Looks like vega is coming between now and June. I think a slide had may release in it. The performance looks like it is between the gtx1080 and Pascal Titan. They are using an updated GCN architecture with better ipc.
Will have to wait for real benches obviously, and firm pricing. The pricing will make/break it, but I guess that's always the case when you aren't the clear performance leader and you have to rely on price/performance to make your business case.
I have a feeling nVidia has a lot of room to come down on 1080 (and 1070) pricing - nVidia is really capitalizing on the fact that they have no competition in this arena, and currently there's no reason they shouldn't command whatever premium price they can get consumers to pay. Vega has to challenge that meaningfully to be able to even enter the marketplace, and (I believe) nVidia has the capability to neutralize or mitigate that price advantage if they should chose to do so.
And nVidia still has a potential 1080Ti up their sleeve. I wouldn't be surprised if nVidia is just sitting waiting on AMD to do something concrete with Vega, then they would drop the 1070 down to about $325, the 1080 down to around $450-500, and the 1080Ti then would be released around $700.
There's also the rumor that nVidia is ready to push the 2000 series (not as Volta, a re-worked Pascal), that ~could~ hit around the same time as Vega. Rumor has it pegged as July/August, which if that's a hard release, given recent trend of soft releases, nVidia could soft release it as soon as Yesterday and just about be on trend, and could be why we haven't seen a 1080Ti. If the WCC source is even close to correct, could spell trouble for Vega - between a 1080 and Titan today could be closer to between a 2070 and 2080 if the speed increases are remotely accurate on the leak, and if the price cuts hit at the same time, that makes a tight window for Vega to make sense on that Price/Performance curve.
The silver lining in that for AMD is that nVidia is looking at pushing HBM2 back until 2018. HBM has the potential to be a game changer, and AMD is clearly leading on that tech at the moment.
That's a lot of conjecture and opinion, but obviously, anything dealing with Vega (and nVidia's possible response) right now is going to be nothing but that.
I think vega is a generation after Polaris. The news we heard earlier is that Vega is ahead of schedule. Instead of being on track for a September release, they are moving in on an earlier release. There has also been news of 2 Polaris GPUs that we never saw. Something tells me AMD cancelled them because they can simply go early with Vega which would make 2 higher end gpus released late 2016 a wasted investment. Similarly I heard nVidia is pushing forward with Volta for an earlier release, maybe September.
From what I understood - Polaris 10 (the ones we have now) were the low-mid range Desktop, and Polaris 11 were going to be mobile versions. Although we are seeing 10's installed in mobile computers (or myab ethe Radeon Pro 455 and 450 are Polaris 11 - those are 2 of the 3 AMD discretes used in the latest Apple models). Vega was going to be hot on the heels of Polaris 10, and fill the mid-high end desktop lineup.
Polaris 10 "launched" in June, and started shipping in July (albeit it took several months for supply lines to finally catch up). Everyone has been expecting Vega "Any Day Now" since then.
nVidia launched Pascal in May (and it took months for supply to firm up there as well), that included the 1080 card.
AMD's Fury X (or 390X, depending on if VRAM is a factor), released in mid-2015, is still their top-performing GPU. Fury was supposed to be the 980 killer, and it was competitive, but it didn't really kill anything, because nVidia pretty much squashed the AMD release by counter-releasing the 980Ti on almost the same date.
And that's about what I think is going to line up here. AMD finally gets something that's competitive, and nVidia just jumps right out and ups the ante with whatever they have in their pocket. The fact that whatever nVidia releases (be it a 1080Ti or a 2080) is faster than Vega (again, speculation) will steal the press - it won't matter that Vega is competeitive versus the 1080 on price and/or performance. The press will just jump on the fact that both companies have released new cards, and nVidia is faster, and that will be the headline.
Why do I think this? Because that's almost exactly what happened with the Fury X and 980Ti. nVidia's had almost a year to come up with some refinement to market as either a 1080Ti or 2080, and there is every reason to think that they have it in their back pocket. The deck looks stacked for another hand to be dealt almost exactly the same this generation as the last. Headlines are all most people read anyway.
Do I think nVidia is bad because of the "Green Tax" that they place on all their products? No, it makes perfect business sense. And they have the capability and technology to make a good product. Of course I wish it were cheaper, but that's me speaking as a consumer, and i can't really blame nVidia for charging what they can get for their product - in lieu of stiff competition, that's what they are supposed to be doing.
Do I think AMD is bad because they can't come up with "the fastest" card? No. GPUs are extraordinarily complicated, and only one company can be "the fastest". With only two companies in the game right now, it takes a good bit to overtake the leader. And it wasn't that long ago that AMD was the leader in it's own right and nVidia was playing catch-up. I don't know if Vega will "disappoint" like Fury did (and that dispointment wasn't that it was a bad card, it's that it couldn't possibly live up to the hype that preceeded it)
Right now I have more interest in what Ryzen is going to do for the CPU market than I do Vega for GPUs. In GPU parlance, AMD and nVidia are nearly neck and neck - right now nVidia is ahead, but that lead could go in the matter of a generation of cards. In the CPU world, Intel has had such a large advantage for a very long time that it's essentially been a one-horse race, to the point that Intel isn't even trying on the CPU side any longer (pretty much every CPU since Sandy has focused more on IGP than CPU), and most people's best hope is that Ryzen catches up to what Sandy Bridge had to offer, more than 5 years ago.
nVidia's had almost a year to come up with some refinement to market as either a 1080Ti or 2080, and there is every reason to think that they have it in their back pocket.
Agreed.
People are buying NVidia's current top end cards really well. Based on latest Steam hardware survey last month GTX 1070 got as many new users as RX 480 and RX 470 did combined. If NVidia were to hype their new and faster products in advance they'd be cutting their current sales.
The big question is what sort of "higher IPC" (as AMD put it) Vega will bring. If it's limited to increased throughput for half-precision and 8-bit integer instructions, that will be a huge disappointment. If it means beefed up fixed-function graphics hardware so that it's not such a bottleneck like it was in Fiji, then that's not nothing, but not really what you hope for as "higher IPC".
If, on the other hand, Vega is markedly better at keeping its shaders busy--that is, if Vega is basically AMD's Maxwell--then it could be a huge hit. Few people realize just how good Maxwell's shader scheduling is; it's far superior to anything that preceded it in GPUs, whether from AMD or Nvidia. GCN already lets you keep all of the shaders busy essentially all of the time in some GPU-friendly compute workloads, so there's nothing to gain on IPC there. The problem is that graphics is not one of those compute-friendly workloads, and isn't even close to it.
I think the GeForce GTX 1080 comparison sets the bar far too low for Vega. If Vega is nothing more than Polaris with everything scaled up to hit the 12.5 TFLOPS claimed for a Radeon Instinct MI25, that would likely beat a GTX 1080 outright without any architectural improvements at all. And let's not forget that consumer graphics cards are usually clocked higher than server cards, which would mean more performance yet. If Vega does improve shader scheduling, then beating a Pascal-based Titan X outright is very possible. Remember that we're probably looking at Vega having more of pretty much every single resource than a Titan X; if it can merely match the Titan X's efficiency in using those resources, Vega wins.
So what can Nvidia do to counter? Some people seem to think that Nvidia will have something up their sleeve just on general principle because some time has passed. But that's not how it works. There are really only three sources of major improvements:
1) Move to a better process node 2) Create a better architecture 3) Make a bigger chip
Option (1) is not happening for Nvidia in time to compete with Vega unless Global Foundries' 14 nm is markedly better than TSMC's 16 nm and Nvidia makes that switch. I'd regard the first half of that as being sufficiently unlikely as to make the second irrelevant.
Option (2) happens from time to time, but I'd regard it as extremely likely that Nvidia can do that before Volta. If you have a new architecture that is far better than Pascal, call it something else.
That leaves option (3), which Nvidia could do. The Titan X is already a big chip, but Nvidia could make an even bigger chip with perhaps 1/3 more resources than the Titan X has. Yields might leave a lot to be desired if they try to do that, though. Nvidia certainly could make another 600 mm^2 chip, though they already tried with the GP100 that was announced last May and still doesn't seem to be available. The rather limited distribution of GP102, even, says nothing good about its yields as it is. A cut-down version of a Titan X showing up at some point seems likely, but that just fills a hole in Nvidia's product stack.
