Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Trouble with Pixinsight and large numbers of subframes

  • Please log in to reply
28 replies to this topic

#1 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 22 February 2018 - 05:43 PM

I now have had two instances of my imaging laptop choking on processing my data. I have a Lenovo E450 laptop with 16GB of memory. I have been doing drizzle integration on large numbers of frames. My latest failure was with 300 frames of DLSR Canon 6D images and Pixinsight froze after doing the Staralignment and Localnormalization. When I tried to do image integration prior to Drizzle integration, PI would get all the way to integrating the final image and stop. It freeze the laptop and I would have to reboot. 16GB is the maximum RAM that I can install on this laptop. I also locked up trying to do a three panel mosaic. I have a hard time seeing how 16GB is not enough memory for this amount of processing. Has anyone else had issue like this? I dont know if it is the size of the individual frames or the number of frames that is driving the issue. I am going to try and see if I can integrate few subframes just as a test. For the mosaic, I could do a two panel but not a three panel image. Any input would be welcome.



#2 Salacious B Crumb

Salacious B Crumb

    Mariner 2

  • *****
  • Posts: 213
  • Joined: 13 Apr 2017
  • Loc: Greater Atlanta Area, GA

Posted 22 February 2018 - 08:02 PM

I have a generic Asus with 16GB RAM also and recently processed over 500 short Ha subs from M33. Mine never choked but it was glacier slow before I added the Swap storage folders. I dedicated 8 GB RAM for them and the change is significant.

 

 

- Sal



#3 50Flat

50Flat

    Lift Off

  • -----
  • Posts: 2
  • Joined: 25 Nov 2013

Posted 22 February 2018 - 08:39 PM

Mine eventually recovers ASUS G75 8MB drizzle 120 images about 15-20 min for each color.   My Pc will sometimes going to sleep mode during the process but wakes up and finishes.  So be sure you turn off sleep mode and let her run.  



#4 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 22 February 2018 - 09:56 PM

I have a generic Asus with 16GB RAM also and recently processed over 500 short Ha subs from M33. Mine never choked but it was glacier slow before I added the Swap storage folders. I dedicated 8 GB RAM for them and the change is significant.

 

 

- Sal

Well, I just left it running and it is actually not crashed. It was just glacial also. Is the swap storage part of window or pixinsight? Can you point me where to learn about this? So you sacrificed 1/2 your ram to the Swap storage?



#5 Kaos

Kaos

    Ranger 4

  • -----
  • Posts: 301
  • Joined: 02 Dec 2015
  • Loc: VA

Posted 22 February 2018 - 10:31 PM

I have had a similar issue on my main PC. That has a i7 6700K processor (4 cores with 8 processors running at 4.3 GHz) and 16 gig of RAM. I have notice that on occasion running significant processes in PI eventually eats up the RAM even when nothing is currently being processed. I can see 6 to 8 gigs of RAM being used by PI when nothing is currently running. I also have noticed that the RAM allocation is continues to increase as more images are opened in the PI desktop. I am not sure whether there is a memory leak or whether logs of prior processes are being stored in the RAM or perhaps both. I noticed it recently when I was running a lot of iterations of NBRGB script without closing the window. It continued to eat up my RAM and eventually bogged my system down to a stop (and nothing generally is able to do it other the Photoshop CC [and that does have a memory leak or allocation issue]). I will likely add more RAM in the near future to my PC. However, I have not had an issue with calibrating, CC, normalization and integration of more than 300 images (as I did just this past weekend). I did notice that the RAM allocation in PI had gone up to almost 6 gig after running through the stack. So either some VERY large logs (and I doubt large enough to eat up that much RAM) are being stored in the RAM or there is a memory leak I am guessing that there is a bug in PI that is causing a memory leak on some processes to occur.

