GeForce GTX 660 Ti Specifications and Launch Date Released

[Casecutter]

$350...?

If they want than to move it needs to be less. That what been Nvivia's equation for the GK104 and it work they better keep that momentum.

 

[Benetanegia]

they will, their will also be a special with 256 bit mem bus called the 660 GTUltra

looks good , they have the upper end well sorted , how about sorting the majority out now or is anyone who games going to have to spend new console money for just the graphics card nowadays, i might and you might but most dont my mates dont at all want to pay more then 200 max for gfx and they think thats high end< back in the real world here Beni

so far to me a prior alleged power user who considerred himself an enthusiast, all nvidias cards have been bordering on ridiculous re:: price

I only buy <$200 cards now too, and know what? That doesn't change even the slightest what is the best thing to do for Nvidia, as a company, in the current situation and it's the situation that drives their decisions and release dates. It's not really all that hard to understand really. The number of 28nm cards that can be sold is limited, very limited. GK106 was rumored to be around 200 mm^2, so they'd be able to sell 3 cards based on it for every 2 GK104 cards that they sell now. Exactly the same applies to HD7850 because it's the same size 218mm^2, so assuming same wafer allocation, an allocation that is completely limited by TSMC output, you'll have only 50% more HD7850's than GTX680's no matter what. Since it costs half as much, you need to sell 100% more, to get the same revenue but you can't, because you can only manufacture 50% more. This would apply to GK106 cards too. And the fact is that there's much more GTX680's sold than HD7850, and much much more 670's in this past month.

So Nvidia is just doing what it's best for them. Believing that Nvidia didn't release a mid-range because of any problem is naive, believeing that not releasing the mid-range is hurting them the slightest is naive. They don't releas it, because they don't need to and because for the time being 7850, HD7700, etc. don't pose a problem to them at all. They'll start worrying when HD7850's starts selling 50% more than they sell GTX680, or when HD77xx starts selling 300% more, until then, they are in the best position they can.

Personally, what I'd like to is to get some awesome $200 card (i.e a $200 GTX680 why not, asking is free), but I live in the real world and in the real world companies sell or try to sell their products at whichever price they want, and we as consumers the best we can do is pay with our wallets and inform other people so they vote with their wallets too. But the point at hand was not about us, it was about why Nvidia didn't release mid-range cards earlier, and it's because that's what it's best for them.

EDIT: The mayority buys what it's best for them in their price range, and don't care about it being a series 6 or 7 or 29. Both AMD and Nvidia have some pretty good cards in the $200 range and none of them really belongs to this gen. GTX560 and 570 and HD68xx and HD69xx cards are much better deals than the new gen almost everywhere, even to this day and Nvidia still has plenty of them, which is reason number 2 (or 1 depending on how you look at it) for not releasing the mid-range earlier.

 

[Xzibit]

And the fact is that there's much more GTX680's sold than HD7850, and much much more 670's in this past month.

Can you provide a link to your facts please ?

 

[Rauelius]

I think AMD's initial pricing for the 7xxx series really hurt them in my eyes. I don't know about all of you, but I was ready for a videocard upgrade when the 7970 came out. $500 was WAY too much, for the little performance increase I would have gotten over my current GTX460-SLI and Radeon 5850-CFX set ups (I've switched between the two depending what game ran best on what GPU). The 7950 was too much of a performance drop for the slight drop in price (the 7950 released at $450). The 7870 made the 7950 pointless, and right now when I was looking to do a GPU upgrade, they cancelled each other out, and I decided to wait longer. If AMD wasn't so greedy with the initial prices, I, and probably a few others out there, would have jumped on them a while ago. Glad I waited for the GTX670 as it's freaking amazing! Can't wait for the GTX660 and replace the GTS450-SLI set up I have in my media PC. Been considering a 7850, but until those get a price-correction to $175, I'll wait for the GTX660 Ti.

