Big Kepler news

[chimaxi83]

Uh.

Is it just me or 50% cores and double the memory bandwidth should bring more than 20-25% increase?

Depends on how much of that will be compute oriented.

 

[Don Karnage]

And I based $600 based on GTX 680 being objectively better than 7970 not subjectively. GTX 680 is no reason to upgrade from a 7970 and in some games is a downgrade. It is like comparing GTX 570 to Radeon 6970 not 580 and 6970. So again, it wasn't exactly wrong

If i had a 7970 i'd be pissed because i wasted more money on an inferior card compared to the 680. Amd fans can say whatever they want about the 680 but its the better card in every aspect.

 

[Edrick]

If i had a 7970 i'd be pissed because i wasted more money on an inferior card compared to the 680. Amd fans can say whatever they want about the 680 but its the better card in every aspect.

Wrong. The 7970 destroys the 680 in GPU Compute. Not everyone only uses a computer for gaming.

 

[Don Karnage]

Wrong. The 7970 destroys the 680 in GPU Compute. Not everyone only uses a computer for gaming.

Yeah that whole 1% who complain about its horrible compute numbers

 

[Genx87]

Yeah that whole 1% who complain about its horrible compute numbers

AMD fans have been waiting for years to make that claim

 

[blckgrffn]

Yeah that whole 1% who complain about its horrible compute numbers

I wouldn't underestimate the number of GTX 580/590 owners who don't ever even use them for games, ever.

Haha, I've seen a GTX 590 combined with a A64 3700+ Clawhammer in the configurations others are using.

That guy had three dual 590's and six dual 580's rigs. All by his lonesome.

I'd be really curious about numbers in this demographic. Obviously, they will never show up in Steam surveys, etc.

AMD fans have been waiting for years to make that claim

That is pretty rich, isn't it?

If I had either a GTX 680 or a 7970 I'd be pretty happy right now

Well, I'd probably be peeved that neither of them are really worth a crap in distributed computing simply from a lack of support standpoint.

 

[aaksheytalwar]

Honestly, if you knew that the 680 would be 5-10% faster at stock and 0% faster once overclocked, only an idiot would have waited 2 months to save $50 and then go for a 680.

I don't care if the 7970 is $50 more, that is irrelevant to me. I just don't want to wait 2-3 months for pretty much the same performance that I can have 3 months in advance.

And people are crying over the 7970 because it is more expensive, had it been from nvidia they would have been perfectly happy. But since AMD is more expensive this time they are crying.

Had the 7970 been around $475, nobody would have minded much. And had they been selling 7970 OCs for $500 which equal 680 in performance, there wouldn't have been a problem. And 100% 7970s reach 1050 core and 1425 mem, 100% of them. It is a matter of just using your mouse.

If of course a person doesn't know how to use a mouse, then I understand.

 

[Don Karnage]

I just have to ask. What programs make use of the 7970's better compute numbers?

 

[BallaTheFeared]

I just have to ask. What programs make use of the 7970's better compute numbers?

lol.. Yeah I was going to ask what kind of performance increase the 7970 gets in CS5.5 but then I remembered it's not supported.

 

[aaksheytalwar]

I am only talking about games. 10% difference at stock with a 3 months delay is not acceptable to me. And practically it is a 0% performance diff once I overclock high enough because I would overclock both. That means a 0-5% performance diff max. I don't mind spending $50 extra to get 97% of the performance 3 months in advance when my card will probably last just about 12-15 months max anyway.

 

[Don Karnage]

I am only talking about games. 10% difference at stock with a 3 months delay is not acceptable to me. And practically it is a 0% performance diff once I overclock high enough because I would overclock both. That means a 0-5% performance diff max. I don't mind spending $50 extra to get 97% of the performance 3 months in advance when my card will probably last just about 12-15 months max anyway.

50 dollars extra to own Amd's flagship just to lose to Nvidia's 660 Ti... I mean 680

 

[Edrick]

50 dollars extra to own Amd's flagship just to lose to Nvidia's 660 Ti... I mean 680

No, more like $50 extra to own a "next gen" card since December instead of waiting to "try" and get one in April.

I am not AMD fan (never owned one), but have to give AMD credit for beating Nvidia out of the gate twice now (fermi anf kepler).

 

[Genx87]

50 dollars extra to own Amd's flagship just to lose to Nvidia's 660 Ti... I mean 680

I guess if this was 3 months ago his point would make sense. Today why would anybody pay more for less?

 

[Genx87]

No, more like $50 extra to own a "next gen" card since December instead of waiting to "try" and get one in April.

I am not AMD fan (never owned one), but have to give AMD credit for beating Nvidia out of the gate twice now (fermi anf kepler).

Who cares, honestly. It was a few months. Both times Nvidia won in performance.

 

[aaksheytalwar]

If I were to buy a card today, it may have been GTX 680. But I won't wait 2-3 months without an official announcement, I can't hang on and stay put with no knowledge of what will happen next.

$50 isn't for AMD's flagship persay

It is to have similar performance 3 months in advance. IMO 3 months of use is worth more than $50 to me

 

[Edrick]

Who cares, honestly. It was a few months. Both times Nvidia won in performance.

4 months in the technology field is a long time...considering refreshes come out anywhere between 8-12 months. Also, when a product comes out 4 months after a competitor's product, most of the time it will perform better (ignoring BD).

If Nvidia was first out of the door with a 680, you can bet your arse that the 7970 would look different than it does today. Just saying. Again, I am not an AMD fan by any means, but I try to be as fair as possible and give credit where it is due.

AND should have dropped the price of the 7970 to $449 once the 680 was released. But I do not work for AMD.

 

[taltamir]

512 bit memory bus?

they used 512 bit memory bus before.
But back then it was coupled to GDDR3 to compete with AMDs 256bit GDDR5.
512bit @ GDDR5 should be impressive.

 

[BallaTheFeared]

4 months in the technology field is a long time...considering refreshes come out anywhere between 8-12 months. Also, when a product comes out 4 months after a competitor's product, most of the time it will perform better (ignoring BD).

If Nvidia was first out of the door with a 680, you can bet your arse that the 7970 would look different than it does today. Just saying. Again, I am not an AMD fan by any means, but I try to be as fair as possible and give credit where it is due.

4 months?

AMD paper launched in Dec, actual shipping took place in January.

It would be an interesting market if Nvidia came out of the gate first with the GTX 660 Ti for $250, I mean at least for consumers... AMD would have probably imploded.

 

[Edrick]

It would be an interesting market if Nvidia came out of the gate first with the GTX 660 Ti for $250, I mean at least for consumers... AMD would have probably imploded.

Things would be a lot different today if that were the case. GK104's for $250 would be real nice. I would have 2 already.

 

[SlowSpyder]

Depends on how much of that will be compute oriented.

Yea, it'll probably be a DP performance monster. It'll be faster than the GTX680 obviously, but I doubt it'll absolutely kill it. And by the time it comes out, we may have an 8970 on its way... good things all around in fall, hopefully.

Yeah that whole 1% who complain about its horrible compute numbers

Why would I be disappointed or pissed off? I've been gaming on my 7970 for months now. If I buy a GTX680, should I be pissed if BigK launches in August and it is a better gaming card? There is always something faster in the works, this is the nature of computer hardware.

And I think you'd be surprised at how many people do some sort of compute work with their cards these days. AMD users enjoy great bitcoin performance, if you factor that in the card will pay for the whole 20 watts more then the GTX680 it uses with money to spare.

And to be honest, I plan on keeping this card for a while. I think the 50% additional memory will be more useful than the out of the box stock clock 10-15% performance advantage the GTX680 has, down the road.

 

[dangerman1337]

512 bit bus? Consdiering the GK104 is 256 bit bus I can't imagine the big one having 51bit bus and the all the HPC stuff that it'll likely be on it and have a TDP of 250.

 

[Edrick]

512 bit bus? Consdiering the GK104 is 256 bit bus I can't imagine the big one having 51bit bus and the all the HPC stuff that it'll likely be on it and have a TDP of 250.

Yea, I wouldn't be surprised to see a 300 TDP.

 

[Jaydip]

People serious about compute use quadro or firepro they don't use a gaming card.It doesn't matter when u start the race it only matters when u finish.Now a days when a 560ti or 6950 is probably enough for most of the console ports this kind cards are getting irrelevant by each day.Also at my place 6990 or 590 is a better buy than a 7970.

 

[Lonyo]

Based on the specs, it should end up having a bigger gap between itself and GK104 as GF110 had to GF114. When comparing both generations, big Kepler has a bigger leap of cores over GK104 AND will have a bigger leap of memory bandwidth. GTX580 was generally 30-40% faster than GTX560ti, so given that I expect big Kepler to be at least 40% faster across the board than GK104.

There is no way the highest end (consumer) video card this chip shows up as will be $499 at release. It will probably be around $649.99 MSRP to start, and will likely (eventually) have two cut down versions that will probably come in at $549.99 and $449.99. This pricing scheme will give room for an updated GK104 based card to come in at around $399.99.

Bigger leap of cores?
384 -> 512
1536 -> 2048
Sure, the number is bigger, but it's still the exact same percentage increase, 33%.

 

[Concillian]

Things would be a lot different today if that were the case. GK104's for $250 would be real nice. I would have 2 already.

Would you?

They can't keep them in stock at current ~$500. What do you think demand would be like at ~$250. Buying one would require an army of browser tabs at different e-tailers with a macro that cycled through the tabs hitting F5 and looking for stock.

I think that's the X factor this generation. TSMC is being pushed on the 28nm mobile SoC front harder than they're being pushed from the 28nm GPU front. There's never been a GPU die shrink that was coupled with the current market demand for SoCs. I just don't think that history is going to be a good predictor this time around. We have to expect prices will start and remain high for longer than "usual".