Jump to content

Windows XP and Pentium 3.2GHz Extreme Edition


Yannis Cheras

Recommended Posts

I have the following system:

MSI 945GCM478 Motherboard (Socket 478)

Intel 945 express chipset

Realtek PCI-E HD Audio chipset (onboard)

Realtek PCI-E Gigabit Ethernet Adapter (onboard)

Asus EN9400 PCI-E with 1GB Memory on card

3GB DDR2 667GHz CL5

4 SATA HDD (2x320GB and 2x500GB all Seagate)

2 IDE ATAPI DVD-RW LG

Pentium 4 3.2GHz Extreme Edition (HT, L2 512KB, L3 2MB)

My problem is that when i replace the previous 2.8GHz HT L2 1MB (Prescot) suprisingly i see a decrease in performance. For example when i watch a movie in my PC or when i play the Galactic Civilizations II Ultimate Edition it's necessary to set both to above normal priority or both have problems to smooth play. I have checked the system with BurninTest pro when i replace the CPU (also i do a fresh install of Windows XP Pro GR 32bit)

and all where OK. I checked the Registry for L3 cache setting in Memory Management but there is not such a thing there. I will be thankful if anyone can help me to solve the performance problem. I'm experienced with Registry tweaks and Windows XP so feel free to tell me advanced modifications.

Thanks in advance

Link to comment
Share on other sites


OK first let's look at a few things.

Old CPU: Pentium 4 2.8GHz (SL7PK-E0)

L2 Cache: 1MB

FSB: 533MHz

New CPU: Pentium 4 3.2GHz (SL6WG-D1)

L2 Cache: 512KB

FSB: 800MHz

System memory: 3GB DDR2 667GHz CL5

So in the old config, you were underclocking your memory from 667MHz to 533MHz. Now with the new CPU one of the following is happening (you'll have to check the BIOS to be sure)

1. New CPU is overclocking memory from 667MHz to 800MHz

2. Motherboard is underclocking the CPU because FSB can't run faster than 667MHz

Both CPUs appear on the CPU compatibility chart:

http://asia.msi.com/index.php?func=prodcpusupport&maincat_no=1&prod_no=1525#menu

So check your BIOS to see what speed the memory and what the ratio for the CPU is. As noted on this list, your ratio should be 16. There is no BIOS update required for either CPU.

If you have any 800MHz RAM, try putting that in there to see if you can see a difference.

Link to comment
Share on other sites

OK first let's look at a few things.

Old CPU: Pentium 4 2.8GHz (SL7PK-E0)

L2 Cache: 1MB

FSB: 533MHz

New CPU: Pentium 4 3.2GHz (SL6WG-D1)

L2 Cache: 512KB

FSB: 800MHz

System memory: 3GB DDR2 667GHz CL5

So in the old config, you were underclocking your memory from 667MHz to 533MHz. Now with the new CPU one of the following is happening (you'll have to check the BIOS to be sure)

1. New CPU is overclocking memory from 667MHz to 800MHz

2. Motherboard is underclocking the CPU because FSB can't run faster than 667MHz

Both CPUs appear on the CPU compatibility chart:

http://asia.msi.com/index.php?func=prodcpusupport&maincat_no=1&prod_no=1525#menu

So check your BIOS to see what speed the memory and what the ratio for the CPU is. As noted on this list, your ratio should be 16. There is no BIOS update required for either CPU.

If you have any 800MHz RAM, try putting that in there to see if you can see a difference.

First thanks for your answer :hello:

One correction for the above:

The 2.8GHz CPU was L2 1MB, FSB 800MHz

The BIOS (ver 1.2) recognize the CPU properly (at least that i see there)

System Info:

Pentium 4 3.2GHz

L2 Cache 512KB

L3 not appear there

At Post shows

Multiplier 16x200=3200MHz

Memory 3072MB 667MHZ (Before BIOS upgrade to 1.1 was show 556MHz)

The same speed appears in Cell Menu (667MHz). How MSI manage to do this when the socket 478 CPU with 200MHz bandwidth for memory!!! In my old mobo with DDR1 RAM was 400MHz with the above 2.8GHz CPU.

