Anyone using an SSD to Run Logos 4 on a Windows Platform?

Page 2 of 2 (34 items) < Previous 1 2
This post has 33 Replies | 5 Followers

Posts 11
Brian McKenzie | Forum Activity | Replied: Wed, Sep 1 2010 9:38 PM

While trying out Logos with my new SSD I found a strange thing. Logos seems to be using less than 1/2 of my computer's CPU capacity even after that hard drive bottleneck has been removed. I have run several exegetical studies of complete chapters and Windows Resource Monitor reports that the SSD drive will run at 3-8MB/sec for a second or so at the start and then drops down to 20-80KB/sec for the rest of the time the exegetic study is built up. However, the CPU never runs full out. Instead it bounces around at about  35-45% for the entire time the study is built up.

It appears that the hard disk is not the slow down in exegetic studies, at least not after the first couple seconds when the hard drive up loads the raw data. What else could be holding the CPU back? Could it be that the Logos programmers have purposely set Logos to use less than 1/2 the CPU power so that the computer will be responsive to the use during the minute or two it takes to build a large exegetical study on an entire chapter? Or is there some other bottle neck holding the CPU back?

A very curious Logos user,

Brian M

Posts 8660
TCBlack | Forum Activity | Replied: Thu, Sep 2 2010 5:53 AM

Brian,

One of the programmers may come along and answer this, but I will say this: 

During the early beta testing we all had some serious questions about multithreading, CPU utilization etc.  and we were assured at the time that Logos will make full use of the CPU as it needs it.  Typically the highest CPU hit will be during indexing (how did THAT go on your SSD?)  After that some of the graphical reports which are harder on your GPU than your CPU will take some effort.

I suspect that what you're seeing is a situation where there is a sudden burst of many threads who are completing their work early, while the remaining threads take longer but use less CPU.

It is just a guess based upon what I now vaguely remember from last years early beta.

Hmm Sarcasm is my love language. Obviously I love you. 

Posts 11
Brian McKenzie | Forum Activity | Replied: Thu, Sep 2 2010 8:49 AM

Thomas:

Thanks for the quick response. Based on your comments I paid more attention to the details of the individual CPU cores loads when starting Logos this morning. And to get a second perspective I had Logos do an exegetical study work up on a new chapter (just in case it had cached results from a previous chapter, or in case Windows had cached the disk access for that previous chapter work up). Here are the results.

During the 20 second start up all 4 of my cpu cores increased their load. They bounced all around in percentage load during the start up. For just under 1 second the combine load of all 4 was about 90% but for most of the 20 seconds the total cpu load was well under 50 percent. As for the disk's bandwidth that was hard to observe because it bounced around even more than the CPU load, but one thing was very clear -- the SSD was not running anywhere near full bandwidth for long enough to register on Windows Resource Monitor's graph. As for the Resource Manager's numeric display of SSD bandwidth, that also bounced around a lot, but it too was not near maximum for any significant length of time. So if the CPU is only running at 90% of load for 1/20 of the start up time and the SSD bandwidth is equally under used, where is the bottleneck. It can't be my graphics card, because for 1/3 of the start up all that is displayed is the Logos 4 logo, for the next 1/3 it is just a blank Logos window with the little Windows wheel rotating and only for the last 1/3 is material being displayed to the screen and even then it is just test that is being written.  Oh, I wonder if the bottleneck could be the RAM. My system has DDR3 which Windows Performance Experience rates as over 7. Start up behavior is a real puzzle.

I am also puzzled by Logos's behavior during its preparation of lengthy exegetical studies although it does appear that there is some advanced load leveling (likely to keep a good portion of CPU available to the user for other tasks). Oh, for these tests, I had Logos work on chapters I have not studied for months so avoid any possibility of caching of previous results.  On my system Logos takes between 30 and 60 seconds to prepare exegetical studies on entire chapters. Here is the pattern I see.

