[SA] News of Nvidia’s Pascal tapeout and silicon is important

[Silverforce11]

This generation there are going to be interesting comparisons - Polaris 10 on 14 LPP vs Pascal on 16FF+. Polaris 11 with HBM2 might be on 16FF+ (due to better yields for larger dies on 16FF+ and TSMC's vast experience making large GPUs 300+ sq mm for more than a decade) and will go up against GP104.

AMD has been very clear, Polaris 10 and 11 are both designed for 14nm FF. Raja Koduri has said it publicly, AMD's presentations have stated it.

It will be Polaris 14nm FF vs Pascal 16nm FF.

 

[tential]

It's hilarious how much of a non issue this is but people are blowing it wildly out of proportion to make their team "look better".

It's quite sad.

 

[Silverforce11]

It's hilarious how much of a non issue this is but people are blowing it wildly out of proportion to make their team "look better".

It's quite sad.

So what are your thoughts with all the tech sites claiming we will get big Pascal in April?

 

[tential]

So what are your thoughts with all the tech sites claiming we will get big Pascal in April?

That was based off a made up road map so I'm putting 0 stock in it until something more concrete shows up.

As we've seen before nvidia does well. Releasing a gtx 980/970 card first. Not seeing why they would rock that boat now....

 

[Silverforce11]

That was based off a made up road map so I'm putting 0 stock in it until something more concrete shows up.

As we've seen before nvidia does well. Releasing a gtx 980/970 card first. Not seeing why they would rock that boat now....

And that's the point of the OP and SA's article, all those tech sites reporting that is pure fud.

They misinterpreted the air cargo manifest and relied on a random user generated roadmap.

ATM we have nothing concrete to suggest it is close to launch. April is not long away and zero leaks of performance, of anything.

Now, back to the bashing of Charlie's article, nothing anyone here has said or linked refutes SA's article. Nothing anyone here has presented suggests Pascal is imminent to release.

The only evidence we do have, suggests its awhile yet due to the fake PCB presented by JHH, a repeat of Fermi woodscrews.

Still, it's not concrete. But if the launch is that imminent, we would already have AIB sample performance data leaked, which always happens before any launch.

So any site that suggests Pascal is coming in April, clearly don't know what they are talking about.

 

[Arachnotronic]

So what are your thoughts with all the tech sites claiming we will get big Pascal in April?

If they don't release it in April, Intel will have a field day with Knights Landing...

 

[Silverforce11]

If they don't release it in April, Intel will have a field day with Knights Landing...

There's this argument and it is a very compelling one:

Intel will struggle to encroach on NV's dominance because the HPC sector and software ecosystem are used to CUDA.

 

[Arachnotronic]

There's this argument and it is a very compelling one:

Intel will struggle to encroach on NV's dominance because the HPC sector and software ecosystem are used to CUDA.

You would be surprised at how much share the relatively crappy 22nm Xeon Phi has already grabbed. NVIDIA would be outright dumb to not respond to 14nm Xeon Phi if it has the capacity to ASAP.

 

[tential]

And that's the point of the OP and SA's article, all those tech sites reporting that is pure fud.

They misinterpreted the air cargo manifest and relied on a random user generated roadmap.

ATM we have nothing concrete to suggest it is close to launch. April is not long away and zero leaks of performance, of anything.

Now, back to the bashing of Charlie's article, nothing anyone here has said or linked refutes SA's article. Nothing anyone here has presented suggests Pascal is imminent to release.

The only evidence we do have, suggests its awhile yet due to the fake PCB presented by JHH, a repeat of Fermi woodscrews.

Still, it's not concrete. But if the launch is that imminent, we would already have AIB sample performance data leaked, which always happens before any launch.

So any site that suggests Pascal is coming in April, clearly don't know what they are talking about.

This is not evidence....

That's my whole point. The fact that people are presenting it as evidence is mind boggling. Like I said, people are blowing a non issue out of proportion....

That whole thing, means NOTHING.

How you're tying that to Pascal coming in April is just a whole other mind boggling jump I don't even get.

In conclusion:
-Using a random roadmap to say "Hey, Pascal is coming in April!" that you find online is just purely idiotic. I won't even debate such a topic.
-Using the Nvidia Drive Press conference to infer Pascal may or may not come in April is also just insane.
-It's a press conference, not a hardware analysis breakdown. They care about being delivered a product that meets expectations. Not that the product that is being held right there is the product that they see. I can't believe I have to even explain this.
STOP DRAWING DEEP DIVE HARDWARE ANALYSIS FROM MARKETING PRESENTATIONS

 

[Silverforce11]

You seem confused @tential

This was about SA's article rebuking all the tech sites that claim Pascal is coming in April.

 

[AtenRa]

NVIDIA may have Pascal test dies in the labs but they definitely DONT have the Pascal SKUs they intend to use on the PX2.

As for the S/A article, I will have to agree here they are 99% correct. The manifest is not about the Pascal dies and the roadmap is not legit, making a conclusion of an April launch based on those two sources will only make it false.

 

[coercitiv]

This was about SA's article rebuking all the tech sites that claim Pascal is coming in April.

And the whole Nvidia Drive Press discussion started on the premise that Nvidia should not and will not show their pre-production GPU silicon to the public, since they have no need to show their hand in exchange for some market excitement on the matter.

All that makes sense, with the sole exception that Nvidia did feel the need to show a pre-production model of the Drive PX2 while ambiguously referring to the GPU chips on that piece of hardware as next gen Nvidia GPU based on the Pascal architecture.

Based on this info:

• is Nvidia late with Pascal? No idea, they might be doing just fine. Information is so irrelevant they might as well be ahead of schedule.
• did Nvidia feel the need to show silicon and call it Pascal in a marketing event? YES.

 

[Goatsecks]

Now, back to the bashing of Charlie's article, nothing anyone here has said or linked refutes SA's article.

Charlie writes with a persistent and overt bias, in particular, he will find a pro-AMD or anti-Nvidia angle for almost all his articles. The perceived validity of what he writes is a reflection of this.

There is barely any solid information regarding the current state of Pascal, so Charlie's criticism at the start of the piece is fair. However, Charlie's own conjecture is based on the same incomplete information; there is nothing to refute.

 

[railven]

Charlie writes with a persistent and overt bias, in particular, he will find a pro-AMD or anti-Nvidia angle for almost all his articles. The perceived validity of what he writes is a reflection of this.

There is barely any solid information regarding the current state of Pascal, so Charlie's criticism at the start of the piece is fair. However, Charlie's own conjecture is based on the same incomplete information; there is nothing to refute.

I use love reading charlie's articles before he went pay-wall. Any AMD/ATI fan would want to read his articles.

More of a reason to take anything he says about Nvidia (or Intel) with huge grains of salt. His Fermi predictions pretty much exploded in his face. He flung so much poop the stuff that stuck didn't distract from the smell of the room.




At this point, I still fail to understand why anyone even cares if Nvidia is "late"? NV was "late" with Kepler and Maxwell2 and they basically destroyed AMD in the process. Intel was late with everything and they're reporting revenue records. Outside of a "LOL they're late" meme. NV can show up a year late, if they deliver a product that dominates what are people going to say then? "Yeah, well they were late. LOLz."

 

[raghu78]

At this point, I still fail to understand why anyone even cares if Nvidia is "late"? NV was "late" with Kepler and Maxwell2 and they basically destroyed AMD in the process. Intel was late with everything and they're reporting revenue records. Outside of a "LOL they're late" meme. NV can show up a year late, if they deliver a product that dominates what are people going to say then? "Yeah, well they were late. LOLz."

The last time Nvidia were very late and had an underwhelming product AMD took market share and pretty much came to parity in market share.

http://www.xbitlabs.com/news/graphi..._on_Discrete_GPU_Market_Mercury_Research.html

This happened after Nvidia dominated the 8800 GTX generation and AMD fought back with HD 4800 and HD 5800 generations. I am not saying it will happen again. But the possibility for AMD to put up a fight and claw back market share exists. Its still a very hard task for AMD as Nvidia is in a much more dominant position today and is a much more stronger company while AMD is in a very bad market share situation and is a much more weaker company today.

If AMD want to be relevant in the GPU market they need to fight back hard
and sustain the pace of innovation. Between 2012 and 2015 AMD did not do much to improve efficiency and paid a heavy price in terms of market share. I hope they learnt a lesson and are more aggressive in their GPU development efforts. The formation of RTG is a good sign and we will have to see over the next 2 years what Raja Koduri does to bring back AMD to a healthy market share.

 

[Gundark]

I'm waiting for April 01 Pascal release.

 

[railven]

The last time Nvidia were very late and had an underwhelming product AMD took market share and pretty much came to parity in market share.

http://www.xbitlabs.com/news/graphi..._on_Discrete_GPU_Market_Mercury_Research.html

This happened after Nvidia dominated the 8800 GTX generation and AMD fought back with HD 4800 and HD 5800 generations. I am not saying it will happen again. But the possibility for AMD to put up a fight and claw back market share exists. Its still a very hard task for AMD as Nvidia is in a much more dominant position today and is a much more stronger company while AMD is in a very bad market share situation and is a much more weaker company today.

If AMD want to be relevant in the GPU market they need to fight back hard
and sustain the pace of innovation. Between 2012 and 2015 AMD did not do much to improve efficiency and paid a heavy price in terms of market share. I hope they learnt a lesson and are more aggressive in their GPU development efforts. The formation of RTG is a good sign and we will have to see over the next 2 years what Raja Koduri does to bring back AMD to a healthy market share.

There was a huge key difference. Price. AMD has already shown they are not willing to go back to being second fiddle (unless, I guess forced). I don't expect AMD to launch first and leave revenue on the table. Now if NV is THAT late I'd expect them to counter with price. But something tells me, they won't have to. (I just get this sneaking suspicion NV is mums because, they aren't threatened.)

If AMD launches big Polaris and it's only 20-30% faster than 980 Ti, I don't see NV worrying. Especially if the two cost similar. Now, if we have a repeat of the golden days where AMD launches at almost 60% the cost of NV's top dog AND outperforms it, NV better be ready for a fight.

I guess, I just don't have the confidence in AMD delivering. Not anymore.

EDIT: THis is why I don't see the good old days returning, and note I defended HD 7970 and it's price increase. But, if this happens again:

I really don't see NV breaking a sweat. AMD has to come out BLAZING. And I just don't think AMD has it left in them.

 

[raghu78]

There was a huge key difference. Price. AMD has already shown they are not willing to go back to being second fiddle (unless, I guess forced). I don't expect AMD to launch first and leave revenue on the table. Now if NV is THAT late I'd expect them to counter with price. But something tells me, they won't have to. (I just get this sneaking suspicion NV is mums because, they aren't threatened.)

If AMD launches big Polaris and it's only 20-30% faster than 980 Ti, I don't see NV worrying. Especially if the two cost similar. Now, if we have a repeat of the golden days where AMD launches at almost 60% the cost of NV's top dog AND outperforms it, NV better be ready for a fight.

AMD's pricing has been aggressive in the R9 290X / R9 290 launch. Unfortunately they messed up on the quality of reference cards and the bitcoin mining rush hurt them badly as resellers resorted to price gouging and when AMD had better supply and pricing came down to MSRP they gained a bit of market share in Q2 2014.

http://jonpeddie.com/press-releases...ket-down-in-q2-nvidia-holds-market-share-lead

But it was too little , too late and Nvidia had GM104 ready in Q3 2014 and destroyed them. AMD's situation in the GPU market is very dangerous. They have never gone below 25% market share (they are at 20% now) and have a unsustainable GPU business without atleast 35% GPU market share. They have no option but to do everything to gain market share.

With Fury X AMD were faced with cost issues as they were launching a GPU which uses 2.5D stacking with HBM for the first time and with a very large GPU die. The fact that AMD could not beat Nvidia with similar die size was the killer blow and it came down to AMD not making enough architectural efficiency improvements. In the past AMD could come within 10-15% of Nvidia's big GPU with a 25-30% smaller die this time they failed badly on perf/transistor and perf/sq mm especially when you consider GTX Titan / 980 Ti OC with Fury X OC.

