The real reasons Microsoft and Sony chose AMD for consoles [F]

[Shaydza]

I would love for AMD to design something similar to the console architecture for the pc.

A fully integrated motherboard with memory, decent 7790 level graphics and a chunk of gddr5 along with a half decent multi core processor in one package.

The motherboard manufacturers can then use the design and put their personal tweaks into it.

In this way they could reduce their sku count and give the manufacturers more flexibility in delivering gaming form factors. This should also lead to reduced cost.

I have been in the PC game for almost 2 decades and frankly the entire industry could benefit from this.

This would allow games and application designers a more uniform platform to code for and allow AMD to stretch out their technology refreshes. This would also improve their overall profitability per technology cycle.

Simply put, we need more integrated and standardised hardware in the market. While there will always be a market for custom builds, I believe there would be a huge market for a good performance pc out of the box, that is attractive and with no compatibility issues.

 

[monstercameron]

I would love for AMD to design something similar to the console architecture for the pc.

A fully integrated motherboard with memory, decent 7790 level graphics and a chunk of gddr5 along with a half decent multi core processor in one package.

The motherboard manufacturers can then use the design and put their personal tweaks into it.

In this way they could reduce their sku count and give the manufacturers more flexibility in delivering gaming form factors. This should also lead to reduced cost.

I have been in the PC game for almost 2 decades and frankly the entire industry could benefit from this.

This would allow games and application designers a more uniform platform to code for and allow AMD to stretch out their technology refreshes. This would also improve their overall profitability per technology cycle.

Simply put, we need more integrated and standardised hardware in the market. While there will always be a market for custom builds, I believe there would be a huge market for a good performance pc out of the box, that is attractive and with no compatibility issues.

probablt this to compete with iris pro.

 

[SiliconWars]

Regarding the WSA argument, is it already proved that the console SoC's manufacturing is going to be rolled into GF and not into TSMC? Not wanna sound pessimistic but is GF really able to take in that demand of 28mn products at this point? (I asume that the real silicon is already being manufactured, if they plan to launch at the end of 2013)

Yeah I've been thinking about that myself and it seems very unlikely that GF's 28nm is ready for such a large SoC.

I woulda tought that, since AMD is going the lego-style silicon design philosophy with jaguar to be the most fab-agnostic as you can be so you can easily switch between manufacturers when needed, TSMC was taking the order for the very first wave of console's SoCs and eventually, to alleviate the WSA burden, GF would take the baton at some point and AMD would settle with them till this next gen becomes EOL.

And that was exactly what my thinking was also. I'd say you are spot-on.

 

[galego]

AMD has been providing solutions for consoles for years. Don't know why this is a surprise.

It is not a surprise for me, because AMD has the technology and experience that none of the competitors had (if really they had real competitors).

What is a surprise is that some people ignores what Sony and Microsoft are saying, and pretend that AMD won in the consoles because Intel and Nvidia were not interested.

 

[Rvenger]

It is not a surprise for me, because AMD has the technology and experience that none of the competitors had (if really they had real competitors).

What is a surprise is that some people ignores what Sony and Microsoft are saying, and pretend that AMD won in the consoles because Intel and Nvidia were not interested.

The real reason is.. they have ATI and they have x86. They can put it all together on one chip and charge one low cost. I can bet you that Intel wasn't even interested. I know that Nvidia was interested but x86 knocked them out of the market. Nvidia can't design an APU with x86 instructions unless they sub it out to Intel, AMD or VIA. Thus there would not be enough profit margin for them to bother traveling that road.

 

[pelov]

O, I agree. I think so far, Read has done a fine job, but long term AMD seems to be sending mixed signals (x84 vs ARM, are they competing at the high-end or not, more custom orders but no real indication of where they're looking to get them). Personally, I think Read already has an idea of where he wants AMD to go, but doesn't quite know how to get there without getting stock holders upset. I could totally be wrong about that but we'll see. I do have a lot more hope with Read at the helm than the last 3 CEO's (even if one was an interim and Dirk really didn't all that bad of a job all things considered).

The market is considerably different now than it was just 2-3 years ago, so AMD's 'mixed signals' should be seen not as indecision, but rather spreading themselves out in such a manner to not get blindsided again. Bear in mind that AMD and Intel both were knocked flat on their butts by the mobile boom and were completely unprepared for it; and in some respects they still aren't, but getting familiar with ARM and making the jump into dense ARM clusters for microservers while offering IP isn't a sign of being undecided, but rather 'let's make sure we're not blindsided yet again.' They're certainly in a better position now then when Read took over, despite the stock being far lower in value. Getting Keller back was a crucial step in that diversification

I too think he's done a fantastic job, but their long term growth has just as much, if not more to do with outside factors out of their control - for instance, whether MS can pick it up and make Windows, and in turn x86, something desirable again. Both Intel (Bay Trail) and AMD (Jaguar) have great chips to power future Windows tablets, but Microsoft hasn't provided the OS and structure around it (apps, maturity, price, products) for that to happen.

 

[sontin]

The real reason is.. they have ATI and they have x86. They can put it all together on one chip and charge one low cost. I can bet you that Intel wasn't even interested. I know that Nvidia was interested but x86 knocked them out of the market. Nvidia can't design an APU with x86 instructions unless they sub it out to Intel, AMD or VIA. Thus there would not be enough profit margin for them to bother traveling that road.

x86 has nothing to do with the designs. It's only relevant because of Windows. And the PS4 is not even using a Windows OS...

nVidia would have build a own CPU. And that was the problem. There would be no market outside of the consoles. So they needed huge gross margins for every chip to make it even worth the trouble.

I too think he's done a fantastic job, but their long term growth has just as much, if not more to do with outside factors out of their control - for instance, whether MS can pick it up and make Windows, and in turn x86, something desirable again. Both Intel (Bay Trail) and AMD (Jaguar) have great chips to power future Windows tablets, but Microsoft hasn't provided the OS and structure around it (apps, maturity, price, products) for that to happen.

Jaguar is not a Tablet CPU. Perf/Watt is so worse that nobody consider the SoCs for tablets.

 

[SiliconWars]

nVidia would have build a own CPU. And that was the problem. There would be no market outside of the consoles.

Yeah that cpu would have been Tegra 4, as you keep telling us it's easily better than Jaguar in perf/W so no reason why that wasn't used by your logic.

On top of that you have Nvidia being absolutely desperate to make Android the new gaming OS. Nvidia would have done everything they could to make this a reality with the new consoles as that would have given Tegra massive traction - the kind of traction they desperately need to save their failing business. There is no way they would put something like margins ahead of that. Margins vs worsening hundred-million dollar losses for Tegra? Yeah that's not a tough one.

You got one thing right though, there is no market for it in or outside of consoles.

 

[pelov]

The decision to go with x86 had to do with developer familiarity and time-to-market. ARM wouldn't have been a poor choice in 1-2 years, but the lack of a true 64-bit architecture right now is what threw them out of that race. nVidia would only have offered an ARM solution, and almost certainly a bog standard ARM core like an A15, and that just isn't going to cut it.

The lead architect of the PS4 went on record stating their reasons for opting to go with x86, and those were:

- developer familiarity
- decreased time-to-market for games
- lower cost of development

This same video didn't go in depth regarding why AMD's Jaguar was picked, but he did say something like "in the end there was only one clear choice." A couple of other things mentioned were:

- developers wanted lots of fast RAM, and preferably GDDR5
- larger portion of the transistor budget spent on the GPU
- 4-8 CPU threads

The video was really long and I found him to be very annoying to listen to, going on at length about stuff nobody cares about, but there were some informative and interesting tidbits in there. I'll try to find the link and post it here

Jaguar is not a Tablet CPU. Perf/Watt is so worse that nobody consider the SoCs for tablets.

w0t?

Jaguar is a great, cheap little chip. It hits that 5-10W threshold, provides great GPU performance, and it doesn't cost a fortune. In contrast, Intel still has to rely on a poor PowerVR graphics solution that has real difficulty driving 1080p displays (or even playing high quality video); and you can forget about gaming on it unless you stick strictly to Metro. Even still, the lack of Win8 tablets has more to do with the incredibly poor Win8 tablet sales than the underlying hardware.

The problem for Intel and AMD in the mobile space is: why would an OEM choose x86 over ARM? The x86 instruction set carries a lot of weight in enterprise and PCs, but it offers no such obvious advantages in tablets/smartphones.

 

[monstercameron]

The decision to go with x86 had to do with developer familiarity and time-to-market. ARM wouldn't have been a poor choice in 1-2 years, but the lack of a true 64-bit architecture right now is what threw them out of that race. nVidia would only have offered an ARM solution, and almost certainly a bog standard ARM core like an A15, and that just isn't going to cut it.