 

Kaos



#6 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 22 February 2018 - 10:51 PM

I have had a similar issue on my main PC. That has a i7 6700K processor (4 cores with 8 processors running at 4.3 GHz) and 16 gig of RAM. I have notice that on occasion running significant processes in PI eventually eats up the RAM even when nothing is currently being processed. I can see 6 to 8 gigs of RAM being used by PI when nothing is currently running. I also have noticed that the RAM allocation is continues to increase as more images are opened in the PI desktop. I am not sure whether there is a memory leak or whether logs of prior processes are being stored in the RAM or perhaps both. I noticed it recently when I was running a lot of iterations of NBRGB script without closing the window. It continued to eat up my RAM and eventually bogged my system down to a stop (and nothing generally is able to do it other the Photoshop CC [and that does have a memory leak or allocation issue]). I will likely add more RAM in the near future to my PC. However, I have not had an issue with calibrating, CC, normalization and integration of more than 300 images (as I did just this past weekend). I did notice that the RAM allocation in PI had gone up to almost 6 gig after running through the stack. So either some VERY large logs (and I doubt large enough to eat up that much RAM) are being stored in the RAM or there is a memory leak I am guessing that there is a bug in PI that is causing a memory leak on some processes to occur.

 

Kaos

Kaos,

 

Thanks. Right now I am integrating 300 16M pixel color DSLR frames using localnormalization and drizzle file. It is not the actual drizzle integration yet and I am actually worried that that will be even worse. It has been running the integration for 24 hours now and it on the last channel. My guess that if it does not crash, it will take another 3-4 hours. This is just not fun.  I went in and set up a SWAP disk of 16GB on my hard drive to see if that helps. From the other posters, and from you, I dont think that it is going to be a huge improvement even buying a new 1500$ laptop with all the bells and whistles. Well see if they swap disk helps first. I am looking at an i7 with a 1TB SSD HD and 2TB HDD and 32GB of RAM as an upgrade, but I would rather put that money into a ASI1600MM-pro!!


  • bsavoie likes this

#7 Salacious B Crumb

Salacious B Crumb

    Mariner 2

  • *****
  • Posts: 213
  • Joined: 13 Apr 2017
  • Loc: Greater Atlanta Area, GA

Posted 22 February 2018 - 10:52 PM

There are a few posts here and at PixInsight forum which will walk you through the process. Just Google Pixinsight Swap files or something similar .

 

 

- Sal



#8 Kaos

Kaos

    Ranger 4

  • -----
  • Posts: 301
  • Joined: 02 Dec 2015
  • Loc: VA

Posted 22 February 2018 - 10:59 PM

 

I have had a similar issue on my main PC. That has a i7 6700K processor (4 cores with 8 processors running at 4.3 GHz) and 16 gig of RAM. I have notice that on occasion running significant processes in PI eventually eats up the RAM even when nothing is currently being processed. I can see 6 to 8 gigs of RAM being used by PI when nothing is currently running. I also have noticed that the RAM allocation is continues to increase as more images are opened in the PI desktop. I am not sure whether there is a memory leak or whether logs of prior processes are being stored in the RAM or perhaps both. I noticed it recently when I was running a lot of iterations of NBRGB script without closing the window. It continued to eat up my RAM and eventually bogged my system down to a stop (and nothing generally is able to do it other the Photoshop CC [and that does have a memory leak or allocation issue]). I will likely add more RAM in the near future to my PC. However, I have not had an issue with calibrating, CC, normalization and integration of more than 300 images (as I did just this past weekend). I did notice that the RAM allocation in PI had gone up to almost 6 gig after running through the stack. So either some VERY large logs (and I doubt large enough to eat up that much RAM) are being stored in the RAM or there is a memory leak I am guessing that there is a bug in PI that is causing a memory leak on some processes to occur.

 

Kaos

Kaos,

 

Thanks. Right now I am integrating 300 16M pixel color DSLR frames using localnormalization and drizzle file. It is not the actual drizzle integration yet and I am actually worried that that will be even worse. It has been running the integration for 24 hours now and it on the last channel. My guess that if it does not crash, it will take another 3-4 hours. This is just not fun.  I went in and set up a SWAP disk of 16GB on my hard drive to see if that helps. From the other posters, and from you, I dont think that it is going to be a huge improvement even buying a new 1500$ laptop with all the bells and whistles. Well see if they swap disk helps first. I am looking at an i7 with a 1TB SSD HD and 2TB HDD and 32GB of RAM as an upgrade, but I would rather put that money into a ASI1600MM-pro!!

