You didn't read what they did they got 2 Ghz out of a 130 nm part first.
Yes, a
DSP. It was definitely nice at the time, but to put that in perspective, TI now have 10GHz fastmips DSPs. If Intrinsity had such enormous competitive advantage, one would expect it would have grown a fair bit larger than 100 people in over 10 years.
Not a Cortex A8 at 1Ghz. You realize also that the samsung hummingbird is the 1 Ghz core that Samsung and intrinsity designed together? And that he A4 is a custom hummingbird?
Yes I realize all that. What I dont realize is what is (or was) so earth shattering about a 1 GHz single core cortex A8? Qualcomm didnt need intrinsity to achieve the same earlier with their 1 GHz snapdragon; well not quite the same, their core is actually a fair bit more powerful than the standard cortex A8. Qualcomm is now up to dual core 1.5 GHz. There is really nothing magical about intrinsity. The best thing about hummingbird is the GPU, even if its an off the shelve PowerVR core.
It's one thing to get a semi custom design from something samsung already did. It's another thing to go out and make yourself a fully custom chip. The risk is so much higher. Margin improvements could quickly become margin eating cost if you screw up or if your sales don't do as well as you thought it would. Buying a off the shelf chip or getting a customized version of another design is another story. You don't have to pay for all/most of the sunk cost to get the chip working in the first place. You just order the numbers you need to meet your sales. Going full out custom chip means you take a big big risk. Unlikely something Jobs would do since this is the guy who refused to pay a dividend because he wants to keep the cash on hand.
We'll see. Some evidence here that Apple is going for full custom chip:
http://www.semiaccurate.com/2010/12/29/evidence-points-apple-designing-arm-laptops/
Although "full custom" can mean anything, you could call Marvell or qualcomm's cores full custom or customized cortex's. They all differ from ARMs design, but have quite a bit in common as well. The same will apply to Apple, MS, and nVidia's designs. They will all borrow from ARMs work and change where its needed to meet their performance or functionality goals. Thats how AMD and Intel work as well, they rarely, if ever start with a complete blank sheet of paper.
To think fusion is as simply and sticking a CPU and GPU into the same chip and that's a over simplification. First generation will looks like this but the point of fusion is proper heterogeneous computing. The Xbox chip is just cost saving integration, and it certainly didn't happen without AMD's help since it's a AMD/ATI GPU.
So what's so custom about the Xenon chips for Xbox? Pray do tell me I'd love to know and hear.
I never said combining a CPU and gpu is simple, did I? What MS did with xenon was actually even more complicated than fusion in some regards, since the new soc had to behave 100% identical to the old chips. It wasnt allowed to be any faster at anything, despite for instance removing the FSB. They had to introduce blocks that mimicked the removed external FSB to slow it down.
But those things aside, I dont see a lot of difference between what amd is doing with llano, combining an existing K10 core with an existing GPU, or what microsoft did with xenon. And yes microsoft, perhaps assisted by IBM, but they did the bulk of the redesign and layout. AMD didnt do a damn thing AFAIK. IBM was probably only involved in removing the FSB. Its a nice learning path, going from integrating off the shelve chips, to codesigning the chips to a mostly inhouse redesign. But of course that could be coincidence and totally unrelated to obtaining an ISA license now. Maybe they got that license to fill an empty spot on Ballmer's office's wall of fame.
It make no business sense to run up your cost by doing more stuff in house when what's out there's good, even great.
Of course it does. You dont give away margin to your supplier and you dont subsidize your competitors. You have the ability to more tightly integrate and control your software and hardware, and you are not at the mercy of a supplier and his roadmap and his execution.
I find it funny how you say that apple buying intrinsity has nothing to do with a competitive advantage but that they get something out of this "vertical integration". If building your own chip gives you no advantage, setting aside the x86 vs ARM debate since we are talking apple's A4, what's the point? How is apple designing their own A4 a good thing if as you put it they get nothing out of it since everyone else have a 1Ghz hummingbird now? Does it help their margins by assuming the fix costs for chip production? Assuming the risk for any potential mistake their own guys made in the custom design?
I never said what you are putting in my mouth now. I was just saying Intrinsity didnt have some magical silver bullet tech that Apple at all costs wanted to avoid getting in competitors hands.. Intrinsity were a competent team with good tech, and Apple needed that because they wanted to build their own socs. It wasnt the other way around; if intrinsity hadnt existed, Apple would have bought another team. They already had. They didnt end up with several CPU design teams by accident.
I don't think Nvidia's likely to be making their own phones in the next 5 years.
A lot of what you say is cherry picking the things you want to think and see and saying look it proves I am right. It doesn't.
To say Microsoft's OS is being challenged by OS from Google, Intel and HP is mixing up OS and target market. You are saying that Android, WebOS and MeeGo is challenging Windows Phone 7? Windows 7? Or what? It seems to me that if you are talking MeeGo, Android and the lot of mobile operating systems it's the other way around. Microsoft's Windows Phone 7 and Windows 7 on tablets and netbooks are challenging Android and Co for the mobile market.
Turning the sentence around changes what exactly? Windows CE has been around for 15 years. HP used to its biggest seller with their ipaqs. And now they have a direct competitor in WebOS. Others have android, meego or whatever. Microsoft competing with HP or HP with Microsoft. Whats the difference?
Clearly for now, the competition in the OS market is mostly in the handheld/tablet market, but that doesnt change the facts, nor does it change the fact they also compete with windows 7 tablets. Nor does it change the reality that ChromeOS will compete with windows 7 on laptops and desktops. Android seems to be going there too, Meego probably as well, and with MS porting windows NT to ARM, they are bound to all compete in increasing segments of the market.
Now if you want to now say this is why MS should integrate vertically. Well why would they do that when the non-vertically integrated android is the dominate OS next to iOS? I mean not building your own hardware and selling it doesn't mean you can't succeed. Further more, Google's not making money directly off Android OS but rather the software package that a vendor can choose to license or not (hence HTC's move towards sense, to make money and avoid paying google). At this point you might say to me well they would make more money if they do an apple and vertically integrate. To which I would remind you they did try to sell their own phone and it didn't seem to have been viable enough for them to continue doing so. What makes you think Nvidia a company with no software platform can come in 5 years and successfully market a phone on their own? Because they make a mobile chip? Well Getrag makes gear boxes so I guess in 5 years time we will be able to buy a Getraq car at a local dealer by your logic.
Im not saying blindly going vertical top to bottom is the best strategy for everyone (especially not in the car industry) nor that its a guarantee for success. Im just trying to open your eyes to the reality happening in the IT industry.
As for android, an OS and soc are very closely tied together. Google may not be designing their own SoC (yet?), they are doing the next closest thing by picking a single SoC design as development platform for each android release. Its the cheaper and easier -as well as more limited- way to gain more control over the component stack that is the end product.