I guess, I just don't have the confidence in AMD delivering. Not anymore.

EDIT: THis is why I don't see the good old days returning, and note I defended HD 7970 and it's price increase. But, if this happens again:

I really don't see NV breaking a sweat. AMD has to come out BLAZING. And I just don't think AMD has it left in them.

One of the advantages unlike when HD 7970 launched is AMD's not building a from scratch architecture. They are making the necessary architectural efficiency improvements but retaining the core GCN elements which make it a very good gaming/compute architecture. I also believe AMD realized their software efforts needed to be much better and they are making the improvements. The Polaris architecture launch is a test as to whether AMD has changed or not.

btw you would be surprised to see what people or companies can do when pushed to the brink of extinction. I think AMD still has the fight. If they disappoint this generation I will write them off and their end won't be far away.

 

[Headfoot]

There's this argument and it is a very compelling one:

Intel will struggle to encroach on NV's dominance because the HPC sector and software ecosystem are used to CUDA.

Even if lots of stuff is written in CUDA, way, way more stuff is written for generic multicore/multisocket x86

 

[railven]

AMD's pricing has been aggressive in the R9 290X / R9 290 launch. Unfortunately they messed up on the quality of reference cards and the bitcoin mining rush hurt them badly as resellers resorted to price gouging and when AMD had better supply and pricing came down to MSRP they gained a bit of market share in Q2 2014.

http://jonpeddie.com/press-releases...ket-down-in-q2-nvidia-holds-market-share-lead

But it was too little , too late and Nvidia had GM104 ready in Q3 2014 and destroyed them. AMD's situation in the GPU market is very dangerous. They have never gone below 25% market share (they are at 20% now) and have a unsustainable GPU business without atleast 35% GPU market share. They have no option but to do everything to gain market share.

The aftermath of the bitmining craze is what lead me to lose faith in AMD. They rode so high at that point. And the marketshare they gained had zero to do with their product and gaming (you and I both know this). When their cards were priced 20-50% over MSRP (even the lowly 280X was encroaching a $400+ price tag) yet selling - it wasn't because of it's gaming performance. But AMD put the gas to the floor and figured they'd ride that wave into revenue nirvana. Then the bubble popped and AMD scrambled to unload, unfortunately for them those cards they sold few months prior were now being sold at super cheap discounts. This sort of painted the perceptive value of AMD cards to me. If the cards aren't generating you money, they are worth a fraction of what you originally paid. NV laughed selling their cards at MSRP, and then laughed some more when AMD used cards fell well below MSRP requiring AMD to price cut themselves just to make NEW cards attractive versus refurbed/used. Ask yourself why a faster 290X with driver improvements were costing almost half of a slower GTX 780? AMD devalued their cards with their knee jerk reaction of dropping MSRP. Then Maxwell came out and welps, we know where the enthusiast/bleeding-edge subset of buyers went.

With Fury X AMD were faced with cost issues as they were launching a GPU which uses 2.5D stacking with HBM for the first time and with a very large GPU die. The fact that AMD could not beat Nvidia with similar die size was the killer blow and it came down to AMD not making enough architectural efficiency improvements. In the past AMD could come within 10-15% of Nvidia's big GPU with a 25-30% smaller die this time they failed badly on perf/transistor and perf/sq mm especially when you consider GTX Titan / 980 Ti OC with Fury X OC.

And this is just another example of AMD mismanaging their products to me. "Let's sell a slower product, that has less memory and uses more power at the same price of out primary competitor." I get the reasoning, but any buyer worth his salt would steer clear of the Fury X outside of just wanting to own one.

Adding injury to insult is as you said Fiji was a monster in the technical aspect of designs. So AMD most likely probably wasn't making bank selling at that price point, going any lower would not be possible. So, woof, again.

One of the advantages unlike when HD 7970 launched is AMD's not building a from scratch architecture. They are making the necessary architectural efficiency improvements but retaining the core GCN elements which make it a very good gaming/compute architecture. I also believe AMD realized their software efforts needed to be much better and they are making the improvements. The Polaris architecture launch is a test as to whether AMD has changed or not.