 

Wow...24 hours!!!! If it is any help to you, my i7 6700k does normalization or drizzle on over 300 frames from my 1600MM in about 35-40 mins smile.gif. Integration of the drizzle files takes longer obliviously. It may take 1-1.25 hours to integrate that many drizzle files, give or take. Sounds like you may be running an i3 or i5 without hyper threading. I can see that taking forever. 

 

Kaos


Edited by Kaos, 22 February 2018 - 11:08 PM.


#9 dhaval

dhaval

    Vendor

  • *****
  • Vendors
  • Posts: 1074
  • Joined: 21 Jul 2008
  • Loc: Round Rock, TX

Posted 22 February 2018 - 10:59 PM

A better route for using PI is to run it on Linux and build a AMD Ryzen machine (8 core, 16 thread). Swap space in Linux is much better handled than in Windoze. One of the reasons why I decided to go the desktop route was because I had a ZWO ASI 1600 in my future (I eventually did buy it). 

 

 

Dhaval


  • tcchittyjr, bsavoie and RDBeck like this

#10 dkeller_nc

dkeller_nc

    Apollo

  • *****
  • Posts: 1407
  • Joined: 10 Jul 2016
  • Loc: Central NC

Posted 23 February 2018 - 12:52 AM

Just to add another data point, I have an ASUS Zenbook with an i7 6700HQ, 16GB of RAM (DDR4 2133) and a 500 GB PCIEx4 SSD.  300 16MP frames from the ASI1600MM-C takes about 30 - 45 minutes to integrate and generate drizzle data, and about 2 hours to process those 300 drizzle files into a master.

 

There are three hardware attributes that really influence PI's speed - hyperthreading and total number of cores, available memory, and whether or not you're using an SSD for swap files.  Since you are using an i5 laptop, you've got 4 cores available (IIRC, virtual cores are disabled on Intel chips except for the i7 and i9 chips).  If your laptop is using a spinning platter instead of an SSD, that will greatly slow down PI, as the integration routine is heavily dependent on fast swap files.

 

However, within your laptop's hardware limitations, there may be some things you can do to speed up the processing.  Specifically, you need to set up 1 swap file for each physical core of your processor (in this case, 4).  Secondly, you'll want to adjust the default stack and buffer sizes in the Image Integration tool.  Here's Juan's instructions on setting the values of these two parameters.  FWIW, I set the buffer size to 1024 MB and the stack size to 12,000 MB.


  • dhaval, bsavoie and cfosterstars like this

#11 pfile

pfile

    Aurora

  • -----
  • Posts: 4534
  • Joined: 14 Jun 2009

Posted 23 February 2018 - 02:45 AM

the linked article above is somewhat incorrect in saying that the pixinsight swap directories are an extension of RAM. this is only true in the most general sense... what the swap directories really are is where pixinsight stores the processing history of every view open on your PI desktop. unlike some software, the undo/redo history is "infinite" - limited only by how much disk space you have.

 

it's unfortunate that juan chose the name "swap" for these directories, because 'swap' also means virtual memory backing store. this leads people to complain that PI has reinvented the wheel by implementing it's own virtual memory system... but it's much simpler than that - it's just the undo history.

 

the reason SSDs or ramdisks help a lot is because before any process runs on a view, PI writes the state of the view to the swap directory, then updates the view in memory with the results of the operation just done. the only problem with using a ramdisk as swap is that generally speaking ramdisks are small, and when your swap disk is full, PI won't be able to do anything more. therefore you can have fewer active views if you're using ram as a swap disk. SSDs are so dang fast that setting PI's swap directories to reside on SSDs gets you the best of both worlds - large size (more views/history) and speed.

 

i don't think ImageIntegration itself is going to be sped up by faster swap file storage. you can prove this to yourself by moving your swap storage to a faster/slower medium and running ImageIntegration to check the runtimes. alternately if you have access to a tool that will tell you what files PI is opening and closing (like fs_usage on OSX), you'll see that ImageIntegration itself does not read or write any swap files. so if you are having super-slow ImageIntegration this is likely because PI has had to allocate so much RAM that it's causing your OS to page out ram to disk. in that case lowering the stack size would probably be helpful. in dkeller's example above PI is going to try to allocate almost 12GB of memory to ImageIntegration, which is fine if you have 32GB of ram but not so hot if you only have 8GB.

