Increase Software Speed

mikec82

Honorable
Feb 7, 2014
24
0
10,510
I am using AABBYY software to take thousands of pdf documents, OCR them, and make them text searchable. It is a 32-bit program. I am running it on a 64-bit Windows 8.1 machine, 32gb RAM, i7 3.5Ghz. It's the only program I have open (other than services/processes). It is still only using 12% of the CPU (84% of CPU is unused). I set the Priority to High for all programs running that were related to AABBYY, but it is still running at the same levels. Any other way to increase the speed?
 

onichikun

Distinguished
Nov 13, 2009
304
1
18,860


Probably not. The high priority is just a suggestion to the scheduler of the OS to schedule the process more often, if you don't have a lot of multitasking going on in the first place, you won't notice much difference.

It is more likely that the program you are running is reaching it's limitations. the 12% CPU usage could be due to the program being a primarily IO bound problem, with the CPU waiting for data to be read from the HD.

It may be possible to increase speed by improving your HDD performance, or cache the files in memory first using something like ramdisk.
 

mikec82

Honorable
Feb 7, 2014
24
0
10,510
Interested in the ramdisk option, but unfamiliar with it. Did a little reading up on it. Dataram makes software for my 32GB machine. Is there a better software out there for ramdisks?
 

onichikun

Distinguished
Nov 13, 2009
304
1
18,860


It appears Dataram is a good option for Windows. If you are OK with using Linux, and if you are able to get AABBYY for Linux for free with your existing licenses, you can setup a Ramfs in Linux for free :)
 

Eximo

Titan
Ambassador
I believe ABBYY Finereader is a single threaded application. An i7 is capable of running 8 threads simultaneously. 12.5% is exactly 1/8. Only way to really improve performance would be an even faster processor and there really aren't any.

Can you install multiple copies?
 

onichikun

Distinguished
Nov 13, 2009
304
1
18,860


Yeah it the process is single-threaded that is going to be a problem, but ABBYY's website seems to say their OCR engine is multithreaded, so there may be an option that the OP is missing to enable multiple threads.
 

mikec82

Honorable
Feb 7, 2014
24
0
10,510
I see a link on multithreading OCR now, but don't know nearly enough about computers to be able to make sense out of it.

http://knowledgebase.ocrsdk.com/article/794

Right now, I am trying out RamDisk. I created a 10GB disk and put 1,000 of the PDFs I want to OCR in it. I went into ABBYY and set the location to create the new OCR documents to also be on the RAMDisk. Lastly, I changed the TMP folder location so it would create the temp files on the RAMDisk. I am getting the exact same speed out of the program. Am I missing something?
 

Eximo

Titan
Ambassador
That knowledge base article is pretty heavy. Apparently you would need to switch to Windows Server and run things through command line, it would take a bit of research to implement that, but it would increase the speed about 8 times.