require really good cooling?

1200mhz on a 7950 is by no means a guarantee like you seem to think.

Only when I slapped on a couple of waterblocks could I do 1200mhz+.

OC is never a 100% certainty. Yours failed to reach 1.2ghz, other uses reached 1.25 or even 1.3ghz. You don't have to go far, just a few overclocking focused forums with their massive threads with benchmarks, screenshot validation included.
 
I can hit 1.2ghz on a stock MSI Twin frozn III heatsink. Paid $309 for the sob too.

This is great news. This shows I was right about the 7950 > 660ti easily. Specially with the price drops.

I mean you can get a MSI Twin frozn III cheaper then the 3gb galaxy they are selling. AND the MSI Twin frozn III comes on a 7970 PCB as well

Do you actually use those clocks 24/7? At anything over 1.1v my cards got really hot and loud.

At even 1100mhz it'll sure beat the pants off of a 660ti, thats for sure.

OC is never a 100% certainty. Yours failed to reach 1.2ghz, other uses reached 1.25 or even 1.3ghz. You don't have to go far, just a few overclocking focused forums with their massive threads with benchmarks, screenshot validation included.

I can just go by what I've seen personally, with two cards. The video card reviewer of this very site seems to have seen similar results. Most people on OCN seem to have similar results. Most people seem to have had similar results with the MSI TF3 here on hardforum. I don't know what else to tell you. 1200mhz and especially 1300mhz is by no means representative of most air cooled 7950s can do especially with acceptable noise.

Just drop it, 100mhz here or there isn't the end of the world at this point. Its not like thats enough to completely change the results.
 
Last edited:
Do you actually use those clocks 24/7? At anything over 1.1v my cards got really hot and loud.

At even 1100mhz it'll sure beat the pants off of a 660ti, thats for sure.

YEs she does use those clocks. (it was my girlfriends b-day gift).

She plays guildwars 2 (when she can) and also skyrim heavily modified and Dayz. Clocks are 1.2ghz GPU 1650 on the memory.

ASIC on the GPU is 89.1%

Edit: like to add, yes the fans get loud, but it does keep it below 82C while playing. I planned to get http://www.newegg.com/Product/Product.aspx?Item=N82E16835186064 soon as well
 
While i don't doubt its the OC results you obtained, your samples are few, just a few cards at most. While there's a massive thread at overclockers.net and other enthusiast OC forums which does show both the 7850 and 7950 average out at ~1.2ghz. I simply feel its not accurate at all to suggest <1.2ghz is the "best", when you put it vs. 1.3ghz kepler, which we do know is its max potential.

1600p could simply be with 4x MSAA, if you feel it lacks the grunt for 8x MSAA. Having no AA for an enthusiast site/review is... ?

In most of the games you tested, the OC cards could handle 8x MSAA at 1080p, easily.

I'm sorry I can't review every single 7950 for you that overclocks as high as humanly possible for this article. We used a great one, it overclocked high, 1195, which is right at 1.2GHz, 5MHz, shy, this is not an average overclock for a 7950, you need a high-end custom cooled and well powered 7950 to achieve this overclock, it is better than average, and I stand by that. I think 1195 is an excellent place to compare 7950's performance at, 1.2GHz, it's a huge overclock from the base of 800 and as the eval showed, the 7950 came out on top a lot in the evaluation.

I also am not able to test at every resolution and combination of AA in an article, the time does not exist for such feats when using real-world run-throughs. I used the resolutions and combinations of AA that make sense for the market segment of this video card. I'm impressed it can run at 2560x1600 with FXAA to begin with, it isn't positioned for that level of gaming, but it can certainly push it. It also goes back to the FXAA vs. MSAA argument, I'd rather go with FXAA and save performance than be dragged down by MSAA at 2560x1600, FXAA provides 4X MSAA like quality with only 2-3% performance hit, plus reduces aliasing on polygon edges, alpha textures, and specular aliasing, it's a no brainier.

At 1080p 4X MSAA is a common setting, it uses more memory bandwidth and is a great equal apples to apples test that everyone can compare their cards to at home. 1080p with 4X MSAA is where I see this card being positioned for, and we found out it can do that, and possibly more. In a future article I'm going to explore this "possibly more" scenerio.
 