 

if you hover your mouse over the text for Buffer Size and Stack Size in ImageIntegration, you'll see some tooltips that talk about how much RAM is going to be used by ImageIntegration. in particular the buffer size setting is going to be multiplied by the number of input images. so if you have 500 input images with 16MB buffer size (the default), you're asking PI to allocate 8GB of ram... again if that exceeds the amount of real RAM you have in the machine, sooner or later your performance is going to be limited by how fast your boot disk is.

 

rob


  • bsavoie likes this

#12 Salacious B Crumb

Salacious B Crumb

    Mariner 2

  • *****
  • Posts: 213
  • Joined: 13 Apr 2017
  • Loc: Greater Atlanta Area, GA

Posted 23 February 2018 - 07:26 AM

24 hours does not sound right. My Asus has an i5 processor also with no SSD drive, just a regular hard disk. Like dkeller_nc's case, my hundreds of 32MB subs took around 45 min to process.

 

 

- Sal


Edited by Salacious B Crumb, 23 February 2018 - 07:26 AM.


#13 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 23 February 2018 - 01:54 PM

Just to add another data point, I have an ASUS Zenbook with an i7 6700HQ, 16GB of RAM (DDR4 2133) and a 500 GB PCIEx4 SSD.  300 16MP frames from the ASI1600MM-C takes about 30 - 45 minutes to integrate and generate drizzle data, and about 2 hours to process those 300 drizzle files into a master.

 

There are three hardware attributes that really influence PI's speed - hyperthreading and total number of cores, available memory, and whether or not you're using an SSD for swap files.  Since you are using an i5 laptop, you've got 4 cores available (IIRC, virtual cores are disabled on Intel chips except for the i7 and i9 chips).  If your laptop is using a spinning platter instead of an SSD, that will greatly slow down PI, as the integration routine is heavily dependent on fast swap files.

 

However, within your laptop's hardware limitations, there may be some things you can do to speed up the processing.  Specifically, you need to set up 1 swap file for each physical core of your processor (in this case, 4).  Secondly, you'll want to adjust the default stack and buffer sizes in the Image Integration tool.  Here's Juan's instructions on setting the values of these two parameters.  FWIW, I set the buffer size to 1024 MB and the stack size to 12,000 MB.

I just went through setting up the swap files per the link. I am running a test and it already seems much faster.



#14 donlism

donlism

    Viking 1

  • *****
  • Posts: 534
  • Joined: 20 Oct 2015
  • Loc: San Jose, CA

Posted 24 February 2018 - 08:27 PM

Are you using Task Manager and/or Resource Monitor to try and get a sense of what's going on?



#15 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 25 February 2018 - 09:27 AM

Are you using Task Manager and/or Resource Monitor to try and get a sense of what's going on?

I am not really sure what I would be looking for.



#16 targetshooter

targetshooter

    Vostok 1

  • -----
  • Posts: 122
  • Joined: 04 Apr 2014
  • Loc: N Richland Hills TX

Posted 25 February 2018 - 11:34 AM

What is your “Page File Size” and what size is the file your working on? In mb.

page file size effects how much of your memory is used.

You don’t want the system swapping your working file between Ram and HD. It is extremely slow as you have found out.



#17 targetshooter

targetshooter

    Vostok 1

  • -----
  • Posts: 122
  • Joined: 04 Apr 2014
  • Loc: N Richland Hills TX

Posted 25 February 2018 - 11:34 AM

Also what operating system are you using?



#18 dkeller_nc

dkeller_nc

    Apollo

  • *****
  • Posts: 1407
  • Joined: 10 Jul 2016
  • Loc: Central NC

Posted 25 February 2018 - 02:01 PM

BTW - Cfoster (Dennis?):

 

There's a couple of things you could consider when looking for better performance out of PI.  The first (and most expensive) is to build a dedicated box with a AMD threadripper 16-core, 64GB of fast RAM, and an M2.2880 fast SSD like the Samsung EVO and run it with Linux.  That's what I'm considering at the moment, and I figure it will run $2500 by the time I include a case, power supply, etc...  Judging from PI's performance benchmarks of such a system and mine, I figure it will be able to integrate 300 subs from the ASI1600MM-C in about 5 minutes, possibly less.

