CPU usage Vs page file usage;When is your processor the

[wantathinkpad]

Hello all.

I had a somewhat random question. How does one know which is the computer bottle neck?

I do have an 5 year old laptop so I am not asking for a miracle. Just wondering is there anyway I can get or make my computer services use more RAM then cpu?

In what ways, or after how many years does the processor become the bottle neck vs the RAM. (You can always add RAM can't add a processor?)

I hope this makes sense. If you need more info ask?

[hunterman223]

In my experiences with desktops and laptops, the ram is almost never the bottleneck, because it usually operates at a much higher speed than everything else. What do you have for a cpu? The hard drive could be a major bottleneck as well, if you want to upgrade that I suggest either a 7200rpm drive,(unless battery life is a concern) or a Samsung HM160HC drive, which is probably the fastest available ATA 5400rpm drive for laptops. Hope this helps. Also, how much ram does your system have? If you have plenty there is a way to disable the page file completely, or use it conservatively, thus reducing the load on the hard drive some.

[wantathinkpad]

Thanks for replying. It was a question out of curiosity. I have an 5 year T43 with an Intel Pentium M 760 2 GHz Cache Memory: 2 MB - L2 cache RAM. I have 1.5GB of ram. The computer is fast enough for want I want to do. The 54000 RPm is about 40% full.
Just when I open alot of firefox windows the CPU usage goes up but everything else stay the same. So I was somewhat curious of what it is. Because the laptop can hold 2 GB or RAM but it is clear that won't help. But like I said I it is 5 years old running well and fast enough for what I want to do. Good to know. I remember when I signed up for this I was told always buy the faster processor, everything else can be changed. I believe this was the fastest for the T43 so no regrets.

[dr_st]

Get some sort of flash blocker for Firefox and you'll be surprised how much CPU usage will go down.

[hunterman223]

Yeah, the bottleneck is definitely not the cpu. That seems like a very nice machine, but with 1.5GBs of memory, you can safely disable the pagefile. To do this, go to control panel -> System. Under the advanced tab, click performance settings. Then, under the advanced tab, click the change button. On your system partition letter(C:), check disable, and click set. Ignore the warning, click ok. Now under startup and recovery, click settings. Make sure "automatically restart" is disabled, click ok, and restart. You should notice a decent performance boost after restarting.

[dr_st]

The "disabling pagefile improves performance" is a fairly old myth.

[hunterman223]

Well it makes since. Juggling stored memory between the slower hard drive, and the faster memory, it seems like that would be a major performance hit. I disable it on all my computers.

[craigmontHunter]

Well it makes since. Juggling stored memory between the slower hard drive, and the faster memory, it seems like that would be a major performance hit. I disable it on all my computers.

I seem to remember - and I may be wrong - that the computer only uses the page file for files and information that it does not need right then, and background services - right now I have several programs running, and my memory usage is at 88mb free out of 256mb. There only seems to be (as far as I can see) 100mb of page file used. This is windows 2000, but the only slowdowns I have are when Firefox loads down the CPU (thank God for Adblock plus) - turning off the page file will not affect performance, but I think it does affect battery life - the computer does not need to access the hard drive when the computer is Idle to run a program?

[Navck]

CPU goes first with 7zip and Freearc while free memory goes down until I hit the pagefile with 3-4 instances running....

Then again, this is from someone who has CPU bottleneck before HDD bottlenecks...

... Then again I really don't have issues when I have to go to the pagefile even with 3GB of RAM on my desktop (No noticeable slowness during multitasking.)

[hunterman223]

This article explains it pretty well. Or try this if you're still skeptical about disabling it completely.

[AMATX]

This article explains it pretty well. Or try this if you're still skeptical about disabling it completely.

Nice. However, I'd recommend you go with larger than 1G for the memory threshold, as in today's environments, 1G is kinda on the small side. MSFT and the pc manufacturers are somewhat known for adding bloatware as time goes on.

In terms of tuning your system, all tuning does is change what you wait on. CPU, memory, disk I/O...take your choice, you'll end up waiting on one or more. The trick, of course, is to eliminate each one, going from one to another, until you reach a point where you've pruned back on all of 'em as much as your setup will allow.

To most effectively do this, you'll need some kind of performance monitor, even if you use the Win default one. Usually easy to see if you're pegging the box w/cpu, so start there. Memory/disk after that.

Easy bang for the buck: 1) add RAM 2) add an SSD

Usually, cpu isn't the bottleneck, cuz the drive is so slow, relative to cpu & memory. If you're drive chugs a bunch, could be a sign memory is constrained. If on WinXP, consider going to Win7, 64 bit, so you can use > 3G of memory.

Any rate, get friendly with the performance monitor, repeat slow scenarios -exactly-, changing -one- thing at a time. Observe result, make adjustment(s). Repeat.

[dr_st]

This article explains it pretty well.

The article was clearly written by someone who has no idea how Windows memory management works. When I see a sentence like "Reading and writing to the paging file requires multiple steps and that takes time" I understand that the person has no concept of what virtual memory really is.

All memory access in Windows uses virtual memory. There is no way you can change it, and while some people think that by disabling the pagefile they will eliminate virtual memory, they are dead wrong. Processes have no normal way of directly accessing physical memory, and thank God for that (well, more accurately, thank the engineers that came up with the concept). It would be very easy to corrupt something, and a lot of it, if it was possible to mistakenly or maliciously access physical RAM directly.

Now what is true is that at any given time, virtual memory may point to either physical RAM or a section of the page file. When there is no page file, it has to point to the physical RAM. The fact that RAM access is faster than disk access is obvious. The assumption that because a pagefile exists, Windows will start using it instead of RAM, all the time, is preposterous.

However, it is also not true (as some claim) that Windows will only use the paging file when it is out of physical memory. The memory management algorithms are more complicated than that, and they tend to page out some blocks, in favor of freeing more physical memory (for processes or cache). To do so effectively they use various heuristics, which take into account how frequently that data has been used, how much total memory the application has requested etc. The goal it to optimize overall system performance, and not just the performance of a single application.

In the end, the algorithms work based on some predictions, educated guesses if you will. And guesses, as educated as they may be, can sometimes be wrong. However, to assume that you can outsmart all these algorithms by turning off the mechanism altogether is a bit naive.

I am by no means an expert on Windows internals, but here's a guy who is: Mark Russinovich, from Sysinternals in the past, and Microsoft at present. Definitely one of the top guys in the world when it comes to understanding Windows and how it works.

Here's his comprehensive guide on virtual memory. Read more specifically the section "How Big Should I Make the Paging File?".

In a nutshell what he says on the topic can be summarized in two points:

* If you know that the peak memory commit you will require will never exceed the amount of physical RAM, you don't need a paging file (although you may want to have one to store crash dumps).
* Improved performance when disabling the paging file may occur in some workloads, but generally all it will do is reduce the amount of available memory.

The suggestion that disabling pagefile improves performance has been circulating the web for quite a few years, but I have yet to see a single well-documented and benchmarked case which shows it to be true for general system performance (you may find some bits of info where people claim, usually even without proof, that it helped some particular task). If you have data to prove it, I will be happy to see it.

The conclusion I draw from all this:

If your usage patterns do not include anything for which a pagefile is required, feel free to disable it. It won't hurt. You may win somewhere and lose somewhere else. But please stop spreading the myth of a "great performance boost" that one will get by doing so.

[AMATX]

^ Good writeup. After all this(!), if you -still- want to fiddle with the pagefile thing(which, as mentioned above, isn't all that likely to help much), then try starting out on the small side. Try a very small pagefile, observe the result. You may find that the small pagefile is not exercised all that much, therefore keeping disk I/O and paging algorithms at a minimum, and you can get most/all of the RAM benefit you originally wanted.

As mentioned in above posts(and is true for all opsys that I'm familiar with), sometimes the pagefile is used to store infrequently used system routines. As such, it really doesn't matter if there's some code/data in the pagefile, if it will not be paged in or out very much. What really matters is that the working set of the app(s) reside in RAM, for immediate use. System code and app code to be used need to be in RAM; everything else is largely irrelevant.

In the pc/Win environment, here's a priority order to consider:

* active app/user pages from active users & apps and necessary system routines reside in RAM for immediate usage.

* IN-active app/system pages from -active- users & apps can reside in the next fastest arena(say, a pagefile on a fast SSD drive), as they'll only cause a momentary blip when retrieved occasionally.

* Pages from IN-active users and IN-active apps can reside in even slower storage(say, a typical hard drive), as these won't affect performance at all until the user/app becomes active in the future. At that point, there will be a slight pause while all of these pages are brought back into the fold, but that's a one-time deal, as now those pages will migrate between RAM and SSD, if necessary.

While it's easy to get worked up over pagefiles and whether you need them, etc., what really matters is what's going on with active code. If you have a paging problem, it's usually -very- noticeable, and easily fixable if you can add some more RAM.

[hunterman223]

The article was clearly written by someone who has no idea how Windows memory management works.

The article itself was written for people who have no idea how windows memory management works. If you did know, why bother reading an article?

All memory access in Windows uses virtual memory.

While that is quite true, it will definitely put less stress on the disk to disable memory paging between the memory and the page file. I have not seen a decent writeup either, but this is what I have always done to all of my systems with the available memory, and to me it seems to work. I was not saying that it is the ultimate answer to boost performance. I suggested it in the hopes that since it has always worked for me, maybe it would work for him as well.

[dr_st]

The article itself was written for people who have no idea how windows memory management works. If you did know, why bother reading an article?

So that I can comment on it. How would I comment on something I never read?

While that is quite true, it will definitely put less stress on the disk to disable memory paging between the memory and the page file.

Put less stress on the disk? Sure, because it will be used less. But since when is this the topic? We were talking about performance.

I have not seen a decent writeup either, but this is what I have always done to all of my systems with the available memory, and to me it seems to work. I was not saying that it is the ultimate answer to boost performance. I suggested it in the hopes that since it has always worked for me, maybe it would work for him as well.

How would you know that it "works" for you, if you always do it for all of your systems? Did you actually compare the performance of your system with and without the paging file? Using actual, quantifiable measurements, not just "to me it feels snappier"?

Like I said, I see nothing wrong with disabling the paging file if your usage does not require it. I just haven't seen a single shred of evidence that it will actually do anything good to the performance of your system (except of course the extra few gigs of space on the hard drive).

[hunterman223]

What I have been trying to say the whole time is, an ide hard disk transfers data at a much slower rate than memory. So therefore, juggling between memory and hard disk doesn't seem very performance wise. Obviously if I disable the page file on all of my systems I must have noticed a decent performance boost. Otherwise, why would I bother. I have not seen a decent article that states "Disabling the Page File Improves Performance!", but common sense and a slight knowledge of Windows and it's memory management techniques tells me that it would.

[dr_st]

What you attribute to common sense and a slight knowledge of Windows memory management I attribute to lack of understanding of Windows memory management, and the placebo effect.

Your only claims, which you repeat over and over seem to be "hard disk is slower than memory, therefore disabling the paging file will boost performance", which is a logical fallacy, and "I've noticed a decent performance boost on all my systems", which is unverifiable and isn't backed up by any empirical data.

You obviously haven't bothered to read into the link I gave and understand it, or even into the comments of myself and others in this thread. I see no point in discussing it any further. We'll just stay in the same spot.

[hunterman223]

We'll just stay in the same spot.

Exactly. I am done trying to explain myself because you obviously will not accept anything other than hard data evidence. Sometimes it is neither possible or necessary to provide that. If you reread the first post, it really doesn't sound like he wanted a the cold hard facts or a huge discussion. It sounds to me like he wanted a quick, simple answer. I suggested to him something that has always worked for me, in the hopes that it might work for him as well. It is almost impossible to measure a speed boost of this manor, because people use their systems in different ways. I don't use mine for a lot of memory intensive applications. My main system has 2gb of memory, and with the page file disabled I still manage to do some light gaming, listen to music, do some programming and website designing, browse the web, and watch movies, never running out of memory. But your mileage may vary. Hopefully now you will understand without me providing solid values or evidence. And again, I also provided an alternative to disabling it completely. Here is the link. That tweak simply tells windows that whenever possible, use memory instead of the page file. I never stated that either of the tweaks that I provided would completely disable virtual memory. I realize that is impossible. Judging your last post, you must take me for a moron or something, and apparently a psychological one. I would appreciate in the future that you keep your opinions and your diagnoses to yourself. I never judged you, so please do not judge me.

[JaneL]

OK, that's enough. There's no reason to start calling people names. Locking thread.