But NV is building an ecosystem that is blocking AMD out at launch of very popular titles. It doesn't matter if AMD improves on GCN if GCN isn't the focal point for popular titles. Again, a dev they're working with put NV's well being before AMDs. That tells me AMD has relatively no say/power in pushing the industry.

All this hooting and hollering of what DX12 is and what it will bring means nothing until the games are using it. And I wouldn't be surprised by the time the games are using NV has a product that dwarves AMD's new GPU. DX12 is only bring AMD to parity to where NV was with DX11. That isn't instilling confidence in me. If NV ends up showing similar gains that AMD got with their new line up, well, then what?

btw you would be surprised to see what people or companies can do when pushed to the brink of extinction. I think AMD still has the fight. If they disappoint this generation I will write them off and their end won't be far away.

This sort of contradicts itself. AMD's back has been up against a wall for a few years now. They botched Hawaii. They botched Fiji.

 

[xthetenth]

You would be surprised at how much share the relatively crappy 22nm Xeon Phi has already grabbed. NVIDIA would be outright dumb to not respond to 14nm Xeon Phi if it has the capacity to ASAP.

This is definitely true, I expect to see big Pascal coming out very early with regards to the rest of the generation. What I don't expect is consumer big Pascal for a while, when they're working to get yields in order and their HPC business needs them. What's interesting is that they didn't have Pascal to put on that demo but still felt the need to say that it was rather than that it would be.

 

[Mondozei]

If AMD launches big Polaris and it's only 20-30% faster than 980 Ti, I don't see NV worrying. Especially if the two cost similar. Now, if we have a repeat of the golden days where AMD launches at almost 60% the cost of NV's top dog AND outperforms it, NV better be ready for a fight.

I guess, I just don't have the confidence in AMD delivering. Not anymore.

EDIT: THis is why I don't see the good old days returning, and note I defended HD 7970 and it's price increase. But, if this happens again:

I really don't see NV breaking a sweat. AMD has to come out BLAZING. And I just don't think AMD has it left in them.

I think you're confusing AMD with the state of FinFET. I doubt NV's GP104 flagship will be much faster than 20-30% of the 980 Ti either. So you're basically holding AMD to a much higher standard and what for?

For AMD to win, it has to basically be much better than it used to be. It's enough to be on-par with NV on the hardware side but their key selling arguments will be better drivers and the free ecosystem, look at the increasing viability of Freesync.

Some, presumably like you, will likely cling to NV no matter what, but there's plenty of people in the middle who will go back to AMD if they get their ship in order and offer a clearly better alternative to NV on issues like driver performance over several years, which they are already doing, and their cheaper/fairer ecosystem approach(Freesync, GPU open etc).

If AMD continues to execute on the software side and eliminates the deficits with NV on the hardware side, that alone will bring them more market share. If they want to have a majority market share, that will take time. Right now, it has to be to regain the lost ground and stabilise. What comes after that, comes after that.

 

[Dribble]

I think you're confusing AMD with the state of FinFET. I doubt NV's GP104 flagship will be much faster than 20-30% of the 980 Ti either. So you're basically holding AMD to a much higher standard and what for?

For AMD to win, it has to basically be much better than it used to be. It's enough to be on-par with NV on the hardware side but their key selling arguments will be better drivers and the free ecosystem, look at the increasing viability of Freesync.

Some, presumably like you, will likely cling to NV no matter what, but there's plenty of people in the middle who will go back to AMD if they get their ship in order and offer a clearly better alternative to NV on issues like driver performance over several years, which they are already doing, and their cheaper/fairer ecosystem approach(Freesync, GPU open etc).

If AMD continues to execute on the software side and eliminates the deficits with NV on the hardware side, that alone will bring them more market share. If they want to have a majority market share, that will take time. Right now, it has to be to regain the lost ground and stabilise. What comes after that, comes after that.

