writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Aug 31, 2012 #1 6.11.0001.402 and 6.11.0001.404 run the six .dmps from my local cache in ~7.288 seconds. Read More: Code: 7.221 seconds to run the kd command on .dmp 1 of 6 10.14 seconds to run the kd command on .dmp 2 of 6 7.393 seconds to run the kd command on .dmp 3 of 6 4.334 seconds to run the kd command on .dmp 4 of 6 3.126 seconds to run the kd command on .dmp 5 of 6 7.319 seconds to run the kd command on .dmp 6 of 6 6.2.8229.0 and 6.2.9200.xxxx run the six .dmps from my local cache in ~39.533 seconds Read More: Code: 1.282 seconds to run the kd command on .dmp 1 of 6 1.285 seconds to run the kd command on .dmp 2 of 6 1.239 seconds to run the kd command on .dmp 3 of 6 1.264 seconds to run the kd command on .dmp 4 of 6 0.849 seconds to run the kd command on .dmp 5 of 6 1.369 seconds to run the kd command on .dmp 6 of 6 6.12.0002.633 runs the six .dmps from my local cache in ~26.407 seconds Read More: Code: 5.428 seconds to run the kd command on .dmp 1 of 6 5.92 seconds to run the kd command on .dmp 2 of 6 5.196 seconds to run the kd command on .dmp 3 of 6 2.142 seconds to run the kd command on .dmp 4 of 6 2.366 seconds to run the kd command on .dmp 5 of 6 5.355 seconds to run the kd command on .dmp 6 of 6 Commands used: "!analyze -v; !sysinfo cpuspeed; !sysinfo SMBIOS; lmtsmn; q" Most of the problem seems to stem from getting the module timestamps. If I run with lm, 6.2.xxxx.xxxx runs each .dmp in ~3 seconds. If I run with lmt, 6.2.xxxx.xxxx runs each .dmp in ~8 seconds. Even 3 seconds is about 4-6 times longer than it runs in 6.11.xxxx.xxx... I cannot find a way to optimize this. Why does 6.2.xxxx.xxxx take so much longer to run than 6.11.0001.402 and 6.11.0001.404 ? Attachments Minidump1.zip 130.9 KB · Views: 4 Last edited: Sep 1, 2012
6.11.0001.402 and 6.11.0001.404 run the six .dmps from my local cache in ~7.288 seconds. Read More: Code: 7.221 seconds to run the kd command on .dmp 1 of 6 10.14 seconds to run the kd command on .dmp 2 of 6 7.393 seconds to run the kd command on .dmp 3 of 6 4.334 seconds to run the kd command on .dmp 4 of 6 3.126 seconds to run the kd command on .dmp 5 of 6 7.319 seconds to run the kd command on .dmp 6 of 6 6.2.8229.0 and 6.2.9200.xxxx run the six .dmps from my local cache in ~39.533 seconds Read More: Code: 1.282 seconds to run the kd command on .dmp 1 of 6 1.285 seconds to run the kd command on .dmp 2 of 6 1.239 seconds to run the kd command on .dmp 3 of 6 1.264 seconds to run the kd command on .dmp 4 of 6 0.849 seconds to run the kd command on .dmp 5 of 6 1.369 seconds to run the kd command on .dmp 6 of 6 6.12.0002.633 runs the six .dmps from my local cache in ~26.407 seconds Read More: Code: 5.428 seconds to run the kd command on .dmp 1 of 6 5.92 seconds to run the kd command on .dmp 2 of 6 5.196 seconds to run the kd command on .dmp 3 of 6 2.142 seconds to run the kd command on .dmp 4 of 6 2.366 seconds to run the kd command on .dmp 5 of 6 5.355 seconds to run the kd command on .dmp 6 of 6 Commands used: "!analyze -v; !sysinfo cpuspeed; !sysinfo SMBIOS; lmtsmn; q" Most of the problem seems to stem from getting the module timestamps. If I run with lm, 6.2.xxxx.xxxx runs each .dmp in ~3 seconds. If I run with lmt, 6.2.xxxx.xxxx runs each .dmp in ~8 seconds. Even 3 seconds is about 4-6 times longer than it runs in 6.11.xxxx.xxx... I cannot find a way to optimize this. Why does 6.2.xxxx.xxxx take so much longer to run than 6.11.0001.402 and 6.11.0001.404 ?
