Originally Posted by Marvin
Clarksfield is rumored to be 35W but we don't know for sure. The current quads are listed as 45W on Intel's processor spec finder:http://processorfinder.intel.com/det...px?sSpec=SLGEJ
Price comes into it too as the 2.26 quad is the same price as the iMac 3.06GHz dual core. I would rather they offered the cheaper 2 GHz quad but if they sell a quad at a lower price than the 3.06GHz, few people would go for the dual core.
I suspect you are right, but isn't that a decision the customer should make. Frankly with the arrival of SL I would see a quad as being a better option for most in a Mini rather than a modest clock rate increase.
In the iMac Apple should certainly offer up the fastest dual core available as some people would benefit from that. For your average user though quads will be the better choice. To me this is like the transistion from single to dual core, it only took a very short time for the OS and apps to catch up and leverage dual core SMP. With SL the transition ought to be even faster on Apples hardware. That is if the quad core SMP hardware ever comes out.
Anyways off my soap box, I'm just trying to reflect on why I consider Apples current desktop hardware to be such a bad value.
The thing about the G5 though is that vectorized code only benefitted Mac users, who are small minority. OpenCL like OpenGL will benefit any user with a compatible GPU on Mac, Linux or Windows as it's an open standard so I would hope that it will gain significantly more usage. Not to mention, the gains on the G5 processors weren't nearly as dramatic. The GPUs have 16-32 stream processors running at 1GHz+and are currently unused completely for non-graphics processing.
I'm not to sure the open standard that is OpenCL will make that much of a difference in adoption of the facility. On the Mac it will be used because it is supported by Apple. Other platforms will have alternative competeing technologies. Besides we are seeing a lot more Mac only apps, so adoption isn't predicated on being able to run on other platforms.
As to those stream processors, it has already been shown that you can overload those and impact your video performance. There is no free lunch here. Obviously on a chip with hundreds of stream processors this is less of a problem.
You're right, it doesn't process run of the mill code and it has to be specific vector code but what you tend to find is that the only places you notice a computer struggling is when you have to do high end processing like image processing, rendering, encoding.
I do very little of that and yet can see my MBP bog down doing traditional things. Image processing is certainly an issue, mostly playback for me, but that is best done on dedicated hardware. The reason I want to see more cores is to speed up the machine when running multiple conventional apps.
As a side note a SSD would likely help me a lot, but that is another thread.
These apps tend to go pretty far with SSE optimization - even Apple's itunes encoder is SSE optimized and it blows away standard MP3 encoders like LAME.
This I don't disagree with but half to point out that not everybody uses their Mac that way. So if that is all the stream processors can be used for, that is vector processing, then they are of limited usefullness. Now the thing is they may not be that limited, but current development seems to be focused on floating point and vector codes.
Here is a test of image processing code with CUDA:http://www.tomshardware.com/reviews/...u,1954-12.html
Using the 8600M GT, same as in the Macbook Pro, they get the slowest GPU measure to be 6 times faster than the Core 2 Duo CPU and at best, 46 times faster. If you scale that, it means the GPU in this case will process this code faster than a quad Mac Pro CPU.
Which is great for that code. I guess my problem is that that is not another system process. It is GPU specific processing that requires a separate development effort for the code.
Now you're right it's specific vector code but developers of image processing apps like Photoshop, Motion, After Effects, Nuke, Mental Ray etc will use it in order to cut render times down dramatically and this will be a massive boost.
Yeah I know which would be great if I used such software.
If Apple drop some code into the Core APIs then the benefits will spread much further such as thumbnail generation of movie clips. We already saw the benefit moving from 10.4 to 10.5 in PDF rendering, which became hardware accelerated.
I fully expect this to happen too. It will do nothing for speeding up XCode though or any number of other apps.
If they use hardware to accelerate their itunes encoders (and by that I mean their Quicktime API) then people are going to notice. Same with ichat and screen sharing, it will process your video frames faster. All the places where it won't show any speedup, people probably wont mind as they are fast enough already.
People will notice. What I don't like is that you basically say where the GPU won't or can't speed things up it doesn't matter as it is fast enough already. This is certainly not the case at all. I don't buy this at all as there are to many apps that need a speed up now where I don't see GPU threads helping at all.