Install hotfix 896256 for performance increase

Operating System, Common Application & ThinkPad Utilities Questions...
Post Reply
Message
Author
WPWoodJr
Junior Member
Junior Member
Posts: 393
Joined: Wed Nov 08, 2006 7:45 pm
Contact:

Install hotfix 896256 for performance increase

#1 Post by WPWoodJr » Thu Feb 15, 2007 10:10 pm

Microsoft's description of this hotfix says the following:
Computers that are running Windows XP Service Pack 2 and that are equipped with multiple processors that support processor power management features may experience decreased performance
To see if you need this hotfix, check the version of hal.dll in \windows\system32 by hovering over the file name. If its at 5.1.2600.3023 or higher, you have this hotfix. I found that although system installer said that I had this hotfix applied, it had failed and I had to install it by hand from the Microsoft website:

http://support.microsoft.com/kb/896256

It may be my imagination but it did seem to speed up the machine a little.

felixcat
Posts: 2
Joined: Fri Dec 22, 2006 3:35 pm
Location: Houston, TX

#2 Post by felixcat » Thu Feb 15, 2007 10:38 pm

Thank you!

I downloaded that hotfix and manually installed it. After my machine reboot, I run Matlab's 'bench' command, and it generate the machine's performance parameters. It is easy to see that after the hotfix, my machine need less time to compute a matrix's LU decomposition, this is a good news.

npish
Sophomore Member
Posts: 200
Joined: Sat Nov 04, 2006 1:46 am
Location: SF, USA

#3 Post by npish » Thu Feb 15, 2007 11:27 pm

pardon my ignorance, but do dual-core processors--such as the T7200 on my T60p--constitute "multiple processors" in this respect?

WPWoodJr
Junior Member
Junior Member
Posts: 393
Joined: Wed Nov 08, 2006 7:45 pm
Contact:

#4 Post by WPWoodJr » Thu Feb 15, 2007 11:32 pm

Yes, Core Duo and Core 2 Duo are multi-processor cores.

iatacs19
Freshman Member
Posts: 115
Joined: Mon Oct 16, 2006 12:36 pm
Location: Washington, DC

#5 Post by iatacs19 » Fri Feb 16, 2007 9:45 am

i will give it a try, it doesn't show up on windows update yet.

gunston
ThinkPadder
ThinkPadder
Posts: 1306
Joined: Thu Sep 01, 2005 8:00 am
Location: Brisbane, QLD AUST
Contact:

#6 Post by gunston » Fri Feb 16, 2007 10:06 am

Does it really help to boost the performance?
How much improvement after the HotFix. :?:
1. T43 2668-B97 14" SXGA+ 1.5G RAM 9cells
2. X60s 1703-CA3 powerful

Zeus
Freshman Member
Posts: 55
Joined: Wed Jan 24, 2007 8:15 am
Location: Pennsylvania, USA

#7 Post by Zeus » Fri Feb 16, 2007 1:47 pm

huh, I needed it too, mine was older. Thanks!

WPWoodJr
Junior Member
Junior Member
Posts: 393
Joined: Wed Nov 08, 2006 7:45 pm
Contact:

#8 Post by WPWoodJr » Fri Feb 16, 2007 2:17 pm

gunston wrote:Does it really help to boost the performance?
How much improvement after the HotFix. :?:
This fix would seem to up the performance of single-threaded tasks when you are using a power profile that uses adaptive CPU throttling (such as "Power Source Optimized") - interesting reading from the hotfix:

Windows XP SP2 is required on computers that have multiple CPUs that support ACPI processor performance states. This requirement includes computers that support the following items:
• Multiple physical sockets
• Multiple-core designs
• Multiple logical threads, such as Intel hyper-threading technology

Because Windows XP was not originally designed to support performance states on multiprocessor configurations, changes are required to correctly realize this support on multiprocessor systems. Windows XP Service Pack 2 includes the required changes to the kernel power manager. These changes make sure that Windows XP correctly functions on multiprocessor systems with processor performance states.

This hotfix also addresses the following issues on computers that have multiple processors that support processor performance states:
• A possible decrease in performance on single-threaded workloads when processor performance states are using demand-based switching.
• The synchronization of the processor Time Stamp Counter (TSC) registers across processors when you use the ACPI Power Management timer on multiprocessor systems.
• ACPI C-state promotion and demotion issues in the kernel power manager.

Update details
Possible decrease in performance during demand-based switching
Demand-Based Switching (DBS) is the use of ACPI processor performance states (dynamic voltage and frequency scaling) in response to system workloads. Windows XP processor power management implements DBS by using the adaptive processor throttling policy. This policy dynamically and automatically adjusts the processor’s current performance state in response to system CPU use without user intervention.

When single-threaded workloads run on multiprocessor systems that include dual-core configurations, the workloads may migrate across available CPU cores. This behavior is a natural artifact of how Windows schedules work across available CPU resources. However, on systems that have processor performance states that run with the adaptive processor throttling policy, this thread migration may cause the Windows kernel power manager to incorrectly calculate the optimal target performance state for the processor. This behavior occurs because an individual processor core, logical or physical, may appear to be less busy than the whole processor package actually is. On performance benchmarks that use single-threaded workloads, you may see this artifact in decreased performance results or in a high degree of variance between successive runs of identical benchmark tests.

This hotfix includes changes to the kernel power manager to track CPU use across the processor package. These changes enable visibility into the true activity level of a CPU complex and therefore help correctly calculate an increased target performance state.

Note This solution favors performance gains over power savings. Although benchmark performance scores may improve, battery life could be negatively affected. Accordingly, this kernel policy change may be disabled by a registry key to allow for maximum flexibility.
How to disable the new performance state policy behavior
Warning Serious problems might occur if you modify the registry incorrectly by using Registry Editor or by using another method. These problems might require that you reinstall the operating system. Microsoft cannot guarantee that these problems can be solved. Modify the registry at your own risk.

After you install the hotfix that is described in this article, you may use registry settings to disable the new performance state policy behavior. To do this, follow these steps:
1. Click Start, click Run, type regedit, and then click OK.
2. Right-click HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager, point to New, and then click Key.
3. Type Throttle for the new key name.
4. Right-click Throttle, point to New, and then click DWORD Value.
5. Type PerfEnablePackageIdle for the value name.
6. Right-click PerfEnablePackageIdle, and then click Modify.
7. In the Value data box, type 0. Make sure that Hexadecimal is selected in the Edit DWORD Value dialog box, and then click OK.

Note You can type 1 in the Value data box to enable the new performance state policy behavior.
8. Quit Registry Editor.

Correct TSC synchronization
On some operating systems, the processor TSC may change the rate at which it counts. Additionally, the processor TSC may stop counting when specific processor power management features are used. On computers that have multiple processors, the TSC is typically the operating system hardware timer that supports calls to the kernel KeQueryPerformanceCounter function. When TSC does not increment monotonically, system components that use the kernel KeQueryPerformanceCounter function may not work correctly. To address this problem, Microsoft makes it possible for the ACPI Power Management Timer to be used as the operating system timer that supports the kernel KeQueryPerformanceCounter function. However, some programs may directly access the TSC by bypassing the Windows timer APIs. The multiple-processor Hardware Abstraction Layer (HAL) makes sure that the TSC registers on all processors on a multiple-processor computer remain closely synchronized. Therefore, access by system software that may be directed to different processors does not return different results. This change makes sure that the multiple-processor HAL continues to correctly synchronize the TSCs across all processors on a computer, even if the ACPI power management timer is used as the operating system hardware timer.

Correct C-state promotion and demotion
This change corrects issues in the kernel power manager to correctly handle processor ACPI C-state promotion and demotion on multiprocessor systems.

absolut 0
Posts: 21
Joined: Tue Oct 31, 2006 2:19 am
Location: Ann Arbor, MI

#9 Post by absolut 0 » Fri Feb 16, 2007 8:24 pm

If this update is so valuable, why hasn't microsoft released it as an automatic update for windows?
Thinkpad T60p Widescreen (8742c6u)
2.33GHz 4MB 667MHz Core 2 Duo (T7600)
2GB PC2-5300 DDR SDRAM
15.4" WSXGA+ (1680x1050)
ATI MOBILITY FireGL V5250 (256mb)
100GB 7200RPM Hard Drive

laundromatt
Sophomore Member
Posts: 154
Joined: Wed Jun 14, 2006 11:19 am
Location: San Francisco, California

#10 Post by laundromatt » Fri Feb 16, 2007 8:29 pm

i didn't have this patch, so i put it on yesterday. i have no objective way to measure this, but i haven't noticed a difference, good or bad.
T60: 2623-D6U
X41: 2526-A29

tomatoeblue
Posts: 15
Joined: Thu Jan 11, 2007 7:54 pm
Location: Vancouver, BC, Canada

#11 Post by tomatoeblue » Fri Feb 16, 2007 8:38 pm

If this update is so valuable, why hasn't microsoft released it as an automatic update for windows?
Because it's only valuable for people looking for high performance from certain kinds of computers. Not all people have multi-processors / multi-cores, and the patch might negatively influence single processor machines.

The patch doesn't actually increase performance in the normal sense. It makes the scheduler more "fair" when allocating cpu time to tasks.

Basically, your computer is not speeding up. It's just giving an application its fair share of processing power when CPU speeds are throttling up and down.

gunston
ThinkPadder
ThinkPadder
Posts: 1306
Joined: Thu Sep 01, 2005 8:00 am
Location: Brisbane, QLD AUST
Contact:

#12 Post by gunston » Fri Feb 16, 2007 9:15 pm

it means the CPU would averagely attribute the application tasks to its duo core/ duo 2 core processor ... in other words, less CPU activities and less heat... :arrow: is this assumption true :?:
1. T43 2668-B97 14" SXGA+ 1.5G RAM 9cells
2. X60s 1703-CA3 powerful

tomatoeblue
Posts: 15
Joined: Thu Jan 11, 2007 7:54 pm
Location: Vancouver, BC, Canada

#13 Post by tomatoeblue » Sat Feb 17, 2007 4:09 pm

No not necessarily. There is nothing about this patch that lowers cpu load or heat levels. In fact, it might increase both.

As quoted from the press release
Note This solution favors performance gains over power savings. Although benchmark performance scores may improve, battery life could be negatively affected.
And that's another reason why it's not a mass autoupdated patch. It's about tradeoffs.

Imagine your computer as a pie. Each process gets to take a piece, and imagine Windows XP calculating how big of a piece each process gets.

Now imagine the pie growing to become much bigger. The sizes of pieces of pie that each application takes must be recalculated to keep everything fair. WinXP by default, does not do this properly with dual core processors, and so some processes are starved, while other processes have much more pie than they know how to deal with. This is inefficient, and is not the best way to use a cpu.

This patch reallocates the pie, so that as the pie grows, bigger pieces are given to hungrier applications (such as video encoding, photoshop, or a 3d game, etc). Of course, these applications need all the processing power they can get, and because they are given more share of the CPU than before, they perform faster.

Since more CPU power is being used, there is actually an overall increase of CPU usage and heat generated. Thus the possibility of lower battery life.

gator
Moderator Emeritus
Moderator Emeritus
Posts: 3401
Joined: Thu Aug 17, 2006 2:28 am
Location: Gainesville, FL

#14 Post by gator » Mon Feb 19, 2007 2:28 am

There is a definite increase in 3D performance with MATLAB with this hotfix. Thanks to the OP for the info.
Now: T60 2613-EKU | T23 2647-9NU | 600X 2645-9FU | HP 100LX
Past: X31 2673-Y13 | T41 2374-3HU | T22 2647-AEU


Rules of the road :thumbs-UP:

avatarz
Posts: 2
Joined: Tue May 09, 2006 10:14 pm

thanks

#15 Post by avatarz » Sat Mar 17, 2007 11:15 am

significant performance increase
`call me crazy.. but seens that the system runs much faster and cooler....
thanks!

avatarz


x60@t7200

spt60
Junior Member
Junior Member
Posts: 250
Joined: Thu Feb 08, 2007 8:59 am
Location: OH

#16 Post by spt60 » Thu Mar 22, 2007 12:32 pm

it failed my Ad-Watch during startup or even slow down the machine as running Ad-Watch afterward. Uninstalled it as the performance did increase but not that much, the heat seems to increase though.

blueblood
Posts: 27
Joined: Sun Mar 04, 2007 4:52 pm
Location: Austin, Texas

#17 Post by blueblood » Mon Mar 26, 2007 10:15 pm

well i applied the update and i can DEFINITELY see performance improvements.

i use matlab a lot and i ran the benchmark in it.
before the patch my computer consistently performed below a 2.1ghz opteron, was 3rd in comparison with the other processors who's data was provided.

after the patch its the top processor by far !
its crunching code a whole lot faster

i have a T60p C2D 2GHZ, 2gb ram

ucsdmike
Posts: 4
Joined: Mon May 28, 2007 1:50 am
Location: San Diego, Calif.

i agree with avatarz

#18 Post by ucsdmike » Mon May 28, 2007 2:57 am

what i've noticed. although it seems like the battery consumption is up ~5 percent on idle (looking at the Current on power manager), it seems to require a lot less power when playing a video. So it all evens out.

Also, the system runs a lot cooler. I'm sure that counts for a lot.

propellen
Sophomore Member
Posts: 171
Joined: Mon Mar 26, 2007 2:50 am
Location: Trondheim, Norway

#19 Post by propellen » Mon May 28, 2007 2:20 pm

Thanks.
Applying this now.

EDIT:
Looks like the system is much cooler indeed.
Thanks alot mate! Very good hotfix.
LENOVO THINKPAD T60 C2D/T5600-1.83G 320GB 2GB 14IN SXGA+ 7 ULTIMATE (UT0FUNO)
TPFanControl

tebore
Freshman Member
Posts: 86
Joined: Sun May 20, 2007 11:09 pm
Location: Toronto, Canada

#20 Post by tebore » Tue Jun 05, 2007 1:40 pm

Just installed the patch and seems to be working but has anyone tried turning off the new states? I noticed a slight drop in idle power turning them off.
T60P 2623-D8U

sokos
Sophomore Member
Posts: 153
Joined: Tue Apr 10, 2007 5:02 pm
Location: Athens, Greece
Contact:

#21 Post by sokos » Sat Jun 09, 2007 3:12 pm

Ok i read the thread (very interesting, thanks) and i decided to do a little testing / benchmarking before and after.

I user two different software, right after rebooting running the benchs. I used:
PC Mark 2005 (Basic Edition) and a little CPU Bench i downloaded from MajorGeeks.com called CPUBench V.4.0.0.6

Before update:
PC Mark 2005: 3275

CPUBench:

MFLOPS: 377,79,
MIPS: 603,44,
PI: 3469,
KPPS: 3690,88,
Lorenz: 438,
Dhrystone: 3536067,
Whetstone: 1005910,
Nblock: 4812,
Queens: 3766,
Matrix: 2640,
Savage: 1359,
B5: 4313,

After Update:
PC Mark 2005: 3285 (Not much of a big gain)

CPUBench:

MFLOPS: 373,19,
MIPS: 603,44,
PI: 3469,
KPPS: 3846,15,
Lorenz: 437,
Dhrystone: 3595828,
Whetstone: 1003890,
Nblock: 4813,
Queens: 3281,
Matrix: 2641,
Savage: 1359,
B5: 4079,

As you can see, performance in CPU has droped a bit in WinXP Pro.

I m keeping the patch for the time being to try notice any further change in system behaviour, i d like to know from other fellow Thinkpaders that installed the patch with some benchmarks and not just "feeling" :)
[/u]
T21 (Sold)
T40 (Sold)
T41 (best machine i ever had, Sold)
Toshiba Satellite (what a mistake, sold)
Thinkpad Z61t (brilliant machine, went to my father)
Got my T61 :)

WPWoodJr
Junior Member
Junior Member
Posts: 393
Joined: Wed Nov 08, 2006 7:45 pm
Contact:

#22 Post by WPWoodJr » Sun Jun 10, 2007 11:05 am

I believe these performance improvements are meant to be apparent under battery power, how did you run your benchmarks? Also, it would be good to include a benchmark that tests both processors.
T60p 2613-CTO, 2.33GHz, 3GB ram, Intel 80gb G2 SSD, H7K 200GB/7200rpm, LG Flexview IPS SXGA+ screen, ATI FireGL V5250
Essential TP Hotfixes and Tweaks

tebore
Freshman Member
Posts: 86
Joined: Sun May 20, 2007 11:09 pm
Location: Toronto, Canada

#23 Post by tebore » Mon Jun 11, 2007 7:39 am

It doesn't matter if it's battery power or AC it's all about multi threaded apps. You take a small hit in single threaded apps (1-5%) but you get a gain (10-20%) in multi threaded apps.
T60P 2623-D8U

sokos
Sophomore Member
Posts: 153
Joined: Tue Apr 10, 2007 5:02 pm
Location: Athens, Greece
Contact:

#24 Post by sokos » Mon Jun 11, 2007 3:35 pm

any good ideas of multithreated apps?

Coz if PCMark2005 doesn't reflect the daily usage of a business user, i wonder what does.. :?

I run Oracle 9i databases on my Thinkpad for testing and development and i haven't seen any difference so far..
T21 (Sold)
T40 (Sold)
T41 (best machine i ever had, Sold)
Toshiba Satellite (what a mistake, sold)
Thinkpad Z61t (brilliant machine, went to my father)
Got my T61 :)

WPWoodJr
Junior Member
Junior Member
Posts: 393
Joined: Wed Nov 08, 2006 7:45 pm
Contact:

#25 Post by WPWoodJr » Tue Jun 19, 2007 12:35 pm

WPWoodJr wrote:I believe these performance improvements are meant to be apparent under battery power, how did you run your benchmarks? Also, it would be good to include a benchmark that tests both processors.
Sorry, I meant to say: This fix would seem to up the performance of single-threaded tasks when you are using a power profile that uses adaptive CPU throttling (such as "Power Source Optimized")

Sokos, did you use adaptive cpu throttling during your benchmarks? Note that the performance improvement will be for single-threaded apps not multi-threaded.

Sorry for the earlier confusing post I had forgotten what this hotfix did.
T60p 2613-CTO, 2.33GHz, 3GB ram, Intel 80gb G2 SSD, H7K 200GB/7200rpm, LG Flexview IPS SXGA+ screen, ATI FireGL V5250
Essential TP Hotfixes and Tweaks

Post Reply
  • Similar Topics
    Replies
    Views
    Last post

Return to “Windows OS (Versions prior to Windows 7)”

Who is online

Users browsing this forum: No registered users and 7 guests