The lead architect of the PS4 went on record stating their reasons for opting to go with x86, and those were:

- developer familiarity
- decreased time-to-market for games
- lower cost of development

This same video didn't go in depth regarding why AMD's Jaguar was picked, but he did say something like "in the end there was only one clear choice." A couple of other things mentioned were:

- developers wanted lots of fast RAM, and preferably GDDR5
- larger portion of the transistor budget spent on the GPU
- 4-8 CPU threads

The video was really long and I found him to be very annoying to listen to, going on at length about stuff nobody cares about, but there were some informative and interesting tidbits in there. I'll try to find the link and post it here



w0t?

Jaguar is a great, cheap little chip. It hits that 5-10W threshold, provides great GPU performance, and it doesn't cost a fortune. In contrast, Intel still has to rely on a poor PowerVR graphics solution that has real difficulty driving 1080p displays (or even playing high quality video); and you can forget about gaming on it unless you stick strictly to Metro. Even still, the lack of Win8 tablets has more to do with the incredibly poor Win8 tablet sales than the underlying hardware.

The problem for Intel and AMD in the mobile space is: why would an OEM choose x86 over ARM? The x86 instruction set carries a lot of weight in enterprise and PCs, but it offers no such obvious advantages in tablets/smartphones.

not entirely accurate, Just remember that the ipad 3/4 uses this IP[sgx 543/544 mp4] and they drive +hd resolution ipad display. Also powervr rogue series should have started sampling[maybe not the right word] among all these chip designers. Powervr isn’t the problem per se, just lintel's implementation [sgx 540 in CT or sgx 544 mp2 in the CT+]

 

[pelov]

not entirely accurate, Just remember that the ipad 3/4 uses this IP[sgx 543/544 mp4] and they drive +hd resolution ipad display. Also powervr rogue series should have started sampling[maybe not the right word] among all these chip designers. Powervr isn’t the problem per se, just lintel's implementation [sgx 540 in CT or sgx 544 mp2 in the CT+]

Oh yea, and they still have the same 3rd-party driver woes that we experienced with the older Atoms from Ye Olde Days of Netbooks.

The hardware isn't bad, although perhaps lacking a bit of graphical power in Clover Trail, it's still sufficient for most use cases. The issue I'm pointing out here is that, regardless of what hardware AMD and Intel have under the hood, a Windows 8 tablet will still struggle with respect to sales. Windows 8 is still too bloated for tablets, it's still too resource intensive, and it hasn't struck a chord with consumers. It takes a lot more horsepower (wattage and price) to get an equivalent experience on a Win8 device than it does on iOS or Android.

 

[sontin]

The decision to go with x86 had to do with developer familiarity and time-to-market. ARM wouldn't have been a poor choice in 1-2 years, but the lack of a true 64-bit architecture right now is what threw them out of that race. nVidia would only have offered an ARM solution, and almost certainly a bog standard ARM core like an A15, and that just isn't going to cut it.

We don't know what nVidia could have done.
They developed a brand new chipset for the Xbox with the best onboard soundchip ever.

The lead architect of the PS4 went on record stating their reasons for opting to go with x86, and those were:

- developer familiarity
- decreased time-to-market for games
- lower cost of development

This same video didn't go in depth regarding why AMD's Jaguar was picked, but he did say something like "in the end there was only one clear choice." A couple of other things mentioned were:

- developers wanted lots of fast RAM, and preferably GDDR5
- larger portion of the transistor budget spent on the GPU
- 4-8 CPU threads
The video was really long and I found him to be very annoying to listen to, going on at length about stuff nobody cares about, but there were some informative and interesting tidbits in there. I'll try to find the link and post it here

So we should take this serious? He is responsible for the hardware. So he will defend it. Maybe we should ask Android and iOS game developers if they need x86.

If x86 would be so important why is it losing it's relevance?

Jaguar is a great, cheap little chip. It hits that 5-10W threshold, provides great GPU performance, and it doesn't cost a fortune. In contrast, Intel still has to rely on a poor PowerVR graphics solution that has real difficulty driving 1080p displays (or even playing high quality video); and you can forget about gaming on it unless you stick strictly to Metro. Even still, the lack of Win8 tablets has more to do with the incredibly poor Win8 tablet sales than the underlying hardware.

Why are you comparing it to Intel? Jaguar is sold in a market which it's not anymore isolated to Windows. The Toshiba Excite with an 1,8GHz Tegra 4 is faster in 3DMark than Temash. The CPU part is even more than 2x faster.

The problem for Intel and AMD in the mobile space is: why would an OEM choose x86 over ARM? The x86 instruction set carries a lot of weight in enterprise and PCs, but it offers no such obvious advantages in tablets/smartphones.

