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

[krumme]

HSA is a feature Jaguar dont support. So that would be amazing.

Even Kaveri is not fully HSA.

You are probably more wrong than right read interview here with ps4 chief architect
http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php

" The 'supercharged' part, a lot of that comes from the use of the single unified pool of high-speed memory," said Cerny. The PS4 packs 8GB of GDDR5 RAM that's easily and fully addressable by both the CPU and GPU.If you look at a PC, said Cerny, "if it had 8 gigabytes of memory on it, the CPU or GPU could only share about 1 percent of that memory on any given frame. That's simply a limit imposed by the speed of the PCIe. So, yes, there is substantial benefit to having a unified architecture on PS4"

" In fact, said Cerny, when he toured development studios asking what they wanted from the PlayStation 4, the "largest piece of feedback that we got is they wanted unified memory."

 

[podspi]

Is HSA a core feature or an SoC feature? I could see HSA being available for the PS4/Xbone if it is part of the SoC, which is obviously different from other Jaguar variants released onto the market...

 

[ShintaiDK]

You are probably more wrong than right read interview here with ps4 chief architect
http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php

" The 'supercharged' part, a lot of that comes from the use of the single unified pool of high-speed memory," said Cerny. The PS4 packs 8GB of GDDR5 RAM that's easily and fully addressable by both the CPU and GPU.If you look at a PC, said Cerny, "if it had 8 gigabytes of memory on it, the CPU or GPU could only share about 1 percent of that memory on any given frame. That's simply a limit imposed by the speed of the PCIe. So, yes, there is substantial benefit to having a unified architecture on PS4"

" In fact, said Cerny, when he toured development studios asking what they wanted from the PlayStation 4, the "largest piece of feedback that we got is they wanted unified memory."

There is nothing in that statement that talks about HSA compute. Just sharing memory. (Something we had since the Amiga days, but the PC lacked.)

 

[krumme]

There is nothing in that statement that talks about HSA compute. Just sharing memory. (Something we had since the Amiga days, but the PC lacked.)

Read the links. Here is an interpretation:
http://www.extremetech.com/gaming/1...ily-modified-radeon-supercharged-apu-design/2

 

[ShintaiDK]

Read the links. Here is an interpretation:
http://www.extremetech.com/gaming/1...ily-modified-radeon-supercharged-apu-design/2

I did, nothing there. And AMD doesnt list Jaguar as HSA either.

 

[krumme]

I did, nothing there. And AMD doesnt list Jaguar as HSA either.

Yes Amd does not do that, and we knew that before.

What is the pupose of hsa and to what degree does ps4 get there?

" Remember, the point of HSA is to allow CPU and GPU to share a common set of pointers and swap data more efficiently.*"

We dont know exactly how its implemented in ps4 but is it totally out of the question - like no apu in ps4 - that you can get the same benefits with the arch decribed by the chief architect?

I mean what should prevent it?

 

[krumme]

Looking at the Interview with Cerny it looks like some of the "real reason" (btw i hate that expression) Amd won the deal is pretty much the requirement for unified memory access - the effecs of greatly simplifying the programming - its lower cost, perhaps a myriad of indie games with good looking ?, and ofcourse the performance benefit of choosing ddr5.

Ps4 is not only better than ps3 because of the initial hardware cost, but because we will see games from day one using the hardware reasonable. It took years for the ps3. The cell days are over. I think this new way of thinking platform is great for us consumers and gamers. The old way was a waste of ressources also leading to lesser gaming investment in the end.

 

[galego]

The consoles don't support HSA.

No.

http://av.watch.impress.co.jp/docs/series/rt/20130325_593036.html

Even Kaveri is not fully HSA.

Kaveri is fully HSA enabled.

 

[blackened23]

Let me post some more thoughts on this. I think intel *had* the capability to produce a SOC for a console, but for whatever reason they didn't - we can speculate on that forever, but the bottom line is we'll never know why.

As a preface, I will say that i've been extremely disappointed by AMD's CPUs for the past several years - I would love for them to return to the Athlon x64 glory days, but their recent releases have left me little hope for that. With that said, one thing that kind of bugs me is that people always discredit AMD to extreme lengths. I think the SOC used in the PS4 looks *very* nice, it is certainly a huge upgrade over the PS3. And then you have those "other guys" who always give AMD way too much credit by saying that Haswell is a failure and what not, and how great the FX chips are. I know you've all seen it. Seeing THOSE posts are annoying to no end, although I won't go into detail on that. You all know what i'm talking about. Me? Well i'm kind of in the middle. I'll give credit where credit is due, and to that end I think the next-gen consoles have a nice SOC. But I still do not like AMD desktop CPUs whatsoever, nor would I ever choose one over an intel CPU currently.

So, with all of that said.....I will give AMD credit here, the SOC in the PS4 does look very impressive regardless of how badly some here want to discredit it. It also will be a huge boon to PC gamers even if you don't want to buy a next generation console - since all of the code base is designed from the ground up using DirectX and x86, this will make all porting for the PC platform trivial. I think the days of dealing with horrible console ports are nearing an end, and i'm pretty excited about that. Anyway, I'm not so sure why some here are going to extreme lengths to discredit AMD - it definitely is a nice SOC being used and will be a good benefit for PC gamers even *if* one doesn't like consoles. This will benefit all of us, IMHO.

 

[mrmt]

So no, its not a simple matter of scaling up, it would have to be redesigned to take advantage of GDDR5, which then needs its own MC, there's no indications Intel can do a 7Gbps GDDR5 MC anytime soon.. or even if they could, it would add to the die space.

Xeon Phi memory controller is GDDR5.

 

[krumme]

Xeon Phi memory controller is GDDR5.

Yes and 4770 have a gpu.
Intel have broadband mobile tech
They can make a phone with a console
They dont even have to pay for tsmc mark up
Intel can do it, they just dont want to!

 

[tential]

AMD probably offered one heck of a deal to secure both consoles. Margins will likely be quite low in comparison though.

I think though that marketing is part of a role too. Getting AMDs name out there to the general public helps quite a bit. Most people have heard of Intel, but I'm guessing very few people heard of AMD outside of tech circles and even fewer consider them as an option. Getting some console gamers to go "I use AMD in my xbox so I also use it in my PC" is definitely going to help a little I think.

Does it make sense? No. But to general consumers, their marketing rarely does when marketing technology.

 

[mrmt]

Yes and 4770 have a gpu.
Intel have broadband mobile tech
They can make a phone with a console
They dont even have to pay for tsmc mark up
Intel can do it, they just dont want to!

Oh, come on Krumme. You are not one of the fanbois, you know better than that. You know what it takes to get a project approved in the corporate decision chain, and you know that happens if your project cannot reach the company's target ROI.

 

[krumme]

Oh, come on Krumme. You are not one of the fanbois, you know better than that. You know what it takes to get a project approved in the corporate decision chain, and you know that happens if your project cannot reach the company's target ROI.

The closer to the top the less the numbers mean.
At some point at the end, especially for huge future strategic direction, its a mix of gut feeling, temper and napkin calculation.
And the project managers at the bottom of the decicion chain, operate within strict depreciation and roi brackets with hugely detailed budgets, running around waiving with their cute project mnagement triangle while they get strangulated by everyone and his brother. Lol.

Go ask Otellini. As said in the recent interview he should have trusted his gut more. Like their gut have a brain. Try using roi bracket argument here. But whatever, they are also hired to move the entire company. In this case, have made the decision to make the apu for iphone even though the economics looked bad.

 

[beginner99]

" Remember, the point of HSA is to allow CPU and GPU to share a common set of pointers and swap data more efficiently.*

No it's not. That is just 1 important feature of HSA but HSA is much more:

HSA incldues a software stack. It's not just hardware.

 

[mrmt]

Go ask Otellini. As said in the recent interview he should have trusted his gut more. Like their gut have a brain. Try using roi bracket argument here. But whatever, they are also hired to move the entire company. In this case, have made the decision to make the apu for iphone even though the economics looked bad.

You have to, Krumme. One thing is to override ROI models and win, another thing is to override and lose.

In any case, think about consoles. What kind of money is there?

 

[krumme]

You have to, Krumme. One thing is to override ROI models and win, another thing is to override and lose.

In any case, think about consoles. What kind of money is there?

Profit for amd no profit for Intel. Plain and simple. Amd couldnt choose differently, they are in no position to do that.

For this kind of technology, ip, competence, heavy business you have to found your business development on what primarily the kind of competence you have at hand and how to change it instead of looking for some artificial roi bracket that doesnt stand the wear of time, and that everyone also is looking for.