For the first 5 to 7 seconds after I click the start arrow to begin the study, very little seems to happen. CPU load increase a very small amount but little else that I can see. Then after that initial hestitation of 5 seconds or so, the SSD gives a major burst (at a rate several MB/sec) that lasts 1/2 to 1.5 seconds. About the same time the CPU load increases to around 25%. Then the SSD load drops 10 K/sec or less (often zero for a second or two at a time).

As for the CPU, its load is low until the SSD has supplied its initial burst of data. Then the CPU settles into a 40% load that is fairly well distributed across all 4 cpu cores. Typically, the first core has a load of about 60%, the 2nd about 40%, the third about 30% and the 4th about 40%. The individual cores bounce around by 10% above or below the above mentioned averages. What is remarkable is that the total of combined load of the 4 cores is much more consistent than the individual loads of the cores. It would appear that if one core is having a temporary up spike of 10% another core is usually having a temporary valley of 5 or 10%. So although the 4 cores bounce up or down by 20%, the total of the 4 cores only bounces up or down by 5% or so. Looks like some sort of intentional load leveling -- and a remarkably good one at that! The only other alternative I can think of is that my RAM can't supply the data to and from the CPU any faster. In any case the CPU is speeding about 60% of its time being idle for the major of the exegetical preparation time. And the SSD hard disk is spending 95-98% of times time being idle.

Oh, one more note: During the exegetical study build up, none of the cpu cores ever hits 100% load. There is a brief moment around the time the SSD uploads data to RAM that the first core hits 70% but then that core settles back to the 60% load range as noted above.

If this load leveling behavior is built into Logos 4's programming code, then I'll just accept that this is as fast as I can get Logos to work. However, if there is some other bottleneck on my computer, I would like to know in case I can fix it.

Thanks again for your response.

Brian M

 

Posts 8660
TCBlack | Forum Activity | Replied: Thu, Sep 2 2010 9:37 AM

Great analysis Brian. 

I'm going to sit back now and twiddle my thumbs while I wait for a comment from Bradley to see what it all means.  Big Smile

Hmm Sarcasm is my love language. Obviously I love you. 

Posts 11
Brian McKenzie | Forum Activity | Replied: Thu, Sep 2 2010 9:44 AM

Hi Thomas:

I've just sent an email to Logos support asking for someone to check out this thread. Hopefully, we will get some clarity in a day or two.

Brian

Posts 3751
BillS | Forum Activity | Replied: Thu, Sep 2 2010 1:04 PM

Brian McKenzie:
if there is some other bottleneck on my computer, I would like to know in case I can fix it.

In an unrelated post on Smart Tags in this thread (http://community.logos.com/forums/p/22045/167803.aspx#167803 - a response to a question whether a 64 bit build of L4 would speed things up), Bradley offered an insight into 32-bit memory utilization that may have some bearing (bold underline mine, not his):

Bradley Grainger:

A 64-bit build of Logos 4 would probably eliminate Out of Memory Exceptions because it would increase the available address space from 2GB to 16TB. (Even if you have 32GB of RAM, Logos 4 is limited to 2GB due to the limitations of 32-bit processes, and the actual limit is closer to 1GB due to the overhead of .NET and other DLLs that are loaded.) However, allowing the application to use an (effectively) unlimited amount of memory isn't necessarily a good idea; on most systems it could cause a huge amount of paging, which would drastically slow down the application. Additionally, slowly leaking memory would slow down each .NET garbage collection, causing the application to run slower and slower over time. Sticking with a 32-bit process is probably better because it requires us to keep our long-term memory usage low, leading to an overall better application.

Furthermore, in a 64-bit process, each pointer (or reference, in .NET) is 64 bits, rather than 32, which (simplistically) doubles the memory usage of the application. The 32-bit process might run just fine on a computer with 2GB of RAM,  but the 64-bit process could require 3-4GB just to achieve the same performance. This may not be a big concern once all new desktop computers are shipping with 8GB+ of RAM, but it's still an issue right now.

