No announcement yet.

Newest Beta VS. RAMDisk

  • Filter
  • Time
  • Show
Clear All
new posts

  • Newest Beta VS. RAMDisk

    I've been seeing BIS has made vast improvements to the engine in terms of its handling of textures and memory hungry resources and figured it might be time to give these improvements a little test. Below are the results along with my testing methodology and as many relevant details as I could possibly think of.

    For those of you who even care to use RAM disks, this thread is for you. For those who don't, well, here is how the newest beta performs compared to something that smokes normal hard disc usage. Hopefully one of you lucky guys that own an SSD can put up some comparisons so we can see how far BIS has progressed.

    Test Hardware & Software
    Motherboard: EVGA NVIDIA 750i SLI FTW
    CPU (Processor): Intel Core 2 Quad Q9550 @ 3.4GHz
    RAM (Memory): Mushkin 2x2GB Redline DDR2 1067MHz
    HDD (Hard Drive Disk)*: Seagate 7200.11 500GB 7200RPM 32MB Cache
    PSU (Power Supply Unit): Corsair TX750 750W
    OS (Operating System)**: Windows Vista 64 Bit SP2
    GPU (Graphics Processing Unit): 2x SLI EVGA NVIDIA 9800GTX+ @ 792 Core, 1965 Shader, 1152 Memory (MHz)
    GPU Driver Version: NVIDIA WHQL 257.21 (June 15, 2010)
    Arma II Beta Version: 71734

    Test Applications: Arma II, Arma II Mark, FRAPS

    Shortcut Target Line***: "C:\Program Files (x86)\Steam\steamapps\common\arma 2\beta\arma2.exe" -nosplash -mod=beta;G:\@RAM disk2;@TGMaps;@ACEM; -cpuCount 4 -winxp -noLand -exThreads 7 -maxmem 4096

    Video Settings:
    Texture Detail: Normal
    Video Memory: High
    Anisotropic: Very High
    Anti-Aliasing: Very High
    Terrain Detail: Normal
    Objects Detail: Normal
    Shadow Detail: High
    Post Process Effects: Disabled
    3D Resolution: 1680 x 1050 (Native)
    Notes: *Primary HDD which contains Arma II and the OS, a second Western Digital Black 640GB 7200RPM 32MB Cache HDD is present within the system.
    **The OS has all of the latest hot-fixes and updates installed as of June 24 2010
    ***To simulate a real playing environment, I've run the TG standard addons in the target line which may cause lower frame rates/scores than in vanilla Arma II.

    RAM Disk Settings
    RAM Disk Software: DATARAM RAMDisk
    Amount of RAM Allocated to RAM Disk: 1510MB
    Files on RAM Disk: plants2_Bush.pbo, plants2_Tree.pbo, Road2.pbo, structures.pbo

    Testing Methodology
    Arma II Mark was run a total of 5 times for each condition in order to net the final results. The first trial for each condition was run off of a clean start of the game with nothing being loaded prior to the test whereas the following trials were all run immediately after with a simple ESC > Restart which simulates a more realistic scenario after playing for a short period of time.

    Test Results
    Before you jump to the results, I'm going to quickly explain what the numbers you are about to see really mean. The trials are the OFP Marks of each test run (5 in all) which is generated by each test within Arma II Mark being added, averaged, then multiplied by 100. For example, Trial 1 under the RAM disk was calculated by Test 1 (30.1769) + T2 (36.7145) + T3 (30.3787) + T4 (46.6926) + T5 (23.0964) / 5 * 100 = 3341.182. I have also taken the time to average out each test individually as some tell more of the performance of memory loading (which would utilize the RAM disk) more than others.

    RAM Disk
    Trial 1: 3341.18
    Trial 2: 3850.38
    Trial 3: 3591.09
    Trial 4: 3632.96
    Trial 5: 3458.20
    Test 1: 30.0899
    Test 2: 36.6336
    Test 3: 29.3964
    Test 4: 45.9137
    Test 5: 36.7045
    OFP Mark: 3574.76
    Without RAM Disk
    Trial 1: 3285.78
    Trial 2: 3817.76
    Trial 3: 3822.31
    Trial 4: 3875.76
    Trial 5: 3768.13
    Test 1: 30.1181
    Test 2: 37.4225
    Test 3: 30.3924
    Test 4: 47.3320
    Test 5: 40.4324
    OFP Mark: 3733.95
    Here we can clearly see that running without a RAM disk is actually beneficial... but how is that possible? Possible explanation below...

    First off, I must say that I was personally quite surprised with the results as I figured things would run smoother loading off of a vastly superior source transfer speed wise. A quick disc benchmark shows exceptional speeds coming from the RAM disk in comparison to my HDD but there is much more to it than simple benchmark speeds.

    The RAM disk would probably have won this little head to head test had I been in possession of more RAM as once the game was running; my system was using about 90% of the total RAM. The game still loads the items on the RAM disk onto the remaining system RAM because a RAM disk is treated the same way a hard drive or solid state drive is. There is nothing telling the game (as far as I know) that the files are already placed on the RAM and thus it loads it in again. This turned out to be a major problem which is quite noticeable in the 3rd test and on where the average FPS began to drop due to the system choking up. Still, the first and second tests which truly demonstrate the strength of the RAM disk don't show a major difference compared to the hard drive anyways.

    The real benefit of the RAM disk is initial loading and caching of textures which is done quite well now by the Arma engine, however, is this benefit of gaining a couple frames per second worth the time required to setup the RAM disk itself? Personally, I don't think it is and it definitely isn't a reason for you to purchase more RAM for this purpose alone. A RAM disk will help the player who frequently pilots through his first flight over of a town but only if there is more than enough RAM to spare in the system.

    Overall, a RAM disk isn't quite what it used to be in the world of Arma but if you already have a monster machine and can't find a use for all that excess RAM; go for a large RAM disk to help those loading times. If you don't have a monster rig though then you'll be just fine with the beta patch and the vast number of add-ons which will help you increase that FPS substantially.

    Below are the results of the first trials which compare the beta against the RAM disk the best as the first trials were straight off of fresh game starts where nothing was loaded into the GPU RAM or the system RAM (the splash screen loaded an empty ocean instead of one of the islands to contribute to this outcome). Test 5 is the major showcase of the RAM disk as test 5 is a camera that descends down on Chernarus from a high altitude at a very fast speed forcing textures and their LODs to load and change very quickly. This means many reads from the HDD or RAM disk and would be the effect a pilot would experience going over a new city/town (aka loading/caching lag). This lag used to be horrific but with the improvements BIS has made it is almost non-existent during actual game-play or even this benchmark.

    w/ RAM Disk:
    w/o RAM Disk:

    Before you throw the idea of use a RAM disk in the gutter (if you haven't already) remember that it isn't pure performance which it provides. The RAM disk allows textures and LODs to load faster thus rendering the game quicker (more prettiness :D). When using the RAM disk I noticed buildings change from that grey block to an actual textured building much faster and although this isn't witnessed much in actual game-play, LOD switching is. When zooming in on Chernogorsk I noticed buildings changing to their higher resolution detail modes MUCH faster when using a RAM disk. So, if you have memory to spare, using it in a RAM disk may also help increase the visual appeal of the game and its immersion factor.

    RAM disks aren't out of the game yet nor do I believe they will be anytime soon but, Arma II has come a very long way since release and maybe now you'll be able to get those friends who shied away due to slow performance to come back and give it another shot. Just to give you an idea of the advancement of Arma II, around the release of patch 1.05 I had run this same benchmark (off of my HDD) and scored about 2800 after a couple hours of strenuous tweaking. Also, back around that time it was a miracle to be able to walk through Chernogorsk with any view distance above 2000 for me but now I can run a view distance over 5000 easily and not suffer from 1 FPS syndrome.

    Phew that was a long conclusion... Hope this is useful and that I haven't left out too much :). Any comments would be greatly appreciated.
    Last edited by Pistolfied; 06-24-2010, 08:36 AM. Reason: Testing methodology

    Damnit Blizzard, fix ZvT already >.<
    In Soviet Russian, Arma admins are nice to you!

  • #2
    Re: Newest Beta VS. RAMDisk

    I tried to get this going in OA but i'm getting this error :/

    no entry 'bin\config.bin/cfgingameUI.mptable.shadow'

    EDIT* Now that i got it working this is really pretty cool. I haven't noticed a serious boost in performance but the LOD's definitely switch faster. I need to play with it more and run some comparison tests.

    Thank you for the post I didn't know that this was even possible until this afternoon. Kudos
    Last edited by Fiddlestx; 07-04-2010, 12:27 AM.




    TeamSpeak 3 Server


    Twitter Feed