While i don't doubt its the OC results you obtained, your samples are few, just a few cards at most. While there's a massive thread at overclockers.net and other enthusiast OC forums which does show both the 7850 and 7950 average out at ~1.2ghz. I simply feel its not accurate at all to suggest <1.2ghz is the "best", when you put it vs. 1.3ghz kepler, which we do know is its max potential.

1600p could simply be with 4x MSAA, if you feel it lacks the grunt for 8x MSAA. Having no AA for an enthusiast site/review is... ?

In most of the games you tested, the OC cards could handle 8x MSAA at 1080p, easily.

The average HD 7950 overclock is mostly around 1.1 - 1.15 Ghz with voltage tweaking. Nobody has compiled a statistical finding but seeing users on OCN and hardforum thats the conclusion we can arrive at. Also at 1.2 Ghz depending on the voltage you are running through the card air cooling starts to hit its limits. the guys who are hitting 1.3 Ghz on Tahiti mostly are watercooling or having custom air cooling like Accelero Xtreme 7970. As for 1600p with 4x MSAA in games like Batman AC the fps won't be really playable and barely if you may say so. Whats the point of comparing numbers which are not really playable. What this review proves is the GTX 660 Ti is a class below the HD 7950 and GTX 670 in terms of on chip resources. So when the going gets tough the HD 7950 and GTX 670 are the only cards which are going to fare well.
 
I'm sorry I can't review every single 7950 for you that overclocks as high as humanly possible for this article. We used a great one, it overclocked high, 1195, which is right at 1.2GHz, 5MHz, shy, this is not an average overclock for a 7950, you need a high-end custom cooled and well powered 7950 to achieve this overclock, it is better than average, and I stand by that. I think 1195 is an excellent place to compare 7950's performance at, 1.2GHz, it's a huge overclock from the base of 800 and as the eval showed, the 7950 came out on top a lot in the evaluation.

I also am not able to test at every resolution and combination of AA in an article, the time does not exist for such feats when using real-world run-throughs. I used the resolutions and combinations of AA that make sense for the market segment of this video card. I'm impressed it can run at 2560x1600 with FXAA to begin with, it isn't positioned for that level of gaming, but it can certainly push it. It also goes back to the FXAA vs. MSAA argument, I'd rather go with FXAA and save performance than be dragged down by MSAA at 2560x1600, FXAA provides 4X MSAA like quality with only 2-3% performance hit, plus reduces aliasing on polygon edges, alpha textures, and specular aliasing, it's a no brainier.

At 1080p 4X MSAA is a common setting, it uses more memory bandwidth and is a great equal apples to apples test that everyone can compare their cards to at home. 1080p with 4X MSAA is where I see this card being positioned for, and we found out it can do that, and possibly more. In a future article I'm going to explore this "possibly more" scenerio.

I think Supersampling is way better then FXAA and MSAA, Its too damn bad it takes a HUGE performance hit. Otherwise I would supersampling all the damn time :/
 
I think Supersampling is way better then FXAA and MSAA, Its too damn bad it takes a HUGE performance hit. Otherwise I would supersampling all the damn time :/

Point of interest, Sleeping Dogs has an in-game SSAA option, but it's SSAA being done in a different way? I haven't had a lot of time this week to look at it yet, but it's a game of interest for me to look at regarding SSAA, it's pretty much the first game in a long long long time I've seen an SSAA option for inside the graphics options. That game with the texture pack plus SSAA brings a 7970 to its knees. Anyway, one of my future projects, I've got a lot piling on my lap suddenly :p
 
Definitely not normal. http://forums.anandtech.com/showthread.php?t=2259333

Lots of MSI TF users. If u have high temps at those voltage, suggest a reseat on the heatsink or reapply some good TIM.

:rolleyes:

Do you own the card? Do you even own a 7950?

I saw the same temps after reseating the cooler. I've seen others mention the same thing in a couple of threads here on hardforum.

I'm done responding to you kid.

