Where they dumping all that money into the Zen code R&D? With the Meltdown craziness, will this be the year we start to see AMD return to the data centre?
PS3 XBOX360 and Nintendo Wii were all IBM chips . They tried a bunch of new architectures but for the most part developer ergonomics was horrible.
So for the PS4 and XBOX One they decided they needed regular computer parts. I'm sure AMD and Intel may have put bids on it but AMD would be the one that would have been extremely motivated to make sales to Microsoft and Sony. As time goes on you can make higher and higher margins since the chips don't change (but on the other hand we are seeing console revisions within the traditional generation). But, I think that the margins will keep on increasing.
It's commonplace for consoles to go through a number of hardware revisions over the lifetime of the console. These revisions are rarely made for performance reasons, but instead would be made to improve reliability, security and manufacturing cost. Setting aside security (which is would be done as reaction to a hack that can't be patched with software), the whole idea is to increase the margins for the manufacturer. Note that the brand owner may, in some cases, sell consoles at a loss, but the manufacturer is very unlikely to be working to the same business model.
As your chip goes from cutting edge process technology($$$) to mid-tier($$) and commodity($) they usually pass that savings along, otherwise they would incentivize going for a mid-cycle refresh and suddenly a large revenue stream for your chip dries up as a product moves to a (potentially) different vendor.
The only other manufacturer of x64 CPUs in Intel, and the only other manufacturer of (competitive) GPUs is Nvidia. Why would Microsoft or Sony incur the high NRE cost to switch to two different suppliers mid-cycle just so they can shave a small amount off their unit cost?
That's not to say AMD wouldn't pass on some of the savings to Microsoft and Sony, I'm sure they would, but it's not likely to be a one-sided deal.
You also always have multiple vendors bidding so if one of them offers a sliding scale you're going to use that as leverage against the other vendors.
It doesn't matter if the potential savings are sunk by the NRE costs of shifting to new vendors before the projected end of the product. You have to look at cost holistically.
When you're talking 15M+ units being able to drop $1-2 from BOM is a huge motivator for things like this. When you get into the millions of dollars worth of savings you can fund a lot of engineering time.
They also may never go through with the switch but use it as leverage for a better deal. Back when I did SoC evals we'd take products right up to the brink of production just to apply pressure. It was almost like a game of chicken between 2-3 vendors to see who blinked first and gave us a better BoM/deal on the core chip. These weren't easy bring-ups but usually were vastly different GPUs + CPU combos. It's very much a thing that happens, at least back when I was involved with stuff like that ~6 years ago.
If you look at the average lifetime of a (successful) console, you're looking at between 5 to 10 years. Within that time, multiple revisions take place, often to replace peripheral components to reduce costs, but in the case of the longest living consoles they will often get a design refresh, including a redesign of the CPU / GPU. I can think of only one example in recent memory where a console switched manufacturers for these core components during its lifetime (that example being the GPU for the Xbox 360, which switched from AMD to IBM). In the case of the Xbox 360, IBM were already manufacturing the CPU, and the redesign combined the CPU and GPU into a single SoC. Aside from that one example, there haven't (to my knowledge) been any other examples of companies switching manufacturers for their core components mid cycle. Contrast this with the PC business, where OEMs will frequently swap between motherboard/CPU/GPU manufacturers, and you'd have to wonder what makes the dynamics in the console market different.
I spent a while in the industry(and shipped some god-awful titles) so I tend to trust the people I know(and don't plan on outing them) but can understand if you don't want to take the word of a random person on the internet.
I know you were probably talking about traditional consoles(xbox/ps/etc) but when you start including portables(with the switch just smashing the 1 year mark at 15M units) then the field opens up significantly. You've got at least 5-6 different GPU vendors and a whole host of companies providing SoCs. Embedded GPUs have been making huge strides and I wouldn't be surprised to see some of the top end of that space start nipping at NVidia/AMD soon. Lots of the people in that domain came from desktop(Adreno is a anagram of Radeon for instance ;)).
