NVIDIA Could Capitalize on AMD GCN Not Supporting Direct3D 12_1

[patrickjp93]

Analysts are known to jump the gun. How many times have they predicted Nintendo going under, for example? Wasn't BlackBerry meant to have gone out of business already too?

Market analysts and financial analysts are very different animals. Market analysts have predicted Nintendo being unable to keep pleasing new generations of gamers. Obviously, they underestimate the nostalgia factor and the fact everyone knows a Nintendo console is not meant to be super expensive or on the leading edge of performance.

 

Market analysts forgot Blackberry still has a big business user following.

 

Financial analysts, having gone into the bowels of the company and releasing a detailed report multiple hundreds of pages long concluding AMD is likely to fail is a slightly different matter.

[bogus]

Let's hope if article, written by someone who was in the early access program for DX12, clears a bit of this fanboyism fueled circle jerking rant - you should read it, for once, this one article actually deserves a click: http://www.bitsandchips.it/52-english-news/5661-clarifications-about-tier-and-feature-levels-of-the-directx-12

 

 

(...)

The first two feature levels roughly coincide to the DirectX 11 levels with the same name (with some differences due the new resource binding model), while feature level 12.0 and 12.1 are new to Direct3D 12.

Despite being pleonastic, it is worth to restate that feature level 12.1 does not coincide with an imaginary “full/complete DirectX 12 support” since it does not cover many important or secondary features exposed by Direct3D 12.

 

In the end, as regards the support of every single capability, it is currently not possible, nor appropriate, to draw up a complete and well-defined table showing the support of on sale hardware.
Unless you name is AMD, INTEL or NVIDIA, you cannot present such report with the drivers currently available on the public channels, nor with non-NDA documentation, therefore everything else is only to be considered as pure rants.

[mr moose]

Nvidia actively "putt the knife in", as it were, would definitely be anti-trust.

 

The point is that becoming a monopoly, aside from being horrific to the consumer and actually against the whole ethos of the free market, would open Nvidia up to an extreme level of scrutiny of their every action. That's something that they do not want. For them it's far simpler to just keep their 75% market dominance.

 

If and I mean IF, Nvidia did something illegal then it would be anti trust, but if they don't then it is just a monopoly, Which has been pointed out in and of itself is not illegal.

 

Everything else is just personal fear, hearsay, assumption or accusation.

[Dabombinable]

If and I mean IF, Nvidia did something illegal then it would be anti trust, but if they don't then it is just a monopoly, Which has been pointed out in and of itself is not illegal.

 

Everything else is just personal fear, hearsay, assumption or accusation.

Yep.

Quite a few people seem to forget that AMD's products started falling behind and as a result consumers spoke with their wallet and didn't buy an AMD product as their next upgrade.

[mr moose]

Yep.

Quite a few people seem to forget that AMD's products started falling behind and as a result consumers spoke with their wallet and didn't buy an AMD product as their next upgrade.

 

Which is how every market works, The problem is accepting that something is second best or not good enough is not human nature and sadly one that is not taught in early education.

[Techhog]

Market analysts and financial analysts are very different animals. Market analysts have predicted Nintendo being unable to keep pleasing new generations of gamers. Obviously, they underestimate the nostalgia factor and the fact everyone knows a Nintendo console is not meant to be super expensive or on the leading edge of performance.

 

Market analysts forgot Blackberry still has a big business user following.

 

Financial analysts, having gone into the bowels of the company and releasing a detailed report multiple hundreds of pages long concluding AMD is likely to fail is a slightly different matter.

Did those analysts fact in that AMD dying would immediately kill the entire console business, giving Nintendo, Sony, and Microsoft huge incentives to help keep them afloat?

[patrickjp93]

Did those analysts fact in that AMD dying would immediately kill the entire console business, giving Nintendo, Sony, and Microsoft huge incentives to help keep them afloat?

The console business in a year doesn't generate even the first 600 million AMD owes January 1st, 2019.

 

Also, Qualcomm or Mediatek could take over. ARM will be strong enough by that point, and Mediatek has licensed AMD's GPU tech and used it several times.

[othertomperson]

If and I mean IF, Nvidia did something illegal then it would be anti trust, but if they don't then it is just a monopoly, Which has been pointed out in and of itself is not illegal.

