Compiling speeds Tests (Fixed link)

Discussion in 'Programming' started by AnonyMouse, Sep 4, 2016.

  1. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    P.S. This test will help to choose the right configuration for heavy games.
    P.P.S. Something was wrong with the old link...

    Hi! That is a test for comparison of compiling speeds on different PC. I made a "game" with 300 sprites, 50 sounds, 10 BG, 3 objects, 2 rooms. Size on disk: 137 mb. Archived size in 7z: 2 mb 630 kb... (I did few test to be sure but that is and it is working, whlie .zip 141 mb!?) If you have problems with the file tell me.

    zip file
    7zip file


    FIRST go to File->Preferences->General and disable "Enable backup on save" and "Enable backup on run/debug/create"!

    The info you should provide:
    CPU:
    RAM:

    GMS v.:
    Cold compile:
    Hot compile
    :
    (some other notes like the HDD/SSD you are using. In my case I have SSD but it is compiled in HDD because in my past tests I found no differences?!)
    GMS version: EA or Stable.
    "Cold compile": when you compilethe game the first time. After that it is making a cache which is deleted after some time or after you restart the PC. If you want to test again in "Cold compile" go to Your Documents/Gamemaker/Cache/CompilingSpeedGMS and delete it.
    "Hot start" is when you compile the game in less than a... unknow time, but soon. The time should be much shorter!


    How to measure the time?
    First method is wait until it is compiled and press "Esc" on the keyboard. Then check on Compile form below "Compile started:" in the very begining and "Compile finished:" Then you can do the math on your own or use site like this https://www.timecalculator.net Dont forget 1 minute contains 60 sec, not 100...
    Second method is with chronometer. Most mobile phones have it but it is harder and the mistake is bigger.

    P.S. Link is fixed.
     
    Last edited: Jan 29, 2018
  2. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    My current main PC:
    CPU: Intel i3-6100 3,70 GHz
    RAM: ADATA 8 GB, DDR4, 2133 mhz
    GMS v.:
    Early Access v. 1.99.505
    Cold compile: 43 seconds (average)
    Hot compile: 4.4 seconds
    GMS v.:
    Stable v. 1.4.1757
    Cold compile: 53 seconds (average)
    Hot compile: 6.5 seconds


    My "office" PC (at my job)

    CPU:
    AMD Pro A10-8750B R7 (3.6 GHz)
    RAM: 4 gb DDR3 1600 mhz
    GMS v.: 1.99.402 Early Access
    Cold compile: 3.4 min (worst 3.52, best 3.36) !!!
    Hot compile: 22 seconds...

    My laptop
    CPU:Intel Core2Duo T7700 (2.40 Ghz)
    RAM: 3 gb ? (32 bit)
    GMS v.: 1.99.402 Early Access
    Cold compile: 2.54 min
    Hot compile: 22 seconds...

    My revived old pc (i3 got some technichal issues... again)
    CPU:Intel Core2Duo E7300 @ 2.66GHz
    RAM: Kingston 2048 MB (DDR2-800)
    GMS v.: 1.99.402 Early Access
    Cold compile: 4.43 min
    Hot compile: 34 seconds...hm...worse than the laptop???

    My backup-revived PC
    CPU: Intel Celeron G3900 (2.8GHz)
    RAM: ADATA 4GB, DDR4, 2133 mhz
    GMS v.: 1.99.505 Early Access
    Cold compile: 1.28 min
    Hot compile: 53 seconds; 33 sec; 17 sec; 8 sec, 7 sec... total mess! Each time I make a test it is different... Probably will test it after few days again.

    Code:
    The amazing thing here are single (multi) core results in cpubenchmark:
    Intel Core2Duo T7700 2.40................0.833.....(1.430)
    Intel Core i3-6100 @ 3.70GHz.............2.107.....(5.512)
    AMD Pro A10-8750B R7  (3.6 GHz)..........1.666.....(5.612)
    Intel Pentium G4400 @ 3.30GHz............1.862.....(3.678)
    I did used that Pentium for a week and results testing my game showed the same or almost the same results with i3... And in single core AMD and G4400 are close, Core2Duo is worse, but in the reality it is much different... I was about to buy AMD Athlon X4 845 (3.8GHz) as back up PC because a lot of problems, yesterday the PC went totally wrong, maybe the PSU...
    Would be nice if someone do that test with another AMD.
     
    Last edited: Oct 28, 2016
  3. Gre_Lor12

    Gre_Lor12 Guest

    Would you mind adding in an .exe too that shows it all as I cannot access GameMaker at the current time. Say no if I am being awkward xD
     
  4. renex

    renex Member

    Joined:
    Jun 23, 2016
    Posts:
    506
    Stumbled upon this and thought it would be funny to try on my little monster.

    CPU: FX-6350 3.5GHz
    RAM: 1x8G DDR3-1866
    HDD: SSD 450mb/s
    GMS v.: 1.4.1763
    Cold compile: 29 seconds
    Hot compile: 7 seconds
     
  5. McWolke

    McWolke Guest

    CPU: Intel i7-4790k 4GHz
    RAM: 2x4GB DDR3-1600
    HDD: SSD (Crucial MX100)
    GMS: Early Access 1.99.505
    Cold Compile: 42 Seconds
    Hot Compile: 5 Seconds
     
  6. The Sentient

    The Sentient Guest

    CPU: AMD FX-8320 3.5Ghz
    RAM: 10GB DDR3
    HDD: SSD (Intel)
    IDE: Visual Studio 2015
    Cold Compile: 3 Seconds
    Hot Compile: 3 Seconds
     
  7. Ziphold

    Ziphold Member

    Joined:
    Jun 22, 2016
    Posts:
    103
    Sorry for filling this thread with garbage, but... what?! Am I the only one noticing something here?
     
