AMD's Kaveri; either a revolution or AMD completely forgot how to make a CPU chip.

Looking through the article, I can see that AMD is really betting everything on possible future applications for its new API’s.

Why I said that, the entire list of new concepts seem interesting for future applications, but at the cost that its new chips have now sunk well below the level of Intel i3’s in various CPU applications (think of that fancy new machine that ends up performing like a bargain-basement E-machine).

The graphics side of things though are a bit better, actually beating Intel in several benchmarks, and who knows about the performance once applications make use of the new API’s that are (essentially) needed for this design, all in all, you’re simply not going to get much out of these transitional chips as they no longer have an actual core-based architecture (as in, the architectures that work good with today’s code). The future for AMD by now seems to either be the arbiter of a new revolution or one holding a chapter 7 liquidation sale, time will tell.

My previous laptop had an AMD chip . . . this was around 4 years ago. I deliberately bought the AMD processor just to give intel some competition. CHIP ran ok…laptop did not fair so well and started to fall apart.

It looks like is in the revolution camp in terms of where they think AMD is heading (the entire concept behind Kaveri being basically the API’s such as HSA and Mantle).

Their argument, this unique marriage of CPU and GPU capability will mean an API that’s easy and inexpensive to program for while allowing greatly increased performance for a lot less money than what you need today. (now if they can significantly bump up the performance of their cores for traditional and legacy programs that will be around for a while longer yet).

Engadget meanwhile has done some HSA driven operations in apps. like Libre Office and found high performance gains, so the tests coming out seem to back up these points.

I for one am interested in how this goes, it’s a revolutionary idea but how they execute it is critical to say the least, one screw-up and their idea might come crashing down.

Here’s who to look out for.

They just god mad funded ( $3.4 m. ) and they will begin on a 64 core ( 8x8 ) “open source” parallel excelerator chip soon. This will definitely be something that can blend… and no more cobbled together drivers.

The future of a computer, is more like a brain…

There is no dedicated anything… You tell a cluster of CPUs to manage a cluster of objects, then send there output to a series of destinations… This way the entire computer is parallel… Now as for writing apps, I would imagine an input something like python, that controlled VERY low level using high level programming.

Part of the reason I like this is because it would make software development a LOT easier (I develop software and frankly it’s a pain to optimize for various hardware combinations)Another reason is that this will result in better performance.