Kepler Nvidia GeForce GTX 780

[Dimitriman]

For this card to match Charlie's description it would have to either:
1- Be faster than 7970, in which case it will likely cost +/- 50$ from $550
2- Be around 570~580 performance and cost 300$

Both those claims above will likely never occur together with Nvidia. Unless they have an incredibly abnormal reason to do so.
Never seen NV have an Ace in the market and undercharge for it.

[boxleitnerb]

So we have two cards with approximately the same die size that could be close in performance. Why is one of them midrange and one highend? Highend in my opinion is reserved for the big chips in the range of 500mm2.
If Tahiti were highend, what will GK110 be, then? Super highend? Ultra highend? Makes no sense. Again: Look at the 8800GT. Same scenario.

[Dimitriman]

So we have two cards with approximately the same die size that could be close in performance. Why is one of them midrange and one highend? Highend in my opinion is reserved for the big chips in the range of 500mm2.
If Tahiti were highend, what will GK110 be, then? Super highend? Ultra highend? Makes no sense. Again: Look at the 8800GT. Same scenario.

You could not make Tahiti 500mm2 without significantly gimping its clock speeds as it already has 210W consumption with 320mm2. Just because it is smaller doesn't mean it is not tuned for high end.

[boxleitnerb]

Then again, what will GK110 be? Or do you honestly believe that GK104 will remain Nvidias top dog? 7950 will offer similar performance to GTX580 at a similar price but lower power consumption. This is perf/$ stagnation and it is not good for the customer.

Fermi was borked. Imagine Nvidia did everything right back then like it appears for Kepler now. Imagine a fixed, more efficient 460 aka 560(Ti) had been released instead of the 460, fighting with the 5870, coming close. This is the exact same situation that we will have in 2 months. Compared to a fixed 480 aka 580, 5870 is not highend. Equally, compared with GK110 a 7970 is not highend. True, there is a time shift in between these two. It changes perception, but not the facts.

[Kuroimaho]

Dunno, seems people are expecting a 300$ card that is faster and consumes less power than a 7970 from what I've been reading...

Those people are in for a rude awakening. If NV has a faster card than a 7970 why would it sell it for 300$ ???

[boxleitnerb]

It will probably not be faster. I expect GK104 around 7970-10% and 160-180W.

[Iconyu]

So we have two cards with approximately the same die size that could be close in performance. Why is one of them midrange and one highend? Highend in my opinion is reserved for the big chips in the range of 500mm2.
If Tahiti were highend, what will GK110 be, then? Super highend? Ultra highend? Makes no sense. Again: Look at the 8800GT. Same scenario.

The 460 has less transistors but is bigger than the 5870 (if I remember correctly), while 260 is bigger and has more transistors than the 4870. Size alone doesn't dictate performance and Nvidia will price as high as they can get for each card, that price is dictated by competition, so $300 is high for a basic mid range card. I'm guessing nvidia are probably expecting a price drop or new performance drivers to go up against the 660GTX launch.

The 8800GT wasn't on the same process as the 8800GTX, the GTX was on 90nm and the GT 65nm. Personally I think Nvidia put in a lot of effort this generation to make sure they got 28nm right, while AMD were making cutbacks.

[SKYMTL]

What I see is people hyping their expectations up to a point where nothing short of mind blowing performance for less than $400 will satisfy them.

To me, that's completely unrealistic.

[boxleitnerb]

Not one of you has given a good reason why. Why should the successor to a 300$ card be priced considerably higher? Why is +60% on a new node completely unrealistic? If you follow this logic, chips will become faster but more expensive. So in 5 years we have +200% performance and +200% price? That is completely unrealistic.
I don't get why you deny that perf/$ will increase considerably - as it did with every new generation eventually...in every market segment.

[Kuroimaho]

What I see is people hyping their expectations up to a point where nothing short of mind blowing performance for less than $400 will satisfy them.

To me, that's completely unrealistic.

Unrealistic is the proper description. Why would you sell a card which is faster than your competition's 550$ card for nearly half the price ? Has it happened before ? Then why would it happen now ?

At this rate no matter how good the next gen NV cards become it will still disappoint people who expected it for half the ATI prices... o.O

[-Boris-]

Computerbase tested with a 4,5 GHz 2600K. The influence of the CPU should diminish at these high settings, but it depends on the benchmark of course. I chose these reviews because they represent many games and have a performance rating for convenience. Can you point out at least 10 games where the 7970 is 60% ahead of the 6970?

I like the reviews of SweClockers.com, but it's in swedish. They try to eliminate bottlenecks and try to isolate the actual perfromance of the cards. They too have an Index, and there you can read that 7970 is 48% faster in 1920x1080 and 2560x1600. The bigger the difference you can get in real world test the less bottlnecks is involved and the closer you are to the real performance difference. They have a number of games where you can see about 60% better performance.

[-Boris-]

Not one of you has given a good reason why. Why should the successor to a 300$ card be priced considerably higher? Why is +60% on a new node completely unrealistic? If you follow this logic, chips will become faster but more expensive. So in 5 years we have +200% performance and +200% price? That is completely unrealistic.
I don't get why you deny that perf/$ will increase considerably - as it did with every new generation eventually...in every market segment.

7970 is a successor to a much cheaper card. But as long there isn't faster chips then new chips are priced around 500$+. When they are releasing an even faster card, Then will 760 drop in price.

[Kristoferr]

Unrealistic is the proper description. Why would you sell a card which is faster than your competition's 550$ card for nearly half the price ? Has it happened before ? Then why would it happen now ?

At this rate no matter how good the next gen NV cards become it will still disappoint people who expected it for half the ATI prices... o.O

True. NVidia, like every other company exists to make profit. On marketing point of view, there is no reason to sell the card as good as 7970 with nearly half price. In other words. It would be completely stupid.
Not going to happen.

[Pantsu]

7970 has high demand as the fastest card, and low supply due to a fresh node that's more expensive to manufacture with. But as long as people are willing to buy at this price, they will sell. Nvidia will do the same, and balance the price according to demand and supply. And demand will of course be affected by what the competition is offering.

[RPGWiZaRD]

The only realistic scenario that exists in my opinion is that:

- If priced $300 is true then:
- If GTX 580 is gonna get EOL it needs to be roughly as fast as GTX 580 ( +/- ~10% )
- GTX 580 performance for $200 dollar less with a bit lower power consumption and DX11.1 = there's your new gen benefits, very fair
- If you can't beat competition in performance despite the competitor was first => start a price war (offer better performance/price ratio)
- $300 GK104 puts pressure on HD 7950 & HD 7970, might have to drop to AT LEAST $400 and ~$500 as I doubt performance difference between GK104 and HD7950 will be big
- Everyone benefits !
- (Compare 8800GT vs 8800GTX except this time $50 expensier, similar scenario has happened before in Nvidia camp)

[tajoh111]

Unrealistic is the proper description. Why would you sell a card which is faster than your competition's 550$ card for nearly half the price ? Has it happened before ? Then why would it happen now ?

At this rate no matter how good the next gen NV cards become it will still disappoint people who expected it for half the ATI prices... o.O

Yes it has happened, but on the other shoe. That product was the 4870.

The gtx 280 and gtx 260 at launch were priced at 650 and 450 respectively. The price of the 4870 was 299 or 280 and was the same speed or hairs faster than the gtx 260. I think the gk104 will be around 7950 ish as far as speed goes. Considering the size of past midranges from NV and their performance, them catching up to mid range AMD high end would not be a surprise(gtx 9800+, gtx 260, gtx460, gtx 560 all performed similarly to the 4850, 4870, 5850 and 6950 respectively).

As far as pricing goes, I think they want to price things in respect to gk110 or gk112 or whatever the high end is. I think at this point in time, the reason the 7970 is selling out is primarily a supply one, however quantities are getting much better quickly and it is getting easier and easier to find one in Canada. Best buy of all places has them at stock, at the 549.99 retail price. NCIX has a decent amount of stock in too. In a month, I have a feeling it will be pretty easy to find a 7970.

I think if gk110 is hard launched, I expect decent quantities for that card considering it will be launched later. There is a point where people don't want to pay anymore and $650 I think was max for a card that is not semi limited edition. During that time we were not in a recession and people had way more money to spend. I think right now the max people will pay for a card is less than 650, if it has decent supply(not a weird halo product like these limited editions(the ultra) or dual cards). Pricing the gtx 660 at 300 dollars leaves room for a $550 or $600 dollar, gtx 680.

[SubZero.it]

What I see is people hyping their expectations up to a point where nothing short of mind blowing performance for less than $400 will satisfy them.

To me, that's completely unrealistic.

+1 What I said the page before

[Lanek]

........... I think right now the max people will pay for a card is less than 650, if it has decent supply(not a weird halo product like these limited editions(the ultra) or dual cards). Pricing the gtx 660 at 300 dollars leaves room for a $550 or $600 dollar, gtx 680.

IF, performance are soo good, i dont know why peoples assume nvidia cant put the GK104: 380 -400 $ - GTX 670: 500$ - GTX680: 570 - 580$.... If really the GK104 perform so well, Why will you Nvidia cares about it? If a price wars occurs ( specially if AMD release a 7950 1.5gb ), they dont want to be to low on MSRP....

[highoctane]

For this card to match Charlie's description it would have to either:
1- Be faster than 7970, in which case it will likely cost +/- 50$ from $550
2- Be around 570~580 performance and cost 300$

Both those claims above will likely never occur together with Nvidia. Unless they have an incredibly abnormal reason to do so.
Never seen NV have an Ace in the market and undercharge for it.

8800gt was an Ace, the card rocked in its day and performed close enough to the 8800gtx for a fraction of the cost, bought one on launch day from NCIX for $189 compared to the $6xx I paid for a 8800gtx.

I really cannot recall anything positive about Nvidia from char-lie so for him to even be able to squeak out any kind of praise is enough to raise some brows and warrant some attention.

We'll all know soon enough, whether it beats or loses to the 7970 is irrelevant, it comes down to price and performance in the end.

[Lanek]

The thing was completely different at this time, the difference between high end and middle end was not the same., todays it will mostly be a 570 and not a 560... the 8800GT ( die shrinked ) was coming with 112SP vs 128SP for the GTS, the clock speed was upped a little bit.. the 90 > 65nm specially used for decrease the price of the GTX.. not really for make a new series. the difference was only of 14tflops Even the difference between the 90>65nm ( ultra, GTX > GS, GT ) was small ( 65nm : 750 vs 690millions of transistors for the 90nm )

As you see the 8800GT ( as the GTX ) was part of the high end SKU, with the same count of transistors, of 690 and then 750 millions.. ( laser cutted )... same as the 570 is for the 580... not a different sku..

8800ultra: 128SP 90nm 690millions transistors _________________ - 612mhz
8800GTX: 128SP 90nm 690 millions transistors. _______ 518Tflops - 575mhz
8800GTS: 128SP 65nm 754millions transistors. _________________ - 650mhz
8800GT: 112SP 65nm 754 millions transistors _______ 504Tflops - 600mhz

[highoctane]

The market demand at the time would agree that the 8800gt was in fact an ace and could be considered underpriced at launch when you consider market demand driving the price up nearly $100 over launch day prices for an extended period of time. What made the 8800gt such a success was the simple idea of price/performance, you got allot of performance for the price. Nobody cared about process, transistors or flops, it was all about price and performance.

There's not allot of sense to me in debating the price, performance & market placement of hardware that we know too little about as of yet.

We'll just have wait until launch to know what Nvidia's next gen is going to bring to the table as far as price/performance is concerned.

[LordEC911]

2- Be around 570~580 performance and cost 300$

Between GTX560Ti 448 and GTX580.
I do agree with you though and hope it is at least around or above GTX570.
Should be pretty competitive with Pitcairn.

[tajoh111]

Between GTX560Ti 448 and GTX580.
I do agree with you though and hope it is at least around or above GTX570.
Should be pretty competitive with Pitcairn.

The gtx 570 and 6970 has been around 330 dollars since the middle of the 2011, maybe even earlier. The gtx 560 448 at $289 hit around the gtx 570 performance as well, heck even the gigabyte gtx 560 SOC hit gtx 570 performance at around $269 and this was 10 months ago. Unless GK104 is well well under 300mm2 while using a new process, it should hit gtx 580 performance (which is at this point, looks like 10% with the 7950) level atleast if the memory controller is fixed.

If from going to gf114 to gk104, Nvidia only gained, 15-20%(gtx 570 performance) while using a new architecture and new process, it would be a massive failure, worse than the 2900xt or the 5800xt in someways because both companies have learned their lessons by now. In addition, with the 6870 going for 180 dollars and gtx 560 ti going for 200, it would be a bad value proposition. Not to mention the gtx 560 launched at $229 and getting gtx $570 performance at $299 would mean worse price to performance even from launch prices to launch price. With a mainstream card, this would be idiotic.

The only unrealistic thing is not so much performance I think from GK104, I think it is more so price. $349 seems like a more realistic price to me.

[zalbard]

IF, performance are soo good, i dont know why peoples assume nvidia cant put the GK104: 380 -400 $ - GTX 670: 500$ - GTX680: 570 - 580$.... If really the GK104 perform so well, Why will you Nvidia cares about it? If a price wars occurs ( specially if AMD release a 7950 1.5gb ), they dont want to be to low on MSRP....

It often makes sense to sell more units cheaper rather than just a few at a high price. If yields are good then this is entirely possible.

[paulbagz]

All these comparisons to the 8800GT makes my head sore.

-PB