jcgriff2 Co-Founder / AdminBSOD Instructor/ExpertMicrosoft MVP (Ret.) Staff member Joined Feb 19, 2012 Posts 21,541 Location New Jersey Shore Sep 1, 2012 #2 A good question. I have noticed the time differences for a while now and 1st thought it was due to slower DSL Internet speed - downloading symbols, but obviously not.
A good question. I have noticed the time differences for a while now and 1st thought it was due to slower DSL Internet speed - downloading symbols, but obviously not.
zigzag3143 Contributor, Sysnative Staff Emeritus Joined Mar 27, 2012 Posts 3,741 Sep 1, 2012 #3 On some DMPS it is even worse. To get any sort of benchmark I have picked a random set of DMPs and use them for all comparisons. Ever try running a full memory DMP?
On some DMPS it is even worse. To get any sort of benchmark I have picked a random set of DMPs and use them for all comparisons. Ever try running a full memory DMP?
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Sep 1, 2012 #4 I had not tried a full .DMP; I'll have to see if I can find my .DMP from when I was running a trial of Avast! on my system and it kept crashing due to scanning a backup of my Application Data junction.
I had not tried a full .DMP; I'll have to see if I can find my .DMP from when I was running a trial of Avast! on my system and it kept crashing due to scanning a backup of my Application Data junction.
U usasma Retired Admin Joined Feb 20, 2012 Posts 2,126 Sep 2, 2012 #5 Just my 2ยข - but things are whizzing along for me (kd.exe version 6.2.8400) now that the Symbols and Symbols Options fields have those check boxes! I'd suggest that we post a "stickie" with a standard group of minidumps - and a kernel dump or two (depends on what John can allow to be stored). Also will need a standard group of settings so that all are using the exact same options (when testing) - but will be able to easily revert to any customized settings that we may have. That way we can use them as "reference" groups - and I'd like to see some with the win32k.sys issues also, as I'm still trying to get something done for that.
Just my 2ยข - but things are whizzing along for me (kd.exe version 6.2.8400) now that the Symbols and Symbols Options fields have those check boxes! I'd suggest that we post a "stickie" with a standard group of minidumps - and a kernel dump or two (depends on what John can allow to be stored). Also will need a standard group of settings so that all are using the exact same options (when testing) - but will be able to easily revert to any customized settings that we may have. That way we can use them as "reference" groups - and I'd like to see some with the win32k.sys issues also, as I'm still trying to get something done for that.
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Sep 2, 2012 #6 Are you talking about a standard group of settings in the apps, or in WinDbg? If we want a standard set of settings in the apps, that can be easily accomplished by saving a .zdn file of your settings through File -> Save As and uploading it to a sticky. Not saying you have to do it; just letting you and others know how it would be done.
Are you talking about a standard group of settings in the apps, or in WinDbg? If we want a standard set of settings in the apps, that can be easily accomplished by saving a .zdn file of your settings through File -> Save As and uploading it to a sticky. Not saying you have to do it; just letting you and others know how it would be done.
U usasma Retired Admin Joined Feb 20, 2012 Posts 2,126 Sep 2, 2012 #7 I'm looking at a standard set - the apps, WinDbg, the dump files, etc. Can you upload a .zdn file of the default settings (I've changed mine too much already)?
I'm looking at a standard set - the apps, WinDbg, the dump files, etc. Can you upload a .zdn file of the default settings (I've changed mine too much already)?
