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

[Phynaz]

Except they are not in the semiconductor business, that's for GloFo and TSMC to worry about, with billions of investments for new nodes.

20% margins for IP and SoC design business after paying for R&D that covers other products of your stack, (ie. these designs will inevitably find their way into consumer APUs) is pretty damn good when it moves in the volumes of hundreds of millions for both consoles.

Not only in terms of raw profit, which with those volumes, would be massive.. but in terms of the entire next-gen development across multiple platform would be running on AMD's ecosystem. Guess what happens when they optimize games for 8 cores and radeon GCN? Pretty sure there's obvious benefits for their core market right there down the road, for many years to come.

AMD certainly is a semiconductor company. They still sell chips, and incur all the costs associated with selling those chips.

 

[Arkaign]

Yeah good luck with a 300W console. Not to mention that's one very large piece of silicon that needs cooling.

Oh for sure. But after all, it really was just like that with 360/PS3 in some ways.

They had high/top end GPUs, and fairly stout 3+Ghz CPUs. For the time, they were really high end even in the PC world. But that made them expensive to make, large power supply, complex PCB, huge cooling requirements. Throw in lead-free solder (hello RROD/YLOD) and you had a prescription for some serious issues. Given how widespread RROD was, I'd be surprised if Microsoft actually turned a meaningful profit on the 360. I know they lost money as a whole on original Xbox.

That said, as an enthusiast, it would have been cool to see a couple of similar consoles today in terms of design (but with better solder of course, I think we're past soldergate controversies by now). Honestly a mobile i7 Haswell + 7970 1.2Ghz or GTX 780, that would have been even more high end than 360/PS3 were at their release date.

With the x86 architecture, and companies using high-end PCs to show off their games at E3, I'm kind of skeptical how well they will actually perform on release. Just as some underestimate the overhead of running Windows/etc in a PC environment, I think some OVERestimate as well (see hype train). And this time, the gap is dramatically larger between PC and console horsepower. With the extremely weak Jaguar CPU, and especially the Xbox on DDR3, I'm thinking that the Xbox One will look like a low end PC with a 7870 in it, and a PS3 will look like a high midrange PC with a 7950 in it.

What really makes the consoles attractive will be the exclusives to the platforms. Things that are also available on PC will almost certainly be better on PC, more so than ever now that x86/GCN is standard on consoles.

 

[insertcarehere]

Whatever. Jaguar is 3.5mm2 on 28nm. Its a performance monster for the mm2 and the cost. 128 bit wide fpu, avx and hsa for next to nothing. A 28nm wafer is what 1500-2000usd?. The marginal production cost of the bare jaguar core with L1 cache is less than half a dollar. Comparing haswell to this is not remotely interesting. Its like proposing a bmw 5 series for public transport in Chinas eastern provinces.

By your logic, should we all use Arm A7s from now on then? You could fit 7-8 of those cores on the same size as a jaguar core.

 

[mrmt]

Whatever. Jaguar is 3.5mm2 on 28nm. Its a performance monster for the mm2 and the cost. 128 bit wide fpu, avx and hsa for next to nothing. A 28nm wafer is what 1500-2000usd?. The marginal production cost of the bare jaguar core with L1 cache is less than half a dollar. Comparing haswell to this is not remotely interesting. Its like proposing a bmw 5 series for public transport in Chinas eastern provinces.

Jaguar will be toast once Celeron Haswell and Silvermont arrives. They won't be able to compete on performance with the former and on costs with the latter. It will be the king of a very small niche. But that's to be expected, AMD is becoming a niche company.

 

[SiliconWars]

Jaguar will be toast once Celeron Haswell and Silvermont arrives. They won't be able to compete on performance with the former and on costs with the latter. It will be the king of a very small niche. But that's to be expected, AMD is becoming a niche company.

I'm going to go out on a limb and say that you probably said the same thing about Bobcat (before it became AMD's best-ever selling chip).

Do I win anything?

Oh btw, Even if AMD never sold another notebook Jaguar it will have been a massive success. It will in fact be the highest selling x86 core in history.

 

[mrmt]

I'm going to go out on a limb and say that you probably said the same thing about Bobcat.

Do I win anything?

Nope, I didn't. 45 and 32nm Atom weren't good enough to push Brazos from the bottom, but SNB and IVB did indeed put pressure on Brazos. AMD was losing Brazos shipments on mobile and had to redirect these chips to desktops because of that. I don't need to tell you how awful are the prices AMD must have been getting for those chips in the desktop market.

Q3 and Q4 game is much harder. Not only Core is much more optimized for mobile, Silvermont is a much better solution than 45nm and 32nm Atoms.

Oh btw, Even if AMD never sold another notebook Jaguar it will have been a massive success. It will in fact be the highest selling x86 core in history.

Maybe you should check your numbers again. Intel ships more per quarter than AMD ships for the entire year. Jaguar is *far* from being the x86 sales champion. That honor probably goes to Sandy or Ivy.

 

[SiliconWars]

Maybe you should check your numbers again. Intel ships more per quarter than AMD ships for the entire year. Jaguar is *far* from being the x86 sales champion. That honor probably goes to Sandy or Ivy.