 

Everything else is just personal fear, hearsay, assumption or accusation.

 

People keep pointing it out despite no one trying to claim that it is illegal. What I have said is that being a monopoly, especially for a company as small as Nvidia, comes with a shit tonne of scrutiny and always the threat of anti-trust -- even if they haven't done anything -- that Nvidia do not want. I can't imagine Nvidia particularly want more than 75%-80% market dominance.

[Opcode]

The console business in a year doesn't generate even the first 600 million AMD owes January 1st, 2019.

 

Also, Qualcomm or Mediatek could take over. ARM will be strong enough by that point, and Mediatek has licensed AMD's GPU tech and used it several times.

The console industry between now and 2019 can easily generate in excess of $1 billion for AMD. With them targeting profitability by the end of this year they will have the opportunity to secure a separate account for covering their debts. Console sales between now and then will easily cover all that's owed in 2019 with quite a bit of surplus. If you look at current console sales revenue for AMD ($2 billion) along with them securing Nintendo's next console coupled with large untapped markets (China). It's safe to say that if AMD took all of their console sales revenue and pushed it aside they can secure enough funds to pay off a massive chunk of their total debt between now and 2020. The problem AMD faces right now is returning to profitability so they don't need to tap into other resources to keep the company moving. Most people don't see AMD going bankrupt by 2020 whether it be bonds or sheer profitability and I personally don't foresee them going under either. As long as they deliver exceptional products over the course of the next few years. However if Lisa Su pulls a "Rory" and goes out and purchases Xilinx just so they can be competitive with Intel in FPGAs then sure, all aboard the lifeboats. I have faith in Lisa's new leadership and have grown to respect her as every time she's in front of the camera she's not spewing bullshit. Hell she didn't even want to acknowledge Fiji at Computex as she sat there holding the monster up for the camera. When she does speak on behalf of the company though she always capitalizes on a few things, two of which being high performance x86 and graphics leadership. Rory needed to be kicked the f' out of there as all he was doing was trying to branch AMD off into all kinds of markets. They need to focus their resources entirely into their strong suits that being x86 and graphics. Now I'm not suggesting that there's no way for AMD to go bankrupt between now and 2020 although there are a lot of routes that the company can peruse to avoid it.

[mr moose]

People keep pointing it out despite no one trying to claim that it is illegal. What I have said is that being a monopoly, especially for a company as small as Nvidia, comes with a shit tonne of scrutiny and always the threat of anti-trust -- even if they haven't done anything -- that Nvidia do not want. I can't imagine Nvidia particularly want more than 75%-80% market dominance.

 

I think that is the issue here, many people are claiming it is illegal. I have responded to about 4 posts now on various threads where people are under the impression (and said outright) that if AMD fold, Nvidia will automatically be broken up. 

[othertomperson]

I think that is the issue here, many people are claiming it is illegal. I have responded to about 4 posts now on various threads where people are under the impression (and said outright) that if AMD fold, Nvidia will automatically be broken up. 

 

Oh really? I'm only responding to the messages addressing me personally because I'm tired and there's 10 pages here. If I'm mistaking you and patrick speaking generally under a quote of something I've said for mis-interpreting the quote then I apologise.

[patrickjp93]

The console industry between now and 2019 can easily generate in excess of $1 billion for AMD. With them targeting profitability by the end of this year they will have the opportunity to secure a separate account for covering their debts. Console sales between now and then will easily cover all that's owed in 2019 with quite a bit of surplus. If you look at current console sales revenue for AMD ($2 billion) along with them securing Nintendo's next console coupled with large untapped markets (China). It's safe to say that if AMD took all of their console sales revenue and pushed it aside they can secure enough funds to pay off a massive chunk of their total debt between now and 2020. The problem AMD faces right now is returning to profitability so they don't need to tap into other resources to keep the company moving. Most people don't see AMD going bankrupt by 2020 whether it be bonds or sheer profitability and I personally don't foresee them going under either. As long as they deliver exceptional products over the course of the next few years. However if Lisa Su pulls a "Rory" and goes out and purchases Xilinx just so they can be competitive with Intel in FPGAs then sure, all aboard the lifeboats. I have faith in Lisa's new leadership and have grown to respect her as every time she's in front of the camera she's not spewing bullshit. Hell she didn't even want to acknowledge Fiji at Computex as she sat there holding the monster up for the camera. When she does speak on behalf of the company though she always capitalizes on a few things, two of which being high performance x86 and graphics leadership. Rory needed to be kicked the f' out of there as all he was doing was trying to branch AMD off into all kinds of markets. They need to focus their resources entirely into their strong suits that being x86 and graphics. Now I'm not suggesting that there's no way for AMD to go bankrupt between now and 2020 although there are a lot of routes that the company can peruse to avoid it.