 

[Benetanegia]

Can you provide a link to your facts please ?

http://store.steampowered.com/hwsurvey/videocard/

Directx 11 GPUs

GTX680 0.73%

GTX670 0.58%

HD7850 0.44%

Notice that the 670 has been in the market for a much shorter period of time, meaning its share gain has been much bigger than the other 2 cards.

 

[Xzibit]

I think AMD's initial pricing for the 7xxx series really hurt them in my eyes. I don't know about all of you, but I was ready for a videocard upgrade when the 7970 came out. $500 was WAY too much, for the little performance increase I would have gotten over my current GTX460-SLI and Radeon 5850-CFX set ups (I've switched between the two depending what game ran best on what GPU). The 7950 was too much of a performance drop for the slight drop in price (the 7950 released at $450). The 7870 made the 7950 pointless, and right now when I was looking to do a GPU upgrade, they cancelled each other out, and I decided to wait longer. If AMD wasn't so greedy with the initial prices, I, and probably a few others out there, would have jumped on them a while ago. Glad I waited for the GTX670 as it's freaking amazing! Can't wait for the GTX660 and replace the GTS450-SLI set up I have in my media PC. Been considering a 7850, but until those get a price-correction to $175, I'll wait for the GTX660 Ti.

I've been waiting for the 660 TI but chances are its gonna be $100 over-priced. 460 and 560 eqivalent were released at $250.

Best guess is
650 @ 250
660 @ 300
660 Ti @ 350

Initial price for AMD was priced on performance vs Nvidias current offering. The current prices would have been much more sutable but why would they under-sell there product when nothing in the market was out to compete against it. There was a 3 month delay in competition.

 

[Nihilus]

I only buy <$200 cards now too, and know what? That doesn't change even the slightest what is the best thing to do for Nvidia, as a company, in the current situation and it's the situation that drives their decisions and release dates. And the fact is that there's much more GTX680's sold than HD7850, and much much more 670's in this past month.

Personally, what I'd like to is to get some awesome $200 card (i.e a $200 GTX680 why not, asking is free),
EDIT: The mayority buys what it's best for them in their price range, and don't care about it being a series 6 or 7 or 29. Both AMD and Nvidia have some pretty good cards in the $200 range and none of them really belongs to this gen. GTX560 and 570 and HD68xx and HD69xx cards are much better deals than the new gen almost everywhere, even to this day and Nvidia still has plenty of them, which is reason number 2 (or 1 depending on how you look at it) for not releasing the mid-range earlier.

Here you go, with $5 to spare.
http://www.techpowerup.com/reviews/Club_3D/HD_7850_RoyalQueen/

http://store.steampowered.com/hwsurvey/videocard/

Directx 11 GPUs

GTX680 0.73%

GTX670 0.58%

HD7850 0.44%

Notice that the 670 has been in the market for a much shorter period of time, meaning its share gain has been much bigger than the other 2 cards.

I would argue that is evidence, but not proof. Not everyone has a boner for Steam.

I've been waiting for the 660 TI but chances are its gonna be $100 over-priced. 460 and 560 eqivalent were released at $250.

Best guess is
650 @ 250
660 @ 300
660 Ti @ 350

From what the specs say, anything less for the 660 ti would be unlikely.

 

[Xzibit]

http://store.steampowered.com/hwsurvey/videocard/

Directx 11 GPUs

GTX680 0.73%

GTX670 0.58%

HD7850 0.44%

Notice that the 670 has been in the market for a much shorter period of time, meaning its share gain has been much bigger than the other 2 cards.

I was hoping you had inside knowledge from Nvidia or AMD since its hard to get the #.

Just so you know the Steam Survey you like to quote as fact is an "Opt-In". Not everyone on Steam is being pooled only those who choose to participate.

Steam Hardware & Software Survey

Steam conducts a monthly survey to collect data about what kinds of computer hardware and software our customers are using. Participation in the survey is optional, and anonymous. The information gathered is incredibly helpful to us as we make decisions about what kinds of technology investments to make and products to offer.