There is also the possibility of higher clock speeds, whether from process node improvements, a respin, or Micron delivering better GDDR5X. Take the same die and clock it higher usually doesn't give you very big gains, though, and a GTX 2080 that is just a GTX 1080 clocked 5% higher isn't a huge improvement. Fixing a badly broken die as happened in moving from the GeForce 400 series to the 500 series can yield larger gains, but the Pascal cards out today don't strike me as badly broken with a ton of room for easy improvements just by fixing the things that are broken.
1) Move to a better process node 2) Create a better architecture 3) Make a bigger chip
If these are the only options to making a faster card, which one does the 400->500 series fall under? Seems like respin isn't on the list: that's not the same thing as a new architecture, and it's not necessarily a bigger chip. We/ve seen it happen a lot with individual cards across a generation as well, from both nVidia and AMD. That's what I think the 1080Ti will be - not really anything new, but at the least better firmware, probably slight hardware tweaks, and that will be enough for it to land firmly ahead of the existing 1080, but not poach the current Titan.
*edit* you talk about that in the last half of your post somewhat. But remember Pascal ~is~ a new architecture. Just because it doesn't seem "badly broken" doesn't mean there aren't ways to boost speed or efficiency, particularly after it's been out in the wild for a bit and a lot more feedback and real-world testing has been performed.
1) Move to a better process node 2) Create a better architecture 3) Make a bigger chip
If these are the only options to making a faster card, which one does the 400->500 series fall under? Seems like respin isn't on the list: that's not the same thing as a new architecture, and it's not necessarily a bigger chip. We/ve seen it happen a lot with individual cards across a generation as well, from both nVidia and AMD. That's what I think the 1080Ti will be - not really anything new, but at the least better firmware, probably slight hardware tweaks, and that will be enough for it to land firmly ahead of the existing 1080, but not poach the current Titan.
*edit* you talk about that in the last half of your post somewhat. But remember Pascal ~is~ a new architecture. Just because it doesn't seem "badly broken" doesn't mean there aren't ways to boost speed or efficiency, particularly after it's been out in the wild for a bit and a lot more feedback and real-world testing has been performed.
It's harder to improve on something that is already pretty good than if all you have to do is to fix something that is broken. The GeForce 400 series was horribly broken and it was obvious in the launch day reviews. The GeForce 1000 series sure doesn't look broken to me. Minor tweaks here and there might get you good enough yields to justify a GP104 salvage part with 17 compute units active rather than the 15 in a GTX 1070, or might add 100 MHz to the clock speed. And I wouldn't be surprised if they do something like that in a GeForce 1100 or 2000 series or whatever that is otherwise mostly rebrands.
But that's minor tinkering around the edges, not a huge jump like going from Kepler to Maxwell (loosely, GeForce 600/700 series to 900 series) or from Maxwell to Pascal (GeForce 900 series to 1000 series). If either of those transitions had gotten only 5-10% performance improvements, Nvidia would have lost the corresponding generation and badly. My basic claim is that a big jump like that is not likely for Nvidia until Volta and/or a 10 nm or smaller process node.
A large jump in top end performance in the first half of this year is probable for AMD, on the other hand, because they surely have a much bigger chip coming and Vega might also be a major architectural improvement over Polaris. But this is partially saying that AMD is more likely to make a big jump than Nvidia because AMD is behind right now.
A large jump in top end performance in the first half of this year is probable for AMD, on the other hand, because they surely have a much bigger chip coming and Vega might also be a major architectural improvement over Polaris. But this is partially saying that AMD is more likely to make a big jump than Nvidia because AMD is behind right now.
This is pretty much all I'm saying. I believe nVidia has something up their sleeve, not because they have to do a lot of engineering work to get there - to deliver a 1080Ti, all they have to do is come somewhere in between the performance of two cards they already have on the market - it needs to beat a 1080, but it doesn't have to beat a Titan X. It doesn't even have to necessarily match a Titan X.
Can they do that, with what they have laying around from their existing Pascal development? Almost certainly. They could ~probably~ do it with existing hardware designs and just some firmware tweaks. It's certainly not hard to imagine, and it doesn't require a new design, a new process node, or anything else to accomplish.
nVidia doesn't have to release Volta or an entire new 2000 series to counter Vega. All they have to do to neutralize Vega, is push a 1080Ti that doesn't cost $1200, lower the price of the existing 1080 and 1070 a bit, and unless Vega greatly outclasses nVidia (which has happened, but doesn't appear to be likely), nVidia keeps the headlines skewed their direction.
That doesn't mean I think Vega will be a bad card. I think Fury was a good card, and I think AMD can easily beat it with Vega, which will make it a contender in the upper-tier space again. I just think the deck is stacked in nVidia's favor at the moment, and the headlines will continue to slant Green, and most of what we see in the press will be more for the benefit of the stock market than anything to do with actual consumers.
Don't matter to me when Vega release i replace my Fury X with another AMD gpu have good experience with AMD from 2009-2017 so far won't change for me anytime soon to other side no way:)
I just hope is good improvement on Fury X i own now thats all im not obsessed with FPS and top contender.
I love ReLive now and drivers AMD deliver!
Vega will be a complete NEW architectural improvement over Polaris!!!
Hope to build full AMD system RYZEN/VEGA/AM4!!!
MB:Asus V De Luxe z77 CPU:Intell Icore7 3770k GPU: AMD Fury X(waiting for BIG VEGA 10 or 11 HBM2?(bit unclear now)) MEMORY:Corsair PLAT.DDR3 1866MHZ 16GB PSU:Corsair AX1200i OS:Windows 10 64bit
Will have to wait for real benches obviously, and firm pricing. The pricing will make/break it, but I guess that's always the case when you aren't the clear performance leader and you have to rely on price/performance to make your business case.
I have a feeling nVidia has a lot of room to come down on 1080 (and 1070) pricing - nVidia is really capitalizing on the fact that they have no competition in this arena, and currently there's no reason they shouldn't command whatever premium price they can get consumers to pay. Vega has to challenge that meaningfully to be able to even enter the marketplace, and (I believe) nVidia has the capability to neutralize or mitigate that price advantage if they should chose to do so.
And nVidia still has a potential 1080Ti up their sleeve. I wouldn't be surprised if nVidia is just sitting waiting on AMD to do something concrete with Vega, then they would drop the 1070 down to about $325, the 1080 down to around $450-500, and the 1080Ti then would be released around $700.
There's also the rumor that nVidia is ready to push the 2000 series (not as Volta, a re-worked Pascal), that ~could~ hit around the same time as Vega. Rumor has it pegged as July/August, which if that's a hard release, given recent trend of soft releases, nVidia could soft release it as soon as Yesterday and just about be on trend, and could be why we haven't seen a 1080Ti. If the WCC source is even close to correct, could spell trouble for Vega - between a 1080 and Titan today could be closer to between a 2070 and 2080 if the speed increases are remotely accurate on the leak, and if the price cuts hit at the same time, that makes a tight window for Vega to make sense on that Price/Performance curve.
The silver lining in that for AMD is that nVidia is looking at pushing HBM2 back until 2018. HBM has the potential to be a game changer, and AMD is clearly leading on that tech at the moment.
That's a lot of conjecture and opinion, but obviously, anything dealing with Vega (and nVidia's possible response) right now is going to be nothing but that.
You are talking high end. AMD is quite competitive in anything below that and how many people have the money to buy a high end graphics card since Nvidia prices are a bit ridiculous?
Personally I feel that Nvidia is acting more like Intel lately and that is NOT good. I have been putting in 480's in a lot of builds lately just because of that. For most gamers that is plenty of HP.
Will have to wait for real benches obviously, and firm pricing. The pricing will make/break it, but I guess that's always the case when you aren't the clear performance leader and you have to rely on price/performance to make your business case.
I have a feeling nVidia has a lot of room to come down on 1080 (and 1070) pricing - nVidia is really capitalizing on the fact that they have no competition in this arena, and currently there's no reason they shouldn't command whatever premium price they can get consumers to pay. Vega has to challenge that meaningfully to be able to even enter the marketplace, and (I believe) nVidia has the capability to neutralize or mitigate that price advantage if they should chose to do so.