In revenue or in profit? In profit the console ventures have net AMD about 80 million so far.

 

Also, you're crazy. The per-unit margins on console sales are barely better than the margins on phone SOCs for everyone but Apple, and the sales numbers are far lower too.

 

Those routes require it beat Intel in performance and get developers on its side for APU software development. The first isn't possible unless Jim Keller created a memory subsystem 50% better than anything Intel has on the market, and without an L3 cache, you'd be dreaming. If AMD puts HBM on all of its Zen CPUs/APUs, it stands a chance of beating the Skylake SKUs with no eDRAM, but just the 40% IPC increase over Excavator puts it squarely at Haswell (Integer Linear Program where the optimization function is an aggregate of scores from open-source benchmarks, normalized of course). Even providing more cores will mean nothing in the consumer segment. We don't have software yet pushing a 2500K, much less a 4790K, unless you do remotely professional work like game streaming, rendering, autocad, etc., and then at that point Intel can drop prices for its enthusiast chips to ensure AMD doesn't gain more than enough to stay alive. Intel could also toss HMC or eDRAM onto every SKU.

 

AMD has a chance to beat Nvidia in graphics if the R 400 series is made at GloFo and is ready to roll out before Pascal, and maybe at that point AMD can make some accelerator sales too. For the 300 series, it's looking like rebrands with slightly better core clocks and some memory speed upgrades. I don't see that being enough to drive sales, even if the Fury XT beats the Titan X (the Nano and Pro won't), and it would have to come with an 8GB frame buffer which isn't possible without using a dual-interposer and thus halving the bandwidth (below that of the GTX 980TI mind you at just 320 GB/s) or Hynix very quickly respinning a double-density 2.5D HBM SKU to put 4 chips on 1 interposer.

 

It's all fine and dandy to hope AMD pulls a hat trick, but at least be honest with yourself and admit it's not remotely likely given what we know and what we've seen.

[Opcode]

In revenue or in profit? In profit the console ventures have net AMD about 80 million so far.

 

Also, you're crazy. The per-unit margins on console sales are barely better than the margins on phone SOCs for everyone but Apple, and the sales numbers are far lower too.

 

Those routes require it beat Intel in performance and get developers on its side for APU software development. The first isn't possible unless Jim Keller created a memory subsystem 50% better than anything Intel has on the market, and without an L3 cache, you'd be dreaming. If AMD puts HBM on all of its Zen CPUs/APUs, it stands a chance of beating the Skylake SKUs with no eDRAM, but just the 40% IPC increase over Excavator puts it squarely at Haswell (Integer Linear Program where the optimization function is an aggregate of scores from open-source benchmarks, normalized of course). Even providing more cores will mean nothing in the consumer segment. We don't have software yet pushing a 2500K, much less a 4790K, unless you do remotely professional work like game streaming, rendering, autocad, etc., and then at that point Intel can drop prices for its enthusiast chips to ensure AMD doesn't gain more than enough to stay alive. Intel could also toss HMC or eDRAM onto every SKU.

 

AMD has a chance to beat Nvidia in graphics if the R 400 series is made at GloFo and is ready to roll out before Pascal, and maybe at that point AMD can make some accelerator sales too. For the 300 series, it's looking like rebrands with slightly better core clocks and some memory speed upgrades. I don't see that being enough to drive sales, even if the Fury XT beats the Titan X (the Nano and Pro won't), and it would have to come with an 8GB frame buffer which isn't possible without using a dual-interposer and thus halving the bandwidth (below that of the GTX 980TI mind you at just 320 GB/s) or Hynix very quickly respinning a double-density 2.5D HBM SKU to put 4 chips on 1 interposer.

 

It's all fine and dandy to hope AMD pulls a hat trick, but at least be honest with yourself and admit it's not remotely likely given what we know and what we've seen.

