Interesting! Benchmark Results for Game Optimization scripts(sort by name ascen)

[Wolfmight]

This is very interesting! Apparently sorting an entire game program directory by name may be slowing down your overall game's performance!

Sure, the averages are close, but check out the max and minimum FPS! The Sort-by optimization caused both to drop tremendously.

View my two attached photos of benchmark results for the PC game: Batman Arkham Asylum.

Game Settings:
Resolution: 1680x1050 (16:10)
Detail Level: Max
Bloom, Dynamic Shadows, Motion Blur, Distortion, Fog Volumes, Spherical Harmonic Lighting, Ambient Occlusion: Yes/Max
Anti-Aliasing: 8x

Test System:

CPU: AMD Athlon64 X2 Dual Core 3.0Ghz 6000+ Windsor
Motherboard: Gigabyte M57SLI-S4 nForce 570 SLI
Ram: 6GB DDR2 400Mhz 5-5-5-18-25-2T
Graphics Card: Radeon HD 4670 512MB GDDR3 PCI Express 2.0 x16
Hard drive: Western Digital Raptor 74GB 10,000rpm Sata 1.5 (53GB Free Space, Windows XP-64 Fully Updated Everything)
Monitor (bragging on this last one): 28" 1080 LCD with desktop at 1920x1200 res O.O


Notes:
-The game includes it's own benchmark tool that loads the same exact tests each time.
-The game has been fully updated to the latest version v1.1
-The DailyDefrag was completed and tested first, then the Optimized Game script (basically doing the same as the daily defrag, except Zone 4 specially optimized the game directory with a full "SortBy File Name Ascending" script, which in theory, was supposed to speed up access times and reduce needle movement by having the entire game in one area of the disk.)


1RST Image = After Daily Defrag
2ND Image = After Sort-by File Name Ascending Optimization of Game Directory 

[amk]

Any reordering may decrease speed. "Strong" optimization need internal game information about using its files, that not accessible for defragmentator. Or very long information collect phase before optimization.

[Wolfmight]

another update, I re-ran a Monthly Optimization and regained the first test's performance with a Max FPS of 192, and Min of 29 with average 52. 
My results go to show the officially provided scripts are still more efficient, because so far, my 3 tests have proven Daily & Monthy defrag scripts to provide better results for Batman Arkham Asylum.
I still need to test them with Fallout 3, because I'm still unsure as to which provides better results for that game, but the Optimization scripts do improve some aspects for certain in Fallout 3's response for certain applications...i.e. much faster load times for certain interfaces.  Actually, I noticed opening and closing interfaces performed better in Batman as well under the sort optimization script.  However - the overall fps did stutter more (which is much more important-i.e. 90%+ of your experience) and that's what sparked my interest in testing these benchmarks today.  I first played the game after a Daily Defrag with perfectly smooth gameplay  (I was already very impressed and didn't want to mess with adding a custom script at the time) and then I used an optimization script to see if I could improve things further, when in fact, I was starting to notice stutters.  Then after running these Daily/Monthy defrags again, the stutters are gone. 

Here are two threads covering the "sort-optimization" theory scripts for faster/smoother PC Games. Also remember, we need more benchmarks in other games to further test these theories and perhaps generate a collected list of which games benefit and which don't.

http://www.mydefrag.com/forum/index.php?topic=2777.0
http://www.mydefrag.com/forum/index.php?topic=2393.0

Perhaps someone should test 3DMark with these methods too. See how your score is effected with 3-5 tests going back and forth (depending on how much time you have between work).

[J.O.A.T.]

To get somewhat correct data from the benchmark it should be run on a newly rebooted machine to avoid having any files etc. in windows internal cache.

[Darlis]

You installed the game on the system partition? Does you script still contain the ImportListFromBootOptimize() in zone 3?

Also the results would be more accurate, if you would run the test at least thrice with every file layout and then average them out.

[poutnik]

Interesting is, what would be the difference, if just created not sorted zone for the game.

[jeroen]

Apparently sorting an entire game program directory by name may be slowing down your overall game's performance!

I am happy that the standard MyDefrag script is performing so well for you. Perhaps your custom script has not included all the files that are used by the game, resulting in some files far away in a different zone and slowing it down. Perhaps your game has files in multiple folders instead of just a single folder, or perhaps you also have to include the DirectX files, the video and audio drivers, the registry, things like that.

[Wolfmight]

To get somewhat correct data from the benchmark it should be run on a newly rebooted machine to avoid having any files etc. in windows internal cache.

I did this, yes.  Fairly new installation of Windows as well with very little software installed on it. Nothing running in the background to potentially use the CPU or Graphics Card.

You installed the game on the system partition? Does you script still contain the ImportListFromBootOptimize() in zone 3?

Also the results would be more accurate, if you would run the test at least thrice with every file layout and then average them out.

Yes.  However, I do have a separate 750GB Sata3.0 drive that I can test the game out on along with others to see a difference in performance.

Apparently sorting an entire game program directory by name may be slowing down your overall game's performance!

I am happy that the standard MyDefrag script is performing so well for you. Perhaps your custom script has not included all the files that are used by the game, resulting in some files far away in a different zone and slowing it down. Perhaps your game has files in multiple folders instead of just a single folder, or perhaps you also have to include the DirectX files, the video and audio drivers, the registry, things like that.

Very true about the DirectX, drivers and registry yes. 

[lh]

Speed is dependent on several things here.  Assuming all files are totally defragmented. 