The future is not about instruction sets, it's about CPU architectures. If you can't compete on price, features and perf/watt than you will die. AMD is learning it with Temash and Kabini right now. We are not living in 2011 where AMD had a huge advantage with Brazos over Intel. Now they are facing Qualcomm, Apple, Samsung, nVidia and other vendors for market share.
And people can't even blame Intel for it. OEMs have now a choice which CPU they want to use. And right now there is not one announcement of a real tablet design with Temash...

 

[jpiniero]

Long term vision is sadly lacking in too many tech companies, even with it being such a fast changing industry. AMD are still in a weak position but they are in a position to grow now, and that is a much better position to be in than what they were previously - trying to hold on to what they had while slowly dying to a 1000 small cuts.

Grow where? They have no relevant tablet or smartphone chip. Servers are more or less dead. As PC sales continue to slide, I am not seeing where AMD plans to turn to, unless it's simply the consoles.

I see AMD spinning off what they can to settle the $2B debt, laying everyone off and just becoming a holding company for the console IP until the console lifecycle ends.

I would love for AMD to design something similar to the console architecture for the pc.

A fully integrated motherboard with memory, decent 7790 level graphics and a chunk of gddr5 along with a half decent multi core processor in one package.

GDDR5 is going to be too expensive.

 

[galego]

x86 has nothing to do with the designs. It's only relevant because of Windows. And the PS4 is not even using a Windows OS...

There is nothing relevant between x86 and Windows.

nVidia would have build a own CPU. And that was the problem. There would be no market outside of the consoles. So they needed huge gross margins for every chip to make it even worth the trouble.

But only after ignoring that Sony and Microsoft received ARM prototype/model and did not pass the benchmark tests, because lacked performance.

Jaguar is not a Tablet CPU.

Agree, it is more than that.

 

[pelov]

I don't see why you're disagreeing with points I never made...

The lead PS4 architect explained why they chose x86, and why developers favored x86. It wasn't because of some magical fairy dust that x86 has, but because of the familiarity and ease of porting between consoles and PCs**. Tablet and smartphone games use different engines and hardware, so adopting ARM was a distant second. Furthermore, ARM's lack of true 64-bit made it a nonstarter.

I was comparing Intel and AMD solely within the Win8 landscape. My additional point was that, regardless of what hardware AMD or Intel came come up with, they're going to be limited by the OS, software, and price that are attributed to Win8 and MS's approach in the tablet space.

The point of OEMs not choosing AMD's Jaguar is a moot one, I feel. This has more to do with Win8's incredibly poor sales in the tablet space as a whole. Take a look at Win RT to see what I mean. OEMs are using more svelte ARM cores yet the products still won't sell, often times getting the chop even before they're released (Samsung, Asus, Acer, HP, Dell and Nokia have all cancelled Win RT products). My point here is that even if Intel and AMD were to make small, cheap x86 cores that were very power efficient (for the record, I don't think this will ever happen), they still wouldn't sell well.

** He went on at length regarding the Cell processor in the PS3. He stated that, although it offered a significant amount theoretical performance, achieving anywhere near the GFLOPs threshold required a massive amount of time, experience, and effort on behalf of the developer; so much so, that even if a hypothetical Power-based chip were to be developed on the Cell architecture, a slower x86 chip would actually be faster in implementation.

 

[sontin]

@galego

You are right: x86 is so great that every market outside of windows is dumping it right now. :lol:

BTW: Sony and Microsoft never received Samples from nVidia or Intel. What they saw would be powerpoint slides of roadmaps, performance targets and prices.

And just no: No sane company would chose Jaguar if CPU performance was a priority.

 

[SiliconWars]

Grow where? They have no relevant tablet or smartphone chip.

A bit like Intel and Nvidia then.

Servers are more or less dead.

Microservers are very much alive and kicking and AMD has a good opportunity to piggyback on ARM's entry here.

I see AMD spinning off what they can to settle the $2B debt, laying everyone off and just becoming a holding company for the console IP until the console lifecycle ends.

Wow, been a while since I read this kind of doom prediction about AMD, or maybe they just get lost and ignored in amongst the constant stock upgrades since the start of the year.

 

[SiliconWars]

BTW: Sony and Microsoft never received Samples from nVidia or Intel. What they saw would be powerpoint slides of roadmaps, performance targets and prices.

The reaction to Nvidia's bid was captured on camera too.

 

[sontin]

I don't see why you're disagreeing with points I never made...

The lead PS4 architect explained why they chose x86, and why developers favored x86. It wasn't because of some magical fairy dust that x86 has, but because of the familiarity and ease of porting between consoles and PCs. Tablet and smartphone games use different engines and hardware, so adopting ARM was a distant second. Furthermore, ARM's lack of true 64-bit made it a nonstarter.