AMD secures $10 for every console sale be their agreement.

 

Between the XBOX One and PS4 there has been over 200,000 units sold world wide.

 

There is no need for AMD to actually beat anyone and still remain in business. That's nothing more than fabricated information that people portray has to happen in order for AMD to survive. In which case AMD has proven that's not entirely true for a decade now (they were in deeper debt before now). The only thing they need to do is be competitive in the markets that their products coexist in. Zen will have L3 so I'm not sure where you get the impression that it wont. Consumer APU wise really doesn't matter as that's only going to account for a very tiny fraction of sales in the consumer market (people are going to be wanting the 8 core 16 thread monsters). Any HPC APU that AMD builds we are pretty confident will come with HBM at that time. I also am interested in how you can claim the solid performance of an architecture without having an ES unit let alone any of the internal workings of the new architecture. Hell even Intel and AMD don't know precisely how their designs are going to perform until they hit fabrication. With Carrizo scratching away at Nehalm and Sandy Bridge a 40% IPC uplift entirely from architecture could be departmental to Skylake offerings as from Sandy Bridge to Haswell Intel has only accumulated a 15-17% IPC uplift. So I wouldn't get your hopes up of Intel crushing AMD without question as no one here could ever possibly know. You can speculate based on track record but really no one was expecting the Athlon 64 (especially Intel). As soon as we establish Excavator performance then we can speculate as to how much of a threat or not Zen will be in the market. We could be looking at a K8 vs P68 era upon us for all we know.

 

I honestly hope AMD does go with GloFo 14nm FinFET process as it's damn impressive thus far. A 85% reduction in leakage compared to their 28nm process tells a hell of a lot about current 28nm process (it's pretty bad). The 300 series will likely be all rebrands although AMD was apparently wise enough to not make it all straight HD 7000 rebrands. A large portion of the GPU's that people will be interested in will only be on their second rebrand (Hawaii/Tonga). Which of course is not as great as entirely new products (other than Tonga XT and Fiji) but it's done with purpose. They will still sell (a lot of consumers won't know the difference) and help AMD further resolve their inventory problem. You don't need 8GB to run 1080p, 1440p or even 4k. Look at how well Vesuvius performs (beats the TITAN X quite often at 4k) with only 4GB of VRAM (dual GPU 4+4=8GB but is still 4GB in DX11). With improved drivers 4GB is of no limitation at this current point in time. By the time 8GB truly matters (e.g. makes a huge difference) the card would seen a new revision with HBM2 (or haven been replaced). Keep in mind HBM2 design wins have been in progress since last year as Hynix stated. I don't think it will take too long for HBM2 to eventually hit volume production. We know AMD will be first in line to get their hands on some.

 

They don't need to pull Intel out of a hat, they just really need to play their cards right. There's a lot more to the puzzle than what financial analysts predict. If AMD launches a new Bulldozer and GCN tanks then sure I have no problems saying all aboard the lifeboats (as I've stated in my previous reply). Although given what the company is projecting for next year and alternative escape routes I don't see the company going under in 2020. They could bankrupt shortly after that but between now and then I like many others don't foresee it as long as they are competitive. You can never rule out the possibility of anything when it remains a mystery.

[patrickjp93]

AMD secures $10 for every console sale be their agreement.

 

Between the XBOX One and PS4 there has been over 200,000 units sold world wide.

 

There is no need for AMD to actually beat anyone and still remain in business. That's nothing more than fabricated information that people portray has to happen in order for AMD to survive. In which case AMD has proven that's not entirely true for a decade now (they were in deeper debt before now). The only thing they need to do is be competitive in the markets that their products coexist in. Zen will have L3 so I'm not sure where you get the impression that it wont. Consumer APU wise really doesn't matter as that's only going to account for a very tiny fraction of sales in the consumer market (people are going to be wanting the 8 core 16 thread monsters). Any HPC APU that AMD builds we are pretty confident will come with HBM at that time. I also am interested in how you can claim the solid performance of an architecture without having an ES unit let alone any of the internal workings of the new architecture. Hell even Intel and AMD don't know precisely how their designs are going to perform until they hit fabrication. With Carrizo scratching away at Nehalm and Sandy Bridge a 40% IPC uplift entirely from architecture could be departmental to Skylake offerings as from Sandy Bridge to Haswell Intel has only accumulated a 15-17% IPC uplift. So I wouldn't get your hopes up of Intel crushing AMD without question as no one here could ever possibly know. You can speculate based on track record but really no one was expecting the Athlon 64 (especially Intel). As soon as we establish Excavator performance then we can speculate as to how much of a threat or not Zen will be in the market. We could be looking at a K8 vs P68 era upon us for all we know.

