Windows tweaks

[Jochenp]

superfetch or superprefetch?

[KoolDrew]

Cooper, the SuperPrefetch tweak does absolutely nothing. If you do not want to take my word for it how about Mark Russinovich?
http://bink.nu/Article4556.bink

You can even run the strings anaylisis yourself. The key is not even recognized by Windows.

I highly recommend changing some of the "tweaks" in the original post. I posted the truths about these "tweaks" and you still didn't change them. It is ok if you keep some things in there, but at least add a few comments. For example, mention for IOPageLockLimit that it does nothing in Windows XP (or anything newer then W2K SP1).

[Jochenp]

All head over to Xteme Overclocking -> Xtreme PI and await my benchmarks that will reveal the truth about tweaks!

[IFMU]

Wowza....
Interesting thread for sure.
KoolDrew? You seem like a very intelligent 16y/o, without a doubt. However, just becuase a book or MS says something, doesnt make it fact. Some things that are done by people on here should not work as far as some books and software makers are concerned, but they do. Real world actions sometimes work when it is thought they shouldnt.
If I may make a suggestion? Keep an open mind, dont dis-believe something right away, find real world proof to one way or the other beforehand.

And please, dont take this as rude or an attack, not meant as so, it just appears from your posts that you prefer to believe the software makers are infallable, regardless of the proof given otherwise.

All just my 2 corroded coppers.

[KoolDrew]

When the book "Windows Internals," several windows developers, and real-world tests back up my claims, it is a fact. The people that recommend "tweaks" have nothing to back them up except "my system feels faster." That has placebo effect written all over it.

Also, when it comes to "tweaks," it is safe to assume that it does nothing for real-world performance, as the majority don't.

[Jochenp]

Check my thread, and you will see some of your 'facts', don't match real-life FACTS
there are 'facts' and facts, and then there are FACTS
I don't want to bash you, but if I have to choose between believing people from Microsoft and SuperPi, my choice goes to SuperPi. (I mean, they say XP is the most stable OS there is )

[KoolDrew]

A SuperPI benchmark doesn't prove anything. I will use the LargeSystemCache tweak as an example. From your benchmark it makes a large difference, but it should not be used for regular use. What it does is change the balance in size of working sets between the OS, processes, and "free" RAM. LargeSystemCache set to 1 permits the OS working set to get bigger at the expense of the other two. It allows the file cache to use a lot MORE ram than processes do, whereas the default is that it competes fairly equally with them. With the "tweak" enabled the file cache will typically use up to 80% of RAM even when free RAM is scarce. This often causes massive amounts of paging of app code and data. This would not be good at all in real-world situations, even though some benchmarks may show an improvement. This is only a good idea if the machine is a server being ONLY used for very lightweight file serving purposes, like an FTP server or a regular old SMB file server. For nearly everyone else it's a terrible idea.

LargeSystemCache determines whether the system maintains a standard size or a large size file system cache, and influences how often the system writes changed pages to disk. Increasing the size of the file system cache generally improves file server performance, but it reduces the physical memory space available to applications and services. Similarly, writing system data less frequently minimizes use of the disk subsystem, but the changed pages occupy memory that might otherwise be used by applications. On workstations this increases paging and causes longer delays whenever you start a new app. Simply put enable this on a file server and disable it on everything else.
- Source

The thing that I find funny is that "tweakers" disable services in order to have more available RAM for their applications, then they recommend this tweak which does the exact opposite.

You can use whatever tweaks you want if it improves your benchmark scores, but I am telling you they won't help in real-world situations (e.g. gaming) and may actually hurt (LargeSystemCache). Just create a .reg file that enables the tweak before you benchmark if you really care about getting a better score. After the benchmark disable it again, as it is a terrible idea to use for normal usage.s

[Jochenp]

Well, it increased my aquamark03 score too, and I'm not noticing any drawbacks in cs:s, so i'll just leave it on, unless you have some hard prove (ie numbers) that it doesnt work

[KoolDrew]

Changing the LargeSystemCache entry to 1 does the same thing as selecting 'system cache' in the screenshot below.



You don't even have to know anything about how Windows functions to know that selecting 'programs' is the best idea for the majority of users. It is common sense.

[goreblast]

I shall add my 2 cents worth.

This thread was started to highlight some on the tweaks that various benchmark gods around the internet use to get that last bit of juice out of their systems. Be it in 3dmark or super pi. The guys that do the benching live by numbers. I have never seen someone saying that their machine feels faster with a certain tweak, everything and I mean everything is backed up with numbers (some might even say that there are too many numbers and screenies in some threads). Therefore, whilst you have read around the subject, if all you have is a Dell Dimension PC and dont benchmark how can you start to lecture to the rest of us here?

I am sure that a lot of what you are saying is correct for real world stable computing - we are not doing that! We are on the edge. Why else would any sane person by 2 or 3 fx-57 chips to get a good one? etc. Rather than trying to knock down all the excellent tweaks that have been stated at the start of this thread, you should either not post at all, or add some that you think will work.

Goreblast

PS: I suppose that Blackviper and TheElderGeek (to name but 2) are wrong when it comes to disabling windows services as well?

[Jochenp]

Blackviper actually doesnt know what he's saying most of the time

[goreblast]

That may be the case, but you know what I am getting at.

[KoolDrew]

I suppose that Blackviper and TheElderGeek (to name but 2) are wrong when it comes to disabling windows services as well?

Most of BV's advice is horribly inaccurate and/or totally unnecessary. Most of his tweaks do absolutely nothing performance-wise, and much more accurate and detailed information about securing services and Windows in general can be found at microsoft.com.

BV's horribly inaccurate definition of virtual memory pretty much sums up how little he knows.

Virtual Memory ~ The name used for the sum of Physical RAM and the Swap File. In other words: Physical RAM + Swap File = Virtual Memory. You cannot "disable" Virtual Memory even if you disable the Swap File. Meaning, 2 GB RAM + 0 MB Swap File = 2 GB Virtual Memory.

Also, if you look at ElderGeeks site he even says this:

Frankly, I don't recommend you change any of the services from the default installation of XP.

The fact is that an unused service will use no CPU time and its memory will be reclaimed as needed. There is no real-world performance benefit from disabling services from the default installation.

Also, I was correcting some of the tweaks that were posted in the original post. If they increase benchmarks socres, fine, but the descriptions about what they do should at least be correct. The ones that do absolutely nothing such as IOPageLockLimit should just be removed from this topic. The key is not even recognized by Windows past W2K SP1.

[scarface_what]

i cant find memory manager under session manager

[Jochenp]

it's memory management

[OmegaMerc]

Question, I followed the steps onto making all services "manual" which resulting in only me having 16 process's. The only thing is that now my windows GUI looks like Windows 98. Also, windows now wants me to re-activate itself wtf? (got to do it through telephone too). Anyone can help?

[Jochenp]

Well, I hope you made a restore point?
When doing these tweaks you should disable one service after the other, reboot, work with it for a day or 2 and see if there's missing anything you need.
If there isn't, go on to the next service.
That way you won't end up getting such supprises

[madgravity34]

Omega, making xp look like 98 isnt bad, its what a lot of tweakers do. Its just because you ended the themes service in windows. I do it on both my bench harddrives so its no big deal. As to the windows activation problem, I have never been hit with that one. Just keep in mind that most of the people who spend a lot of time getting rid of services do not use thier systems 24/7, rather, they are only meant to bench and report scores and therefore, if you are using your system 24/7, I would recommend not ending any of the services unless you are pretty careful. Personally, I have all my services disable except 5 and use my system all the time, but then again, I only bench, tweak, overclock, and surf the net on my system at home. I have others for school work, listening to music, and gaming.

[OmegaMerc]

Reactivated systems to default and disabled some I already know I dont need, so now im back up to 23 which isn't a big deal considering system specs. After I re-actiavted my theme service I was able to reactivate windows xp without the use of the phone (yeah i know, wtf?). Anyways this is my 24/7 system, but im waiting arrival of my HTPC parts to make that my 24/7 system and this my gaming / work.

[KoolDrew]

Before you do any "tweaking" you should know what you are doing. Research what each service does before messing around with it. Then you can make an educated decisions as to whether or not you want to disable it.

The advice given in the original post telling you to set every single one to manual is terrible.