It's not clear that running a 64-bit process delivers obvious immediate performance gains. In some cases, loading and processing 32-bit data incurs a performance penalty.

As I understand the gist of his comment, 32 bit may be optimum for now given inherent limitations of a .NET framework & other DLLs. All that makes me wonder if you've hit the upper limit of performance in that same framework, by essentially minimizing all the other wait times.

Grace & Peace,
Bill


MSI GF63 8RD, I-7 8850H, 32GB RAM, 1TB SSD, 2TB HDD, NVIDIA GTX 1050Max
iPhone 12 Pro Max 512Gb
Fire 10HD 64GB 7th Gen

Posts 11
Brian McKenzie | Forum Activity | Replied: Thu, Sep 2 2010 2:43 PM

Hi Bill:

Thanks for pointing me to the advantage of 32-bit applications as noted by Bradley. I will keep that in mind for other software that I buy in the future.

As for Logos 4, I don't think 32-bits is the limiting factor at least on my computer. After reading your post I shut my computer down completely and restarted (a cold boot rather than sleep or hibernate). On boot up here is the memory usage according to Window 7's Resource Monitor (memory tab):

In Use = 1.5GB, Modified=0.16GB, Standby=0.68, Free=5.8GB. I assume Standby are items that  Windows has prefetched from the hard drive just in case I open it. I am sure Outlook is prefetched since I open it a lot. Perhaps Logos 4 is also prefetched as well).

After loading Logos 4 the memory use is as follows:

In Use = 1.94GB (up 440MB), Modified=unchanged, Standby =1.13(up 450MB)

Then I had Logos do an exegetical study for all of Titus 1 which took about 50 seconds, a typical time and the CPU load was also around 45% which is about 5% higher than I observed in my prior tests using chapters from Romans and 1 Peter. As for memory, after the exegetical study was complete the memory was as follows:

In Us = 2.01GB(up 70MB), Modified unchanged and Standby=1.16GB (up 30MB).and Free RAM was 4.8GB.

In summary, it does not appear that Logos 4 (on my computer) is using anything close to the 2 GB limit for 32-bit applications. Instead it appears to take only 500MB or so from the above observations.

Since there is a possibility Logos 4 was already partly in the "standby" portion of RAM, I had a look at the Windows Resource Monitor (Memory tab) and saw that the Logos4.exe "process" has a 402,566 KB "working set" commitment with 86,624KB of hat being "shareable" and  315,948KB is "private".  So it looks like Logos4 is a compact program that is not pushing any 32 bit limits. My assumption is that the "shareable" portion of the RAM usage might be portions of the NET framework that Logos uses but would also be usable by other programs.

Blessings to all,

Brian

Posts 11
Brian McKenzie | Forum Activity | Replied: Wed, Oct 13 2010 9:14 AM

Just finished re-indexing after installing new books. The Logos icon in the tool tray estimated it would take 95 minutes, but it finished in 29 so indexing looks to be about 3 times faster. My CPU average load was about 25-30% so the bottle neck appears to still be the disk drive, but at least the SSD is a faster than what I had before.

Posts 236
Michael March | Forum Activity | Replied: Wed, Oct 13 2010 6:22 PM

Ok, just a word here.  I was so worried about my new laptop's ability to run logos 4 before it arrived after getting caught up in the race for speed that I tried to cancel the order (unsuccessfully).  When the laptop got here (lenovo x201) I just couldn't help trying it out, even though I thought I'd probably be sending it back.   Now when I see the actual speed this laptop works logos without a dedicated graphics card and "only" 6 meg of ram and a paltry 7200 rpm hd, I have to admit that this is truly more than fast enough for me.  