 

I honestly hope AMD does go with GloFo 14nm FinFET process as it's damn impressive thus far. A 85% reduction in leakage compared to their 28nm process tells a hell of a lot about current 28nm process (it's pretty bad). The 300 series will likely be all rebrands although AMD was apparently wise enough to not make it all straight HD 7000 rebrands. A large portion of the GPU's that people will be interested in will only be on their second rebrand (Hawaii/Tonga). Which of course is not as great as entirely new products (other than Tonga XT and Fiji) but it's done with purpose. They will still sell (a lot of consumers won't know the difference) and help AMD further resolve their inventory problem. You don't need 8GB to run 1080p, 1440p or even 4k. Look at how well Vesuvius performs (beats the TITAN X quite often at 4k) with only 4GB of VRAM (dual GPU 4+4=8GB but is still 4GB in DX11). With improved drivers 4GB is of no limitation at this current point in time. By the time 8GB truly matters (e.g. makes a huge difference) the card would seen a new revision with HBM2 (or haven been replaced). Keep in mind HBM2 design wins have been in progress since last year as Hynix stated. I don't think it will take too long for HBM2 to eventually hit volume production. We know AMD will be first in line to get their hands on some.

 

They don't need to pull Intel out of a hat, they just really need to play their cards right. There's a lot more to the puzzle than what financial analysts predict. If AMD launches a new Bulldozer and GCN tanks then sure I have no problems saying all aboard the lifeboats (as I've stated in my previous reply). Although given what the company is projecting for next year and alternative escape routes I don't see the company going under in 2020. They could bankrupt shortly after that but between now and then I like many others don't foresee it as long as they are competitive. You can never rule out the possibility of anything when it remains a mystery.

And what did AMD spend in R&D for the console APU designs, and what of the developer support it lent to Microsoft and Sony? The actual ROI is very slim.

 

I hadn't heard of any L3 except on the HPC APU part, and I had seen only the core diagrams showing L2 cache. I just found the following showing the 4-core cluster with L3 cache. That's a relief. I was originally very disappointed in Keller not including a decent L3 cache. http://www.kitguru.net/components/cpu/anton-shilov/more-amd-zen-cpu-details-emerge-quad-core-units-inclusive-cache-high-speed-interconnects/

 

It's not the deepness of the debt, but the amount due when relative to what AMD is likely to possess in liquid cache. AMD would have to become extremely competitive, despite having a more expensive foundry than either TSMC or Intel. Intel can play the price war clear into the dirt if pushed. While Intel does have a vested interest in AMD surviving, if it gets caught being anti-competitive trying to keep it alive, it'll spell legal ruin for both companies, because even without any coordinated conspiracy, AMD would be benefiting from crime and would have to return any perceived gains deemed by a court of law.

 

And on the GPU side, Nvidia's actually gunning to take AMD down completely. Nvidia can play the price war, meaning AMD has to win in performance and perf/$ to win over enough customers, and if AMD is producing at GloFo instead of TSMC, that's going to limit their ability to do that once again for expense reasons. I'm only stating that AMD's back against a wall with a bullet already in its shoulder because it's effectively true.

 

No, most people are going to want dual-core, quad-core, and APU options. That is the bulk of consumer demand. The enthusiasts will want to play with the new 8-core, 12-core, and 16-core beasts, but that's a small and dying minority, especially since DX 12 will be bringing a boost to older hardware which will dampen sales for all 3 companies.

