4GB GTX680 on the way

Feb 19, 2009
10,457
10
76
A 4gb gtx680 should end any reason to have 7970s at comparable prices, as it should really scream in multi-card and multi monitor setups.

But given the bad pricing so far on 28nm, it's prolly going to go at $600.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
A 4gb gtx680 should end any reason to have 7970s at comparable prices, as it should really scream in multi-card and multi monitor setups.

But given the bad pricing so far on 28nm, it's prolly going to go at $600.

Increasing the amt. of RAM doesn't increase bandwidth. The 7970 is still going to win there. It's going to be a really rare situation where 2GB of RAM isn't enough.
 

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
Interestingly enough, one of the slides shows 8+8 power connectors. I wonder if that will be a different model? It shows the TDP at 300w!! That could be amazing...
 

SolMiester

Diamond Member
Dec 19, 2004
5,330
17
76
Interestingly enough, one of the slides shows 8+8 power connectors. I wonder if that will be a different model? It shows the TDP at 300w!! That could be amazing...

Just had it pointed out to me that 8+8 = 375TDP, 300 is only 6+8.....lots of OC room.....and if the memory goes up to 7ghz, we will really start to see some benefit...
 

Skurge

Diamond Member
Aug 17, 2009
5,195
1
71
Looks like a pointless card, the 680 is already faster, more memory would fix it's one achilies heel which is bandwidth, just make it more expensive and 2GB of ram is more than enough, 3GB on the 7970 is overkill.

Although thats what some nvidia fans said about 2GB, now they are excited about 4GB? Whatever, at least its not as bad as that 6GB 7970.
 

Grooveriding

Diamond Member
Dec 25, 2008
9,117
1,266
126
Looks like a pointless card, the 680 is already faster, more memory would fix it's one achilies heel which is bandwidth, just make it more expensive and 2GB of ram is more than enough, 3GB on the 7970 is overkill.

Although thats what some nvidia fans said about 2GB, now they are excited about 4GB? Whatever, at least its not as bad as that 6GB 7970.

You mean wouldn't I assume ? :)
 

Stayfr0sty

Senior member
Mar 5, 2012
465
0
0
Increasing the amount of RAM on a card is just a marketing gimmick.
Hence you see low to mid range cards with so and so amount of RAM and the unaware/ignorant n00b thinks its gonna be a beast cause of all the RAM.
What matters is the memory architecture and bandwith and shader count, period.
 

kidsafe

Senior member
Jan 5, 2003
283
0
0
The practical limitations for OCing both the 680 and 7970 are probably the 28nm process itself and air/water-cooled temperatures. 12 power phases and 8+6 PEG is likely completely unnecessary unless you're on LN2, and then both cards have already been OC'd to 1800-1900MHz.
 

dust

Golden Member
Oct 13, 2008
1,328
2
71
Increasing the amount of RAM on a card is just a marketing gimmick.
Hence you see low to mid range cards with so and so amount of RAM and the unaware/ignorant n00b thinks its gonna be a beast cause of all the RAM.
What matters is the memory architecture and bandwith and shader count, period.

Droolfest for the retailers:thumbsup:
 

VulgarDisplay

Diamond Member
Apr 3, 2009
6,188
2
76
6+8 pin as well, should OC better than reference.

Should do better, but I've been reading that the gtx680 doesn't scale as well as the 7970 with clock speed?

Either way, I'm really damn excited to see what big kepler brings to the table. 7970 will be a decent placeholder for that card lol.
 

aaksheytalwar

Diamond Member
Feb 17, 2012
3,389
0
76
You might not believe me now, but I am 80-90% certain that the big kepler will be available in nearly all parts of the world no sooner than September or October at the very earliest and would increase performance over a stock 680 by about 15-25%. No idea about its oc room but hopefully won't be worse there either

So the big Kepler will be a bit faster than a highly Oced 680, that is about it. You will see this when it happens and the chance of this statement going wrong is just 10%.
 

Cookie Monster

Diamond Member
May 7, 2005
5,161
32
86
The practical limitations for OCing both the 680 and 7970 are probably the 28nm process itself and air/water-cooled temperatures. 12 power phases and 8+6 PEG is likely completely unnecessary unless you're on LN2, and then both cards have already been OC'd to 1800-1900MHz.

For the GTX680 it could make a difference seeing as the stock PCB circuitry looks almost bare minimum.. and I should know since I do that sort of work ;)
 

Grooveriding

Diamond Member
Dec 25, 2008
9,117
1,266
126
The practical limitations for OCing both the 680 and 7970 are probably the 28nm process itself and air/water-cooled temperatures. 12 power phases and 8+6 PEG is likely completely unnecessary unless you're on LN2, and then both cards have already been OC'd to 1800-1900MHz.

My 2 cards are getting here today. I'm going to see what difference temperature makes to the GPU boost feature. Everything I've read has consistently mentioned temps as playing a big role in defining at what point the card throttles back, 72C is the number I keep hearing.

I ordered a few waterblocks from EK directly, hopefully they come before the weekend.

Going to do a few different benches on air using 3dmark, furmark and some crysis_gpu loops to track temp, volts, clocks and power use. Will do them at stock and then whatever the max overclock I can get without crossing the throttle threshold.

Will repeat them all once I get my blocks in and installed, as well as seeing if much lower temps open the GPU boost & overclock up more. Hoping to find a significant difference so I feel better about ordering the blocks! :D
 

BoFox

Senior member
May 10, 2008
689
0
0
My 2 cards are getting here today. I'm going to see what difference temperature makes to the GPU boost feature. Everything I've read has consistently mentioned temps as playing a big role in defining at what point the card throttles back, 72C is the number I keep hearing.

I ordered a few waterblocks from EK directly, hopefully they come before the weekend.

Going to do a few different benches on air using 3dmark, furmark and some crysis_gpu loops to track temp, volts, clocks and power use. Will do them at stock and then whatever the max overclock I can get without crossing the throttle threshold.

Will repeat them all once I get my blocks in and installed, as well as seeing if much lower temps open the GPU boost & overclock up more. Hoping to find a significant difference so I feel better about ordering the blocks! :D

Most sites say about 70C is when the card refuses to hit the highest "Boost" step, and that over 70C does not really affect the lower steps. Simply increasing the power threshold for boost to the max makes the card stay at 2nd highest boost step all the time. At least that's how some review sites explained it, IIRC..
 

ViviTheMage

Lifer
Dec 12, 2002
36,189
87
91
madgenius.com
Increasing the amt. of RAM doesn't increase bandwidth. The 7970 is still going to win there. It's going to be a really rare situation where 2GB of RAM isn't enough.

Yeah, I have no issues with being an early adopter of the 2GB version...really, what resolutions need more then 2GB? I know there are people that would need it, but who, and what resolutions?
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
The 4GB version will most likely be way overpriced just like the 3GB 580 was.

You can argue 2GB over 1GB pretty easily, but no way you can argue 4GB over 2GB. Maybe in 3-4 years, but certainly not today.
 

AdamK47

Lifer
Oct 9, 1999
15,635
3,410
136
I'd be all for it if doubling the memory also ment going from 256-bit to 512-bit.
 
Feb 19, 2009
10,457
10
76
I'm pretty sure users with 3 x 30 inch monitors would appreciate the extra vram for their tri or quad SLI.. there's always a few of those "too rich to know value". ;)
 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
Should do better, but I've been reading that the gtx680 doesn't scale as well as the 7970 with clock speed?

Either way, I'm really damn excited to see what big kepler brings to the table. 7970 will be a decent placeholder for that card lol.


You've been reading people who aren't able to discern the actual stock operating frequency. The same people then assume Kepler scales poorly based on the idea that is somehow manages to never boost past 1006MHz at stock.