And nVidia still has a potential 1080Ti up their sleeve. I wouldn't be surprised if nVidia is just sitting waiting on AMD to do something concrete with Vega, then they would drop the 1070 down to about $325, the 1080 down to around $450-500, and the 1080Ti then would be released around $700.
There's also the rumor that nVidia is ready to push the 2000 series (not as Volta, a re-worked Pascal), that ~could~ hit around the same time as Vega. Rumor has it pegged as July/August, which if that's a hard release, given recent trend of soft releases, nVidia could soft release it as soon as Yesterday and just about be on trend, and could be why we haven't seen a 1080Ti. If the WCC source is even close to correct, could spell trouble for Vega - between a 1080 and Titan today could be closer to between a 2070 and 2080 if the speed increases are remotely accurate on the leak, and if the price cuts hit at the same time, that makes a tight window for Vega to make sense on that Price/Performance curve.
The silver lining in that for AMD is that nVidia is looking at pushing HBM2 back until 2018. HBM has the potential to be a game changer, and AMD is clearly leading on that tech at the moment.
That's a lot of conjecture and opinion, but obviously, anything dealing with Vega (and nVidia's possible response) right now is going to be nothing but that.
WccfTech lol..yeh we believe that site to be true:P
Hope to build full AMD system RYZEN/VEGA/AM4!!!
MB:Asus V De Luxe z77 CPU:Intell Icore7 3770k GPU: AMD Fury X(waiting for BIG VEGA 10 or 11 HBM2?(bit unclear now)) MEMORY:Corsair PLAT.DDR3 1866MHZ 16GB PSU:Corsair AX1200i OS:Windows 10 64bit
Comments
I have a feeling nVidia has a lot of room to come down on 1080 (and 1070) pricing - nVidia is really capitalizing on the fact that they have no competition in this arena, and currently there's no reason they shouldn't command whatever premium price they can get consumers to pay. Vega has to challenge that meaningfully to be able to even enter the marketplace, and (I believe) nVidia has the capability to neutralize or mitigate that price advantage if they should chose to do so.
And nVidia still has a potential 1080Ti up their sleeve. I wouldn't be surprised if nVidia is just sitting waiting on AMD to do something concrete with Vega, then they would drop the 1070 down to about $325, the 1080 down to around $450-500, and the 1080Ti then would be released around $700.
There's also the rumor that nVidia is ready to push the 2000 series (not as Volta, a re-worked Pascal), that ~could~ hit around the same time as Vega. Rumor has it pegged as July/August, which if that's a hard release, given recent trend of soft releases, nVidia could soft release it as soon as Yesterday and just about be on trend, and could be why we haven't seen a 1080Ti. If the WCC source is even close to correct, could spell trouble for Vega - between a 1080 and Titan today could be closer to between a 2070 and 2080 if the speed increases are remotely accurate on the leak, and if the price cuts hit at the same time, that makes a tight window for Vega to make sense on that Price/Performance curve.
The silver lining in that for AMD is that nVidia is looking at pushing HBM2 back until 2018. HBM has the potential to be a game changer, and AMD is clearly leading on that tech at the moment.
http://wccftech.com/nvidia-pascal-volta-gpu-leaked-2017-2018/
That's a lot of conjecture and opinion, but obviously, anything dealing with Vega (and nVidia's possible response) right now is going to be nothing but that.
Similarly I heard nVidia is pushing forward with Volta for an earlier release, maybe September.
Polaris 10 "launched" in June, and started shipping in July (albeit it took several months for supply lines to finally catch up). Everyone has been expecting Vega "Any Day Now" since then.
nVidia launched Pascal in May (and it took months for supply to firm up there as well), that included the 1080 card.
AMD's Fury X (or 390X, depending on if VRAM is a factor), released in mid-2015, is still their top-performing GPU. Fury was supposed to be the 980 killer, and it was competitive, but it didn't really kill anything, because nVidia pretty much squashed the AMD release by counter-releasing the 980Ti on almost the same date.