  8. GMWolf

    GMWolf aka fel666

    Joined:
    Jun 21, 2016
    Posts:
    3,496
    I do not think this is a fair project to test with:
    This is all just texture page stuff. (notice how hot compile is very short...).

    I would recomend throwing sound into the mix, as they can take time to convert (Add a good mix of compressed and streamed sounds)

    Also code! Lots of code! I would be very interested to see how code affects performance, and also the YYC!

    Yeah @The Sentient ... why?
     
  9. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    Fel666, there are sounds. 49 sounds.
    Code - I am not sure but I thing its affect is zero. The reason for testing is to see what is better when selecting the hardware - CPU single/multicore, RAM speed, HDD/SSD. My game contains a lot of graphics and the compiling time is huge so I reduced the graphics with static basic sprites and it is much better now for testing. I cant compare RAM speed now because I have only DDR4 2133. Also seems there is a difference between GMS 1.4 and GMS 2 in multicore but I will see that in Monday in my other PC with i3.
    Ziphold, GMS 1.4 is affected mainly by singlecore speed so I doubt AMD could compete with Intel. I am not a fan/enemy of any of them but Intel has much, much better singlecore results. So maybe here we can see some false results. And there are some other reasons for the speedchange, some tempfolders maybe.
     
  10. GMWolf

    GMWolf aka fel666

    Joined:
    Jun 21, 2016
    Posts:
    3,496
    well, that is most certainly true of previous generations. and although they are still a little bit weaker on the ipc and clock speeds, Ryzen sure has closed the gap. Would be interested to know their performance when compiling.
     
  11. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    I doubt that, cpubenchmark.net shows Ryzen is weak in singlecore the same way as other AMD CPUs. For example

    Intel Core i7-6700 @ (3.40 GHz) - 2,142
    Intel Core i7-4770K @ 3.50GHz - 2,258
    Intel Core i7-7700 @ 3.60GHz - 2,348
    Intel Core i7-7700K @ 4.20GHz - 2,582
    Intel Core i7-8700K @ 3.70GHz - 2,729
    According to cpubenchmark the best AMD in singlecore is
    AMD Ryzen Threadripper 1900X - 2,054 ($477.47)
    which is a bit worse than
    Intel Pentium G4600 @ 3.60GHz - 2,057 ($90.90)
    and far away from those above and even i3 and i5 have good results and their prices are...
    Probably in multicore competition on compiling GMS2 could be different but I have no data.

    P.S. Heh, just noticed the link to the file is invalid... Will reupload it soon.
     
  12. GMWolf

    GMWolf aka fel666

    Joined:
    Jun 21, 2016
    Posts:
    3,496
    Huh ok, so seems like there is still around a 25% difference in single core performance.

    But please don't compare the price when looking a single core performance...

    So fair enough ryzen won't be quite as fast as intel for GMS1 compiling.

    You know what is missing from this conversation though? Storgar!
    In my experience moving to an SSD made a huge difference. This tells me comoilico isn't really CPU bound, but storage bound.
    I guess that's because most of time is spent on resources.

    It's why I wanted to see a code heavy project, as it stands everything is probably being eclipsed by the resource processing.
     
  13. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    There is huge difference when you load the application or windows. Compiling time is the same and I tested that and I am testing it again on my other PC with windows and GMS and project on HDD and on SSD. And even that SSD is faster than that on my main PC. Loading project in the application is faster but that is not related to the compiling.
    P.S. Link is fixed.
     
    Last edited: Jan 29, 2018
  14. GMWolf

    GMWolf aka fel666

    Joined:
    Jun 21, 2016
    Posts:
    3,496
    Oh yeah, compiling will always be done on you main C drive. So if your c drive is an SSD< it doesnt matter where the project is.
    I tried it on the same configuration with and without and SSD, and it was quite a bit faster. (Sorry, no figures, this was a while ago when I added my SSD to my PC.)
     
  15. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    No, I mean when there is no ssd. I made a test on flash drive, cant remember now did I put assets and cash and projects directory there but compiling time was impressive slow.
     
  16. JackTurbo

    JackTurbo Member

    Joined:
    Oct 19, 2016
    Posts:
    828
    What sort of flash drive? Was it usb?

    Could have been the usb transfer speed bottlenecking the process.
     
  17. GMWolf

    GMWolf aka fel666

    Joined:
    Jun 21, 2016
    Posts:
    3,496
    Your other post said drive did not relate to compile time?
     
  18. AnonyMouse

    AnonyMouse Member

    Joined:
    Jun 23, 2016
    Posts:
    106
    Yes, SSD has no effect, but flash drive has. Maybe it is like the RAM - if it is not enough the speed is less but when it is more the speed is standard.
    Flash like this, the same brand/
    [​IMG]

    I am curious about PCI-E disk with very high speed but it is too expensive.
    Today tested my i3 Dell with GMS 2 and seems the CPU is not 100% intensive as it was in my Celeron. Will test it again but it is becoming a total mess.
     

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice