Nvidia GeForce GTX 650 And 660 Review: Kepler At $110 And $230

[guner451]

It would have been interesting to see how performance varied with increases/decreases in tessellation for these cards as I think that changes the look of some games substantially.

 

[blazorthon]

[citation][nom]jimbaladin[/nom]The 7870 can be had for $239 at newegg (after rebate...I know), but also comes with Sleeping Dogs free. That makes it a clear winner here - I think Nvidia needs to drop prices on the 660 by about $30-$40 in order to compete.[/citation]

The 7870 can sometimes be found for $200 AR. $240 is a price that can be found even without a rebate.

 

[Guest]

anyone else notice pcpers bullshit skyrim results on the 660 xD

 

[jaquith]

Actually, running a n i7-920 and 680 Lightning ATM. That doesn't mean I'm an AMD hater, either, it simply means it's the hardware I currently use in my main machine. I tend to swap graphics cards as the mood suits me.

Then what's the real reason with all of this? Does TH receive monies from AMD now? What's your angle here? What were you instructed to do? What if any compensation was behind this? What's your motivation? Frankly, then with a MSI GTX 680 you'd know or should know a lot more than I do about Adaptive vSync. Based on your own review you don't follow your own advice and that make you a hippocrite or self-loathing.

Your best guess sucks. You simply can't acknowledge that other opinions are just as valid if they conflict with your own, I suppose.

See above or you shouldn't be writing articles. Opinions aren't the issue, it's blatant disregard of the 'center,' and using (1) O-N-E thing (setting) of countless other permutations to push your narrow point of view. You're hurting others here, not me and apparent not yourself (GTX 680).

That I find 8x MSAA more important than adaptive Vsync doesn't make me wrong, Jaquith. it doesn't make me biased, either. It means I disagree with you, and I find it interesting that you can't accept this without concocting a conspiracy theory.

Then why only MSAA; see my prior posts of all of the different AA's. Frankly, the forthcoming TXAA supported games makes MSAA look horrible in comparison.

I haven't once called you an AMD hater or an Nvidia lover because your opinion doesn't sync with my own. Why do you feel the need to do otherwise?

I never said you were either? Clearly you're bias or again see above -- paid, instructed, etc.

Perhaps you feel that you can cheapen someone's counter-argument by suggesting they're biased, instead of arguing the topic?

What topic? You ignored all of mine that were inarguable; Jaquith = 10 & Don = 1. And I'm biased. I said >4xMSAA nVidia's got a problem, but other forms it's flip-flopped or the impact (frame cost) in comparison is negligible -- so I use other forms of AA with better results - simple. Otherwise 16xMSAA on AMD & 32x__AA on nVida, you purposely picked the biased version on nVidia why not visa-versa on AMD???

Again, IF I felt you were doing an impartial review I'd do as I said before 'That'a Boy' & Thumbs up.' There's so much frigging hate recently caused by TH choices it befuddles me; WTH is going on there?? It's by the News & Articles that TH chooses to run, last year none of this -- TH is cutting their nose off to spite their face.

Hey, I've seen many of the unpublished articles and have prevent a few major in your face incidence from occurring. I get folks will see what they see and in the end make some really bad decisions and any of these comments are ignored by only a tiny fraction of folks.

 

[Regor245]

This might be a conversation better placed in the forums, but as a side note, if money is too tight, you could go for a Llano build and have a Hybrid CF setup to make use of the Radeon 6670 since you already have it. A Phenom II CPU (assuming that your motherboard supports them, given that you have an Athlon II, it probably does) could also be used even with your current memory (Phenom II supports DDR2 and DDR3, just not both at the same time) and motherboard. So long as you stick with an AMD APU/CPU, you can reuse parts of your current build so that you can afford better upgrades with the parts that you replace, although if you get the budget to get an i5, that would probably be much more ideal CPU performance for future-proofing as well as current CPU-bottle-necked gaming.

The Phenom II will be a cheaper upgrade since i dont have to buy another mobo+memory, But A6-3670k is a good choice and i can get a little boost with my 6670 by crossfiring it. I could sell my old cpu/mobo/ram to get an A55/75 mobo with ram.

Well i will wait a little, once my scooter gets sold in cash, i could buy a used i5 2400 and an H61 with Ram.

thanks again.

 

[blazorthon]

[citation][nom]diceman203777[/nom]anyone else notice pcpers bullshit skyrim results on the 660 xD[/citation]

I just looked at them and they look accurate. What's wrong with them?

 

[blazorthon]

[citation][nom]jaquith[/nom]Then what's the real reason with all of this? Does TH receive monies from AMD now? What's your angle here? What were you instructed to do? What if any compensation was behind this? What's your motivation? Frankly, then with a MSI GTX 680 you'd know or should know a lot more than I do about Adaptive vSync. Based on your own review you don't follow your own advice and that make you a hippocrite or self-loathing.See above or you shouldn't be writing articles. Opinions aren't the issue, it's blatant disregard of the 'center,' and using (1) O-N-E thing (setting) of countless other permutations to push your narrow point of view. You're hurting others here, not me and apparent not yourself (GTX 680).Then why only MSAA; see my prior posts of all of the different AA's. Frankly, the forthcoming TXAA supported games makes MSAA look horrible in comparison. I never said you were either? Clearly you're bias or again see above -- paid, instructed, etc.What topic? You ignored all of mine that were inarguable; Jaquith = 10 & Don = 1. And I'm biased. I said >4xMSAA nVidia's got a problem, but other forms it's flip-flopped or the impact (frame cost) in comparison is negligible -- so I use other forms of AA with better results - simple. Otherwise 16xMSAA on AMD & 32x__AA on nVida, you purposely picked the biased version on nVidia why not visa-versa on AMD???Again, IF I felt you were doing an impartial review I'd do as I said before 'That'a Boy' & Thumbs up.' There's so much frigging hate recently caused by TH choices it befuddles me; WTH is going on there?? It's by the News & Articles that TH chooses to run, last year none of this -- TH is cutting their nose off to spite their face.Hey, I've seen many of the unpublished articles and have prevent a few major in your face incidence from occurring. I get folks will see what they see and in the end make some really bad decisions and any of these comments are ignored by only a tiny fraction of folks.[/citation]

TXAA has comparable video quality to some levels of MSAA. It simply makes much less of a performance impact to get that high-quality AA effect.

Disagreeing with Cleeve and the others is one thing that you and any other person have every right to do, but you're really trying to make your point in a very offensive and occasionally extremist way. How can you expect a positive response to what you're saying when you say it in such a negative way? Simply saying that you think that doing tests with other types of AA and other settings in addition to the MSAA tests would help to make a more complete picture of the performance of these cards than mostly focusing on high levels of MSAA does or something like that would have probably gotten you much better results.

Also, you are not winning any argument. Your side of this conversation so far has been mostly personal attacks with a few good points spread far and wide intermixed with a little inaccuracy and highly opinionated claims. I mean that without any intent of offending you, but you seem to have no intent of trying to not offend anyone here.

 

[cleeve]

Then what's the real reason with all of this? Does TH receive monies from AMD now? What's your angle here? What were you instructed to do? What if any compensation was behind this? What's your motivation? Frankly, then with a MSI GTX 680 you'd know or should know a lot more than I do about Adaptive vSync.

My instruction is to write an article based on hardware testing, nothing more. To assume there is more implies your bias, not mine.

You certainly have the right to think that adaptive Vsync is the best thing since sliced bread. Honestly, to me, it really isn't a game changer. Are you saying it's impossible to value high levels of AA over adaptive Vsync? Really, is that your platform here? That any opinion other than yours implies bias or incompetence?

Based on your own review you don't follow your own advice and that make you a hippocrite or self-loathing.

How in the world is that your deduction. I have never, ever once in my life said that Nvidia sucks. Radeons and GeForces are both quite good in my opinion. There are no bad cards anymore, only bad prices.

You ignored all of mine that were inarguable; Jaquith = 10 & Don = 1

You're just not making sense at all. I've heard your arguments and simply don't agree. you don't have a monopoly on rightness.

I have the ability to understand that your opinion is valid even if it doesn't agree with my own.
On the other hand, you appear to feel the need to throw accusations of bribery and hypocrisy.

Your world must be very small, you have my pity.

 

[Master Bob]

Wish they would have thrown in a 7950, that's the card i'm looking at.

This was focusing on mid-ranged cards. The 7950 is too high to be in here.

 

[tomfreak]

[citation][nom]Cleeve[/nom]The price difference is too large to compare the 7770/7750 with the 7850/660. The only reason I included the 650 in the high detail chart was to give a reference point for the difference in ability between the 660 and 650.It'd be nice to include every possible combination, but we don't have infinite time, and I like to focus the charts on what makes sense. That means similar price points.[/citation]well the thing is 6870 shouldnt be there from the first place. It is no where near 660/650. Had the time use for the 6870 be replace by AMD 7700. It would have been a lot easier to compare the result. 650 are a card with shader bound. It has enough memory bandwidth and ROP. The real problem is the Shader. I would like to see how it perform on higher/ultra preset against the 7750/7770. the existence of 6870 didnt really help me.

 

[jaquith]

My instruction is to write an article based on hardware testing, nothing more. To assume there is more implies your bias, not mine.

Your world must be very small, you have my pity.

Don your review is a pure hatchet job, it's that simple. Frankly, it's the only one I could find where MSAA was a deliberate con job. You're seeing yourself and yes it's small. :lol:

 

[blazorthon]

[citation][nom]Tomfreak[/nom]well the thing is 6870 shouldnt be there from the first place. It is no where near 660/650. Had the time use for the 6870 be replace by AMD 7700. It would have been a lot easier to compare the result. 650 are a card with shader bound. It has enough memory bandwidth and ROP. The real problem is the Shader. I would like to see how it perform on higher/ultra preset against the 7750/7770. the existence of 6870 didnt really help me.[/citation]

The 6870's placement offers context for people considering upgrading from Radeon 6800 or similarly performing models and also shows us how Radeon 6800's enhanced VLIW5 GPU handles MSAA compared to the competition. It might not help you in particular, but it is reasonable to include it because it might help some readers.

 

[cleeve]

[citation][nom]jaquith[/nom]Don your review is a pure hatchet job, it's that simple. Frankly, it's the only one I could find where MSAA was a deliberate con job. You're seeing yourself and yes it's small.[/citation]

Vitriol. Ouch! You mad, bro?

Call it a hatchet job if spite is your thing, and call it a con job if it makes you feel validated.

But here's the deal: it was translated into multiple languages and read by millions. Literally.

Your trolling is limited to English, on this forum... and most people can't even see it here because it's been voted down so much.

Think happy thoughts! Your baseless accusations, blind brand loyalty and and self-righteous rhetoric are making a difference. Really.

 

[cleeve]

[citation][nom]Tomfreak[/nom] the existence of 6870 didnt really help me.[/citation]

Well, it's priced within $30 of the 7850 (now as low as $200), so I think it's a valid add.

To each his own.

 

[pizzlemynizzle]

So a Radeon 7870GHz is $230 after MIR. Same price as the GTX 660. Seem like a no brainer at that price point. The radeon clearly performs much better and isn't bogged down by a restricted memory bus. The $300 range is kind of a toss up. The HD 7950 slightly outperforms the GTX 660 Ti overall and is also slightly more expensive. So pick your poison there. For the guy who wants a $170-$180 price range gpu, catch a 7850 on sale, gtx 560 ti 448 core on sale, or get a used gtx 570, gtx 580 or hd 7850 or hd 7870. Ebay is a great place to get used computer components. Those cards will all be about a year old, and these cards are made to last a lot longer than that. You would be upgrading in a couple years anyway.

 

[EzioAs]

Then again, I find 120 Hz motion compensation on televisions irritating as hell when its applied to 24 FPS movies...

Really? I actually like it very much, but I'm sure not everyone would agree. I even use a software called SVP (Smooth Video Project) on my PC to produce higher framerates (60fps on my monitor) than the standard 24/30 fps and actually once I've tried it, there's just no going back to 24fps. Although the software is a little bit too demanding to play 1080p files on my current rig (C2Q, 4GB RAM and GTX460).

I encourage you guys to try it

 

[abitoms]

i think i see a Geforce GTX 780 in one of the pages !!!!

Where?
page: 'OpenCL: Image Processing (Basemark CL)'
Chart: 'Image Blur' (first chart)
URL: http://www.tomshardware.com/reviews/geforce-gtx-660-geforce-gtx-650-benchmark,3297-21.html

I am thinking GTX 780 doesn't yet exist - or am I mistaken?

 

[blazorthon]

i think i see a Geforce GTX 780 in one of the pages !!!!

Where?
page: 'OpenCL: Image Processing (Basemark CL)'
Chart: 'Image Blur' (first chart)
URL: http://www.tomshardware.com/reviews/geforce-gtx-660-geforce-gtx-650-benchmark,3297-21.html

I am thinking GTX 780 doesn't yet exist - or am I mistaken?

Good catch, but I think that it's supposed to say GTX 480 where it says 780.

 

[de5_Roy]

my two cents on jwow vs the world:
adaptive vsync is nvidia-developed technology. testing both radeons and geforces-using-adaptive vsync on would be unfair. it's like asking to turn on physx to test both radeon and geforce cards.
fxaa is also developed by nvidia afaik. amd's fxaa competitor is mlaa.
http://www.tomshardware.com/reviews/anti-aliasing-nvidia-geforce-amd-radeon,2868.html
http://www.tomshardware.com/reviews/anti-aliasing-performance,3065.html
so benchmarking both radeons and geforces with nvidia-favoring tech enabled would be unbalanced. same with using amd-favoring tech. the games that favor nvidia also make use of driver command list.
imo, toms benches are more balanced and make fairer comparisons.
curiously, "test settings" info seems to be missing from the 660ti roundup article... or did i miss it..
however, individual owners of geforce 6xx and radeon hd 78/79xx cards are likely to play games with modes like fxaa, physx, adaptive vsync, mlaa, txaa turned on and gain fps. this will be especially prevalent in useless arguments like "my geforce/radeon plays better than your radeon/geforce"....without going into detailed explanation.

 

[kristi_metal]

Can't believe graphic boards can create so much passion
Holy Wars against the Unfaithfulls )

 

[army_ant7]

This might be a conversation better placed in the forums, but as a side note, if money is too tight, you could go for a Llano build and have a Hybrid CF setup to make use of the Radeon 6670 since you already have it. A Phenom II CPU (assuming that your motherboard supports them, given that you have an Athlon II, it probably does) could also be used even with your current memory (Phenom II supports DDR2 and DDR3, just not both at the same time) and motherboard. So long as you stick with an AMD APU/CPU, you can reuse parts of your current build so that you can afford better upgrades with the parts that you replace, although if you get the budget to get an i5, that would probably be much more ideal CPU performance for future-proofing as well as current CPU-bottle-necked gaming.

Just want to clarify blaz, for anyone who might be misled, and I'm certain you're aware of this (and I have guesses about what you really meant) but just happened to word it in a way that could mislead. All desktop APU's (I know of) and any Phenom II's (and some Athlon II's since there are FM1 versions) are not socket/motherboard compatible. You could use Phenom II's on an AM2+, AM3, or an AM3+ mo-bo's as I believe, but you can only use current APU's on FM1 and FM2 mo-bo's (and those two not being cross-compatible either, though I remember from a TH news article, talk of a possible adapater from FM2 to 1). (Again, info for those who may stumble upon this without knowing before-hand. Maybe you could edit it if it isn't too much trouble. Sorry. Sorry. Just had the need to point this out to be safe. )


With that aside, I just want to point out how Nvidia's (SLI) drivers are relatively impressive compared to AMD's. It topped 7 of the 8 game tests. Hm... Just thought now that scaling should be considered. Let me have a take at making scaling percentages here...............................................Whew... That took some time. Here:
(min%/avg%)...HD 7850 CF......GTX 660 SLI
BM:AC............61%/166%........36%/179%
BF3................196%/195%......181%/189%
C2.................200%/183%......170%/179%
D:SD .............205%/201%......171%/192%
MP3...............159%/158%.......213%/176%
M2033............160%/192%......161%/162%
ES5:S.............145%/157%......173%/175%
WoW:MoP........123%/157%.....162%/169%
Overall...........156%/176%..158%/178%

Hm... Wait a sec. Woah! They're pretty much on par with overall scaling (with these games). Interesting... So this shows that it (GTX 660) won because it was plainly better in performance to begin with, which isn't totally true as shown by the single-GPU test of Skyrim (where the HD 7850, but lost in dual-GPU). So that shows how the scaling percentages can be spread out in a way that it makes one card (GTX 660 in this case) seem really good in dual-GPU tests overall.