Yep and Jaguar will sell at least 200 million in consoles. I did say highest selling core, so yeah I cheated.

Even on a SoC basis it could very well become the highest selling x86 chip ever. That's counting IVY/SB from top to bottom, from the extreme editions to the lowest Celerons.

 

[spat55]

Don't fool yourself. The only real reason is performance per dollar.

This plus add in a APU with the GDDR5 8GB in the PS4, it will make a lot of bang for how much it costs, and should run much cooler than anything else.

 

[mrmt]

Yep and Jaguar will sell at least 200 million in consoles. I did say highest selling core, so yeah I cheated.

Even on a SoC basis it could very well become the highest selling x86 chip ever. That's counting IVY/SB from top to bottom, from the extreme editions to the lowest Celerons.

Still doesn't make up for what you are saying here. Intel chips over 350MM of chips per year, it's not hard to fathom more than 400 MM of IVB or SNB systems out there. Put 3 as the core average and yes, you have far more Sandy or Ivy Cores than AMD will ship on its wildest dream, unless you are believing that China will open itself to consoles and that they will be a success there.

 

[SiliconWars]

I do believe China will open up, but even at 200 million consoles Jaguar will sell far, far many more cores than any Intel core will. 8x200 million - 1.6 billion, the vast majority of Intel chips sold are dual cores or dual+HT.

It's a pretty silly argument but not quite as absurd as your claim of it being "king of a very small niche". You didn't think that one through very well.

 

[mrmt]

It's a pretty silly argument but not quite as absurd as your claim of it being "king of a very small niche". You didn't think that one through very well.

Didn't I or didn't you? Even if the consoles sell 400 million of units combined in a 8 years life span, and I think I'm being fairly generous here, Intel will be selling over 3 billion notebook/desktop systems over the same period, mobile systems will be in excess of 6 billion, so yes, consoles *are* a niche. Unless you consider 400 million lagging edge node units something relevant when compared to a 3-4 billion bleeding edge units.

 

[SiliconWars]

Didn't I or didn't you? Even if the consoles sell 400 million of units combined in a 8 years life span, and I think I'm being fairly generous here, Intel will be selling over 3 billion notebook/desktop systems over the same period, mobile systems will be in excess of 6 billion, so yes, consoles *are* a niche. Unless you consider 400 million lagging edge node units something relevant when compared to a 3-4 billion bleeding edge units.

Let me add the bit you missed then we can see how it's important.

... Unless you consider 400 million lagging edge node units that has already been paid for something relevant when compared to a 3-4 billion bleeding edge units

While Intel is shooting their money cannon at R&D and new fabs over the coming 8 years, AMD just forgets Jaguar ever existed as a steady stream of income just keeps rolling in and paying WSA from some ancient, backward (cheap) node somewhere in Dresden or Taiwan, and your dream of AMD dying becomes even further from reality.

 

[mrmt]

While Intel is shooting their money cannon at R&D and new fabs over the coming 8 years, AMD just forgets Jaguar even existed as a steady stream of income just keeps rolling in and paying WSA from some ancient, backward (cheap) node somewhere in Dresden or Taiwan, and your dream of AMD dying becomes even further from reality.

I don't disagree with that, but you seem to overestimate the size of this stream of money. PS3 didn't make IBM or Nvidia rich, and XB360 didn't save AMD GPU division bacon, as they have been limping around break even point since I can remember.

Embedded business is naturally low margin, and it has to be. No company would offer such a long contract if they weren't sure that costs would be at the very bottom of the possibilities. And given that both companies lost a lot of money with their last consoles, you can be sure that cost will be heavily watched on this console generation.

This may change a bit if China opens itself up to consoles, but you can be sure that MSFT and Sony will get the lion's share here. In the end, what we have is a nice business, steady stream of money but guaranteed to be low margin, just like all other embedded business.

I don't dream of AMD dying, but you seem to dream that AMD will come back to the competitive days again, and sorry, you'll be disappointed. The company as you like has already died. Just look at their roadmap three years ago and look at the roadmap now. Look at the size and the focus of their engineering team. No bleeding edge company worth its stock price creates an embedded business division and puts it as the mainstay of the company. The fact that AMD did says a lot about the future of the company.

 

[SiliconWars]

I don't dream of AMD dying, but you seem to dream that AMD will come back to the competitive days again, and sorry, you'll be disappointed. The company as you like has already died. Just look at their roadmap three years ago and look at the roadmap now. Look at the size and the focus of their engineering team. No bleeding edge company worth its stock price creates an embedded business division and puts it as the mainstay of the company. The fact that AMD did says a lot about the future of the company.

You can say the same thing about Intel. Intel is trying to become a smartphone company and is happy to abandon its desktop roots in order to make it a reality. That's why Atom is the #1 chip for them now and we won't see 14nm desktop parts for another 18 months at least.

AMD on the other hand is changing into a gaming company, which is something I can handle quite easily.

 

[sontin]

While Intel is shooting their money cannon at R&D and new fabs over the coming 8 years, AMD just forgets Jaguar ever existed as a steady stream of income just keeps rolling in and paying WSA from some ancient, backward (cheap) node somewhere in Dresden or Taiwan, and your dream of AMD dying becomes even further from reality.