The suprising thing is that in when i watch a movie or playing a game and i have task manager open in both system tray and task bar shows extreme low CPU activity ( no more than 8% in movies currently play). But the playing is not smooth any more until i increase the priority to above normal. I don't understand that. I'm confusing.

Link to comment
Share on other sites

I'm not really certain how the memory works on this type of P4. My P4 experience has only been on D850M series (both V and D) boards from Intel which uses RAMBUS. So on those, the RAM runs at 800MHz but FSB at 400MHz. But RAMBUS has direct access to the CPU so the speed is different in relation to DDR2.

However, since you can fix it in Task Manager, I'm going to bet it is not just a hardware difference involved. But I too have seen changes in performance when changing memory speeds.

Link to comment
Share on other sites

Increasing the priority just means you're giving more CPU execution time to that process compared to others at a lower priority (which is basically anything but core windows system functions if we're talking about increasing a process priority to anything above normal priority).

To give you some insight into how Windows works, and possible reasons why your changes are "fixing" things, you need to understand how Windows schedules threads to run on the processors available to it. It is important to remember that Windows deals in threads when it comes to scheduling, not processes - that means that when you increase the priority of a process, you are also increasing the default priority of each thread running in that process to the same priority as the process itself (well, in most situations - I'm avoiding some more obscure thread scheduling scenarios that change this in order to keep this simple).

  • On a Windows system, each running thread gets a quantum that tells how long it can run on the CPU before it's swapped out for another thread. When a thread's current quantum has ended, if there's another thread that has a higher priority, or an equal priority and has not run in a set amount of time, it's context will be saved, and then swapped out for the context of a different thread - that thread then starts it's quantum on the CPU, and the cycle continues.
  • When a thread reaches the end of it's current quantum, the scheduler checks to see if there are any other threads that have a higher priority that are ready to execute (and again, on Windows, really only critical base system functions run above normal priority). Anything with a higher priority will be scheduled to run, and the current thread's context will be swapped out and saved until the next time it has a chance to run. If nothing is found ready with a higher priority, the next check is made.
  • The scheduler then checks to see if any threads at the same priority level that are ready to execute - any other thread on the ready list at the same priority at this time is scheduled, and the current thread's context is swapped out and saved until the next time it has a chance to run. If nothing is found ready with a higher priority, the next check is made. If nothing is found ready with the same priority level, the last check is made.
  • The scheduler last checks to see if there are threads at any priority level that have not executed in some length of time, and need their priority boosted so that they're given time to run (otherwise, low priority threads would never run). If anything is found on the list to have not executed in a long length of time, it's priority is temporarily boosted to a higher one, the current thread's context is swapped out and saved, and that new thread is scheduled. Since most things on a Windows system will run normal priority, this doesn't happen often on a system that isn't incredibly busy.
  • If all three of those checks find no threads that supersede this one at that time, it gives the currently running thread another quantum and it continues running.

I'm postulating, along the lines Tripredacus is, that this might come down to something other than just a processor or system difference, especially if the new hardware replacing the older hardware is at least as powerful - this is likely more to do with things running on the box at the same time (if increasing the priority of a process "fixes" a performance issue, this is 99% of the time the most likely culprit).

Link to comment
Share on other sites

You said you did a fresh install of Windows XP. A couple of questions: Did you install all the hardware drivers? Did you install the latest DirectX and updates from Microsoft?

If you indeed have an SL6WG-D1 processor as Tripredacus said then you don't have the Extreme Edition and there is no L3 cache.

I'm using a 3.2 Northwood right now and this machine has no problems what so ever with video so as others said the problem is most likely with software and not hardware.

But just to be on the safe side go inside the case and unplug everything pull the ram and then put everything back. Some time things come loose during an upgrade.

Link to comment
Share on other sites

You said you did a fresh install of Windows XP. A couple of questions: Did you install all the hardware drivers? Did you install the latest DirectX and updates from Microsoft?

If you indeed have an SL6WG-D1 processor as Tripredacus said then you don't have the Extreme Edition and there is no L3 cache.

I'm using a 3.2 Northwood right now and this machine has no problems what so ever with video so as others said the problem is most likely with software and not hardware.

