Questions on multitasking in Windows 7

Topic

New Reply

Ok multitasking is great in these modern times of multiprocessors and Win 7 management directives, but what about the times when multitasking is not a priority? When one basically wants one task to get done as quickly as possible and many times, using a program that is not very multi-processor aware?

Using affinity only seems to distribute the workload more evenly among the processors but does very little to actually increase the utilization of cpu time overall. Same lack of utilization when priority is given high or real time status (because its not competing with anything else of significance already).

So its the multitasking reserve that I’d like to adjust, make it 10% or less instead of up to 45% or so in some instances (program dependent?) with 30% seemingly the most common reserve approximation. Any way to do that on a dynamic or static basis?

For example when I’m using DVDFlick on my quad, its terrible in Win 7; processing is only 14 frames per second and as mentioned affinity and priority have little to no effect. If I boot to XP on the same system and use DVDFlick on the same file, 45 frames per second processed, no adjustments made, distribution to all 4 processors excellent and 2-10% reserve only.

That’s what I want from Win 7 if possible.

Reply | Quote

Replies

Leave processor affinity alone, set the app to high or realtime in “set priority”.
Do not change core os components priority settings, only installed apps.

It’s also possible that you have a specific program that is better coded to XP than Windows 7, give it another version update or two to catch up.

Reply | Quote

Leave processor affinity alone, set the app to high or realtime in “set priority”.
Do not change core os components priority settings, only installed apps.

It’s also possible that you have a specific program that is better coded to XP than Windows 7, git it another version update or two to catch up.

As I described, that does little to nothing since with nothing else of any significance running, there is nothing to prioritize over. I have 8 Win7 workstations presently and all with multi-core (haven’t tested the single cores yet) behave more or less the same with several different up to date programs (Studio 14, WinX HD Video Converter, VideoStudio X3, Picasa 3 making a video slideshow) in that they top out in the 60-80% CPU utilization during any rendering processes in Win7. DVDFlick on the Quad-core was just an example of the least utilization of them all.

I’m sure one of the things that makes Win 7 feel so “smooth” is that reserve of 20-40% for other tasks but it also makes it a “chugger” on long rendering projects when single-purposed. So my only good solution so far is that when the job (1) needs to get done, XP will get me right into the 90-98% utilization by default. I even wiped a perfectly good Win7 install and put XP on to verify the performance difference (again, with the understanding of the narrow focus of “performance” in this case where I’m just looking to scrape up all the CPU cycles I can for one program).

Reply | Quote

I also do rendering, in VirtualDubMod, but I have not noticed a slowdown. The opposite actually, I gain about 15-30 min
over XP, albeit my processor never rises above 25- 45% usage.
Intel Q9650@3GHz

So it may very well be the written code. VirtualDub is older than dirt, but it’s well written and well supported by a large community.

Reply | Quote

That doesn’t make a lot of sense to me since project processing time is project processing time and the comparison has to be on a project basis to eliminate other factors such as length and the amount of filter application. A platform may be inefficient or the program not be optimized to access CPU time on a consistent basis and show inverse efficiencies when measured against CPU time I suppose (but it would have to be some pretty awful inefficiencies!), but in general, a program with higher CPU access utilization must by default, process quicker.

One also has to be aware of other bottlenecks in the system that will affect CPU utilization as well such as right now I’m rendering a project at the same time I’m writing large files from an optical disc, to the same drive, so that drive and I/O channels are very busy and “starving” the CPU to the point that the poject is only able to get about 12-20% CPU usage where normally with this type of project and this program here in Win7 I would get about 58-64% (98% in XP). The time factor is appropriately reflected in the estimated time to completion of the project being some 14 hours (and rising at the moment). Of course as soon as the optical drive activity ceases, remaining time will be quickly projected downward as CPU usage for the project returns to normal for Win7.

Reply | Quote

CPU utilization is not necessarily the best barometer for measuring whether a program is running better under one OS than another. It is merely an indicator. CPU utilization can be affected by several factors such as:

Configuration – different page file sizing could cause more I/O to be done on a large file. When I/O is going on at some point the CPU is not being utilized.
OS differences – Coding differences in operating systems can make some processes more efficient. Memory utilization can be changed causing processing differences.
Underlying library differences – updates to various run time libraries can change processing characteristics.
Driver/BIOS updates – any updates that change the OS interface to the hardwarcan make a big difference in utilitization.

Other factors in the types of jobs you mentioned may be more significant – frames per second or straight wall clock time.

All that said, you need to do other debugging to accurately discover what is happening on the system. Analyzing performance issues is a time consuming pain in the rear.

Joe

--Joe

Reply | Quote

CPU utilization is not necessarily the best barometer for measuring whether a program is running better

Of course not but when paired with comparative job completion times it does give a very good directly comparable relationship to CPU utilization.

Configuration – different page file sizing could cause more I/O to be done on a large file. When I/O is going on at some point the CPU is not being utilized.
OS differences – Coding differences in operating systems can make some processes more efficient. Memory utilization can be changed causing processing differences.
Underlying library differences – updates to various run time libraries can change processing characteristics.
Driver/BIOS updates – any updates that change the OS interface to the hardwarcan make a big difference in utilitization.

Indeed, all those and more, but despite that, I don’t get the gut feeling that Win7 is coming up short in any of those departments. I still feel it is a straight OS management protocol that keeps more CPU utilization capacity in reserve for other tasks that may be started because once in a while I will get a video rendering project started, start a Levelator (levels audio) task and begin a audio conversion project, and I will get 98% CPU utilization overall…Also having tested it on so many different systems with different configurations, I would expect to find more variation if governed more by other factors affecting CPU utilization.

Its gotten me interested in testing the single core units now which up until now have been ignored because who wants to render on a single core these days right? Cuz it knocks hell out any mutitasking prospect for that computer while its underway (philosophical point drifting by) though the FX line of AMDs almost seem like they are eeking out some form of multitasking even at 100% CPU utilization for a project. How will Win 7 handle a rendering project under a single core? Hmmmm.

Reply | Quote

Have you tried running the program at realtime priority? Many OS processes run at a higher priority that user mode processes. I wonder if even more do with Win7. That could make it appear smoother without doing much work.

Joe

--Joe

Reply | Quote