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Sep 2, 2012 #8 defaultNoLoss.zdn will not overwrite your headers, footers, signatures, or template files with the defaults. default.zdn will overwrite all files with the defaults. Make sure to save your current settings to a .zdn file if you want to revert back to your current use of the apps. Attachments defaultNoLoss.zip 2.4 KB · Views: 5
defaultNoLoss.zdn will not overwrite your headers, footers, signatures, or template files with the defaults. default.zdn will overwrite all files with the defaults. Make sure to save your current settings to a .zdn file if you want to revert back to your current use of the apps.
jcgriff2 Co-Founder / AdminBSOD Instructor/ExpertMicrosoft MVP (Ret.) Staff member Joined Feb 19, 2012 Posts 21,541 Location New Jersey Shore Sep 2, 2012 #9 It may be beneficial to list system specs here as well. User OSWindbg/ KD VersionSystem Mfr Model NumCPURAMHDDJaidynMjcgriff2 Windows 7 x64 SP1 6.11, 6.12, 6.2 HP dv7-1020us Intel P7350 Core2 Duo @ 2 GHz 4 GB DDR2 500 GB SATA @ 5400 rpm Shintarousasma writhziden zigzag3143 Last edited: Sep 3, 2012
It may be beneficial to list system specs here as well. User OSWindbg/ KD VersionSystem Mfr Model NumCPURAMHDDJaidynMjcgriff2 Windows 7 x64 SP1 6.11, 6.12, 6.2 HP dv7-1020us Intel P7350 Core2 Duo @ 2 GHz 4 GB DDR2 500 GB SATA @ 5400 rpm Shintarousasma writhziden zigzag3143
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Sep 3, 2012 #10 UserOSWindbg/ KD VersionSystem Mfr Model NumCPURAMHDDJaidynMjcgriff2Windows 7 x64 SP16.11, 6.12, 6.2HP dv7-1020usIntel P7350 Core2 Duo @ 2 GHz4 GB DDR2500 GB SATA @ 5400 rpmShintarousasmaWindows 8 Release Preview x646.2.8400.0Asus P6T Deluxe v2 moboIntel Core i7-920 (1st gen) @ 2.67GHz 12 gB DDR2W-D Velociraptor (300 gB SATA) @ 10,000 rpmwrithzidenWindows 7 x64 SP16.11, 6.12, 6.2VAIO VPCF232FX/BIntel Core i7-2670QM CPU @ 2.20GHz8 GB DDR3500 GB SATA @ 7200 rpmzigzag3143Win 8 PRO RTM 6.2.9200.16384 HP DV-7 4010TX Intel Core i7-920 (1st gen) @ 2.67GHz 8 Gigs 500@7200 32gig SSD Last edited by a moderator: Sep 3, 2012
UserOSWindbg/ KD VersionSystem Mfr Model NumCPURAMHDDJaidynMjcgriff2Windows 7 x64 SP16.11, 6.12, 6.2HP dv7-1020usIntel P7350 Core2 Duo @ 2 GHz4 GB DDR2500 GB SATA @ 5400 rpmShintarousasmaWindows 8 Release Preview x646.2.8400.0Asus P6T Deluxe v2 moboIntel Core i7-920 (1st gen) @ 2.67GHz 12 gB DDR2W-D Velociraptor (300 gB SATA) @ 10,000 rpmwrithzidenWindows 7 x64 SP16.11, 6.12, 6.2VAIO VPCF232FX/BIntel Core i7-2670QM CPU @ 2.20GHz8 GB DDR3500 GB SATA @ 7200 rpmzigzag3143Win 8 PRO RTM 6.2.9200.16384 HP DV-7 4010TX Intel Core i7-920 (1st gen) @ 2.67GHz 8 Gigs 500@7200 32gig SSD
U usasma Retired Admin Joined Feb 20, 2012 Posts 2,126 Sep 3, 2012 #11 Edited my info into writhziden's post.... Will be updating to Win8 RTM as soon as I get some free time! Last edited: Sep 3, 2012
Edited my info into writhziden's post.... Will be updating to Win8 RTM as soon as I get some free time!
zigzag3143 Contributor, Sysnative Staff Emeritus Joined Mar 27, 2012 Posts 3,741 Sep 3, 2012 #12 Also added specs into writhzidens post. Similar results to what he has experienced.
jcgriff2 Co-Founder / AdminBSOD Instructor/ExpertMicrosoft MVP (Ret.) Staff member Joined Feb 19, 2012 Posts 21,541 Location New Jersey Shore Sep 3, 2012 #13 I simply listed a few names when creating the table; many are missing. Please feel free to provide system info & we'll add it. UserOSWindbg/ KD VersionSystem Mfr Model NumCPURAMHDD https://www.sysnative.com/forums/sh...e-Time-Disparity?p=25906&viewfull=1#post25906 Thanks!