But just to be on the safe side go inside the case and unplug everything pull the ram and then put everything back. Sometimes things come loose during an upgrade.

My CPU is indeed Extreme Edition. I use the CPU-Z utility and the results published here: http://valid.canardpc.com/show_oc.php?id=1355609

Drivers:

1. Intel INF Update Utility 9.1.1.1025 22-12-2009 for 945GC express chipset. Same as previous install of Windows.

2. NVIDIA GeForce,ION Driver Release 257.21 WHQL 2010.06.15 Windows XP for NVIDIA 9400 PCI-E graphic card. In previous install was 196.x

3. Realtek HD Audio Codecs R2.49 11-6-2010 for PCI-E onboard audio. In previous install was an older version. I don't remember the version.

4. Realtek WinXP 32,64 and Win2K Auto Installation Program (SID 1412117) 5.760 22-6-2010 for PCI-E onboard ethernet controler. In previous install was an older version. I don't remember the version.

5. DirectX End-User Runtimes (June 2010). In previous install was an older version

The XP pro Greek with SP3 is original and all the updates from microsoft is installed.

My intention is to install 2x2GB DDR2 RAM just in case is needed by BIOS to address memory space for the PCI-E devices. Especially for graphic card with 1GB memory on card. Any ideas about that?

Link to comment
Share on other sites

There's one thing I would avoid in your setup, you are doing double channel memory with two modules of different sizes.

If you want to do double channel, the modules have to be exactly the same type and capacity.

Either put in 2x1GB or 2x2GB.

Mind that Windows XP only uses about 3.2GB of the 4GB installed, unless you have the 64 bit XP version.

Its kinda weird also to see a 945 chipset supporting socket 478, that socket is obsolete.

Time to move on to a new platform.

Cheers!

Edited by Escorpiom
Link to comment
Share on other sites

There's one thing I would avoid in your setup, you are doing double channel memory with two modules of different sizes.

If you want to do double channel, the modules have to be exactly the same type and capacity.

Either put in 2x1GB or 2x2GB.

Mind that Windows XP only uses about 3.2GB of the 4GB installed, unless you have the 64 bit XP version.

Its kinda weird also to see a 945 chipset supporting socket 478, that socket is obsolete.

Time to move on to a new platform.

Cheers!

There is no option in BIOS to change the dual channel feature. It appears in POST automatically but i am not sure if really the mobo use it that way for the reasons you mention above. For the use of 4GB in XP 32bit is known that can be done with /3GB switch in boot.ini but you need applications who knows how to utilize this extra memory. I'm afraid that the whole design of that mobo is a little buggy. But i have a second PC (Pentium Core 2 3.016GHz socket 775 :thumbup ) for my heavy hobby :P , video editing. The whole matter here is to determine if the decrease in performance is software or hardware related. I have installed Proccess Explorer and Autoruns to help me with that problem but i am not sure in what informations in Proccess Explorer to concentrate my attention. I will appreciate the help of anyone who will spend some time to write an answer for me. Thanks in advance.

Link to comment
Share on other sites

There's one thing I would avoid in your setup, you are doing double channel memory with two modules of different sizes.

If you want to do double channel, the modules have to be exactly the same type and capacity.

Either put in 2x1GB or 2x2GB.

Mind that Windows XP only uses about 3.2GB of the 4GB installed, unless you have the 64 bit XP version.

Its kinda weird also to see a 945 chipset supporting socket 478, that socket is obsolete.

Time to move on to a new platform.

Cheers!

There is no option in BIOS to change the dual channel feature. It appears in POST automatically but i am not sure if really the mobo use it that way for the reasons you mention above. For the use of 4GB in XP 32bit is known that can be done with /3GB switch in boot.ini but you need applications who knows how to utilize this extra memory. I'm afraid that the whole design of that mobo is a little buggy. But i have a second PC (Pentium Core 2 3.016GHz socket 775 :thumbup ) for my heavy hobby :P , video editing. The whole matter here is to determine if the decrease in performance is software or hardware related. I have installed Proccess Explorer and Autoruns to help me with that problem but i am not sure in what informations in Proccess Explorer to concentrate my attention. I will appreciate the help of anyone who will spend some time to write an answer for me. Thanks in advance.