Yes It doesnt matter much for Intel they didnt get the deal and they probably knew arm was not in the running. Its comodity market (after i wrote "its" my s4 phone proposed "comodity" and then "market" - no shit )

(Edit: it proves my phone has a gut)

 

[Charles Kozierok]

Just to add one more thing in here about corporate decision-making -- its quality and foresightedness is often overestimated, especially in the computing world. The entire history of technology is filled with companies making decisions that seemed smart at the time and turned out to have been horrendous.

As just one example, the entire x86 architecture was actually based on earlier CPUs that evolved from a custom design Intel did for another company. Intel was late with the design, so the customer decided they didn't want to buy the product, and reverted the IP back to Intel to save ... $50,000.

 

[krumme]

No it's not. That is just 1 important feature of HSA but HSA is much more:

HSA incldues a software stack. It's not just hardware.

Getting ps4 on the market is a huge step to get legs and arms for that future software stack. Its the strategic important factor in this deal for amd and intel. I guess its practically hsa attention vs avx2 attention?

 

[Piroko]

Some posts in this thread make it look like Intel had the money, patents, designs and both production- and development capacity for the right time to market to step into next gen consoles, but chose not to because 'because.'. That seems like bad business practice to me even if it were a low margin product.

My guess is that time to market was the biggest factor for AMD and against both Intel (improved Crystalwell) and Nvidia/Arm (>32bit memory address space).

 

[ShintaiDK]

Getting ps4 on the market is a huge step to get legs and arms for that future software stack. Its the strategic important factor in this deal for amd and intel. I guess its practically hsa attention vs avx2 attention?

But the PS4 is not HSA. And it wont take HSA anywhere.

And HSA is currently on the fasttrack to end like 3Dnow!. There is still no specification for HSA that 3rd party can use.

 

[ShintaiDK]

Some posts in this thread make it look like Intel had the money, patents, designs and both production- and development capacity for the right time to market to step into next gen consoles, but chose not to because 'because.'. That seems like bad business practice to me even if it were a low margin product.

My guess is that time to market was the biggest factor for AMD and against both Intel (improved Crystalwell) and Nvidia/Arm (>32bit memory address space).

Are you saying Intel should jump into everything, as long as there is "any" margins?

While Intel could, you would have to move resources from other more profitable areas. You dont just hire a 1000 Ph.Ds and expect 100% output after a week in all the chain.

What you do is to look at your resources and see where they make the most money. You would not take an extra job for a dollar an hour would you? I mean...any money right? Profit?

 

[krumme]

Just to add one more thing in here about corporate decision-making -- its quality and foresightedness is often overestimated, especially in the computing world. The entire history of technology is filled with companies making decisions that seemed smart at the time and turned out to have been horrendous.

As just one example, the entire x86 architecture was actually based on earlier CPUs that evolved from a custom design Intel did for another company. Intel was late with the design, so the customer decided they didn't want to buy the product, and reverted the IP back to Intel to save ... $50,000.

Nice story.
We tend to like forecasting models that shows a liniar effect cause relation. They are easy to present in the books, projects and investments and difficult to argue against.

The problem is they dont grasp the only thing we know about the future is, thats its unpredictable. That doest prevent people everyday all over the world to complain about project not keeping budget, that it was because of the requirement specs not beeing detailed enough.

What companies is having huge troubles with is having a culture and mentality that is ready to change and adapt when the markets demand it. Its difficult to know eg. What x86 evolved into, but its manageable to create an adaptive organisation that moves in time.

Look how utterly slow amd and intel have been here. Apple offered intel a golden goose with the iphone opportunity, amd even managed the unthinkeble and sell the goose to qualcomm for nothing.

 

[Phynaz]

Some posts in this thread make it look like Intel had the money, patents, designs and both production- and development capacity for the right time to market to step into next gen consoles, but chose not to because 'because.'. That seems like bad business practice to me even if it were a low margin product.

Business' walk away from deals everyday for a myriad of reasons.

 

[krumme]

Some posts in this thread make it look like Intel had the money, patents, designs and both production- and development capacity for the right time to market to step into next gen consoles, but chose not to because 'because.'. That seems like bad business practice to me even if it were a low margin product.

My guess is that time to market was the biggest factor for AMD and against both Intel (improved Crystalwell) and Nvidia/Arm (>32bit memory address space).

It is bad business practice. Its still prevalent in sectors with bad competition and its old business thinking. In competitive businesses if its in your strategic scope meaning you have the eg. tech, competence and there is profit, you go for as it expands your business.

And yes its good you mention ttm here in this thread as that is surely a major factor.