 

A less expensive option for you that would make a gigantic difference in performance without breaking the bank would be to replace the spinning platter in your laptop with a SATA II 500GB SSD, then use an external WD passport 3 or 4 TB drive for archiving.  This is what I do if I'm processing on my laptop - perform the processing with the data stored on the SSD, then off-load it to the WD harddrive for archiving.  You can plug your laptop's model number into Crucial's website, and be assured that the SSD model will fit your laptop (and if any form adapters are required).  Should cost you about $150 for the SSD, and another $100 for a USB 3.0 4TB WD passport.


  • cfosterstars likes this

#19 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 25 February 2018 - 03:55 PM

BTW - Cfoster (Dennis?):

 

There's a couple of things you could consider when looking for better performance out of PI.  The first (and most expensive) is to build a dedicated box with a AMD threadripper 16-core, 64GB of fast RAM, and an M2.2880 fast SSD like the Samsung EVO and run it with Linux.  That's what I'm considering at the moment, and I figure it will run $2500 by the time I include a case, power supply, etc...  Judging from PI's performance benchmarks of such a system and mine, I figure it will be able to integrate 300 subs from the ASI1600MM-C in about 5 minutes, possibly less.

 

A less expensive option for you that would make a gigantic difference in performance without breaking the bank would be to replace the spinning platter in your laptop with a SATA II 500GB SSD, then use an external WD passport 3 or 4 TB drive for archiving.  This is what I do if I'm processing on my laptop - perform the processing with the data stored on the SSD, then off-load it to the WD harddrive for archiving.  You can plug your laptop's model number into Crucial's website, and be assured that the SSD model will fit your laptop (and if any form adapters are required).  Should cost you about $150 for the SSD, and another $100 for a USB 3.0 4TB WD passport.

BTY - Its Chris.

 

I dont think I want to do the first. I just would rather put more money into the rig right now. I want to upgrade from my DLSR to a 2nd coolled camera for my SCT and galaxy season.

 

What I am considering is getting a new laptop anyway. My E450 is an older wndows 7 system and it has a cracked case, limited to 16GB of memory and its a i5 core. I found a ASUS gaming laptop with Intel i7-7700HQ, 32GB RAM, 1TB NVMe SSD + 2TB HDD, NVIDIA GeForce GTX 1050 4GB, 15.6" Full HD, Windows 10 Home for 1400$. It will have both a SSD and a 2TB drive for archiving. I have multiple 4TB drives for storage of raw data. This should do much better than my E450. 



#20 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 25 February 2018 - 04:05 PM

What is your “Page File Size” and what size is the file your working on? In mb.

page file size effects how much of your memory is used.

You don’t want the system swapping your working file between Ram and HD. It is extremely slow as you have found out.

Where would I find this information?



#21 targetshooter

targetshooter

    Vostok 1

  • -----
  • Posts: 122
  • Joined: 04 Apr 2014
  • Loc: N Richland Hills TX

Posted 25 February 2018 - 07:03 PM

It is in the operating system info. It’s easy to get to but I’ll have to check at a cpu. So at workin the am I’ll get back to you.

That’s also why I asked about the OS? Windows 7 or 10 or between. Your not using Mac I am guessing?



#22 targetshooter

targetshooter

    Vostok 1

  • -----
  • Posts: 122
  • Joined: 04 Apr 2014
  • Loc: N Richland Hills TX

Posted 25 February 2018 - 07:07 PM

This is for Windows 7, 8 & 10

DO NOT CHANGE ANYTHING YET.

 

Right click on Computer and open Properties.
Select Advanced System Properties
Click Advanced tab
Under Performance, click Settings
Under Performance Options, click Advanced tab
Here under Virtual memory pane, select Change
Uncheck Automatically manage paging file size for all drives
Highlight your system drive
Select Custom size
Change the Initial size value and the Maximum size value to a higher value
Click Set
Finally Click Apply/OK all the way.


  • cfosterstars likes this

#23 dkeller_nc

dkeller_nc

    Apollo

  • *****
  • Posts: 1407
  • Joined: 10 Jul 2016
  • Loc: Central NC

Posted 25 February 2018 - 08:19 PM

 

BTW - Cfoster (Dennis?):

 

There's a couple of things you could consider when looking for better performance out of PI.  The first (and most expensive) is to build a dedicated box with a AMD threadripper 16-core, 64GB of fast RAM, and an M2.2880 fast SSD like the Samsung EVO and run it with Linux.  That's what I'm considering at the moment, and I figure it will run $2500 by the time I include a case, power supply, etc...  Judging from PI's performance benchmarks of such a system and mine, I figure it will be able to integrate 300 subs from the ASI1600MM-C in about 5 minutes, possibly less.

 

A less expensive option for you that would make a gigantic difference in performance without breaking the bank would be to replace the spinning platter in your laptop with a SATA II 500GB SSD, then use an external WD passport 3 or 4 TB drive for archiving.  This is what I do if I'm processing on my laptop - perform the processing with the data stored on the SSD, then off-load it to the WD harddrive for archiving.  You can plug your laptop's model number into Crucial's website, and be assured that the SSD model will fit your laptop (and if any form adapters are required).  Should cost you about $150 for the SSD, and another $100 for a USB 3.0 4TB WD passport.

BTY - Its Chris.

 

I dont think I want to do the first. I just would rather put more money into the rig right now. I want to upgrade from my DLSR to a 2nd coolled camera for my SCT and galaxy season.

 

What I am considering is getting a new laptop anyway. My E450 is an older wndows 7 system and it has a cracked case, limited to 16GB of memory and its a i5 core. I found a ASUS gaming laptop with Intel i7-7700HQ, 32GB RAM, 1TB NVMe SSD + 2TB HDD, NVIDIA GeForce GTX 1050 4GB, 15.6" Full HD, Windows 10 Home for 1400$. It will have both a SSD and a 2TB drive for archiving. I have multiple 4TB drives for storage of raw data. This should do much better than my E450. 

 

Yeah, no question that setup will be a massive improvement.  While in general it's preferable to build a desktop for high performance computing (laptops CPUs are slower to save power), if you're going to buy a laptop anyway, might as well kill two birds with one stone.  



#24 cfosterstars

cfosterstars

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1526
  • Joined: 05 Sep 2014
  • Loc: Austin, Texas

Posted 25 February 2018 - 09:50 PM

This is for Windows 7, 8 & 10

DO NOT CHANGE ANYTHING YET.

 

Right click on Computer and open Properties.
Select Advanced System Properties
Click Advanced tab
Under Performance, click Settings
Under Performance Options, click Advanced tab
Here under Virtual memory pane, select Change
Uncheck Automatically manage paging file size for all drives
Highlight your system drive
Select Custom size
Change the Initial size value and the Maximum size value to a higher value
Click Set
Finally Click Apply/OK all the way.

Yes I did this and set the initial size to 8GB and max to 16GB



#25 targetshooter

targetshooter

    Vostok 1

  • -----
  • Posts: 122
  • Joined: 04 Apr 2014
  • Loc: N Richland Hills TX

Posted 26 February 2018 - 08:18 AM

 

This is for Windows 7, 8 & 10

DO NOT CHANGE ANYTHING YET.

 

Right click on Computer and open Properties.
Select Advanced System Properties
Click Advanced tab
Under Performance, click Settings
Under Performance Options, click Advanced tab
Here under Virtual memory pane, select Change
Uncheck Automatically manage paging file size for all drives
Highlight your system drive
Select Custom size
Change the Initial size value and the Maximum size value to a higher value
Click Set
Finally Click Apply/OK all the way.

Yes I did this and set the initial size to 8GB and max to 16GB

 

So this means once you get to a 8mb file it starts using the HD as memory and is swapping your working file between the HD and RAM.

You can try entering "0" and see if it helps. This tells the processor to use all of the memory for your working file. 

Try it with minimum items open while you are editing.

Just write down what it is now so you can put it back if you don't see a difference or it crashes. It shouldn't crash. This page file is used normally with 32GB, 64GB or more. I have 32GB on my work CPU and my files are normally in the 20-25MB size.

Good Luck, I hope it helps.




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics