blazorthon
Glorious
- Sep 24, 2010
- 14,022
- 21
- 43,265
[citation][nom]reaper123[/nom]Although i love ATI/AMD i dont consider myself a fanboy of either company.Just wanted to point that out first. The thing is that if you compare the 7970 with the Gtx 680 at same clock rates they perform pretty much the same in gaming. However there are a few things to consider:The 7970 is a computation beast and is better for 3D designing /3D applications compared to the gtx 680. However the so called Gtx 680 was initialy the Gtx 660ti/670 and the correct equivalent of the 7970 is supposed to be the Big Kepler Gtx 685. Am pretty sure that Nvidia people are rubbing their hands from the stunt they pulled trickering people to pay double the money for that gcard. Seeing that their midrange kepler card was able to outperform a stock 7970 they changed the label to 680 so they could rip us off and pull the big kepler card in the unlike event AMD produces something to rival Nvidia.[/citation]
Nvidia used GK104 to compete with AMD in the high end instead of GK100 because AMD switched from gaming oriented chips to compute oriented chips while Nvidia switched from compute oriented chips to gaming oriented chips. Did anyone mock Nvidia for needing massive dies (greater than 500mm, the GF100/GF110 from the GTX 580, 570, 560 TI 448 core was 530mm2 if I remember correctly) just to keep up with or barely surpass AMD's much smaller dies (Cayman was something like 375mm2 or 374mm2 and fought with the 530mm2 GF110)?
The reason for Nvidia winning now is obvious: AMD doesn't want to use massive dies like Nvidia has in the past. Achieving high compute performance requires extra die space and that extra die space also requires extra power. This is because compute oriented chips often have more compute oriented hardware such as more complex scheduling hardware than a gaming oriented chip. For example, Kepler's schedulers are a lot simpler than GCN's and Fermi's.
Pitcairn is far less compute oriented than Tahiti and has extremely similar gaming performance per mm2 of die space and per watt to Kepler. If AMD made a Pitcairn-style chip about the size of the GK104 (just short of 300mm2), then it would perform very similarly to GK104.
The fact that the 7970 is so close to the 680 despite it having a much more compute oriented GPU tells us that AMD managed to merge gaming and compute performance far better than Nvidia did with Fermi, especially with the 7970 being over three times faster than the GTX 580 for dual precision compute despite it being only one process node ahead of the 580. That the 7970 is still in line with it's performance difference between it and Nvidia's last generation top single GPU card when compared to the last process node jump in addition to it's huge compute performance jump is nothing short of spectacular for AMD.
That Nvidia decided to take the easy road with a gaming oriented GPU after making a living off being compute-heavy is exactly the opposite. However, for gaming performance, Nvidia still did make the proper jump when comparing it to the last process jump.
Here're the examples:
Radeon 4000 and GTX 200.
Radeon 4870 is right behind the GTX 285.
Radeon 4870X2 and GTX 295 are roughly equal.
next process.
GTX 480 roughly equals the 295 in gaming performance. the 5870 is right behind the 4870X2, 295, and 480.
Same process, next generation
Radeon 6970 finally roughly equals the 4870X2, 295, and 480. GTX 580 surpasses them all by about one tier's worth of gaming performance.
next process.
Radeon 7970 is right behind the 6990 and 590. GTX 680 is in line with the 6990 and 590.
So, despite there being some odd differences, the gaming performance is still on the same trend as it was before. The differences being how computationally oriented each card is and as another result of that, the power efficiency of each card. Also consider that despite the 7970's much higher TDP, even Guru3D, someone known for being a little Nvidia biased, showed the 7970 hardly using much more power than the 680 during gameplay! So, AMD managed a better unity of gaming and computational performance than Nvidia did AND did it, AMD managed this with a better ratio of power efficiency between the compute heavy cards and the gaming heavy cards of the same generation, and even a site called out for being pro Nvidia shows this in their charts from their review of the 680.
Considering how games are becoming more compute performance reliant, AMD might turn out to be the winner of this after all. Which company wins will probably depend on just how long it takes the most compute reliant games to come out. If they are out around the time of the Radeon 8000's arrival, then AMD will do FAR better. If they aren't out until the Radeon 9000s or later (it's possible), then Nvidia may win, at least for a while.
Of course, this is only going by the variables that have already been presented and assumes that neither manufacturer will screw something up or do something else that is exemplary. Anything could happen between now and then.
With the 680 only having half of the GTX 580's compute performance and also having a huge memory bandwidth bottleneck, this would go very poorly for Nvidia buyers. Of course, Nvidia would either respond by making another compute oriented architecture or by getting the game developers to lean against compute, so it's hard to say exactly how this will go down. As of right now, things look good for AMD no matter what happens.AMD had great head start on Nvidia and is still the only one of the two to have their next generation cards in consistently good supply right now.
With Nvidia Kepler cards almost completely MIA and AMD's price cuts, Radeon 7000 looks like an excellent buy right now. At least for a while, this does not bode well for Nvidia if they can't get more cards out and in good supply regardless of how compute heavy games are.
As of right now, AMD is the only option for next generation cards and AMD also now has decent prices on their cards. If Nvidia doesn't get back up soon, then many of the people who want a next generation card will have already gone with AMD. This could leave Nvidia relying on late adopters just because the early adopters had no choice but to either wait, or buy AMD now. With AMD's compelling performance, prices, and power usage compared to the current cards, who could blame them for not wanting to wait even longer?
Nvidia used GK104 to compete with AMD in the high end instead of GK100 because AMD switched from gaming oriented chips to compute oriented chips while Nvidia switched from compute oriented chips to gaming oriented chips. Did anyone mock Nvidia for needing massive dies (greater than 500mm, the GF100/GF110 from the GTX 580, 570, 560 TI 448 core was 530mm2 if I remember correctly) just to keep up with or barely surpass AMD's much smaller dies (Cayman was something like 375mm2 or 374mm2 and fought with the 530mm2 GF110)?
The reason for Nvidia winning now is obvious: AMD doesn't want to use massive dies like Nvidia has in the past. Achieving high compute performance requires extra die space and that extra die space also requires extra power. This is because compute oriented chips often have more compute oriented hardware such as more complex scheduling hardware than a gaming oriented chip. For example, Kepler's schedulers are a lot simpler than GCN's and Fermi's.
Pitcairn is far less compute oriented than Tahiti and has extremely similar gaming performance per mm2 of die space and per watt to Kepler. If AMD made a Pitcairn-style chip about the size of the GK104 (just short of 300mm2), then it would perform very similarly to GK104.
The fact that the 7970 is so close to the 680 despite it having a much more compute oriented GPU tells us that AMD managed to merge gaming and compute performance far better than Nvidia did with Fermi, especially with the 7970 being over three times faster than the GTX 580 for dual precision compute despite it being only one process node ahead of the 580. That the 7970 is still in line with it's performance difference between it and Nvidia's last generation top single GPU card when compared to the last process node jump in addition to it's huge compute performance jump is nothing short of spectacular for AMD.
That Nvidia decided to take the easy road with a gaming oriented GPU after making a living off being compute-heavy is exactly the opposite. However, for gaming performance, Nvidia still did make the proper jump when comparing it to the last process jump.
Here're the examples:
Radeon 4000 and GTX 200.
Radeon 4870 is right behind the GTX 285.
Radeon 4870X2 and GTX 295 are roughly equal.
next process.
GTX 480 roughly equals the 295 in gaming performance. the 5870 is right behind the 4870X2, 295, and 480.
Same process, next generation
Radeon 6970 finally roughly equals the 4870X2, 295, and 480. GTX 580 surpasses them all by about one tier's worth of gaming performance.
next process.
Radeon 7970 is right behind the 6990 and 590. GTX 680 is in line with the 6990 and 590.
So, despite there being some odd differences, the gaming performance is still on the same trend as it was before. The differences being how computationally oriented each card is and as another result of that, the power efficiency of each card. Also consider that despite the 7970's much higher TDP, even Guru3D, someone known for being a little Nvidia biased, showed the 7970 hardly using much more power than the 680 during gameplay! So, AMD managed a better unity of gaming and computational performance than Nvidia did AND did it, AMD managed this with a better ratio of power efficiency between the compute heavy cards and the gaming heavy cards of the same generation, and even a site called out for being pro Nvidia shows this in their charts from their review of the 680.
Considering how games are becoming more compute performance reliant, AMD might turn out to be the winner of this after all. Which company wins will probably depend on just how long it takes the most compute reliant games to come out. If they are out around the time of the Radeon 8000's arrival, then AMD will do FAR better. If they aren't out until the Radeon 9000s or later (it's possible), then Nvidia may win, at least for a while.
Of course, this is only going by the variables that have already been presented and assumes that neither manufacturer will screw something up or do something else that is exemplary. Anything could happen between now and then.
With the 680 only having half of the GTX 580's compute performance and also having a huge memory bandwidth bottleneck, this would go very poorly for Nvidia buyers. Of course, Nvidia would either respond by making another compute oriented architecture or by getting the game developers to lean against compute, so it's hard to say exactly how this will go down. As of right now, things look good for AMD no matter what happens.AMD had great head start on Nvidia and is still the only one of the two to have their next generation cards in consistently good supply right now.
With Nvidia Kepler cards almost completely MIA and AMD's price cuts, Radeon 7000 looks like an excellent buy right now. At least for a while, this does not bode well for Nvidia if they can't get more cards out and in good supply regardless of how compute heavy games are.
As of right now, AMD is the only option for next generation cards and AMD also now has decent prices on their cards. If Nvidia doesn't get back up soon, then many of the people who want a next generation card will have already gone with AMD. This could leave Nvidia relying on late adopters just because the early adopters had no choice but to either wait, or buy AMD now. With AMD's compelling performance, prices, and power usage compared to the current cards, who could blame them for not wanting to wait even longer?
Facebook
Twitter
Instagram