The average HD 7950 overclock is mostly around 1.1 - 1.15 Ghz with voltage tweaking. Nobody has compiled a statistical finding but seeing users on OCN and hardforum thats the conclusion we can arrive at. Also at 1.2 Ghz depending on the voltage you are running through the card air cooling starts to hit its limits. the guys who are hitting 1.3 Ghz on Tahiti mostly are watercooling or having custom air cooling like Accelero Xtreme 7970. As for 1600p with 4x MSAA in games like Batman AC the fps won't be really playable and barely if you may say so. Whats the point of comparing numbers which are not really playable. What this review proves is the GTX 660 Ti is a class below the HD 7950 and GTX 670 in terms of on chip resources. So when the going gets tough the HD 7950 and GTX 670 are the only cards which are going to fare well.


Yup, thats mostly what I've seen as well.
 
Point of interest, Sleeping Dogs has an in-game SSAA option, but it's SSAA being done in a different way? I haven't had a lot of time this week to look at it yet, but it's a game of interest for me to look at regarding SSAA, it's pretty much the first game in a long long long time I've seen an SSAA option for inside the graphics options. That game with the texture pack plus SSAA brings a 7970 to its knees. Anyway, one of my future projects, I've got a lot piling on my lap suddenly :p

O wow did not know that....Last game I remember to have an option for SSAA was Witcher 2.

Also I just now loaded up Guild Wars 2, and notice they have a SSAA option as well....Might want to look into that game as well.

Edit: Might have to buy Sleeping dogs now lol
 
Point of interest, Sleeping Dogs has an in-game SSAA option, but it's SSAA being done in a different way? I haven't had a lot of time this week to look at it yet, but it's a game of interest for me to look at regarding SSAA, it's pretty much the first game in a long long long time I've seen an SSAA option for inside the graphics options. Anyway, one of my future projects, I've got a lot piling on my lap suddenly :p

It would be great to see an article on Sleeping Dogs SSAA quality and performance. just like the max payne 3 article. does SSAA implementation in Sleeping Dogs provide much better image quality and at what cost. ;)
 
It would be great to see an article on Sleeping Dogs SSAA quality and performance. just like the max payne 3 article. does SSAA implementation in Sleeping Dogs provide much better image quality and at what cost. ;)

I don't understand why you can't disable FXAA in that game. There really isn't a need for it with SSAA.
 
The average HD 7950 overclock is mostly around 1.1 - 1.15 Ghz with voltage tweaking. Nobody has compiled a statistical finding but seeing users on OCN and hardforum thats the conclusion we can arrive at. Also at 1.2 Ghz depending on the voltage you are running through the card air cooling starts to hit its limits. the guys who are hitting 1.3 Ghz on Tahiti mostly are watercooling or having custom air cooling like Accelero Xtreme 7970. As for 1600p with 4x MSAA in games like Batman AC the fps won't be really playable and barely if you may say so. Whats the point of comparing numbers which are not really playable. What this review proves is the GTX 660 Ti is a class below the HD 7950 and GTX 670 in terms of on chip resources. So when the going gets tough the HD 7950 and GTX 670 are the only cards which are going to fare well.

That thread on OCN has users which have ~1.2vcore achieving ~1.2ghz or greater. The others <1.2ghz are all less on the vcore or even less than stock voltage. Yes i own a 7950 and it's @ 1.2ghz.

The point here is in some of those games at 1600p, 4x MSAA is doable. Especially at 1080p, 8xMSAA have already been shown by other review sites to be fine. The problem here is [H] paints the 660ti as being within ~10% of a 670, well yes.. if you game at 4x MSAA 1080p. It looks very different across the board when you game at 1080p with 8x MSAA or 1600p with 2/4x MSAA. There's a trade-off for the 196bit bus and less ROPs, and its in AA performance. By not exposing that and painting a rosey scenario, its not offering both sides for potential buyers.

But I do appreciate [H] for their real benchmarks using gameplay and the fps graphs.. however, the FXAA vs MSAA issue is definitely not the favor of FXAA where consistent blur renders high-res textures looking like low-res is not a trade-off most of us gamers are willing to hop on board just yet. MSAA is still the standard, for a good reason.
 
I don't understand why you can't disable FXAA in that game. There really isn't a need for it with SSAA.

If you run an Nvidia card, You can run FXAA with any game basically since its in the drivers. Not sure if Nvidia driver option is better then in-game options though.

I know I have ran SSAA with FXAA at the same time with Skyrim....not sure if it helps since SSAA looks so damn badass anyway lol
 
If you run an Nvidia card, You can run FXAA with any game basically since its in the drivers. Not sure if Nvidia driver option is better then in-game options though.

I know I have ran SSAA with FXAA at the same time with Skyrim....not sure if it helps since SSAA looks so damn badass anyway lol

Sleeping Dogs runs FXAA by default and there is no way to turn it off.
 
I stand by my assertion that at 1200 MHz a 7950 is an exceptional overclock and is certainly above average. One I tested hit 1150 with a small voltage boost, it might have went higher but it wasn't mine so I stopped pushing it. I would consider 1100 core on a HD 7950 an average overclock. Any higher and it is just the icing on the cake.

For me this review is very fair and unbiased, [H] went above and beyond to ensure all the cards tested were all good samples. Frankly if someone purchases a HD 660Ti they can be assured they are getting a very good mid range card. They shouldn't feel bad about their purchase, or feel they were shafted.

Competition is good.
 
BTW, I do have more planned, one more article for 660 Ti. It is the "super high AA" article a lot of you have been wanting. GPU clock for clock, 192-bit bus the only difference, where is the 660 Ti wall and how does it directly compare to GTX 670 with 256-bit bus. 1600p 8xaa/4xaa, 1080p 8xaa/4xaa, stay tuned......
 
I stand by my assertion that at 1200 MHz a 7950 is an exceptional overclock and is certainly above average. One I tested hit 1150 with a small voltage boost, it might have went higher but it wasn't mine so I stopped pushing it. I would consider 1100 core on a HD 7950 an average overclock. Any higher and it is just the icing on the cake.

For me this review is very fair and unbiased, [H] went above and beyond to ensure all the cards tested were all good samples. Frankly if someone purchases a HD 660Ti they can be assured they are getting a very good mid range card. They shouldn't feel bad about their purchase, or feel they were shafted.

Competition is good.

http://www.techpowerup.com/reviews/Sapphire/HD_7950_Flex/31.html

Even at stock volts, most 7950 are capable of 1050mhz core. With ~1.2 to 1.25vcore, you think 1.1ghz is the average? Kepler at 1.3ghz is exceptional, because there's not voltage tweaking, they tend to reach 1.2ghz "average".

So in essence, its putting a max clocked 660ti vs a slightly above average 7950.. and not put in through tests where its low ROPs and 196bit bus is exposed. Fair? Informative? At least make the comments so readers will understand the strengths and weaknesses of the product.
 
BTW, I do have more planned, one more article for 660 Ti. It is the "super high AA" article a lot of you have been wanting. GPU clock for clock, 192-bit bus the only difference, where is the 660 Ti wall and how does it directly compare to GTX 670 with 256-bit bus. 1600p 8xaa/4xaa, 1080p 8xaa/4xaa, stay tuned......

Perfect! Much more informative.
 
While i don't doubt its the OC results you obtained, your samples are few, just a few cards at most. While there's a massive thread at overclockers.net and other enthusiast OC forums which does show both the 7850 and 7950 average out at ~1.2ghz. I simply feel its not accurate at all to suggest <1.2ghz is the "best", when you put it vs. 1.3ghz kepler, which we do know is its max potential..

You are familiar with the concept of selection bias, right? People with crappy overclocks are much less likely to advertise that fact on a thread than people who get great overclocks. Plus you pretty much have to discard the overclocks that come from users who are watercooling, or who have made physical or BIOS changes to the card to achieve the overclocks they are advertising. Most users are simply not giong to make those concessions to squeeze the last possible bit of performance out of a card. If you want to talk max possible overclock (by any means) versus max possible overclock, that is one thing, but to include those results in a discussion of average" overclocks is not proper.
 
Perfect! Much more informative.

This super high AA article was always in the plans, from the very beginning, we've had a schedule of priority first, and that was to first cover overclocking thoroughly, now that it is done, I can move on to the whole memory bus high AA testing article. There's only one of me and he hasn't had much sleep this past week :p
 
Does that drop on the 660 in Batman also happen if you pull back on the memory clock?
 
Does that drop on the 660 in Batman also happen if you pull back on the memory clock?

Going to find out in the next article, haven't tested the card yet at 4X MSAA in Batman at 1080p in the launch eval, but this next article will have that tested so we will see.
 
Back
Top