GT3E on MacBook Pro Retina?

[Mr MM]

It would be pretty awesome if they could do that. Maybe only on the high end model, as it would probably add a bit of cost.

i would like to point out to exhibit A

I`m willing to bet that 33% higher TDP isn`t 33% higher temperatures. Its only worst case scenario

EDIT: Sigh: Its a mixed picture.

 

[dusk007]

Just a question why would the 4850 turbo higher than the 4800?

Relative to the base clock. Absolute, as I said, they would end up at about the same clock.
Both CPUs have 47W and some Temp limit. If 20 of the 40 EUs are power gated and the remainder under little or no load the 4 cpu cores would probably end up with about the same absolute clock speeds. Not the 4850 with 500 Mhz less.

Yes they both replace the same CPU and they both will end up delivering almost the same CPU performance unless the 4850's GPU is under heavy load.

OpenCL may benefit a lot from the shared memory and Cache of the CPU. It is also doubtful that they would ever actually pair a 4850 with a dGPU so there would be either or. Also it depends which GPU is active which brings us back to imperfect switching.
Some loads may run faster with more processing of a dedicated GPU some OpenCL stuff would work much better on the more integrated IGP due to memory sharing and cache sharing. It really comes down to the driver with that kind of stuff.
In any case that "feature" is more of a future thing that may become more relevant in the coming years but isn't really now.

 

[Freyqq]

i would like to point out to exhibit A

Doesn't mean much at all. Bf3 is a very gpu heavy game. The CPU doesn't have to work very hard with a 680m in there. Would have been more helpful if they maxed out the CPU for 15 min instead.

 

[Mr MM]

Relative to the base clock. Absolute, as I said, they would end up at about the same clock.
Both CPUs have 47W and some Temp limit. If 20 of the 40 EUs are power gated and the remainder under little or no load the 4 cpu cores would probably end up with about the same absolute clock speeds. Not the 4850 with 500 Mhz less.

Yes they both replace the same CPU and they both will end up delivering almost the same CPU performance unless the 4850's GPU is under heavy load.

OpenCL may benefit a lot from the shared memory and Cache of the CPU. It is also doubtful that they would ever actually pair a 4850 with a dGPU so there would be either or. Also it depends which GPU is active which brings us back to imperfect switching.
Some loads may run faster with more processing of a dedicated GPU some OpenCL stuff would work much better on the more integrated IGP due to memory sharing and cache sharing. It really comes down to the driver with that kind of stuff.
In any case that "feature" is more of a future thing that may become more relevant in the coming years but isn't really now.

YES, HSA is not something that consumer load will be affected, server and scientific on the other hand have been using that cache for awhile and the gpu as well, though not in the same package.

In the end it doesnt make any sense that the 4850 will ever perform the same as the 4800. Its simple

There are only 3 notebooks that uses muxes, without those you cant power gate effectively the igpu cores (they need to be on), so in the end the 4800 will always come ahead. And its pretty lame to count on turbo to off set such a disparity on clocks despite the even still 200mhz short of max turbo

Doesn't mean much at all. Bf3 is a very gpu heavy game. The CPU doesn't have to work very hard with a 680m in there. Would have been more helpful if they maxed out the CPU for 15 min instead.

that would be interesting, but BF3 is actually pretty intensive when it comes to cpu and it likes 4 cores. Best way to run that game is with a quad with speeds of over 3.4ghz

which in turn for the point that Im making here is nill, the temp is what counts, and you can see that the same toasty i7 3520m gets as hot as quads, they are in the end in a variation area, very different from the other 35w chips in there

 

[dusk007]

There are only 3 notebooks that uses muxes, without those you cant power gate effectively the igpu cores (they need to be on), so in the end the 4800 will always come ahead. And its pretty lame to count on turbo to off set such a disparity on clocks despite the even still 200mhz short of max turbo

What needs to stay on is very little though for Optimus and the like to work and the same for the 4800. Intel said that in GT3 half the array can be completely shut down to make it effectively a GT2. At which point it would just be the same amount of GPU running at the same load in both chips. I don't see what difference it makes then how many notebooks use muxes.
What is left running for Optimus is also very little. It is just a framebuffer copy. It can power gate almost everything. The GPU can be almost completely off only the display output needs to be on. Since Optimus isn't exactly new, I am sure Intels power management has planned for that operation.
Lets wait and see who is right. I bet the 4850 does a lot better than the low base clock would imply in most work loads. It is not a 4750 and not cheap. I don't see why Intel shouldn't reallocate the GPUs power budget if the GPU doesn't need it.