What does a GPU do that a faster CPU can't??

[AppleGoat]

In other words, how much of a loss is the NVIDIA GT320m...outside of gaming. I realize neither computer is optimized for gaming and the new Intel Sandy Bridge IGP is only marginally less competent (well, at least as long as it's paired to a strong CPU), but after reading the CNET review, I began thinking. Gaming does not interest me, but seeing the frame-rate discrepancy between models on certain games and settings, what else does this carry over to. Is video playback marginally better on the 2010 models? Certain rendering?

 

[impulse462]

With the right technologies, like OpenCL, if the power of the GPU can be harnessed for everyday CPU tasks, the computer will be much more efficient.

At NVISION (08?) They showed a demonstration of this.
They had a machine that automatically produced a smiley face made of dots in around 20 seconds (around that time I don't remember that long ago)

They then took another machine that drew the entire Mona Lisa in a very short amount of time (3-5 seconds)

This was shown as a metaphor of the power a CPU is capable of compared to the GPU.

As for real world tests, I wouldn't know much outside of gaming as I used to be a gamer, but as for the "theory" the GPU is much more powerful than the CPU.

 

[dta]

I heard GPU's can handle certain mathematical operations much faster because of their specialized design. Some scientists make simulations just to run in the GPUs. See the video here :http://www.nvidia.com/object/race-for-better-science.html

 

[1BadMac]

GPU's can be leveraged to perform complex parallel computations - parallel being the keyword there. CPU's are serial processing, meaning they do a single task at a time. Hyperthreading / Dual Core makes this a little better (1 task / 2 task per core, depending on CPU architecture. i5/i7's with HT are two tasks per core).

GPU's, however, can run several tasks, up to HUNDREDS of tasks, in parallel. This depends on the GPU architecture, amount of memory, etc...

Gizmodo has a good "GPGPU for beginners" blog post you can check out.

http://gizmodo.com/#!5252545/giz-explains-gpgpu-computing-and-why-itll-melt-your-face-off

 

[AppleGoat]

Thanks for the interesting responses. I guess I'm really asking, if the graphics on the 2010 13" are, in fact, better than the 2011 13". Outside of gaming, what else will be compromised in the newer models. Yes, I know, more likely than not, marginally compromised. But, what are those things?

 

[1BadMac]

Thanks for the interesting responses. I guess I'm really asking, if the graphics on the 2010 13" are, in fact, better than the 2011 13". Outside of gaming, what else will be compromised in the newer models. Yes, I know, more likely than not, marginally compromised. But, what are those things?

They are almost identical for the purpose you are discussing.

The HD 3000 in the new 13" has 12 "execution units". Still looking for what that translates into for cores. The 320m has 6 multiprocessors and 48 cores.

As more programs start to figure out how to leverage OpenCL for more than pure graphics rendering, more can be offloaded to the GPU. More near term items would be video editing tasks, etc.

What are you missing out on right now? Nothing really. The performance is (on the weak side) for both of these graphics processors.

 

[alust2013]

Thanks for the interesting responses. I guess I'm really asking, if the graphics on the 2010 13" are, in fact, better than the 2011 13". Outside of gaming, what else will be compromised in the newer models. Yes, I know, more likely than not, marginally compromised. But, what are those things?

There's basically nothing. I suppose if you do CAD work or 3D modeling, you may notice a slight difference. However for any standard uses, including HD video, you won't notice a thing. My original unibody MacBook can do 1080p without breaking a sweat, and the GPU is half as powerful as either of the two newer ones.