In any case, I haven't heard of such incidents wrt/ the MacBook or MacBook Pro. Personally, my MacBook Pro (advertised as 2 GHz) runs anywhere between 1 and 2 GHz, just as expected.
Thats not really how it works, is it? The frequency at which the processor runs doesn't scale based on the load you put on the chip (meaning you have to get a 100% load to be running at 2GHz, and a 50% load is running at 1GHz). Modern day processors can scale back the frequency when idle, but it is not a direct relation to the % of CPU being used.
That % from top, is just a figure of time that each processes are using the CPU, not what % of the CPU is being utilized. So if a process uses 100% of the CPU, but only for a few microseconds, because the rest of the time the processor was idle, the % wont hardly budge. And within most programs, so little is being done, because it is blocking for i/o (user, disk, memory, etc), that makes for a lot of idle time. There are some programs that can stream their data in nicely, and can use high percentages of time of the CPU. But that is not to say that those processes are getting more GHz than a process utilizing 30% of the CPU time.
Is there a way for me to "assign" WoW to another processor core?
OS X does all of this automatically. There are tons of processes running all the time on your machine. Even if all you have open is WOW and the Finder, there are probably 20-30 other processes (each with multitudes of threads) all running (or sleeping, depending on their functions). The OS handles passing off the processes to the cores, and does its best to make for maximum utilization of your available resources.
OS X does all of this automatically. There are tons of processes running all the time on your machine. Even if all you have open is WOW and the Finder, there are probably 20-30 other processes (each with multitudes of threads) all running (or sleeping, depending on their functions). The OS handles passing off the processes to the cores, and does its best to make for maximum utilization of your available resources.
All multiprocessor operating systems do this automatically, but others do allow the user to assign a specific task to specific processors. This can be useful for performance because otherwise the kernel scheduler would just assign the load equally between the processors. Assigning a processor affinity to a task would reduce the amount of cache thrashing.
I push it hard plenty, it never goes beyond 100% though.
Leopard will hopefully change this. It's rumored to be better at threading.
Most programs just use one core and at that, the most you should get is 2GHz. Using both cores, it would go up to nearer 4GHz. This is why the Intels are benching so close to the G5s, even the quad.
Programs that are multi-threaded tend to use two processors so a quad mostly goes at 2x2.5 ~ 5GHz and the Intels at 2x2Ghz ~ 4GHz and as mentioned, you can't really go by clock speed alone even though they are a rough guide if the machines are built well enough, so they come pretty close.
The best way to push the machine is to learn how to multitask. If you need to rip/burn a DVD then don't wait until you've finished what you are doing. Set it to work in the background and it will use one core while you work with another.
I ripped all my Family Guy episodes to ipod video fromat on the quad G5 and I did 3 at a time while I was doing my office work because quicktime only uses about one processor. I probably should've used isquint or something but I didn't know about it at the time.
Leopard will hopefully change this. It's rumored to be better at threading.
Most programs just use one core and at that, the most you should get is 2GHz. Using both cores, it would go up to nearer 4GHz. This is why the Intels are benching so close to the G5s, even the quad.
It's pretty unfortunate that many of those benchmarks don't mention the CPU load, I mean, what you see is what you suggested, that the software in question isn't smart enough to spread the load over more processors. This doesn't have much to do with the OS right now, most of the load is computation by the software in question. I don't think there is any OS that can break up a thread to spread around the load, that has to be programmed in.
I don't think there is any OS that can break up a thread to spread around the load, that has to be programmed in.
Right. All that multiprocessing OSes like OS X can do is use load-balancing to put other threads on the other processor(s). This keeps the busy process from having to get swapped in and out every 10 milliseconds as it would be if there were only 1 processor in the system. When the timeslice is up, the OS can see that there aren't any threads waiting in the run queue for that busy processor, or if there are, they can be moved to the other processor's queue, and leave the busy process to take the next timeslice without the overhead of a process switch.
When they get into an infinite loop, I can use my other processor to end that process with 't3h quickness'
great for php programming while drunk
CHEERS!
I doubt this is true; these aren't OS 9 days anymore. The kernel will manage processes and share time between them; unless your php program is preempting the kernel (and it isn't), you can always kill the infinite looping program, independent of # of processors.
I doubt this is true; these aren't OS 9 days anymore. The kernel will manage processes and share time between them; unless your php program is preempting the kernel (and it isn't), you can always kill the infinite looping program, independent of # of processors.
This depends whether or not the infinite loop process is holding a kernel funnel lock or not. If it is you are generally screwed. Luckily this doesn't happen often because most of the boneheaded software isn't running as a kernel extension/driver.
Comments
Originally posted by Chucker
In any case, I haven't heard of such incidents wrt/ the MacBook or MacBook Pro. Personally, my MacBook Pro (advertised as 2 GHz) runs anywhere between 1 and 2 GHz, just as expected.
Thats not really how it works, is it? The frequency at which the processor runs doesn't scale based on the load you put on the chip (meaning you have to get a 100% load to be running at 2GHz, and a 50% load is running at 1GHz). Modern day processors can scale back the frequency when idle, but it is not a direct relation to the % of CPU being used.
That % from top, is just a figure of time that each processes are using the CPU, not what % of the CPU is being utilized. So if a process uses 100% of the CPU, but only for a few microseconds, because the rest of the time the processor was idle, the % wont hardly budge. And within most programs, so little is being done, because it is blocking for i/o (user, disk, memory, etc), that makes for a lot of idle time. There are some programs that can stream their data in nicely, and can use high percentages of time of the CPU. But that is not to say that those processes are getting more GHz than a process utilizing 30% of the CPU time.
Originally posted by Orionetheus
Is there a way for me to "assign" WoW to another processor core?
OS X does all of this automatically. There are tons of processes running all the time on your machine. Even if all you have open is WOW and the Finder, there are probably 20-30 other processes (each with multitudes of threads) all running (or sleeping, depending on their functions). The OS handles passing off the processes to the cores, and does its best to make for maximum utilization of your available resources.
Originally posted by kupan787
OS X does all of this automatically. There are tons of processes running all the time on your machine. Even if all you have open is WOW and the Finder, there are probably 20-30 other processes (each with multitudes of threads) all running (or sleeping, depending on their functions). The OS handles passing off the processes to the cores, and does its best to make for maximum utilization of your available resources.
All multiprocessor operating systems do this automatically, but others do allow the user to assign a specific task to specific processors. This can be useful for performance because otherwise the kernel scheduler would just assign the load equally between the processors. Assigning a processor affinity to a task would reduce the amount of cache thrashing.
Originally posted by Orionetheus
I push it hard plenty, it never goes beyond 100% though.
Leopard will hopefully change this. It's rumored to be better at threading.
Most programs just use one core and at that, the most you should get is 2GHz. Using both cores, it would go up to nearer 4GHz. This is why the Intels are benching so close to the G5s, even the quad.
Programs that are multi-threaded tend to use two processors so a quad mostly goes at 2x2.5 ~ 5GHz and the Intels at 2x2Ghz ~ 4GHz and as mentioned, you can't really go by clock speed alone even though they are a rough guide if the machines are built well enough, so they come pretty close.
The best way to push the machine is to learn how to multitask. If you need to rip/burn a DVD then don't wait until you've finished what you are doing. Set it to work in the background and it will use one core while you work with another.
I ripped all my Family Guy episodes to ipod video fromat on the quad G5 and I did 3 at a time while I was doing my office work because quicktime only uses about one processor. I probably should've used isquint or something but I didn't know about it at the time.
Originally posted by Marvin
Leopard will hopefully change this. It's rumored to be better at threading.
Most programs just use one core and at that, the most you should get is 2GHz. Using both cores, it would go up to nearer 4GHz. This is why the Intels are benching so close to the G5s, even the quad.
It's pretty unfortunate that many of those benchmarks don't mention the CPU load, I mean, what you see is what you suggested, that the software in question isn't smart enough to spread the load over more processors. This doesn't have much to do with the OS right now, most of the load is computation by the software in question. I don't think there is any OS that can break up a thread to spread around the load, that has to be programmed in.
When they get into an infinite loop, I can use my other processor to end that process with 't3h quickness'
great for php programming while drunk
CHEERS!
I don't think there is any OS that can break up a thread to spread around the load, that has to be programmed in.
Right. All that multiprocessing OSes like OS X can do is use load-balancing to put other threads on the other processor(s). This keeps the busy process from having to get swapped in and out every 10 milliseconds as it would be if there were only 1 processor in the system. When the timeslice is up, the OS can see that there aren't any threads waiting in the run queue for that busy processor, or if there are, they can be moved to the other processor's queue, and leave the busy process to take the next timeslice without the overhead of a process switch.
Originally posted by slughead
Thank God for single processor programs.
When they get into an infinite loop, I can use my other processor to end that process with 't3h quickness'
great for php programming while drunk
CHEERS!
I doubt this is true; these aren't OS 9 days anymore. The kernel will manage processes and share time between them; unless your php program is preempting the kernel (and it isn't), you can always kill the infinite looping program, independent of # of processors.
Originally posted by Orionetheus
I push it hard plenty, it never goes beyond 100% though.
No, what you think is pushing it isn't. Very few people actually push their computers much anymore, despite what their ego's say.
Originally posted by Orionetheus
Is there a way for me to "assign" WoW to another processor core?
No.
Originally posted by chych
I doubt this is true; these aren't OS 9 days anymore. The kernel will manage processes and share time between them; unless your php program is preempting the kernel (and it isn't), you can always kill the infinite looping program, independent of # of processors.
This depends whether or not the infinite loop process is holding a kernel funnel lock or not. If it is you are generally screwed. Luckily this doesn't happen often because most of the boneheaded software isn't running as a kernel extension/driver.