• Position on disk is the most important with most HDs.  As the 'front' of the drive is fastest and the end of the drive the slowest.  Sometimes by as much as half as fast.
• Clustering/ordering is also important.  With the default scripts that clustering is by name.  If done correctly this minimizes head movement.  Which produces latency.
• Is it in the cache?  If you can get this the access time is very low.  Then position and order does not matter at all (same thing with an SSD).

If you look at what I did with my boot zone at one point using procmon.  You could do something similar. 
http://www.mydefrag.com/forum/index.php?topic=2686.0

You could sort the list instead of 'total' access time like I did but total number of accesses (both read and write).  Putting the file that has most accesses first in a zone.  Then pick where you want the zone of sorted files.

[poutnik]

Speed is dependent on several things here.  Assuming all files are totally defragmented. 

• Position on disk is the most important with most HDs.  As the 'front' of the drive is fastest and the end of the drive the slowest.  Sometimes by as much as half as fast.
• Clustering/ordering is also important.  With the default scripts that clustering is by name.  If done correctly this minimizes head movement.  Which produces latency.
• Is it in the cache?  If you can get this the access time is very low.  Then position and order does not matter at all (same thing with an SSD).

I would say

• absolute positions make speed of transfer
• relative positions make speed of access
• there is need to distinguish between clustering/ordering. Most of the speed can be done by being on fast tracks and close each other, not necesserily for being ordered. Ordering expects the files are read one after other without disturbation. It is quite rare case.
  If disk is working with other files, MFT or folders, the effect of sorting is lost.
• Concerning sort by boot order, it can be disturbed by accessing not yet read part of MFT and folders, or if some files are accessed multiple times, like registry.
• Cache decreases the effect of disk optimization, but finally the disk has to be accessed.(speaking about classical disks). Only if disk is not accessed at all, transfer and access speed does not matter at all.

[TheAnonymous]

With 6GB of memory I'd expect a lot of data would be in the page cache regardless..


I wouldn't be surprised if the severe degredation in performance was related to something entirely different.

[BloodySword]

A good programmes game notices if there is plenty of ram and while loading the maps, the files will be stored in memory for faster access. How is it with 64 bit? Are there games based on 64 bit out there? Games written in 32 bit might have big problems due to the 2 GB limit for any process.

I don't know much about games because I'm not such a gamer...

[TheAnonymous]

A good programmes game notices if there is plenty of ram and while loading the maps, the files will be stored in memory for faster access. How is it with 64 bit? Are there games based on 64 bit out there? Games written in 32 bit might have big problems due to the 2 GB limit for any process.

I don't know much about games because I'm not such a gamer...

The page cache is OS controlled, as it's part of the kernal. When files are accessed on the disk, they are placed into unused memory for faster access, but that data is cleared if memory is needed for more urgent things. With the NT kernal, cached memory is still reported as "free" memory, so it causes confusion on how memory management works in Windows. That's what I meant.

As for the game itself utilizing memory, I've been PC gaming since 2000 and for the past two years I haven't seen a single game htat uses over 1GB of memory for it's working set.

So, after a reboot, when all memory is cleared, I'd expect game performance to suffer for a little while as data has to be put into memory, but after exiting the game, part of it will reside in the page cache. So, for the above benchmarks, I'm wondering if he did a reboot between each defragment. If he didn't reboot, the first benchmark he did would get lower results as some data has not been placed into the page cache. Where the second benchmark would get better results regardless of what he did.

[Kasuha]

Unless your game is able to show you something like FPS graph or histogram, the only reasonable data you get from the benchmark is average framerate which is 50~52 - that's pretty consistent to me and probably given by your monitor refresh rate.
Both minimum and maximum framerate are matter of a single frame. It tells nothing except that your PC choked at some random point on something that may be completely unrelated to your disk.

[Wolfmight]

Unless your game is able to show you something like FPS graph or histogram, the only reasonable data you get from the benchmark is average framerate which is 50~52 - that's pretty consistent to me and probably given by your monitor refresh rate.
Both minimum and maximum framerate are matter of a single frame. It tells nothing except that your PC choked at some random point on something that may be completely unrelated to your disk.

Good point.  A detailed line graph would prove more useful, allowing us to see "all" of these chokes in the FPS to see if we can find a more obvious difference in the outcome performance of these sort methods.    I'm not sure if Fraps has this kind of stockmarket line graph function lol, but surely another program might.  .

Well it would be better to record videos of these benchmarks as well to give people their own judgment of the actual performance to better analyze this theory of hard disk performance.
I do remember the 2nd benchmark choking a lot more though, and that's where a video would come in handy. But the footage would best be record externally (as high of quality as possible) and not from a software based method on the same computer in order to preserve genuine performance.  Although screen capturing software can record footage of gameplay and clearly present different levels of performance for each system (youtube proves this) they all require resources from the same system in order to accomplish their recording process

Also, some lesser popular games may hold poor programming that causes performance problems (more choke issues) and never become patched due to lack of publicity and sales.  Many times a ported game is left alone in the dark as well, as long as they game "does" allow some to complete it fully. (ouch)  This particular game, Batman Arkham Asylum, is fairly popular at the moment and the software company has released a huge (100mb+) patch for the game, so I wouldn't consider faulty programming to be near as much of an issue.  The game really does run much smoother than a lot of PC games actually, perhaps it utilizes more innovative programming than usual for PC games lol.