 

We have the cycle counts for every single instruction for Excavator released in the latest edition of the AMD x86 manual. It doesn't take much effort to build a tester program that builds a theoretical CPU in qEMU and runs open source benchmarks to collect scores, normalize, give the fitness function result, and then optimize the inputs (the cycle counts) where the constraints are the cycle counts must be integers, must be greater than or equal to 1 cycle, and the end count average must result in no greater than 40% greater IPC over the base Excavator counts. Barring any magnificent memory subsystem upgrades, the best AMD can do is Haswell. With an associative cache system and a 3rd level, I can rerun the numbers for you if you'd like. In terms of IPC, AMD can do no better than Haswell unless they exceed a 40% IPC gain. That's just mathematical fact. Now, as per clock speeds, we don't know what will come, but I doubt we'll see much more than 3.5GHz for 8 cores, which is about what AMD will be facing for the 6960X and 7960X anyway. Intel doesn't know the performance exactly, but it can tell you the scaling with clock speeds rather easily before production even begins.

 

Athlon 64 only picked some low-hanging fruit that Intel wasn't expecting AMD to pick so fast. That era is long past. It's a slog from here onward.

 

You overestimate AMD's GPU brand power which has evaporated in recent years, and without boosting the rebrands to GCN 1.2, there's no way it can match Maxwell's performance. With the given MSRPs, there's no way AMD will sell most of its cards either without an immediate price drop.

 

You do need more than 4GB to run 4K. That's no longer in dispute. Whether you should need it or not is hotly debated, and I'm on the side of "shouldn't," but that has no bearing on the reality many games use more than 4GB of VRAM when pushed even at 1440p.

 

You're looking at old benches if the 295x2 still beats the Titan X at anything.

 

You're operating under a delusion that it's still a mystery. Intel sees the writing on the wall. Do you honestly think they'd hold back in the consumer segment on Skylake if they thought AMD had the goods necessary to compete in 2016 when Intel's skin is in the game beyond simple sales?

[Techhog]

The console business in a year doesn't generate even the first 600 million AMD owes January 1st, 2019.

Also, Qualcomm or Mediatek could take over. ARM will be strong enough by that point, and Mediatek has licensed AMD's GPU tech and used it several times.

Only strong enough to match current consoles. Though, if AMD dies most of us will have to move to consoles anyway so I guess it doesn't matter how powerful they are...

Also, my implication was that those companies would bail AMD out, not that they would use money generated from the console business. This is LTT though. I'm sure you're fine with $2000 single GPU cards if it means the end of consoles.

[patrickjp93]

Only strong enough to match current consoles. Though, if AMD dies most of us will have to move to consoles anyway so I guess it doesn't matter how powerful they are...

And those companies wouldn't bail AMD out as quickly as you might think. All Nvidia has to do is offer a souped up Denver-based quad core with a bigger iGPU based on Maxwell or Pascal.

Also, my implication was that those companies would bail AMD out, not that they would use money generated from the console business. This is LTT though. I'm sure you're fine with $2000 single GPU cards if it means the end of consoles.

Consoles are the bane of good PC gaming. We

wouldn't best jaded if studios sucked less at porting games. Or Hell, maybe start on PC first and design it right the first time! That's not too much to ask honestly...

AMD dying would not bring on $2000 graphics cards. Nvidia would quickly have its hands full with an Intel armed with all of AMD's graphics IP and likely Raja Koduri and Mark Papermaster to lead the graphics team.

[mr moose]

Oh really? I'm only responding to the messages addressing me personally because I'm tired and there's 10 pages here. If I'm mistaking you and patrick speaking generally under a quote of something I've said for mis-interpreting the quote then I apologise.

 

When I initially quoted you it was because I read from your post that you were saying a monopoly is illegal, I had not seen any prior posts that indicated you didn't believe that to be the case, not that that means there were none, just that that's how I interpreted your post.

[othertomperson]

When I initially quoted you it was because I read from your post that you were saying a monopoly is illegal, I had not seen any prior posts that indicated you didn't believe that to be the case, not that that means there were none, just that that's how I interpreted your post.

 

I said it was undesirable and promotes more scrutiny into anti-trust. I didn't say it was illegal.

 

Nvidia isn't that huge a company, either, to deal with this as well as some of the huge pseudo-monopolies such as the ISPs etc. They're doing better than AMD but I can't imagine they have that much money to just throw into court cases if they did arise -- whether they were actually behaving illegally or not.

[Techhog]

Consoles are the bane of good PC gaming. We

wouldn't best jaded if studios sucked less at porting games. Or Hell, maybe start on PC first and design it right the first time! That's not too much to ask honestly...

AMD dying would not bring on $2000 graphics cards. Nvidia would quickly have its hands full with an Intel armed with all of AMD's graphics IP and likely Raja Koduri and Mark Papermaster to lead the graphics team.

