Just asing Ivy Bridge CPUs can handle heterogeneous computing. Such as sharing 64bit address space between the CPU and GPU ? Currently the CPU needs to copy data to the GPU, load & unload for any kind of GPGPU work.. inefficient.
But I think this will end with next gen AMD GCN2 and Nvdia Maxwell.
So the point is, is it safe to guess Ivy Bridge can fully utilize a future GPUs coming next year?
Or will I need a Haswell / Skylake/ whatever..?
Asking because I don't want to buy another CPU next year.
No, HSA is a AMD patent at this point in time already noticable in a APU setup but as of yet HSA is to early so Intel may address that in the future but right now only AMD is concerned in parallel computing.
Thanks for the post!
I was hoping Maxwell can do this as well with a descrete video card (Intel or AMD CPU) - just to have more choices.
I'll need to look at AMD more closely if I want GPGPU APU to function properly with the CPU and GPU (cores) sharing the same 64bit address space. Then just wait a LONG time for software to catch up.
In the mean time, if I want this functionalty with any highish-end CPU and a descrete card, I'll need to wait a few years.. maybe this is only possible in a post PCIe ERA ! Skylake, Excavator, whatever.
AMD's entire roadmap is based on HSA, Vision is future mantra indicative of AMD and partners efforts into parallelism of computing. A A10 does certain rendering jobs up to 4x as fast as the i7 3770k where workloads are optimized, some Adope and Photoshop apps show this off. If HSA is your thing then just looking into AMD is a prospect as AMD are heading in the right direction now. Just keep a look out.