[Twin]

@KoolDrew: I trust you here, you're backing up your facts very well, even tho I am certain some things "just work better" anyway due to theorists (grammar? word?) always think detailed and divide each tweak into it's own use and function, and I bet 2 random tweaks can perform very well together as they weren't thought out to to.

BUT, I'd very much like something else from you, than just telling what's not working, and that's what really IS working.
I bet you only change 2-3 things after a clean xp sp2 install, and what would those things be? Please make a mini xp tweak guide, ppl here can laugh or applaud you or whatever, it's for my personal use.

@Others; flame away, I won't bother.

[KoolDrew]

I change a lot of things when I install Windows. However, this is not for performance reasons. It is for personal preference. Some things may increase performance like using the Windows Classic look, but I don't do this for performance. I do it because I prefer it.

The fact is that most of these "tweaks" you run across on the internet don't increase performance. Instead of blindly following a tweak guide you should do research to make sure you know what everything does before applying them. Then you can decide what you prefer.

[Cooper]

Followed one tweak:

Windows XP Services

Here is a way to figure out which services to set as Automatic and which to set as manual or disabled.

Step1: Set ALL your services to *manual* setting.
Step2: Reboot computer and wait for windows XP to load. XP will boot pretty slowly, since it has to turn on each service
separately. Use your computer for a bit, doing what you normally do during a computing session. This will allow any
other services to activate (such as DHCP or other services that didn't load during the boot process).
Step3: Go back to the list of services, and see which ones are Started. Change the services that are marked as Started, to
*Automatic*. This way windows will automatically load all of the services that you normally use, and the ones that
you don't use will stay unloaded.

Tryed it. And after reboot I couldn`t play any music - audio drivers were disabled. Also my system locked several times w/o any notice in events viewer or any issues with HW.

Please be aware what services you turn off !!!

SuperFetch tweak enables one great Vista feature. Read more here

[Newblar]

Without Virtual Memory Windows would not function You are referring to the pagefile which is not the same thing. Setting the PF size to the same amount of RAM you have is not a good idea. For the majority of the users it is best to leave the PF system managed, but to find the optimal size you need to run your most intensive applications like you would in a normal day or w/e. Then measure PF usage using perfmon. This is the only way to measure Pf usage using built-in tools. If you want something else this will work.

Once you have figured your PF usage multiply this number by 4. That number should be set as your initial size. For the max multiply the number you just calculated by at least 2.

Many people will recommend setting a "fixed" pagefile, and the reason they do this is because they want to prevent fragmentation. With my recommendation the PF will never have to resize anyway because the initial size is high enough for your needs. However if it does come a time where it needs to resize it can. Setting a fixed size removes this "safety net."

Also you should never disable the pagefile. Many people will also recommend this to improve performance, but all this will do is hurt you. First of all you have to realize the pagefile is NOT the only file involved with paging. All exe's and dl''s are also. So when you do disable the pagefile you have to keep all "private" data in RAM and only code and mapped files can be paged. So even if the "private" stuff has not been touched in hours. So this will cause more paging of code. It also means that paging cannot be correctly balanced. It will also cripple the file cache and slow down code execution. Also not to mention some applications will fail without a pagefile.

Also in theory moving any file to the outer cylinders of the drive will help, but in real-world performance it will not. This only makes a difference in sustained transfer rate and the PF IO is never in buffers of more than 64KB at any given time. This is basically the same reason why fragmentation is taken way too seriously by a lot of people. When a executable is read so are all the libraries it needs. Of course loading all of it in RAM would be a waste of memory so it only loads the immediately necessary parts. This is called demand paging. So regardless if the file is fragmented or not the head will have to move anyway, paging a bit here and there. Seek time is everything when it comes to HDD performance so you want your most frequently-accessed stuff together. So the pagefile should be on the most-used partition and the least-used HDD. This will give you best performance. However whether you will benefit from it not will depend on the amount of RAM you have and your usage patterns.

I multiply the amount of my ram x 1.5 to get the initinal pf which for me in 720. I keep the maximum pf at the same amount too. I have integrated video so 512 - 32 = 480 x 1.5 = 720

I dont think your 16.. -_-

[Jochenp]

Why not? Can't people of 16 know something?