And losing every other market they are in. :thumbsup:

 

[mrmt]

You can say the same thing about Intel. Intel is trying to become a smartphone company and is happy to abandon its desktop roots in order to make it a reality. That's why Atom is the #1 chip for them now and we won't see 14nm desktop parts for another 18 months at least.

AMD on the other hand is changing into a gaming company, which is something I can handle quite easily.

Intel is making a planned transition, it's not that their house was on fire like AMD's.

I'm glad you'll still be able to cheerlead for AMD once their consumer line goes bust.

 

[SiliconWars]

And losing every other market they are in. :thumbsup:

It'll be fun comparing PS4 and Xbox revenues to Shield, that's for sure.

Intel is making a planned transition, it's not that their house was on fire like AMD's.

I'm glad you'll still be able to cheerlead for AMD once their consumer line goes bust.

Intel's house is on fire and when it finally hits the fan it's going to make AMD's problems look tiny in comparison.

 

[sontin]

It'll be fun comparing PS4 and Xbox revenues to Shield, that's for sure.

Yeah, because it's the only device with Tegra 4.

 

[Arkaign]

It'll be fun comparing PS4 and Xbox revenues to Shield, that's for sure.



Intel's house is on fire and when it finally hits the fan it's going to make AMD's problems look tiny in comparison.

LOL

http://newsroom.intel.com/community...up-24-percent-239-in-annual-eps-up-19-percent

http://www.inquisitr.com/486126/intel-profits-11-billion-in-2012/

Yeah, if that's a house on fire, light me up. They are really looking good in process transitions to mobile devices. By holiday 2014 timeframe, you'll see more and more mobile devices with Intel inside. Having their R&D and fabs under one 'roof' so to speak along with enormous resources gives them capabilities and efficiencies that many companies only wish they could achieve. Using outside fabs, you have to cut them in, along with a longer delay to market and all kinds of headaches. This is why AMD post-ATI buyout is so laggy and inefficient in comparison to what they were when they were truly the captain of their own ship.

The most you could say is that with components becoming cheaper over time, and the PC market contracting, that Intel's massive profits will become less massive. But that isn't exactly a condemnation of Intel more than it is a reflection of the direction of semiconductor dynamics as an economic reality.

 

[Exophase]

In any case, the second reason an OOE processor doesn't change the need to reserve a core is for the security reasons listed above. Only by completely blocking user threads from executing on the same core as the OS can you guarantee that those user threads aren't trying to read the OS's memory pages (there's supposed to be process isolation, but of course intentional glitches can be used to get past that). This is no different for Jaguar than it was for Cell. You're still going to need to dedicate a core to the OS if you want to sandbox the user environment for security purposes.

I can't think of a reason why a hypervisor in any sane design needs to run on a different core from the guest code. If anything that's more of a pain since you have to add communication between the cores.

I don't know what PS3's setup was, but it's possible that their security worked by being able to prohibit IO or some other privileged access on a per-core basis, and that the hyperviser core then acted as a restricted abstraction layer to accessing those privileged resources. Normal virtualization support like what's available on Jaguar wouldn't need anything like this, any core can safely elevate to hypervisor privilege and the hypervisor should be capable of running on all of them.

 

[krumme]

Can you move this war talk to another thread?

 

[mrmt]

Intel's house is on fire and when it finally hits the fan it's going to make AMD's problems look tiny in comparison.

Why is Intel house on fire? Because AMD's is on fire? AMD and Intel aren't in the same league. AMD could have made it if they had played the right cards circa 2003, but that ship has sailed and they are what they are now, and that means that they cannot compete with Intel in an equal footing.

 

[beginner99]

Intel's house is on fire and when it finally hits the fan it's going to make AMD's problems look tiny in comparison.

I might be wrong but once silvermont hits first jaguar will become obsolete (Baytrail-T) and then Qualcoms and Co. house will be on fire (Baytrail-M). Just look at the Razr I phone that beat Razr M in perfromance and battery life using old Atom uArch on an older process than ARM chip in Razr M.

You can laugh at me later if I'm wrong but IMHO the only way intel can fail is with pricing. silvermont else will smash ARM in perfromance/watt.

 

[NTMBK]

Jaguar will be toast once Celeron Haswell and Silvermont arrives. They won't be able to compete on performance with the former and on costs with the latter. It will be the king of a very small niche. But that's to be expected, AMD is becoming a niche company.

We've still not seen pricing on the Silvermont chips, which is the important thing- but yeah, I expect them to be the competition for Jaguar. Haswell? Meh.

Bobcat is still in an absolute ton of laptops, or at least it is around here. If you stroll into a PC store, the bulk of the affordable laptops/craptops are running Bobcat chips, then transitioning up to i3s as you get into more mid-range pricing. I expect Jaguar will do just as well.

As for IB putting pressure on Bobcat- gee, who'd have thought that the 40nm chip would have some trouble with the 22nm chip? Bobcat sold absolute gangbusters for AMD.

 

[Arkaign]

Apologies Shintai, my comment is directed at the post above yours, lol.