Man! I just have to say seeing percentages close to and actually 200% (theoretical performance) is just BEAUTIFUL! But it even goes past that. Hehehe... Sorry. Optimization nerd here. :lol: But I did think that if I did encounter a percentage past 200%, it might be because there's something "wrong" with the drivers, the game itself, the system, etc. (maybe intentional with the first two? Haha! Just kidding around. I'm no conspiracy theorist, but I guess it's possible, but I'm not saying it is true or likely. There is that obvious "flaw" with Batman: Arkham Asylum, and I remember this was brought up in an older article.); or there's a bottle-neck when the card is alone, but the bottle-neck gets alleviated in a CF/SLI setup (if that's possible), aside from the fact that I might've just made an error. Anyway, this same concept may apply to any of the percentages (even if they're not at all close to the theoretical performance), I'm thinking.

Anyone feel free to criticize or point out errors with these percentages. It may be very mildly off since I didn't take note of significant figures (which is important for the accuracy of scientific data) and I also rounded off the the percentages for each test and then again with the overall, though this itself contradicts the significant figures rule and I think we can deal with mild accuracy errors (maybe, just maybe, +/-1% difference more or less).

 

[abraham_mammogram]

[citation][nom]jaquith[/nom]Then why only MSAA; see my prior posts of all of the different AA's. Frankly, the forthcoming TXAA supported games makes MSAA look horrible in comparison. I never said you were either? Clearly you're bias or again see above -- paid, instructed, etc.What topic? You ignored all of mine that were inarguable; Jaquith = 10 & Don = 1. And I'm biased. I said >4xMSAA nVidia's got a problem, but other forms it's flip-flopped or the impact (frame cost) in comparison is negligible -- so I use other forms of AA with better results - simple.[/citation]


perhaps I am mistaken, but isn't TXAA only supported on nvidia cards? How would you perform a direct apples to apples comparison of two cards, when one of the cards (AMD) does not support the feature in question?

 

[blazorthon]

[citation][nom]army_ant7[/nom]Just want to clarify blaz, for anyone who might be misled, and I'm certain you're aware of this (and I have guesses about what you really meant) but just happened to word it in a way that could mislead. All desktop APU's (I know of) and any Phenom II's (and some Athlon II's since there are FM1 versions) are not socket/motherboard compatible. You could use Phenom II's on an AM2+, AM3, or an AM3+ mo-bo's as I believe, but you can only use current APU's on FM1 and FM2 mo-bo's (and those two not being cross-compatible either, though I remember from a TH news article, talk of a possible adapater from FM2 to 1). (Again, info for those who may stumble upon this without knowing before-hand. Maybe you could edit it if it isn't too much trouble. Sorry. Sorry. Just had the need to point this out to be safe. ) With that aside, I just want to point out how Nvidia's (SLI) drivers are relatively impressive compared to AMD's. It topped 7 of the 8 game tests. Hm... Just thought now that scaling should be considered. Let me have a take at making scaling percentages here...............................................Whew... That took some time. Heremin%/avg%)...HD 7850 CF......GTX 660 SLIBM:AC............61%/166%........36%/179%BF3................196%/195%......181%/189%C2.................200%/183%......170%/179%D:SD .............205%/201%......171%/192%MP3...............159%/158%.......213%/176%M2033............160%/192%......161%/162%ES5:S.............145%/157%......173%/175%WoW:MoP........123%/157%.....162%/169%Overall...........156%/176%..158%/178%Hm... Wait a sec. Woah! They're pretty much on par with overall scaling (with these games). Interesting... So this shows that it (GTX 660) won because it was plainly better in performance to begin with, which isn't totally true as shown by the single-GPU test of Skyrim (where the HD 7850, but lost in dual-GPU). So that shows how the scaling percentages can be spread out in a way that it makes one card (GTX 660 in this case) seem really good in dual-GPU tests overall.Man! I just have to say seeing percentages close to and actually 200% (theoretical performance) is just BEAUTIFUL! But it even goes past that. Hehehe... Sorry. Optimization nerd here. But I did think that if I did encounter a percentage past 200%, it might be because there's something "wrong" with the drivers, the game itself, the system, etc. (maybe intentional with the first two? Haha! Just kidding around. I'm no conspiracy theorist, but I guess it's possible, but I'm not saying it is true or likely. There is that obvious "flaw" with Batman: Arkham Asylum, and I remember this was brought up in an older article.); or there's a bottle-neck when the card is alone, but the bottle-neck gets alleviated in a CF/SLI setup (if that's possible), aside from the fact that I might've just made an error. Anyway, this same concept may apply to any of the percentages (even if they're not at all close to the theoretical performance), I'm thinking.Anyone feel free to criticize or point out errors with these percentages. It may be very mildly off since I didn't take note of significant figures (which is important for the accuracy of scientific data) and I also rounded off the the percentages for each test and then again with the overall, though this itself contradicts the significant figures rule and I think we can deal with mild accuracy errors (maybe, just maybe, +/-1% difference more or less).[/citation]

I meant that two good options for Regor245 were keeping his/her current motherboard and memory, upgrading the CPU and graphics, or keeping his/her Radeon 6670 and upgrading the motherboard to FM1 socket, memory to DDR3, and the CPU to a compatible A6 or A8 APU.

Depending on the game and settings, the 660 is too close to the 7850 or the 7870 or one of the two are too close to it to really say that either 7800 card necessarily beats the 660 regardless of being in SLI/CF or single GPU comparisons. Honestly, as an overclocker, I'd take advantage of the 7850's and 7870's superior overclocking headroom, so I'd grab one of them instead of the 660 for that reason, but to say that most non-heavy overclockers will probably be pleased with the 660 in most games at most reasonable settings configurations for the 660.

As for CF versus SLI, well, again, I like to overclock and in that usage, the 660 would generally lose by considerably large margins, even in SLI and even if it scales a little better.

 

[tomfreak]

[citation][nom]blazorthon[/nom]The 6870's placement offers context for people considering upgrading from Radeon 6800 or similarly performing models and also shows us how Radeon 6800's enhanced VLIW5 GPU handles MSAA compared to the competition. It might not help you in particular, but it is reasonable to include it because it might help some readers.[/citation]No it doesnt help me at all. it doesnt show me how the limited 384 kepler core perform. That GTX650 is shader bottleneck compared to its own ROP/Memory bandwidth. I'll like to know how serious is this when we put it on a serious test.

to put this as a simple conclusion I think this review article have put more focus on 660.(upgrade for 6870 user)

Would love a separate article just for 650 where it takes accounts with the old card user such as 9800GT/GTX200 series to choose between 7700/460 or 650.

 

[demonhorde665]

[citation][nom]RealiBrad[/nom]I always find it funny how people say they are a fan of this or that in the computer world. With how quickly things change, being a fan does nothing. Example: When AMD came out with their cpu that was better than the current Intel cpu's at the time, fans of Intel still bought Intel even though there was no reason to. At the time, AMD beat Intel in most every way. Then, Intel came back and beat the hell out of AMDs chips. In the tech world, you can only be a fan for a generation or so, because next round everything changes. A company that had horrible quality and or performance might dominate next time. Things change too quickly to stick with a brand forever. People need to be a fan of quality, performance and innovation in probably that order.[/citation]

also some people choose to be fans of a company based off ethics as well , sure brand X may not bea s good as brand A right now , but at least company X did not go around and pay people to not carry brand A. or brand X didn't try to make a monoploy like brand A did for example.

now im not really like this but i know many people that refuse to buy intel because intel has done multiple unethical buisness deals trying to control the cpu market. in fact if intel had not done these dirty deals it's likely AMD would be on a much more level playing field today, fincially, as it stands Amd just doesn't have the finances to compete in reasearch and development.