There is no option in the bios. In your case you have only two DIMM slots, so the only way to test this is taking out the 1GB module and test again with only the 2GB module installed.

As for the switch in boot.ini, that is only used for old applications that cannot run with a large amount of memory.

I recall the issue with Visual Foxpro compiled applications for example, old versions would not run if the memory wasn't limited in boot.ini. In your case it has nothing to do with the actual problem. Having 4GB installed doesn't require the /3GB switch, Windows will run anyway but only recognizes about 3.2GB of memory instead of 4GB.

If any proposed solution fails, you can try to start with a clean installation:

- Use latest BIOS from MSI ver. 1.2 update date 2010-06-10

1. This is AMI BIOS release

2. This BIOS fixes the following problem of the previous version:

Fixed system report incorrect HDD size after disabled LBA mode.

Improved VGA compatibility. <<this could apply to your config

Fixed system report incorrect memory size during POST.

- clean XP SP3 install

- Chipset and video drivers (Intel and Nvidia)

Do not install anything else and test again.

If the problem persist, it's the hardware. I agree with you that the platform could be buggy. Like I said it's not common to see socket 478 teamed up with 945 chipset.

Cheers!

Edited by Escorpiom
Link to comment
Share on other sites

Increasing the priority just means you're giving more CPU execution time to that process compared to others at a lower priority (which is basically anything but core windows system functions if we're talking about increasing a process priority to anything above normal priority).

To give you some insight into how Windows works, and possible reasons why your changes are "fixing" things, you need to understand how Windows schedules threads to run on the processors available to it. It is important to remember that Windows deals in threads when it comes to scheduling, not processes - that means that when you increase the priority of a process, you are also increasing the default priority of each thread running in that process to the same priority as the process itself (well, in most situations - I'm avoiding some more obscure thread scheduling scenarios that change this in order to keep this simple).

  • On a Windows system, each running thread gets a quantum that tells how long it can run on the CPU before it's swapped out for another thread. When a thread's current quantum has ended, if there's another thread that has a higher priority, or an equal priority and has not run in a set amount of time, it's context will be saved, and then swapped out for the context of a different thread - that thread then starts it's quantum on the CPU, and the cycle continues.
  • When a thread reaches the end of it's current quantum, the scheduler checks to see if there are any other threads that have a higher priority that are ready to execute (and again, on Windows, really only critical base system functions run above normal priority). Anything with a higher priority will be scheduled to run, and the current thread's context will be swapped out and saved until the next time it has a chance to run. If nothing is found ready with a higher priority, the next check is made.
  • The scheduler then checks to see if any threads at the same priority level that are ready to execute - any other thread on the ready list at the same priority at this time is scheduled, and the current thread's context is swapped out and saved until the next time it has a chance to run. If nothing is found ready with a higher priority, the next check is made. If nothing is found ready with the same priority level, the last check is made.
  • The scheduler last checks to see if there are threads at any priority level that have not executed in some length of time, and need their priority boosted so that they're given time to run (otherwise, low priority threads would never run). If anything is found on the list to have not executed in a long length of time, it's priority is temporarily boosted to a higher one, the current thread's context is swapped out and saved, and that new thread is scheduled. Since most things on a Windows system will run normal priority, this doesn't happen often on a system that isn't incredibly busy.
  • If all three of those checks find no threads that supersede this one at that time, it gives the currently running thread another quantum and it continues running.

I'm postulating, along the lines Tripredacus is, that this might come down to something other than just a processor or system difference, especially if the new hardware replacing the older hardware is at least as powerful - this is likely more to do with things running on the box at the same time (if increasing the priority of a process "fixes" a performance issue, this is 99% of the time the most likely culprit).

I configure the proceess explorer to show Context Switches and CSwitch Delta. From the number of Hardware Interrupts and Deferred Procedure Calls it is obviously that a have a big problem. In internet pages i see to refer that as interrupt and DPC storms / CPU usage problem. I download and i'm intented to use some debuging utilities to see what driver/hardware issue is causing the problem. Thanks for your answer. It led me to unexplored (to me) paths.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...