Originally Posted by wizard69
This is meaningless commentary on a product that hasn't even had time for software to achieve normal optimization. Not to mention that you need operating system support to fully leverage the capability. In a nut shell it is far too early to pass judgement on HSA via app testing. It will likely take a couple of years for everything to develop properly.
Your nuanced command of the English language is simply awe inspiring. Lesser mortals might have thought that the phrases "The jury is out" and "It’s impossible to evaluate the long-term potential of HSA at this point" means the same thing as "it is far too early to pass judgement" but you evidently have teased out the impressive differences between those phrases.
The point here though is that the reasoning behind HSA is very sound and the architecture melds well with Apples directions in operating system and library design. Even then I would never expect GPU acceleration to work for everything. GPU acceleration is about leveraging the strengths of the GPU when you can.
The point here is that HSA is unproven in the way that matters: shown to provide significant real world performance advantages.
The fact that Haswell implements a huge cache chip for Iris Pro is evidence that there is a bandwidth problem. Every bit of rational benchmarking on the net indicates clearly that more bandwidth is need for both Intel and AMDs processors.
The fact that Haswell implements a huge L4 cache is evidence that a 128MB cache chip was judged sufficient to address GPU memory bandwidth concerns in this generation and possibly the next one.
You have yet to provide any benchmarks, rational or otherwise. That both Intel and AMD will increase bandwidth is pretty much a given. That there is a current "bandwidth problem" isn't.
Maybe a Mini built into a keyboard! Even in a more conventional design I can see Apple going this route.
1980 called and they'd like their computer design back. A NUC or Mac Mini is sufficiently small and I prefer my keyboards light and not connected to a power line. Or any line for that matter.
I disagree, the Mini in my mind is the machine most exposed to suffer from the success of the iPad.
It is simply bizarre that you think that powerful desktop machines used as servers and HTPCs are "most exposed to suffer from the success of the iPad".
I hear this all the time and frankly I see it as total BS. The Mini brings many customers into the fold that might not otherwise consider a Mac. Each Mini sale is a potential future iMac, Mac Pro or laptop sale.
Only if the Mini is not so good as to make the iMac a so-so upgrade. The cost deltas are huge.
It is amazing that in another post you can claim that Mini sales are dismal and in this post you can claim the Mini brings many customers into the fold. In any case only Apple has the number on sales and possibly the data on whether sales go to new customers or existing Apple owners.