If you disagree with something I've said, by all means explain why.
Interesting.
> "It's obvious from your posts that you don't have any particular knowledge of AMD's console dealings, or even console dealings in general."
I never claimed to.
> "You only speak to generalities which are well known among even casual observers of the industry. None of what you've said specifically refutes the original claim."
It's precisely because these market generalities do not match with what was proposed that I called it into question.
Developing new silicon is a risky endeavour. Even going from one process node to a smaller process node without changing the design architecturally is fraught with problems, resulting in lower yield until the manufacturing kinks are worked out. Why would a chip manufacturer agree to a deal where they're expected to take on increased risk for a lower reward? It makes no business sense. That's what I was calling into question.
> "I know you don't have any grounds to speak with any authority on the matter, and you know it too if you're honest with yourself"
I may not personally know the same people vvanders knows, but I can put myself in the shoes of a businessman running a chip manufacturer. I enjoyed debating with vvanders, and aside from leaving out the dig at the people they know in the industry, I'd do it again.
If you know more, which you may well do, the way to comment here is to share some of what you know in a way the rest of us can learn from.
Intel graphics solutions were (and remain) sub-par, and Nvidia has no x86 solutions (and their ARM chips would have been underpowered).
A lot of that has to do with it being a handheld hybrid, as you point out, but these compromises would have been a lot harder to stomach if it had been a dedicated set-top console like its direct competitors.
If the only concern is resolution and not quality or gameplay, then the Switch doesn't stack up, but I think the sales numbers point to the Switch being a smash hit with consumers so those compromises seem stomachable from my perspective.
(Anecdotally: I wish the graphics in BotW were better, at least in terms of draw distance, but then I take the Switch out of the dock and take it to work to play Mario Kart with my coworkers and it seems like a fine tradeoff to me.)
And while this time the Nintendo gimmick seems aimed more at the core gamer crowd (Take your console game anywhere!) I'm still not convinced this won't be turning out the same way like the Wii did.
Btw: If you want BotW with better graphics you can play it using a Wii U emulator, all the 4k BotW your eyes can handle.
That said, I love the switch, and BotW has quickly become one of my all time favorite games. My only problem is resisting the urge to kick the kids off the switch so I can play. ;)
Original Doom was already running on the SNES like 25 years ago.
Getting something to run is not really that big of a hurdle as you can always scale down resolution/graphical fidelity/frames per second. The question after that rather being: Who would want to play an obviously inferior version of the game? Because that's exactly what these Switch versions of Skyrim and Doom are, inferior.
You can see that in 2016, their console (+ other stuff) was making an operating income (basically revenue - cost of business not including R&D) roughly the same size as their losses in their main CPU&GPU business. This is before their ~400 operating losses in other categories. I think R&D gets rolled into that? I'm not sure
I'm guessing razor-thin margins to get the contracts.
Despite spinning off, GloFo and AMD remained tightly aligned businesswise, and AMD signed agreements such as minimum silicon wafer usage.
But by 2013, AMD had sold its headquarters in a "reverse mortgage" to remain solvent. https://arstechnica.com/information-technology/2013/03/amd-s...
AMD had almost a full decade without making money.
In order to make the fab seem like something someone would want to buy, AMD tied their futures into it by entering into a wafer supply agreement where they will buy certain amounts of GloFo:s production whether they need it or not.
This agreement was a millstone around AMD's neck during the worst years. It means that AMD absolutely wants to sell at least a certain amount of silicon, even at negative margin if necessary.
The article says the PS3 cost of goods was $800 per unit at launch, but it is worth remembering that the PS3 was an investment in two strategic initiatives for Sony. One was the PS3 standard, that they would earn future software royalties on. The other, and at the time maybe more important, was the BluRay standard.
The changes game developers had to make to handle cell allowed them later on to support multicore CPUs and GPGPU much easier.