AMD A10-5800K preview - iGPU side only

[pelov]

Ummm...no. Again, see my link.

There's a very easy way to prove this:

Find a GPU-accelerated openCL Intel benchmark.

I'll spare you the trouble: it doesn't exist.

Intel's openCL acceleration is the same as Intel's license with ARM holdings: no plan to use it.

Here's what I mean

GPU

CPU+GPU

AMD's APP, in use above, accelerates the CPU performance of both, but with the weight sitting mostly on the CPU it shows little benefit for GPU acceleration. So while AMD's and Intel's processors benefit from the improved CPU performance, AMD's GPU acceleration isn't able to beat Intel's i3s in the benchmark, though it does close the gap.

- just a bit more info

The benefits vary on the application involved and its implementation. Currently it's quite limited, as is the case with Adobe, where certain filters provide a very significant benefit whereas others show zero improvement. So this really isn't a method of AMD closing the performance gap; it would take a monumental change in software and development for that to happen. It makes far more sense where TDP limitations mean high-performing CPUs are nonexistent, allowing the use of the GPU to leverage more computing power. Basically, it shows more promise in HPC and mobile than it does on the desktop, unless you're talking workstation.

- some additional info

Here are some OpenCL benchmarks from the Intel Ivy Bridge CPU. Being compared though is AMD's APP SDK, which does support running OpenCL on x86 CPUs, to Intel's CPU-based OpenCL SDK for Linux. To some surprise, AMD's Accelerated Parallel Processing SDK when using the Ivy Bridge CPU is actually faster than the Intel OpenCL SDK on the same hardware.

I can't find the link, but the above scenario for Linux and AMD's APP performing better than Intel's SDK is true in Windows as well. This has more to do with AMD spending more time on openCL for both CPU and GPU in order to squeeze out as much as they possibly can.

 

[Dark Shroud]

Some of you guys are being seriously obtuse here.

An APU is a CPU with an in die GPU that is able to General-purpose computing aka GPGPU.

Intel's IGP does not function as a GPGPU, so my i7-2600k is not an APU.

If VIA & Nvidia signed a partnership to embed Nvidia GPUs onto the same die of a VIA x86 CPU that would be an APU.

AMD's CPUs do not have IGPs so they are called CPUs.

Some people in the industry have referred to Intel's CPUs with IGPs as iCPUs trying to clarify the difference. Intel prefers CPU-EGP.

 

[Leon55ia]

I don't want to brag but lets be honest $100 is very little if you keep that PC for 2 years, that is about $4 per month.

 

[Blitzvogel]

I don't want to brag but lets be honest $100 is very little if you keep that PC for 2 years, that is about $4 per month.

It's not always easy just dropping a few hundred dollars, but I would agree that you do often get what you pay for, and for many, a $400 A8 or A10 Trinity equipped PC would do the trick for light gaming, media, and non-intensive work. As well as the benches have been portraying Trinity, I'm entertaining the idea of a small low profile Trinity machine for light gaming and HTPC stuff.

 

[CHADBOGA]

Its simple, the APU's were designed to accelerate apps with the most efficient and highest performing method available(CPU,gGPU, 3rd party IP, or a combination of the 3), ergo Accelerated Processing Units. Naturally inetl can't call there's APUs they don't have the same abilities. But really its irrelevant what inetl wants to call them, the industry already knows them as APUs and the term is already ubiquitous.

Your dyslexia appears to have returned.