That's a lot of "if's" and against that are the fact AMD is a smaller company with smaller dev teams which have been gutted by several rounds of layoff's. Software wise AMD could never hold a candle to Nvidia even before they gutted the teams with layoffs, so that's a given that Nvidia will be ahead again (i.e. AMD will probably be a year+ behind on levels of DX12/Vulcan driver optimisation again on the next gen of gpu's). However for hardware...

It might be that samsung's 14nm for smaller chip's is available for AMD before TSMC's 16nm is there for nvidia. In that case AMD might get a leg up for smaller chips - low end gpu's, mid range laptop stuff by a month or two. I'd be very surprised if samsung can beat TSMC to big chips, I'd be fairly surprised if they are even going to attempt big chips. Almost certainly both AMD and nvidia will turn to TSMC for the larger chips and there nvidia is going to be first due to the fact they are a larger customer willing to pay more (as they can sell those chips in HPC for big $$$).

That's going to be much later in the year not just because it'll be hard for TSMC to make the big chips, but because as the spec's for GDDR5X and HBM2 just got finalised and the memory won't be available for a while.

 

[MrTeal]

The aftermath of the bitmining craze is what lead me to lose faith in AMD. They rode so high at that point. And the marketshare they gained had zero to do with their product and gaming (you and I both know this). When their cards were priced 20-50% over MSRP (even the lowly 280X was encroaching a $400+ price tag) yet selling - it wasn't because of it's gaming performance. But AMD put the gas to the floor and figured they'd ride that wave into revenue nirvana. Then the bubble popped and AMD scrambled to unload, unfortunately for them those cards they sold few months prior were now being sold at super cheap discounts. This sort of painted the perceptive value of AMD cards to me. If the cards aren't generating you money, they are worth a fraction of what you originally paid. NV laughed selling their cards at MSRP, and then laughed some more when AMD used cards fell well below MSRP requiring AMD to price cut themselves just to make NEW cards attractive versus refurbed/used. Ask yourself why a faster 290X with driver improvements were costing almost half of a slower GTX 780? AMD devalued their cards with their knee jerk reaction of dropping MSRP. Then Maxwell came out and welps, we know where the enthusiast/bleeding-edge subset of buyers went.

You're making the assumption that the large increases in selling price for the GCN GPUs translated into increased revenue for AMD. Additional units obviously would increase revenue, but without knowing more you can't say that Newegg selling a 290 for $600 put any more money in their pockets than one Newegg sold at $400. Unfortunately for them, there's not a whole lot they can do about that as there really wasn't a way to segment mining sales from gaming sales. The abrupt collapse of the market and the flood of used cards is something AMD would have had limited options to counteract. They could have kept MSRP up at $400/$550, but by the time the glut of used/refurb cards cleared out Hawaii would have been facing GM204, and that's a losing battle at the same prices.

I wouldn't exactly call Hawaii botched either, other than the loud reference blower, and the review clock speed issue. Hawaii competed very well against GK110, and it wasn't until a year later when GM204 was released that Hawaii really started losing out.

 

[railven]

I think you're confusing AMD with the state of FinFET. I doubt NV's GP104 flagship will be much faster than 20-30% of the 980 Ti either. So you're basically holding AMD to a much higher standard and what for?

Am I? Where did I say that? I'm saying AMD has a bigger mountain to climb than NV. NV can sit 5% faster than AMD and have no issues. AMD had a card that was almost 20% faster than it's rival card costing 10% more and they still lost market share.

For AMD to win, it has to basically be much better than it used to be. It's enough to be on-par with NV on the hardware side but their key selling arguments will be better drivers and the free ecosystem, look at the increasing viability of Freesync.

AMD has to change its perception. They've been the bargain brand for years. They try to say "no we're also premium" and then fall flat on their face.

IE, for AMD to try to get away charging what NV is charging, they have to be at minimum on par. But instead they trout out a slower, hungrier, with less memory card and try to sell "but it's a technical marvel, HBM!" HA.

Some, presumably like you, will likely cling to NV no matter what, but there's plenty of people in the middle who will go back to AMD if they get their ship in order and offer a clearly better alternative to NV on issues like driver performance over several years, which they are already doing, and their cheaper/fairer ecosystem approach(Freesync, GPU open etc).

Que? Amigo, you don't know me. I've only given NV money for a card I bought for myself twice in my 20+ years of PC gaming. I'd gladly go back to AMD. I still prefer their products. But like people here love calling NV buyers sheep, it be more sheepish to buy an inferior product (using the criteria of performance at my resolution, power consumption in my usage, and total VRAM available to me) as being blind.

GPUOpen can suck an egg if it doesn't get put into the games I play. Why do I care if AMD is improving the future for all gamers if the games I play they completely ignore? To make you, and what it seams the average AMD user, I need to suck it up and deal with inferior performance for the games I play just to promote AMD over NV? Screw that. And that doesn't translate to "hooraah NV is the best!" My time and money are both limited. I'm going to buy the product that makes best use of both.

If AMD continues to execute on the software side and eliminates the deficits with NV on the hardware side, that alone will bring them more market share. If they want to have a majority market share, that will take time. Right now, it has to be to regain the lost ground and stabilise. What comes after that, comes after that.

Here is my question. What happened to the AMD of 2012-2013? When they were partnering up with game titles on the ying yang? They were securing titles like Bioshock, Battlefield, Thief, FFXIV, Tomb Raider? Is it the lack of funding? Because if it is, they are in for a long road to turn this around. They can improve their software stack all they want, but when NV rolls up with a bag of dollars and devs are creating exceptions for them, AMD is boned.

You're making the assumption that the large increases in selling price for the GCN GPUs translated into increased revenue for AMD. Additional units obviously would increase revenue, but without knowing more you can't say that Newegg selling a 290 for $600 put any more money in their pockets than one Newegg sold at $400.

No, that's what you're interpreting. I said the inflated price hurt AMD because they didn't see any of it. And when the bubble burst that new inventory they brought to market to capitalize on the bitmining craze was cannibalized by the used market. So AMD got shafted.
1) by inflated [EDIT:vendor] prices that they didn't see a dime off, which put the cards squarely in the hands of miners, not gamers
2) miners who would later unload these cards at discount prices stealing sales away from the fresh batch of cards from AMD (remember the cards were also in short supply and were often selling fast [to miners] so when a plethora of used cards hit the scene, they sold rather well].
3) NV's price cut to cards directly competing with them

My mentioning of AMD raising the price was to draw a comparative to how I felt they botched Fiji. They raised the price of Fiji going with the "premium" concept. When they had a product that would actually sell, ie bit miners would have bad stupid amounts of money for those cards) they left the MSRP low and allowed vendors to rake in the dough. But Fiji's turn around, they had no issues trying to fleece their biggest supports (ie again, selling something on many metrics inferior to their rival at the same price point).

Unfortunately for them, there's not a whole lot they can do about that as there really wasn't a way to segment mining sales from gaming sales. The abrupt collapse of the market and the flood of used cards is something AMD would have had limited options to counteract. They could have kept MSRP up at $400/$550, but by the time the glut of used/refurb cards cleared out Hawaii would have been facing GM204, and that's a losing battle at the same prices.

Or, they could have upped MSRP, got more money from it, that would have countered when the bubble burst. I mean, Nvidia already had a card on the market for $1,000? Hey AMD :
"For the bitminer who wants the best hash, introducing the AMD Radeon Bitminer Edition! MSRP $1,000!"

I bet it would have sold out. When the bubble burst (or supply got higher) bring out the 290X. I mean, what does it matter if only a handful of gamers were lucky enough to get a 290X before the vendors jacked up the price?

I wouldn't exactly call Hawaii botched either, other than the loud reference blower, and the review clock speed issue. Hawaii competed very well against GK110, and it wasn't until a year later when GM204 was released that Hawaii really started losing out.

When your card is selling easily for almost double MSRP and you aren't seeing a dime of it - yeah, you botched it. Who ever over at NV said "we can sell big Kepler for $1,000 to a market that doesn't exist" probably got a promotion. Meanwhile, AMD HQ is salivating at their cards selling out but they're only see a fraction of that money.