And that's about what I think is going to line up here. AMD finally gets something that's competitive, and nVidia just jumps right out and ups the ante with whatever they have in their pocket. The fact that whatever nVidia releases (be it a 1080Ti or a 2080) is faster than Vega (again, speculation) will steal the press - it won't matter that Vega is competeitive versus the 1080 on price and/or performance. The press will just jump on the fact that both companies have released new cards, and nVidia is faster, and that will be the headline.
Why do I think this? Because that's almost exactly what happened with the Fury X and 980Ti. nVidia's had almost a year to come up with some refinement to market as either a 1080Ti or 2080, and there is every reason to think that they have it in their back pocket. The deck looks stacked for another hand to be dealt almost exactly the same this generation as the last. Headlines are all most people read anyway.
Do I think nVidia is bad because of the "Green Tax" that they place on all their products? No, it makes perfect business sense. And they have the capability and technology to make a good product. Of course I wish it were cheaper, but that's me speaking as a consumer, and i can't really blame nVidia for charging what they can get for their product - in lieu of stiff competition, that's what they are supposed to be doing.
Do I think AMD is bad because they can't come up with "the fastest" card? No. GPUs are extraordinarily complicated, and only one company can be "the fastest". With only two companies in the game right now, it takes a good bit to overtake the leader. And it wasn't that long ago that AMD was the leader in it's own right and nVidia was playing catch-up. I don't know if Vega will "disappoint" like Fury did (and that dispointment wasn't that it was a bad card, it's that it couldn't possibly live up to the hype that preceeded it)
Right now I have more interest in what Ryzen is going to do for the CPU market than I do Vega for GPUs. In GPU parlance, AMD and nVidia are nearly neck and neck - right now nVidia is ahead, but that lead could go in the matter of a generation of cards. In the CPU world, Intel has had such a large advantage for a very long time that it's essentially been a one-horse race, to the point that Intel isn't even trying on the CPU side any longer (pretty much every CPU since Sandy has focused more on IGP than CPU), and most people's best hope is that Ryzen catches up to what Sandy Bridge had to offer, more than 5 years ago.
no you should speak as corporate representative as youre obviously not a consumer and you dont want responsible company and want to be cheated.
People are buying NVidia's current top end cards really well. Based on latest Steam hardware survey last month GTX 1070 got as many new users as RX 480 and RX 470 did combined. If NVidia were to hype their new and faster products in advance they'd be cutting their current sales.
If, on the other hand, Vega is markedly better at keeping its shaders busy--that is, if Vega is basically AMD's Maxwell--then it could be a huge hit. Few people realize just how good Maxwell's shader scheduling is; it's far superior to anything that preceded it in GPUs, whether from AMD or Nvidia. GCN already lets you keep all of the shaders busy essentially all of the time in some GPU-friendly compute workloads, so there's nothing to gain on IPC there. The problem is that graphics is not one of those compute-friendly workloads, and isn't even close to it.
I think the GeForce GTX 1080 comparison sets the bar far too low for Vega. If Vega is nothing more than Polaris with everything scaled up to hit the 12.5 TFLOPS claimed for a Radeon Instinct MI25, that would likely beat a GTX 1080 outright without any architectural improvements at all. And let's not forget that consumer graphics cards are usually clocked higher than server cards, which would mean more performance yet. If Vega does improve shader scheduling, then beating a Pascal-based Titan X outright is very possible. Remember that we're probably looking at Vega having more of pretty much every single resource than a Titan X; if it can merely match the Titan X's efficiency in using those resources, Vega wins.
So what can Nvidia do to counter? Some people seem to think that Nvidia will have something up their sleeve just on general principle because some time has passed. But that's not how it works. There are really only three sources of major improvements:
1) Move to a better process node
2) Create a better architecture
3) Make a bigger chip
Option (1) is not happening for Nvidia in time to compete with Vega unless Global Foundries' 14 nm is markedly better than TSMC's 16 nm and Nvidia makes that switch. I'd regard the first half of that as being sufficiently unlikely as to make the second irrelevant.
Option (2) happens from time to time, but I'd regard it as extremely likely that Nvidia can do that before Volta. If you have a new architecture that is far better than Pascal, call it something else.