I simply listed a few names when creating the table; many are missing. Please feel free to provide system info & we'll add it. UserOSWindbg/ KD VersionSystem Mfr Model NumCPURAMHDD https://www.sysnative.com/forums/sh...e-Time-Disparity?p=25906&viewfull=1#post25906 Thanks!
U usasma Retired Admin Joined Feb 20, 2012 Posts 2,126 Sep 3, 2012 #14 Suggest that we stabilize testing on a certain version of the app. Then, let's set up a procedure for testing. Y'know: Step 1 - Save your settings to a zdn file Step 2 - Import writhziden's default zdn file Step 3 - Run these dumps (groups of 10, 30, and 100?) <attachment goes here> Step 4 - Post back with times for 10, 30, and 100 dumps in seconds Step 5 - Import your saved zdn file to return to your usual work style
Suggest that we stabilize testing on a certain version of the app. Then, let's set up a procedure for testing. Y'know: Step 1 - Save your settings to a zdn file Step 2 - Import writhziden's default zdn file Step 3 - Run these dumps (groups of 10, 30, and 100?) <attachment goes here> Step 4 - Post back with times for 10, 30, and 100 dumps in seconds Step 5 - Import your saved zdn file to return to your usual work style
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Sep 3, 2012 #15 I would suggest standardizing with BETA: Sysnative BSOD Apps 2.0.0.6 since all the save and load functionality is working properly by that version. Note The select all / deselect all misses the last three items in Output Options in all of the versions released to this point. This will be fixed in a later release. A link to 10, 30, and 100 as well as a kernel .DMP: Mike P's SkyDrive Public Folder.
I would suggest standardizing with BETA: Sysnative BSOD Apps 2.0.0.6 since all the save and load functionality is working properly by that version. Note The select all / deselect all misses the last three items in Output Options in all of the versions released to this point. This will be fixed in a later release. A link to 10, 30, and 100 as well as a kernel .DMP: Mike P's SkyDrive Public Folder.
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Sep 3, 2012 #16 Procedure: Place the .dmps in the folder with the SysnativeBSODApps.exe application file. Run SysnativeBSODApps.exe When finished, hit OK to exit out; do not record the time given by the apps since it will be erroneous due to error/warning messages that pop up. Open outputDmps\<this set's timestamp>\SysnativeBSODApps\dmpOptions\timeBetweenRuns.txt Scroll down or search for "seconds to runDmps()". Copy and paste that line into your table. I will be adding my table to this post when I am finished running all five versions I have to test. Also, you may want to run through all the .dmps once with one of the versions to get the symbols set up for the test run. UsernameWindbg/kd.exe Version10 dmps30 dmps100 dmps1 Kernel DMPwrithziden6.11.0001.40263.662 seconds to runDmps()157.306 seconds to runDmps()225.593 seconds to runDmps()3.944 seconds to runDmps()writhziden6.11.0001.40461.571 seconds to runDmps()150.951 seconds to runDmps()227.752 seconds to runDmps()2.549 seconds to runDmps()writhziden6.12.0002.633108.673 seconds to runDmps()361.733 seconds to runDmps()754.292 seconds to runDmps()3.487 seconds to runDmps()writhziden6.2.8229.0109.81 seconds to runDmps()350.586 seconds to runDmps()797.359 seconds to runDmps()4.14 seconds to runDmps()writhziden6.2.9200.16384116.966 seconds to runDmps()350.712 seconds to runDmps()778.028 seconds to runDmps()4.812 seconds to runDmps() Another interesting one: This is with the following kd command: Code: kd -c "q" -z "<pathToDmpFile>" All it does is open the .dmp through the kernel debugger to see if it is a .dmp; no real commands are run other than to quit after opening the .dmp, but it is much slower in 6.2 than previous versions... UsernameWindbg/kd.exe Version10 dmps30 dmps100 dmps1 Kernel DMPwrithziden6.11.0001.4021.994 seconds to getDmpsList()1.621 seconds to getDmpsList()1.578 seconds to getDmpsList()1.648 seconds to getDmpsList()writhziden6.11.0001.4041.826 seconds to getDmpsList()1.651 seconds to getDmpsList()1.939 seconds to getDmpsList()1.815 seconds to getDmpsList()writhziden6.12.0002.6332.826 seconds to getDmpsList()1.782 seconds to getDmpsList()2.168 seconds to getDmpsList()2.26 seconds to getDmpsList()writhziden6.2.8229.014.402 seconds to getDmpsList()14.731 seconds to getDmpsList()15.126 seconds to getDmpsList()12.896 seconds to getDmpsList()writhziden6.2.9200.1638412.514 seconds to getDmpsList()11.826 seconds to getDmpsList()12.701 seconds to getDmpsList()11.953 seconds to getDmpsList() Attachments Windbg Processing Time Comparison.zip 11.3 KB · Views: 3 Last edited: Sep 3, 2012
Procedure: Place the .dmps in the folder with the SysnativeBSODApps.exe application file. Run SysnativeBSODApps.exe When finished, hit OK to exit out; do not record the time given by the apps since it will be erroneous due to error/warning messages that pop up. Open outputDmps\<this set's timestamp>\SysnativeBSODApps\dmpOptions\timeBetweenRuns.txt Scroll down or search for "seconds to runDmps()". Copy and paste that line into your table. I will be adding my table to this post when I am finished running all five versions I have to test. Also, you may want to run through all the .dmps once with one of the versions to get the symbols set up for the test run. UsernameWindbg/kd.exe Version10 dmps30 dmps100 dmps1 Kernel DMPwrithziden6.11.0001.40263.662 seconds to runDmps()157.306 seconds to runDmps()225.593 seconds to runDmps()3.944 seconds to runDmps()writhziden6.11.0001.40461.571 seconds to runDmps()150.951 seconds to runDmps()227.752 seconds to runDmps()2.549 seconds to runDmps()writhziden6.12.0002.633108.673 seconds to runDmps()361.733 seconds to runDmps()754.292 seconds to runDmps()3.487 seconds to runDmps()writhziden6.2.8229.0109.81 seconds to runDmps()350.586 seconds to runDmps()797.359 seconds to runDmps()4.14 seconds to runDmps()writhziden6.2.9200.16384116.966 seconds to runDmps()350.712 seconds to runDmps()778.028 seconds to runDmps()4.812 seconds to runDmps() Another interesting one: This is with the following kd command: Code: kd -c "q" -z "<pathToDmpFile>" All it does is open the .dmp through the kernel debugger to see if it is a .dmp; no real commands are run other than to quit after opening the .dmp, but it is much slower in 6.2 than previous versions... UsernameWindbg/kd.exe Version10 dmps30 dmps100 dmps1 Kernel DMPwrithziden6.11.0001.4021.994 seconds to getDmpsList()1.621 seconds to getDmpsList()1.578 seconds to getDmpsList()1.648 seconds to getDmpsList()writhziden6.11.0001.4041.826 seconds to getDmpsList()1.651 seconds to getDmpsList()1.939 seconds to getDmpsList()1.815 seconds to getDmpsList()writhziden6.12.0002.6332.826 seconds to getDmpsList()1.782 seconds to getDmpsList()2.168 seconds to getDmpsList()2.26 seconds to getDmpsList()writhziden6.2.8229.014.402 seconds to getDmpsList()14.731 seconds to getDmpsList()15.126 seconds to getDmpsList()12.896 seconds to getDmpsList()writhziden6.2.9200.1638412.514 seconds to getDmpsList()11.826 seconds to getDmpsList()12.701 seconds to getDmpsList()11.953 seconds to getDmpsList()
Tekno Venus Senior Administrator, Developer Staff member Joined Jul 21, 2012 Posts 7,199 Location UK Sep 3, 2012 #17 I have typed up an Excel spreadsheet to use. It has areas for user system specs and for dmp processing time. This excel sheet can be filtered and sorted to make life easier at the end. I will let Mike decide what app version and settings to use. If you need the file in another format, (open office, Excel 2003, .csv etc,) please ask. Stephen Attachments Windbg Processing Time Comparison.zip 9.8 KB · Views: 6
I have typed up an Excel spreadsheet to use. It has areas for user system specs and for dmp processing time. This excel sheet can be filtered and sorted to make life easier at the end. I will let Mike decide what app version and settings to use. If you need the file in another format, (open office, Excel 2003, .csv etc,) please ask. Stephen
Tekno Venus Senior Administrator, Developer Staff member Joined Jul 21, 2012 Posts 7,199 Location UK Sep 3, 2012 #18 And Mike just posted some other steps as I composed my post... Feel free to use my table in your steps Mike.
And Mike just posted some other steps as I composed my post... Feel free to use my table in your steps Mike.
niemiro Senior Administrator, Windows Update Expert Staff member Joined Mar 2, 2012 Posts 8,772 Location District 12 Sep 4, 2012 #19 Just in case I do any testing. I started off with some profiling yesterday, although I have yet to track down anything concrete. niemiro Windows 7 x64 SP1 6.11.1.404, 6.12, 6.2.8229.0 Self built Intel Core i7 2600k @ 3.4GHz 16GB DDR3 1600MHz 64GB Crucial M4 SSD 1TB Western Digital Caviar Black (Sata III)
Just in case I do any testing. I started off with some profiling yesterday, although I have yet to track down anything concrete. niemiro Windows 7 x64 SP1 6.11.1.404, 6.12, 6.2.8229.0 Self built Intel Core i7 2600k @ 3.4GHz 16GB DDR3 1600MHz 64GB Crucial M4 SSD 1TB Western Digital Caviar Black (Sata III)
writhziden Administrator, .NET/UWP Developer Staff member Joined May 23, 2012 Posts 2,923 Location Colorado Oct 24, 2012 #20 Can we re-open this discussion? I have been running 6.2 due to using an older system that requires me to use 32-bit Windows, and I do not have a 32-bit version of 6.11 to run. 6.2 has been driving me nuts. it sees "symbol file could not be found for win32k.sys" so the apps think the symbols are wrong, but never find the right ones because the win32k.sys issue is always there even using online symbols. I much prefer the message that the timestamp could not be verified since that implies the downloaded .dmp has an issue rather than the local symbols are missing in most cases. I cannot change the apps because then they would fail for 0x124 BugChecks if symbol files could not be found. The good .dmps still take ~twice as long, and the win32k.sys issue causes them to take up to 13 times as long... Ludicrous... Why is the newest version of the kernel debugger so much slower than the old 6.11 version? Why does it take so much longer to sort the driver list by timestamp? Why does it take so long just to open a .dmp to see if it is a .dmp without running a command? All questions that need to be asked...
Can we re-open this discussion? I have been running 6.2 due to using an older system that requires me to use 32-bit Windows, and I do not have a 32-bit version of 6.11 to run. 6.2 has been driving me nuts. it sees "symbol file could not be found for win32k.sys" so the apps think the symbols are wrong, but never find the right ones because the win32k.sys issue is always there even using online symbols. I much prefer the message that the timestamp could not be verified since that implies the downloaded .dmp has an issue rather than the local symbols are missing in most cases. I cannot change the apps because then they would fail for 0x124 BugChecks if symbol files could not be found. The good .dmps still take ~twice as long, and the win32k.sys issue causes them to take up to 13 times as long... Ludicrous... Why is the newest version of the kernel debugger so much slower than the old 6.11 version? Why does it take so much longer to sort the driver list by timestamp? Why does it take so long just to open a .dmp to see if it is a .dmp without running a command? All questions that need to be asked...