Next year nVidia will bring Kepler to Android. You can even build an ARM PC with Titan today. The Unreal Engine 3 and 4 have support for different plattforms. That is not argument for x86. It would be sad if console developers can't port their games to ARM, because they will lost a huge market in a few years.

The reaction to Nvidia's bid was captured on camera too.

Na, that was the reaction to the CPU performance. :awe:

 

[pelov]

BTW: Sony and Microsoft never received Samples from nVidia or Intel. What they saw would be powerpoint slides of roadmaps, performance targets and prices.

It was extremely unlikely that MS would have picked nVidia again after their quarrel regarding pricing with the original Xbox and the NV2A. MS was really unhappy with the way that went down and vowed not to use nVidia for their consoles in the future.

 

[pelov]

Next year nVidia will bring Kepler to Android.

I'm talking about the CPU here.

You can even build an ARM PC with Titan today.

What?

The Unreal Engine 3 and 4 have support for different plattforms.

It's a lot more complicated than 'here are two engines to drive games for multiple platforms.' Many big game studios have their own engines/platforms and would never license from somebody else. Just because an alternative is available doesn't make it a viable alternative

That is not argument for x86. It would be sad if console developers can't port their games to ARM, because they will lost a huge market in a few years.

I don't think you understand what 64-bit means.

I agree with you that game developers have to focus on ARM and iOS/Android, but we're talking about two different segments of the market. Most apps downloaded on tablets and smartphones are games, and there's a huge market for that, but that market doesn't consist of $59.99 titles requiring 6GB of GDDR5.

 

[Sweepr]

A bit like Intel and Nvidia then.

NVIDIA is inside my Nexus 7. I bet this tablet model sold more than all AMD-based tablet offerings combined. Sounds relevant to me. I woudn't call Intel irrelevant until we see how Bay Trail plays out.

 

[sontin]

I'm talking about the CPU here.

And the CPU part is not the important piece. You only need to compile the part of the engine/game to ARM. It can't be so hard when every physics libary has support for different architectures (PhysX even for a GPU...)

It's a lot more complicated than 'here are two engines to drive games for multiple platforms.' Many big game studios have their own engines/platforms and would never license from somebody else. Just because an alternative is available doesn't make it a viable alternative

Ubisoft will have an engine for every plattform in the future:

"I expect maybe in three or four years from now to be able to have mostly the same engine running on tablet and the main console," he said. "On The Division, we are using the same assets on the console and the tablets. In the future, it will be easier for us to provide this kind of experience."

http://www.oxm.co.uk/57498/ubisoft-...engine-on-tablets-and-console-in-three-years/

Epic supports ARM with the UE3+. SoCs will support OpenCL and full OpenGL4.x next year. You will see the same games on different plattforms: PC (independent Steam games), Android, Xbox/PS4.

The remake of Gina Sister will be on Android in a few weeks. And this is a port from the PC version. You know x86...

 

[pelov]

I think what you're getting at is that a developer should have the same game running across all platforms?

If you're implying that the same title that sells for $60 on the PC/Xbox/PS4 and requires 4-8GB of DDR3/GDDR5 to run at 60FPS at 1080p is going to somehow run on the same engine and lower-tier hardware on a tablet then I don't know what to tell you... I think the fact that you're ignoring that whatever nVidia could come up with on the CPU side would have been limited to 4GB-LPDDR and would somehow be a good choice for a console speaks volumes, frankly.

The point of the next round of consoles was to make it easier to develop for, not to make it a living nightmare.

 

[sontin]

I think what you're getting at is that a developer should have the same game running across all platforms?

A lot of games will run on all plattforms.

If you're implying that the same title that sells for $60 on the PC/Xbox/PS4 and requires 4-8GB of DDR3/GDDR5 to run at 60FPS at 1080p is going to somehow run on the same engine and lower-tier hardware on a tablet then I don't know what to tell you...

In three years since 2010 nVidia increased the GPU performance 20x over Tegra 2. In three years from now we could have 1 - 2 TFLOPs SoCs.
With 64bit ARM CPUs you can even use the same amount of memory.
The ARM world is not standing still like consoles.

I think the fact that you're ignoring that whatever nVidia could come up with on the CPU side would have been limited to 4GB-LPDDR and would somehow be a good choice for a console speaks volumes, frankly.

Huh? nVidia could have build their own 64bit low latency core (CPU). They doing it with the ARM instruction set.