[SA] GK110 aka GTX 680 release date: Late Q3 '12

Page 9 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

GaiaHunter

Diamond Member
Jul 13, 2008
3,697
397
126
Let me take a stab at this die size argument.

What Keysplayr is trying to say is AMD's 1st gen 28nm GPU is bigger than AMD's 1st gen 40nm GPU.
The die size being bigger from 1st gen to 1st gen will in effect make the die size bigger for the 2nd gen 28nm compared to 2nd gen 40nm.

Keysplayr is not comparing 2nd gen 40nm to 1st gen 28nm which has the later slightly smaller thanks to the process shrink.

To have any noticable increase in performance from 1st gen 28nm to 2nd gen 28nm, AMD will most likely have to add to the die, so 2nd gen 28nm will most likely grow in size to be bigger than Cayman. This shows a trend that AMD's GPUs are getting bigger over time.

IMO, I think AMD is trying to keep their die size around the same for each generation from here on out. Leave it around 200-220W for the high end single GPU, the overclockers can get some extra performance out of the chip, and the dual chip board will have enough headroom to be quite a bit better.

They getting bigger because they are also packing more GPGPU stuff, admittedly still completely irrelevant for most of us that only want them to play games.

Additionally it has a 384-bit memory controller instead of a 256-bit controller.

It is slightly bigger, it has 40% (on average) more gaming performance at stock than previous gen AMD card but the GPGPU performance is much higher (not that I as a consumer give a toss about it atm and for the foreseeable future it still seems unlikely I will). Of course that previous gen is actually 10% bigger - compared to the 5870, that is somewhat smaller (we talking about 5% difference in die sizes), the 7970 is 60%+ faster at relevant resolutions.

5870: 334 mm^2
6970: 389 mm^2
7970: 352 mm^2

Pitcairn should offer performance around 6970 levels at die sizes of ~250 mm^2 according to rumours.
 
Last edited:

Arzachel

Senior member
Apr 7, 2011
903
76
91
Keysplayr said:
o......m.......F......g........

Maybe I can put this in a pop up book form for some of you.

Stop deflecting and answer this.

MrTeal said:
By that logic, accounting for manufacturing process Sandy Bridge at 216mm^2 is much larger than McKinley at 421mm^2 since it is on a smaller node and has four and a half times as many transistors.

The whole argument is dumb. If you had said "since the 4XXX series" or said that the chips have more transistors at the same die area it would make sense, but for the argument as it is, die shrinks are meaningless, they don't magic away the measurements you make or something equaly inane.


Charlie commented at SA that AMD's next gen seems to be coming out a few months after GK110. Looks like AMD isn't refreshing Tahiti, they're aiming to get Sea/Canary Islands out as fast as possible. GK110 might be a top dog for even less than Tahiti.
 
Last edited:

Keysplayr

Elite Member
Jan 16, 2003
21,211
50
91
Stop deflecting and answer this.



The whole argument is dumb. If you had said "since the 4XXX series" or said that the chips have more transistors at the same die area it would make sense, but for the argument as it is, die shrinks are meaningless, they don't magic away the measurements you make or something equaly inane.


Charlie commented at SA that AMD's next gen seems to be coming out a few months after GK110. Looks like AMD isn't refreshing Tahiti, they're aiming to get Sea/Canary Islands out as fast as possible. GK110 might be a top dog for even less than Tahiti.
Haserath explained it beautifully. Far better than I did. As AMD strives to make their GPUs more GPGPU oriented like nvidias architecture is, AMDs die sizes will continue this growing trend. Simple.
 

tviceman

Diamond Member
Mar 25, 2008
6,734
514
126
www.facebook.com
Charlie commented at SA that AMD's next gen seems to be coming out a few months after GK110. Looks like AMD isn't refreshing Tahiti, they're aiming to get Sea/Canary Islands out as fast as possible. GK110 might be a top dog for even less than Tahiti.

