http://www.thehindubusinessline.com/industry-and-economy/info-tech/we-are-at-the-dawn-of-the-pc-era-says-amd-india-chief/article4137439.ece?homepage=true&ref=wl_home
I found that very interesting. Apparently in India, APU sales for desktop systems are going incredibly well.
Everyone is saying the desktop market is dying, but you know what? It is just dying in America and possibly Europe. China and India are still huge, developing markets for desktop PCs. These people dont' have good gaming computers or anything remotely powerful. They're not going to want some crappy tablet for consuming content. They're going to want something far more powerful to get into games. We can see that very, very clearly in that article in regards to APUs doing so well.
You want my opinion? AMD dropped Steamroller in 2013 to save cash and then fling APUs and AMD FX into developing markets like India and China and to reap massive profits from those markets. That's 2 billion people in those markets, much much larger than the American market.
If AMD India has a chip that's selling like hot cakes, and they have virtually no competition when it comes to a cheap integrated solution with a powerful GPU, they have no reason to spend money half assing something and trying to half-fix something that's majorly broken. AMD's main goal right now isn't to beat Intel, it's to actually make a profit. The best way to do that is to leverage existing intellectual property and make the most profit you can out of that property. That's exactly what AMD is doing in India.
Reap the rewards of developing markets, have Jim Keller takes the reigns on a new architecture for 20nm TSMC process when the company is profitable, and then bam, AMD is in good shape.
AMD has a big chance to turn profitable by leveraging India and China and the other developing markets. They can then possibly come out with a stronger Steamroller and catch Intel with their pants down as Intel chases ARM. The performance gains from the larger x86 cores then trickle down and improve performance on APUs and smaller cores. If Trinity was still using BDver1 cores, it wouldn't be a good product. It's not bad at all with PD, and with better x86 cores it could be an extremely good product.
If anything, I would say big x86 core R&D at AMD is on hiatus, not cancelled. Charlie @ S|A is saying Intel is going to give up on the enthusiast market as well. AMD would have the enthusiast market by the balls if they released unlocked CPUs on an enthusiast platform while Intel shovels out chips soldered onto motherboards.
Just give it time, Intel is going to shoot itself in the foot by chasing things like Medfield. Even if Medfield CPU performance is good, their GPU is god-awful. It's about 40% slower than Adreno 225 and Adreno 320 humiliates it. So do all the PowerVRs. And as we've seen with Larabee, Intel can't make a good GPU to save its life. There's a huge reason why all those articles and reviews that say "OMG MEDFIELD" are usually always lacking in GPU benchmarks.
This is like pre-K8 era. Intel is off chasing something irrelevant (pre-K8 it was chasing clock speed instead of performance) while neglecting the thing they do best (kick ass x86 high end CPUs).
People are blinded by their "Intel makes awesome x86 CPUs so everything they do is magical and perfect" goggles. Itanium, their motherboards, Larrabee, Atom, etc have all been commercial or complete failures. The only thing Intel has going for itself right now is its fabrication process and good x86 CPUs, and they seem pretty content with getting out of the high performance x86 CPU market and trying to cram an architecture that's not designed for low power into a low power form factor to compete with an architecture that's specifically designed for low power form factors.
I don't see AMD dying at all. 2013 will be very rough for them, but beyond that they have a huge future ahead of them. They could turn the company around by leveraging existing IP in emerging markets and then filling the void Intel left in the enthusiast market, hopefully by creating a platform that shares a common socket between server and desktop CPUs as opposed to enthusiast desktop and APU sharing the same socket. Hell, just put all of them on the same socket. Who the hell wouldn't want a platform where you could choose between ARM APU, x86 APU, enthusiast CPU, or 16 core/20core server CPU? All the while your competition is selling a motherboard with a soldered on CPU.
I just hope Rory Read and company are smart enough to at least see what I'm seeing, even if they have a different or better solution than I do.
Facebook
Twitter
Instagram