I'm sure that 5 years ago, if I said that there would be a $1000 single GPU card and a $3000 dual GPU card, you would have called me crazy. Intel is not going to compete on the high-end, and the low-end will be gone. Thus, Nvidia will increase prices significantly and there won't be anything to do about it. Same goes for Intel. There's also the chance that someone will buy AMD and end the cross licensing deal, thus bringing an end to x86 and PC gaming very abruptly. Also, we need consoles. With Nvidia convincing devs to make games with features which require a $650 graphics card to get 60FPS on ultra at 1080p, PC games simply won't be accessible enough to justify high budgets.

 

But if you want to believe that AMD dying and taking consoles with them would bring sunshine and rainbows to us all, be my guest. Also considering putting you on my ignore list due to being living satire...

[bogus]

I just don't get this speech... there are only 3 main players in the graphics processor business, something that is required pretty much for every device with a user interface... yet somehow people think that there will only remain 2 players in the market IF AMD goes bankrupt...

I mean, come on... even if the dGPUs market has a low cap, it's still in the order of the billions, and they have nice margins, so... yeah if you think no one would want to dip on that and would leave it just for NVIDIA... it's a bit naive. Same goes for CPUs/Custom SoCs/APUs.

So just trust me on this, even in the worst case scenario (wich is AMD going bankrupt) we will be more then fine.

[othertomperson]

AMD dying would not bring on $2000 graphics cards. Nvidia would quickly have its hands full with an Intel armed with all of AMD's graphics IP and likely Raja Koduri and Mark Papermaster to lead the graphics team.

 

It would not, but that's because AMD are already not producing products in any meaningful way at the top end. Hopefully the Fury will end up changing this, but we also need Zen to compete with the 5960X tier product. £800 for a CPU, another £800 for a GPU. The death of AMD would serve nobody.

[patrickjp93]

It would not, but that's because AMD are already not producing products in any meaningful way at the top end. Hopefully the Fury will end up changing this, but we also need Zen to compete with the 5960X tier product. £800 for a CPU, another £800 for a GPU. The death of AMD would serve nobody.

It would serve everybody. In order for Nvidia to compete against Intel in CPUs in the short and medium terms is to go cheaper than Intel. The same is true for Intel going in graphics in the short and medium term. In the long term where prices go will all depend on how the two compete in performance.

[othertomperson]

It would serve everybody. In order for Nvidia to compete against Intel in CPUs in the short and medium terms is to go cheaper than Intel. The same is true for Intel going in graphics in the short and medium term. In the long term where prices go will all depend on how the two compete in performance.

 

It depends. I'm currently on what is essentially a re-badged cut down Xeon with no iGPU at all, a fair few people are still using FX CPUs with no GPU part. On this forum in particular the need for a dedicated GPU is so strong that I think people have legitimately no reason to care if iGPUs are now strong enough to kick out the need for a GT 740 or Nvidia Ion or something. Maybe we will reach the point where AMD are able to put a fully enabled flagship GPU on their APU core and suddenly we're in a whole new ball game but until that happens I don't blame people for not caring when Intel have had free reign over basically every decent performance CPU since Nehalem.

[patrickjp93]

It depends. I'm currently on what is essentially a re-badged cut down Xeon with no iGPU at all, a fair few people are still using FX CPUs with no GPU part. On this forum in particular the need for a dedicated GPU is so strong that I think people have legitimately no reason to care if iGPUs are now strong enough to kick out the need for a GT 740 or Nvidia Ion or something. Maybe we will reach the point where AMD are able to put a fully enabled flagship GPU on their APU core and suddenly we're in a whole new ball game but until that happens I don't blame people for not caring when Intel have had free reign over basically every decent performance CPU since Nehalem.

A lot of industry experts are saying 98% of central processor dies will be GPU by the end of the decade, so it is possible

[othertomperson]

A lot of industry experts are saying 98% of central processor dies will be GPU by the end of the decade, so it is possible

 

Eventually. Anyone with a Sandy Bridge i5 will tell you there comes a point when raw CPU power stops mattering too much. Right now, though, the CPU part of the APUs available are just too weak to want to use as pure CPUs and the GPUs are too weak to make gaming much fun, and ultimately that's what people care about.