That leaves option (3), which Nvidia could do. The Titan X is already a big chip, but Nvidia could make an even bigger chip with perhaps 1/3 more resources than the Titan X has. Yields might leave a lot to be desired if they try to do that, though. Nvidia certainly could make another 600 mm^2 chip, though they already tried with the GP100 that was announced last May and still doesn't seem to be available. The rather limited distribution of GP102, even, says nothing good about its yields as it is. A cut-down version of a Titan X showing up at some point seems likely, but that just fills a hole in Nvidia's product stack.
There is also the possibility of higher clock speeds, whether from process node improvements, a respin, or Micron delivering better GDDR5X. Take the same die and clock it higher usually doesn't give you very big gains, though, and a GTX 2080 that is just a GTX 1080 clocked 5% higher isn't a huge improvement. Fixing a badly broken die as happened in moving from the GeForce 400 series to the 500 series can yield larger gains, but the Pascal cards out today don't strike me as badly broken with a ton of room for easy improvements just by fixing the things that are broken.
*edit* you talk about that in the last half of your post somewhat. But remember Pascal ~is~ a new architecture. Just because it doesn't seem "badly broken" doesn't mean there aren't ways to boost speed or efficiency, particularly after it's been out in the wild for a bit and a lot more feedback and real-world testing has been performed.
But that's minor tinkering around the edges, not a huge jump like going from Kepler to Maxwell (loosely, GeForce 600/700 series to 900 series) or from Maxwell to Pascal (GeForce 900 series to 1000 series). If either of those transitions had gotten only 5-10% performance improvements, Nvidia would have lost the corresponding generation and badly. My basic claim is that a big jump like that is not likely for Nvidia until Volta and/or a 10 nm or smaller process node.
A large jump in top end performance in the first half of this year is probable for AMD, on the other hand, because they surely have a much bigger chip coming and Vega might also be a major architectural improvement over Polaris. But this is partially saying that AMD is more likely to make a big jump than Nvidia because AMD is behind right now.
Can they do that, with what they have laying around from their existing Pascal development? Almost certainly. They could ~probably~ do it with existing hardware designs and just some firmware tweaks. It's certainly not hard to imagine, and it doesn't require a new design, a new process node, or anything else to accomplish.
nVidia doesn't have to release Volta or an entire new 2000 series to counter Vega. All they have to do to neutralize Vega, is push a 1080Ti that doesn't cost $1200, lower the price of the existing 1080 and 1070 a bit, and unless Vega greatly outclasses nVidia (which has happened, but doesn't appear to be likely), nVidia keeps the headlines skewed their direction.
That doesn't mean I think Vega will be a bad card. I think Fury was a good card, and I think AMD can easily beat it with Vega, which will make it a contender in the upper-tier space again. I just think the deck is stacked in nVidia's favor at the moment, and the headlines will continue to slant Green, and most of what we see in the press will be more for the benefit of the stock market than anything to do with actual consumers.
I just hope is good improvement on Fury X i own now thats all im not obsessed with FPS and top contender.
I love ReLive now and drivers AMD deliver!
Vega will be a complete NEW architectural improvement over Polaris!!!
Hope to build full AMD system RYZEN/VEGA/AM4!!!
MB:Asus V De Luxe z77
CPU:Intell Icore7 3770k
GPU: AMD Fury X(waiting for BIG VEGA 10 or 11 HBM2?(bit unclear now))
MEMORY:Corsair PLAT.DDR3 1866MHZ 16GB
PSU:Corsair AX1200i
OS:Windows 10 64bit
Personally I feel that Nvidia is acting more like Intel lately and that is NOT good. I have been putting in 480's in a lot of builds lately just because of that. For most gamers that is plenty of HP.
WccfTech lol..yeh we believe that site to be true:P
Hope to build full AMD system RYZEN/VEGA/AM4!!!
MB:Asus V De Luxe z77
CPU:Intell Icore7 3770k
GPU: AMD Fury X(waiting for BIG VEGA 10 or 11 HBM2?(bit unclear now))
MEMORY:Corsair PLAT.DDR3 1866MHZ 16GB
PSU:Corsair AX1200i
OS:Windows 10 64bit