And people thought cayman was going to blow Fermi out of the water when all they actually did was match gtx480 performance.
 

maddie

Diamond Member
Jul 18, 2010
5,151
5,537
136
Haserath explained it beautifully. Far better than I did. As AMD strives to make their GPUs more GPGPU oriented like nvidias architecture is, AMDs die sizes will continue this growing trend. Simple.
You know this how?
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
It's only logical, because in the HPC and the professional market is the big bucks. AMDs GPU division has not been as profitable as Nvidia for years.
 

formulav8

Diamond Member
Sep 18, 2000
7,004
522
126
From what we know now, im not convinced Cayman was suppose to be a 32nm design. I believe the 32nm design was scraped and a new 40nm design (Cayman) took its place when they learned that TSMC changed its plans. I believe GCN Tahiti is the 32nm design evolution in to 28nm process.


Well AMD are the only ones that outright knows. But the rumored 1920 shaders(don't remember the places I read that at) originally for Cayman sounds about right if it was released as a 32nm part compared to what it ended up being as a 40nm part with only 15xx. Also the release of Cayman would have put it around the 32nm release timeframe if it wasn't canceled? So I do think 32nm was their target process for Cayman. But I of course can't prove it. (I also think Tahiti was meant for a matured 32nm process).
 
Last edited:

AtenRa

Lifer
Feb 2, 2009
14,003
3,362
136
Well AMD are the only ones that outright knows. But the rumored 1920 shaders(don't remember the places I read that at) originally for Cayman sounds about right if it was released as a 32nm part compared to what it ended up being as a 40nm part with only 15xx. Also the release of Cayman would have put it around the 32nm release timeframe if it wasn't canceled? So I do think 32nm was their target process for Cayman. But I of course can't prove it. (I also think Tahiti was meant for a matured 32nm process).

I dont believe they intended to change to VLIW-4 at 32nm. I dont believe they EVER wanted to go with the VLIW-4. I believe they wanted to go with GCN at 32nm. They were forced to change to VLIW-4 in order to keep the die size under the 400mm2 mark at 40nm. Except Cayman (HD6970/50) no other chip was created with VLIW-4 to this day and i dont expect any in the future.

But yes, only AMD knows ;)
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
The gtx580 is EOL, which is why it's price isn't dropping. Look at it from this point of view - AMD released a card 13 months after the gtx580 was released, is ever so slightly faster, and costs more than what gtx580's have been bought for in the last several months.

That is not a win at all for consumers. The hd7950 is a really good card that is currently overpriced.

Wait, how is that not a win for consumers? I like how you word it as "slightly" and don't imply the same for the cost.

Which consumers are you referring to? The ones that own GTX 580s? Why does everyone generalize everyone owns top tier to make these absurd claims about "who wins and who loses?"

I'm a consumer, and this is a definite win for me: 10% more cost for 20% more performance. How do I lose?

Just answer this question and follow-up:
If you have a $450-550 buying budget, right now, and want to buy, right now, which would you buy: HD 7950, GTX 580, or HD 7970?

And of your selection, do you feel you lost out on something?
 

Stingercjc

Member
Sep 26, 2006
44
0
0
Found these specs on the net - looks like it beats a 7970 in everything!

oi7nep.jpg





Just trying to lighten up the mood in this thread. Things were getting pretty intense.
 

tviceman

Diamond Member
Mar 25, 2008
6,734
514
126
www.facebook.com
Wait, how is that not a win for consumers? I like how you word it as "slightly" and don't imply the same for the cost.

Which consumers are you referring to? The ones that own GTX 580s? Why does everyone generalize everyone owns top tier to make these absurd claims about "who wins and who loses?"

I'm a consumer, and this is a definite win for me: 10% more cost for 20% more performance. How do I lose?

Just answer this question and follow-up:
If you have a $450-550 buying budget, right now, and want to buy, right now, which would you buy: HD 7950, GTX 580, or HD 7970?

And of your selection, do you feel you lost out on something?

With this kind of logic, higher performance video cards should be escalating to higher price brackets with each new release. If you are happy buying tech hardware right now that is 5-10% faster for as much money, if not more money, than a competing product has been selling at for the last 14 months, then God bless you. Businesses love your anti-competitive attitude.

I, on the other hand, will patiently wait until the performance range that both of these products are currently in trickle down to lower price points.
 

formulav8

Diamond Member
Sep 18, 2000
7,004
522
126
I dont believe they intended to change to VLIW-4 at 32nm. I dont believe they EVER wanted to go with the VLIW-4. I believe they wanted to go with GCN at 32nm. They were forced to change to VLIW-4 in order to keep the die size under the 400mm2 mark at 40nm. Except Cayman (HD6970/50) no other chip was created with VLIW-4 to this day and i dont expect any in the future.

But yes, only AMD knows ;)

I still think it was intended to be a 38xx to 48xx situation. Cayman for maturing 32nm and gcn on 32nm after maturing. :p I guess we will never know what exactly thier original plan was... :cool:
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
With this kind of logic, higher performance video cards should be escalating to higher price brackets with each new release. If you are happy buying tech hardware right now that is 5-10% faster for as much money, if not more money, than a competing product has been selling at for the last 14 months, then God bless you. Businesses love your anti-competitive attitude.

I, on the other hand, will patiently wait until the performance range that both of these products are currently in trickle down to lower price points.

Wait, you didn't even answer my question and attacked my position. Nice. So basically, you admit defeat by refraining from answering?

Keep in mind, not everyone has a GTX 580 sitting in their rig. Those of us who don't have a GTX 580 - I don't see how this is a lost for us.

I've openly asked for competition, is it my fault nVidia isn't delivering (they can easily price cut the GTX 580 to make it more attractive)?

So, you're accusing me of being anti-consumer yet you're making excuses for nVidia and would tell people to wait for nVidia products if they wanted to buy now? Interesting.
 

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
Personally, I'm quite disgusted as a value seeking consumer that the GTX580 has barely dropped in price. Someone must be buying them even today at those prices, if they weren't selling any at all then the price would have to come down. It's not a matter of who is happy with prices, it's a tug of war where some of the consumers happen to be tugging for the corporate team.

With this kind of logic, higher performance video cards should be escalating to higher price brackets with each new release. If you are happy buying tech hardware right now that is 5-10% faster for as much money, if not more money, than a competing product has been selling at for the last 14 months, then God bless you. Businesses love your anti-competitive attitude.

I, on the other hand, will patiently wait until the performance range that both of these products are currently in trickle down to lower price points.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Personally, I'm quite disgusted as a value seeking consumer that the GTX580 has barely dropped in price. Someone must be buying them even today at those prices, if they weren't selling any at all then the price would have to come down. It's not a matter of who is happy with prices, it's a tug of war where some of the consumers happen to be tugging for the corporate team.

AMD users argument -------------- >
1) 7970 is cheaper or same price as 3gb GTX 580
2) Nvidia is at fault for not lowering their prices after 1.5 years
3) 7970 performs 20% better at stock and up to 80% better when oc'ed than 580
4) Nvidia has a history of overcharging and ripping customers off
5) Kepler will be released in 2021

NVIDIA users argument -------------->
1) 7970 is too expensive! How dare AMD charge the same as NV
2) 7970 is *only* 20% faster than a 580 at stock
3) AMD is charging too much
4) AMD is abandoning their sweet spot strategy!
5) NV is justified in keeping the price high
6) AMD has a history of charging reasonably, screw them for not this time!
7) 8970 will have a die the size of montana

I predict this dead horse will be beat dead, dead, dead for a few months
 
Last edited:

NIGELG

Senior member
Nov 4, 2009
852
31
91
AMD users argument -------------- >
1) 7970 is cheaper or same price as 3gb GTX 580
2) Nvidia is at fault for not lowering their prices after 1.5 years
3) 7970 performs 20% better at stock and up to 80% better when oc'ed than 580
4) Nvidia has a history of overcharging and ripping customers off

NVIDIA users argument -------------->
1) 7970 is too expensive! How dare AMD charge the same as NV
2) 7970 is *only* 20% faster than a 580 at stock
3) AMD is charging too much
4) AMD is abandoning their sweet spot strategy!
5) NV is justified in keeping the price high
6) AMD has a history of charging reasonably, screw them for not this time!
7) 8970 will have a die the size of montana