At some point we're just doing benchmarks that have little or no meaning.  I admit to being something of a computer geek and also to a love for fast to the extent that I ALWAYS overpay for the latest processor that never really gives enough bang for the extra buck.  But seriously--in actual use, that is to say, real life research, working on a sermon, trying to understand the Word even better, does 10, or even 20 seconds honestly make a difference (consider now that you can already check the first resource while you're waiting this extra 20 seconds for 50 more resources to open)?

I am the first to admit that checking for speed and benchmarks and windows experience indices are really fun, but in actual, practical, sermon preparation, I spend FAR enough time reading articles and resources that a ten second burst here or there makes no real difference at all.

Not trying to start a fight or anything, just a dose of reality perhaps...

Windows PC - Android Phone - Surface Pro 4

Posts 11
Brian McKenzie | Forum Activity | Replied: Thu, Oct 14 2010 7:27 AM

Michael:

Very good points. The money saved on not getting an SSD could buy quite a few books. Also, my SSD reports show that the speed up by switching to an SSD is not as dramatic as one might expect, so this may be helpful to others who might wish to save money by using a traditional hard drive.

I agree with you that for a long working session when studying a passage for the first time, Logos's long build up does not matter at all, especially when one can start reading the first commentary while Logos finishes the build up. But I frequently want to check a single fact about a passage I know well so I am not interested in re-reading the first commentary Logos makes available. In these cases the 60 second time it used to take Logos to re-build a passage study or exegetical study from a traditional hard drive always felt like an irritation since it was only going to take me 3-5 seconds to verify that single fact. And if I want to verify a fact from 3 different passages, I found Logos's 60 second build ups to be very, very irritating. Now that the SSD has reduced the wait to "just" 20 seconds per study build up Logos's speed has become tolerable. This of course is a personal thing. Others might well have a study pattern better suited to Logos's speed characteristics with a traditional hard disk.

Glad to hear you are fully happy with the 7,200 speed HD.

Blessings,

Brian

Posts 236
Michael March | Forum Activity | Replied: Thu, Oct 14 2010 8:22 AM

I see your point Brian, and it's a good one.  I actually had a second bible study program for quick word searches and lookups just because of that slowness for a simple task.  Now it seems I just leave logos up and running.  But I agree with your point when you're just trying to find a bible passage, L4 is a bit cumbersome to open up.

Windows PC - Android Phone - Surface Pro 4

Posts 11
Brian McKenzie | Forum Activity | Replied: Thu, Oct 14 2010 9:12 AM

Michael, like you I used to use another Bible program for quick look ups and searches rather than opening Logos. Since moving to Windows 7, my non-Logos Bible program does not work. Rather than upgrading that other program, I used the saved money to help pay for part of the SSD cost. I was expecting the SSD to make Logos very snappy. In reality it did not help as much as expected, although I am pleased enough I would do it again. -- Oh, if I were to do it again I might cut my cost in half by getting a small SSD just for Logos and leaving the operating system and other programs on a regular 7,200 hard drive.

Blessings,

Brian

Posts 9
John Hopcraft | Forum Activity | Replied: Tue, Jul 17 2012 6:59 PM

I just got a new  desktop with an SSD (64bit win 7, i7 processor 16gb RAM, 256 gb SSD Dual monitors)  L4 runs just beautifully in this new system.  The SSD has made a very worthwhile difference.

Posts 27232
Forum MVP
Dave Hooton | Forum Activity | Replied: Tue, Jul 17 2012 8:34 PM

John Hopcraft:

I just got a new  desktop with an SSD (64bit win 7, i7 processor 16gb RAM, 256 gb SSD Dual monitors)  L4 runs just beautifully in this new system.  The SSD has made a very worthwhile difference.

It does indeed. I got an Intel 180 GB SSD and put it into my laptop. According to the Windows Experience Index the speed of memory is limiting things and the SSD has the highest rating (the HDD had the lowest rating).

Dave
===

Windows 10 & Android 8

Page 2 of 2 (34 items) < Previous 1 2 | RSS