So its a fact that more GTX 680 and GTX 670 user participate/opt-in on the steam survey then Radeon HD 7000 series users.

Its a nice survey to point out but to come to the conclusion of sales from that survey alone is an enormous stretch.

I would like to know the units sold from each camp it would be interesting to see.

 

[Benetanegia]

Here you go, with $5 to spare.
http://www.techpowerup.com/reviews/Club_3D/HD_7850_RoyalQueen/

I said GTX680. HD7850 is not an upgrade over my card by my standards.

I would argue that is evidence, but not proof. Not everyone has a boner for Steam.

Semantics. And I'd like for someone to ever provide actual proof or evidence to show why they think Steam survey is not an accurate source, instead of the typical: Ha! You have to opt in and last time I didn't opt in, so it's flawed." and other similarly stupid claims.

EDIT: ^^ LOL. Posted as I was writing... see?

@Xzbit: Which chip exactly or which part of the chip in GK104 and Pitcairn, has the purpose of immediately take control of your head and respectively make you op-in or not, in that order?

 

[Xzibit]

I said GTX680. HD7850 is not an upgrade over my card by my standards.



Semantics. And I'd like for someone to ever provide actual proof or evidence to show why they think Steam survey is not an accurate source, instead of the typical: Ha! You have to opt in and last time I didn't opt in, so it's flawed." and other similarly stupid claims.

EDIT: ^^ LOL. Posted as I was writing... see?

@Xzbit: Which chip exactly or which part of the chip in GK104 and Pitcairn, has the purpose of immediately take control of your head and respectively make you op-in or not, in that order?

My guess would be the one that makes me belive a Opt-In survey has a direct corolation to units sold.


Steam survey is an accurate source for what its representing.

The issue is when someone (In this case u) uses it to make a direct comparison to sales figure/units sold. Thats just plain laughable.

 

[Benetanegia]

My guess would be the one that makes me belive a Opt-In survey has a direct corolation to units sold.


Steam survey is an accurate source for what its representing.

The issue is when someone (In this case u) uses it to make a direct comparison to sales figure/units sold. Thats just plain laughable.

Again provide evidence that Steam survey is biased toward any particular card or brand or stop the nonsense.

You can say it's not 100% accurate, as in reality would be: 0.70% for the 680 and 0.50% for HD7850. I'd still say the 680 has sold much more and I would still be right. Steam has over 40 million users, the opt in is offered to 1/12th of the users (randomly chosen), so that by the end of the year the sample size is equal to the number of users. Unless you can provide proof that thousands upon thousands of people are biased towards a certain election (opt-in or not) based on their elected hardware, please, stfu and simply accept that Steam survey is about the most perfect survey ever made, based on sample size alone.

 

[Nihilus]

I was hoping you had inside knowledge from Nvidia or AMD since its hard to get the #.

Just so you know the Steam Survey you like to quote as fact is an "Opt-In". Not everyone on Steam is being pooled only those who choose to participate.

He has used this 'evidence' before, glad to know I am not alone on this logic.

Semantics. And I'd like for someone to ever provide actual proof or evidence to show why they think Steam survey is not an accurate source, instead of the typical: Ha! You have to opt in and last time I didn't opt in, so it's flawed." and other similarly stupid claims.

EDIT: ^^ LOL. Posted as I was writing... see?

@Xzbit: Which chip exactly or which part of the chip in GK104 and Pitcairn, has the purpose of immediately take control of your head and respectively make you op-in or not, in that order?

Alright big guy let's think about this now... The GTX 670 and GTX 680 are high end cards. They will have more users than the 7800 series that can afford Steam and games that use Steam. If you you get a GTX 670 or 680, it's more than likely you are a hard-core gamer. If you show any kind of correlation between sales of the cards and Steam usage, I will gladly shut my mouth.

 

[Benetanegia]

Alright big guy let's think about this now... The GTX 670 and GTX 680 are high end cards. They will have more users than the 7800 series that can afford Steam and games that use Steam. If you you get a GTX 670 or 680, it's more than likely you are a hard-core gamer. If you show any kind of correlation between sales of the cards and Steam usage, I will gladly shut my mouth.

And that's why the first cards on the list are 560, 460 and 5770. All of them high-end cards which are more likely to be on Steam. Right.

And 550, 450, 540M, highest-end of the super enthusiast market.

 

[Nihilus]

Yes, agreed they are not high end cards. But those are also some of the highest selling cards ever. Even at a lower percentage of users than say a GTX 580, they would still have alot more users on Steam by far.

 

[Benetanegia]

Yes, agreed they are not high end cards. But those are also some of the highest selling cards ever. Even at a lower percentage of users than say a GTX 580, they would still have alot more users on Steam by far.

Or they simply sell at a similar rate as shown there...

If what you say was true we would see much more 7850.

The most played games are not precisely very demanding games. It's games like Counter Strike (1.6 and Source), TF2, MW 1,2,3, Dota 2, etc. I'd think there's far more evidence pointintg to Steam users in general NOT being enthuiasts. There's definitely NOT 40 million enthusiasts out there.

Or you can post evidence of why Steam is more crowded with enthusiast gamers than others. Why is there a higher percentage of high-end card owners in Steam? There's arguably more "gaming" card owners but both $250 and $500 cards are gaming cards. No one pays $250 for solitaire and web browsing.

We can post "What ifs" all they long, provide a proof of Steam Survey being biased or simply accept is the most accurate source we have access to.

EDIT: And btw you can be piss poor and have an Steam account. Forget about Steam sales which often sell you games for $2 and such. There's like 3 dozen free games right now and 2 of them are on the most played list almost everyday.

 

[Xzibit]

Again provide evidence that Steam survey is biased toward any particular card or brand or stop the nonsense.

You can say it's not 100% accurate, as in reality would be: 0.70% for the 680 and 0.50% for HD7850. I'd still say the 680 has sold much more and I would still be right. Steam has over 40 million users, the opt in is offered to 1/12th of the users (randomly chosen), so that by the end of the year the sample size is equal to the number of users. Unless you can provide proof that thousands upon thousands of people are biased towards a certain election (opt-in or not) based on their elected hardware, please, stfu and simply accept that Steam survey is about the most perfect survey ever made, based on sample size alone.

Wow, you are mentaly screwed up.

Even if you take what your dillusional mind has to offer it still doesnt have any solid facts to units sold as you were try'n to imply.

On the one hand your saying its an accurate survey for Steam users yet state its 1/12th offered to random users. Of those how many Opt-In you just recount there votes until years end until it equals Steams user base #. Obvious people who dont bother to Opt-In will never be counted and Opt-Ins will be counted multiple times that way.

If you can't understand how in-accurate that is. I cant help you..


You should be a political polester. Dead people vote too. Heck you can count there vote twice or as many times as you need for them to side with your view.

 

[HumanSmoke]

Or they simply sell at a similar rate as shown there...
If what you say was true we would see much more 7850..

Makes pretty good sense.
Everyone seems to point at a train of false logic; Midrange cards are more accessible than high end (cost) -> Pitcairn is midrange -> Pitcairn #'s must be better than GTX6__/or whatever.

HD 7870/7850 initial sales weren't actually that great as I remember. The HD 7850 offers similar performance to the HD 6950 but MIR's and non-reference OC'ed factory cards made the older series the better buy. A common refain from reviews (and forums) were that the prices were to high and that the buying market was largely nullified by the price hike from the previous series in the product stack (HD 6850/6870), and by competition from inventory clearouts of HD 6950/6970.

On the one hand your saying....

I thought it was fairly obvious. Steams user base has the ability to partake of the survey. UNLESS the owners of a particular card are more inclined to opt-in than any other it would stand to reason that survey sample is representative across all sectors....unless of course there is some undergound organization that gets in touch with (say) HD 7850 owners and instructs them to not partake as some sort of convoluted conspiracy to skew the only month-to-month numbers available.

I'd already posted previously a summary of discrete graphics cards sold by market segment (published by Mercury Research) it would also tend to supoort -roughly- the Steam segmentation in the past:

(Info available on the Investor Village boards amongst others- see their boards for the latest summaries)

 

[ViperXTR]

closely eyeing on the 660 series to replace my ol GTX 460 SE D:

 

[Nihilus]

I'd already posted previously a summary of discrete graphics cards sold by market segment (published by Mercury Research) it would also tend to supoort -roughly- the Steam segmentation in the past:
http://i.imgur.com/rqrm6.jpg
(Info available on the Investor Village boards amongst others- see their boards for the latest summaries)

Finally, this is all I asked for! This is a bit dated, but I will concede with this new EVIDENCE. Number of Amazon GPU reviews is probably more accurate a measure than Steam. At least the percent of people who write reviews/card is more consistent between various price ranges.

 

[Xzibit]

I thought it was fairly obvious. Steams user base has the ability to partake of the survey. UNLESS the owners of a particular card are more inclined to opt-in than any other it would stand to reason that survey sample is representative across all sectors....unless of course there is some undergound organization that gets in touch with (say) HD 7850 owners and instructs them to not partake as some sort of convoluted conspiracy to skew the only month-to-month numbers available.

I'd already posted previously a summary of discrete graphics cards sold by market segment (published by Mercury Research) it would also tend to supoort -roughly- the Steam segmentation in the past:
http://i.imgur.com/rqrm6.jpg
(Info available on the Investor Village boards amongst others- see their boards for the latest summaries)

Share % wasnt the arguement he was making. He was saying Steam Survey % reflect units sold as fact. Which i'm still pondering and laughing at because Steam doesnt take into account all GPU configurations and it points it out. It's like he has it in his head that every single person who opt-in one month wont opt-out the next or a new person will want to opt-in or not with the same configuration that someone else opted to balance it out month to month. Too many variable to consider it as a be all end all Fact.

If we were to translate to share % the numbers are still way off. They just dont match up.

Like i said i'm not disputing what the Steam Survey is representing but the fact that Benetanegia is using it as a fact basis for his theory to mean something totaly different.

Finally, this is all I asked for! This is a bit dated, but I will concede with this new EVIDENCE. Number of Amazon GPU reviews is probably more accurate a measure than Steam. At least the percent of people who write reviews/card is more consistent between various price ranges.

This is more up-to-date

JPR has more broken down numbers but you have to buy in to receive the charts

 

[Hilux SSRG]

Best guess is
650 @ 250
660 @ 300
660 Ti @ 350

That's my thinking as well for the prices when the cards are initially released. If these leaked specs are true for the 660 Ti, then what would the 660 non-Ti look like?

Instead I think these are specs are for the 660 non-Ti

and the 660Ti is actually:

6 SMX
1152 shaders
256 bit
2 gb memory

 

[alwayssts]

Late to the party commenting on this one, but it really should be 75% of 680 performance. Kind of a clever way they did it.

1536sp @ 1112mhz needs 6008ghz/256-bit.

75% (192-bit) of 1536sp with 1344 is the median of 670 clocks (~953mhz).

24 ROPs becomes a limiting factor around 1400sp when at the same clock...so a good mix there. If you figure 8 ROPs is equivalent to 1SM for yields, this obviously makes sense as the shaders do not need the ROPs, nor the memory bandwidth when at a low clock (which again helps yields).

With how GK104's memory controller scales, into the give-or-take 7ghz range, with a clock range on the 670 around the give-or-take 1100mhz range, that fits pretty well.

With >150w it will beat the 7870...and more importantly perhaps...a <150w 8850 if 1536sp (as one would assume that would be limited to 1050-1100mhz). With <150w it would still beat 7870, but potentially similar to an 8850. It will interesting to see how it clocks, and where it will typically run (mhz-wise).

I still don't think 7950 is part of the equation when you look at things with a fair eye, as overclocked that is always going to just flat-out be a stronger product and the stock clock is ridiculous to use as a comparison. That said, either way the mid-range (good-enough 1080p) market is about to get a much-needed kick in the balls. Odds are 660ti will be priced/perfed between 7870/7950, and amd will adjust prices to reflect that.

I really like the idea of 229/279/329. That almost sounds...I dunno...REASONABLE (finally)!

 

[Benetanegia]

Share % wasnt the arguement he was making. He was saying Steam Survey % reflect units sold as fact. Which i'm still pondering and laughing at because Steam doesnt take into account all GPU configurations and it points it out. It's like he has it in his head that every single person who opt-in one month wont opt-out the next or a new person will want to opt-in or not with the same configuration that someone else opted to balance it out month to month. Too many variable to consider it as a be all end all Fact.

If we were to translate to share % the numbers are still way off. They just dont match up.

Like i said i'm not disputing what the Steam Survey is representing but the fact that Benetanegia is using it as a fact basis for his theory to mean something totaly different.


This is more up-to-date

JPR has more broken down numbers but you have to buy in to receive the charts

You have no idea how statistics work. So I laugh at you. And nowhere did I say that the numbers for each month are added up. Whether they are added or not is irrelevant anyway, because each month the opt-in candidates are taken from the same random pool of 40++ million users. Any card from any segment is as likely to appear twice and even if only 1% opt-in which is deliriously low, your sample size is more than 33k++ people. A more reasonable cypher of 1 out of ten people opt-in and your sample size is 300k++ people. If you knew anything about statistics or simple math, you should have realized by now that it's impossible to introduce any bias towards any segment. This is something humansmoke already told you and you decided to ignore. You are also willingly ignoring the fact that if Steam Survey were to be innacurate it could also be skewing the results in HD7850's favour and not only the other way around and that in reality the GTX680 might have actually sold more comparatively than Steam says, so your point is wrong no matter what.

And I laugh at you for believing that JPR, Mercury research and the likes are so much more accurate, when their numbers are produced by taking into account shipping (and not actual retail sales, for all we know retailers could be having to thrw away certain cards several months later...) to certain retailers and distributors and then extrapolated. The sample size is way smaller than Steam and it's only more accurate to compare low-end cards and integrated, which Steam cannot really take into account.

 

[Desert Eagle]

A lot of good info in this thread. Thanks to all the contributors. I have suspected from the get-go that the current GTX 680 was originally intended to be the GTX 670 and the GK110 would be the GPU for the GTX 680 but when Nvidia saw the performance of Tahiti they shuffled things a bit so, as has been said here, they could make a bigger profit and the spectre of inadequate yields for the GK110 (due to the much higher transistor count) could be averted until the 28 nm process was better refined.

 

[Casecutter]

Not to get into the whole Steam argument, just try to wrap my mind around those numbers. So if such percentages are representative of the 40 million Steam that would mean:
.70% of 40 million = 28,000 while .58% = 23,200 or a total of 51,200 unit (670/680) just to those on Steam. So let's say April, May, June Nvidia has sold 17,066 to just the steam community each of those months?

Nvidia own chart provide about end of April showed shows they delivered about 4,200 units of the GTX580 first couple of weeks of that release. They show they had delivered 60% more 680's units to their AIB’s than they had 580's. Basically their chart shows that by early May they were close to delivering a total of 7000 units globally to AIB's. That not onto actual users hand it take 4-6 week to take a chip get on the card box and get them into the channel.

Those numbers are way too far apart?