I predict this dead horse will be beat dead, dead, dead for a few months
8)Nvidia has PhysX,CUDA AND 3D....Therefore it merits a premium....and with the tons and tons of GPU physX titles each year it's worth it....
9)AMD drivers sucks all kind of ass....



o_O.
 

notty22

Diamond Member
Jan 1, 2010
3,375
0
0
GTX 580's are sporting rebates and discounts. Also you can get the premium models for the 450-500 that AMD is getting for the 7950, which is not that much faster or equal at stock speeds. Many gamers don't care whether they are getting 28nm or 40nm. Or they care as much as understanding they are getting the top model in a given lineup, or the second tier card etc.
450.00 after rebate.
http://www.newegg.com/Product/Product.aspx?Item=N82E16814162073
http://www.overclock.net/t/1159043/galaxy-gtx-580-thread-58nlh5hs3pxz o/c discussions here
c5d64c11-a697-40eb-8a05-f2267244ecc3.jpg



73857076.jpg
 

maddie

Diamond Member
Jul 18, 2010
5,151
5,537
136
Wait, you didn't even answer my question and attacked my position. Nice. So basically, you admit defeat by refraining from answering?

Keep in mind, not everyone has a GTX 580 sitting in their rig. Those of us who don't have a GTX 580 - I don't see how this is a lost for us.

I've openly asked for competition, is it my fault nVidia isn't delivering (they can easily price cut the GTX 580 to make it more attractive)?

So, you're accusing me of being anti-consumer yet you're making excuses for nVidia and would tell people to wait for nVidia products if they wanted to buy now? Interesting.
I agree with your argument.

At any point in time, when someone is upgrading their video card, the actually available cards on sale are the only possible options, or you can wait.

Again, for the rabid naysayers, and (please no BS about traditional pricing, unknown future cards, etc, etc).

If buying now, which card will you choose NOW to beat the 7850 and 7870?
 
May 13, 2009
12,333
612
126
8)Nvidia has PhysX,CUDA AND 3D....Therefore it merits a premium....and with the tons and tons of GPU physX titles each year it's worth it....
9)AMD drivers sucks all kind of ass....



o_O.

I actually use 3D on a regular basis. PhysX /shrug hey it's a free feature and adds something to the game.

AMD still doesn't have a whql driver for the 79XX series. Take whatever you want out of that. I can't give opinions on AMD products or drivers. :sneaky:
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
I actually use 3D on a regular basis. PhysX /shrug hey it's a free feature and adds something to the game.

AMD still doesn't have a whql driver for the 79XX series. Take whatever you want out of that. I can't give opinions on AMD products or drivers. :sneaky:

I actually do all those too. :D HD 5870 + 9800 GTX+ hybrid build.

Note: If anyone cares, the nVidia issue of the Display Stopped Responding persist in my rig even though the AMD card is the primary card. It's actually kind of funny haha.
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
8)Nvidia has PhysX,CUDA AND 3D....Therefore it merits a premium....and with the tons and tons of GPU physX titles each year it's worth it....
9)AMD drivers sucks all kind of ass....



o_O.

9) Yeah the 1 physx title produced in the past 11 months, definitely sounds worth a 150$ premium o_O
10) AMD drivers killed my puppy and burned my house down
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
Notty: where's the 3gb GTX 580 that is 450$ after discount? I don't think a 1.5gb card would do too great in 3d surround ;)

I'm adding to the dead horse, sorry, carry on everyone
 

NIGELG

Senior member
Nov 4, 2009
852
31
91
9) Yeah the 1 physx title produced in the past 11 months, definitely sounds worth a 150$ premium o_O
10) AMD drivers killed my puppy and burned my house down
LOL...I was being sarcastic:cool:.Just echoing some of the known pro NV arguments....
 

96Firebird

Diamond Member
Nov 8, 2010
5,739
334
126
I actually do all those too. :D HD 5870 + 9800 GTX+ hybrid build.

Note: If anyone cares, the nVidia issue of the Display Stopped Responding persist in my rig even though the AMD card is the primary card. It's actually kind of funny haha.

Who would have thought, a system with two different video cards that require two different drivers has problems. :awe: