Yes there is a lot of assumption grand central will just make everything multi-threaded and that isn't the case. Part of it however is making it easier for developers to make their code multi-threaded.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2009 Mac Pro in hand - unboxing etc.
- Thread starter WonderSausage
- Start date
- Sort by reaction score
Yes there is a lot of assumption grand central will just make everything multi-threaded and that isn't the case. Part of it however is making it easier for developers to make their code multi-threaded.
!!??
I know this is off topic, but I sure hope this statement is wrong. Simple applications dont' need the power of multiple cores - it's exactly programs like photoshop that do!
Well, Photoshop is a simple application in the sense that its threads could well use some combing. The amount of elbow grease required to do that however may currently be prohibitive.
OTOH there are computations (some of them very simple to program) that no amount of technology will ever multithread.
The distinction has nothing to do with pro vs amateur or consumer.
Well I went looking on the forum to verify what I remembered and here are the quotes from an Adobe engineer:
"Grand Central is designed for smaller applications with simple data flow. It's sort of the beginners guide to multiprocessor programming, and not designed to handle professional applications. This was made quite clear at their WWDC presentations."
"Photoshop already splits what it can among the available processors. But not all tasks are splittable that way, not all of them will speed up (many would slow down), and there are higher level issues to deal with if you really want to make the best use of additional processors."
Jim
"Grand Central is designed for smaller applications with simple data flow. It's sort of the beginners guide to multiprocessor programming, and not designed to handle professional applications. This was made quite clear at their WWDC presentations."
"Photoshop already splits what it can among the available processors. But not all tasks are splittable that way, not all of them will speed up (many would slow down), and there are higher level issues to deal with if you really want to make the best use of additional processors."
Jim
2x2.26 GB scores up:
http://browse.geekbench.ca/geekbench2/chart/115594
EDIT -- also the quad-core 2.66
http://browse.geekbench.ca/geekbench2/chart/115594
EDIT -- also the quad-core 2.66
After Effects CS4
In addition to the great Cinebench and Geekbench results, I think it would be instructive to see Adobe After Effects CS4 render results since it uses all available cores in "multiprocessing" mode.
You can download a full working Trial version from the Adobe site. And there is a sample project called Total Benchmark that we use with it. If some of you are willing to run it, I'll provide links and instructions here.
In addition to the great Cinebench and Geekbench results, I think it would be instructive to see Adobe After Effects CS4 render results since it uses all available cores in "multiprocessing" mode.
You can download a full working Trial version from the Adobe site. And there is a sample project called Total Benchmark that we use with it. If some of you are willing to run it, I'll provide links and instructions here.
It must be reporting something wrong unless I'm really misreading. The X7460 is a 6 core processor and the IBM 3850 M2 only supports 4 of them.
Seriously?
I bought my 3.2GHz mac pro a month and a half ago in January and am anything but jealous. Sure, the top end 2009 mac pro is insane but so is the price - double what I paid for my 3.2GHz. The 2.66 is over £1k more expensive and the 2.26GHz is either on par or slower than my 3.2GHz.
Delighted is the word I'm looking for
There is something I don't understand when I see the GB score for memory perf... look :
Quad 2,66 (2009) : 4407/4055 - QPI 4,8 GHZ
Octo 2,26 (2009) : 3061/3019 - QPI 5,8 GHz
Octo 2,93 (2009) : 5075/4636 - QPI 6,4 GHz
There is something wrong with the 2,26, normally it should be around 4500/4500
or maybe the Octo 2,26 (low GHz) can't go up in bandwidth ...
Oh and one big difference between the Core i7 and the Xeon 3500/5500 is the floating point performance.
It seems that the FP is much higher on the later. Dell Studio XPS 435MT and Quad 2,66 w3520
Quad 2,66 (2009) : 4407/4055 - QPI 4,8 GHZ
Octo 2,26 (2009) : 3061/3019 - QPI 5,8 GHz
Octo 2,93 (2009) : 5075/4636 - QPI 6,4 GHz
There is something wrong with the 2,26, normally it should be around 4500/4500
or maybe the Octo 2,26 (low GHz) can't go up in bandwidth ...
Oh and one big difference between the Core i7 and the Xeon 3500/5500 is the floating point performance.
It seems that the FP is much higher on the later. Dell Studio XPS 435MT and Quad 2,66 w3520
Makes me wonder how much faster mine would have been.
Only marginally, I'm sure. I had your exact specs plus a GT 120.
What happened to your order? I thought you were gung ho to get the 2.93 octa?
I saw some Nehalem Geekbench Scores on the Geekbench Browser but some of them are 32 bit scores. (search on "MacPro4,1")
You can't compare 32 bit runs to 64 bit runs. Plus the 64 bit version squeezes out the maximum performance from the Nehalem.
If you send me a private email, I can tell you how to get a 64 bit score if you can't afford to buy a Geekbench 64 bit license -- though for what you paid for the Nehalem, you can afford it. ;-)
Warning: this is the same tactics dishonest mortgage brokers use(d)
Now, Mr. barefeats, how's that MP review coming along?
I wonder if the quotee actually gave an example of something PS does that is not amenable to multithreading—I would venture an educated guess that they did not. (And yes, I realize that splitting the Gaussian blur of a 2Mp image into 10000 threads, that is, for 5000 cores, may indeed result in a slowdown.) In fact, I hope that Adobe will be demonstrating the second quoted paragraph increasingly inaccurate.
By "higher level issues" I think the quotee meant the task of isolating and supplying the data to the right thread, and that multithreading neighbourhood-sensitive filters requires a sound algorithm—you cannot expect a general-purpose multitasking API to take care of the problem for you automagically.
Yes, I did read what you wrote. Did you forget what you wrote?
You said that Grand Central makes it easier to use multiple cores. That is only part of the story.
You did not include the part about Grand Central greatly improving the organizing, assigning, and scheduling of tasks amongst available cores. That is where there will be a gain for applications that already support hyperthreading and multiple cores.
S-
Pity...
Oh well, my x4450 benchmark has been bested. I guess I'll have to try one of the new 6 core x4450's one of these days or fork out the money for the 64-bit version.
FWIW, it was up there for quite a while.
Oh well, my x4450 benchmark has been bested. I guess I'll have to try one of the new 6 core x4450's one of these days or fork out the money for the 64-bit version.
FWIW, it was up there for quite a while.
Can someone post the 2.93 quad core benchmarks. I'm curious how it compares to the eight core units.
Please advise
I just received my 2 x 2.93-GHz quad-core and we are setting up my rig, but it smells awful, like something is going bad. I think it comes from the power supply (keeping my fingers crossed). Has anyone else noticed the bad odor?
I just received my 2 x 2.93-GHz quad-core and we are setting up my rig, but it smells awful, like something is going bad. I think it comes from the power supply (keeping my fingers crossed). Has anyone else noticed the bad odor?
Making no sense....? You lost me...
Looks like from the chart, my 2007 3.0GHz Octo smacked that basic 2.26GHz Octo. Unfortunately the same thing can't be said when it comes to memory bandwidth on the real-world test...
The smell of fresh electronics, awesome!
I'm guessing you mean a somewhat "dry" smell that's best described as electrical and metal.
Of course, there's the odd chance your MP flew over on the same flight as a couple of stressed out Jaks in mating season...