EVGA

Frametime and memory speed correlation?

Author
GangsterFresh
New Member
  • Total Posts : 10
  • Reward points : 0
  • Joined: 2018/07/15 11:56:42
  • Status: offline
  • Ribbons : 0
2018/10/28 19:28:56 (permalink)
Okay so quick backstory:

For the last 2 months of my life this issue of microstutter has been consuming me. Dumped a ton of money on my rig only for it to microstutter every few seconds. This issue happens in all games that heavily depend on the CPU such as Fortnite,LOL, and CSGO. While the issue does not happen in games that depend on the GPU such as AC:Origins,farcry, tombraider etc... So that right there is what is really confusing. Naturally I thought my CPU was the issue so I went ahead and replaced my 4690k with a 4790k (worst and most useless upgrade of my life) and the issue stayed so this made me conclude that the issue was not on my CPU's fault. So then I went through crazy amount of time diagnosing issues left and right and hell I even made a list of the stuff I have already tried and that did not work:

Went back to older Windows 10, 2016 LTSB 1607 version

turned off all 3rd applications and only kept monitoring program and game up *

did the whole windows 10 memory standby clearing (I really thought this was going to fix it) but sadly even when I clear memory standby and memory cache resets, the problem is still there so I doubt its a windows 10 issue * * *

disabled full screen optimizations *

turned gsync off/on *

tried uncap vs cap fps *

fullscreen vs windowed *

Hell I even started using vsync with the input delay and still no help. *

used ULMB mode,fast sync everything... *

Thought it was my monitor with its Gamemode and turned that off and nothing *

Turned off xbox game DVR *

played around with different refresh rate 120,144, overclocked to 165 hz and nothing worked. *

Turned windows into high performance and same issue.

Tried on another PC in the same house and same issue

Overclocked and stock speed my CPU and cache

So when I tried on another PC this made me think that hey the issue is probably not my hardware and its either power or internet because that's the only 2 things my 2 Pcs share. Its not the internet because when I unplugged the ethernet cable the microstutters were still there so then it must be power right? Nope I used my laptop and its not plugged to a power socket and same issue. So the fault has to be on these games and not me or am I crazy? I guess most people don't complain about microstutters and frametimes is because they don't spot it and so even though this issue is happening to everyone who plays on the CPU bound games, most don't even notice the stutter to even report it.

So I had a lot of freetime and decided to play around with my ram speeds and this was shocking, heres a frametime graph of around 10 mins for each graph while running Fortnite:

800 MHz ram speed:

1600 (XMP) mhz ram speed:

1800 mhz ram speed:

So at speeds of 800 mhz the frametime spikes were at around 26-30 MS and at speeds of 1600-1800 they ranged from 16-20 MS, so i can't believe i am saying this because how the heck could ram speed cause this but is it possible that the higher the ram speed is the less and less you will notice stutters? FOR CPU BOUND GAMES ONLY Again at ultra settings triple A graphics games i was not seeing this issue, but when i played these multiplayer games i started seeing it, idk it has to do with texture filtering or assets loading or something along those lines. But yea what are your guys's thoughts on this? Should i even bother buying 3000 MHz ddr4 ram compared to my 1600 ddr3 ram or is this a more serious issue and i should do more tests? if i should do more tests then please link em below. Thanks! Also any forums or subs suggestions to ask about this issue? I could just simply copy paste my issue so more people could see it and hopefully we can find a solution to this.

My rig:



Monitor: XB271HU 1440p 144z g-sync monitor

GPU: 980 ti hybrid

Firestrike score:

CPU: 4790k@4.5 GHz

ram: Corsair Vengeance DDR3 1600 C9 4x4GB
post edited by Sajin - 2018/10/28 19:38:37
#1

4 Replies Related Threads

    batboy88
    Superclocked Member
    • Total Posts : 234
    • Reward points : 0
    • Joined: 2018/03/07 16:01:58
    • Status: offline
    • Ribbons : 0
    Re: Frametime and memory speed correlation? 2018/10/29 19:22:48 (permalink)
    I don't know...use Passmark as a reference or example or something. Intervals of frame dippage and peaks etc. Still hit some good 322fps, 265 in DX11 etc 95 in dx12. Nothing wrong here man.
    post edited by Sajin - 2018/10/29 19:28:20

    7700k/5.2ghz-1.34v/ FTW K/ custom Loop/ Crucial/Micron Ballistix- 3200mhz/ Vega 56 Reference/ Cooler Master GXII Pro 750w
     
    2600x/MSI X470 4.3ghz
    #2
    arestavo
    CLASSIFIED ULTRA Member
    • Total Posts : 6916
    • Reward points : 0
    • Joined: 2008/02/06 06:58:57
    • Location: Through the Scary Door
    • Status: offline
    • Ribbons : 76
    Re: Frametime and memory speed correlation? 2018/10/29 21:41:18 (permalink)
    Frametime is just another way of looking at FPS. 60 FPS = 16.6 milliseconds (ms), 120 FPS = 8.33 ms, 144 FPS = 6.94 ms. Or, 1000 divided by w/e your FPS is.
     
    https://www.mvps.org/directx/articles/fps_versus_frame_time.htm 
     
    Now, by neutering your CPU by lowering the memory speed to 800, yeah your FPS and by extension your frametime will go down. Your CPU has to feed your GPU, and in CPU limited games that will bottleneck the FPS/frametime.
     
    My suggestion would be to increase the resolution of the games that give you these issues, via Nvidia's DSR, to move the bottleneck from the CPU back over to the GPU.
     
    FWIW, dual channel memory setups suffer MUCH more from lower memory speeds than quad channel setups. I actually tested this out with Rise of the Tomb Raider about two years ago with quad channel memory from 1333MHz up to 2133MHz on my old X79 platform with a 980 Ti. There were measurable differences, but nowhere near the difference that dual channel setups had.
     
    Also, make sure that your CPU/memory overclocks are stable. It is possible that they aren't after being run for several years, and may need more voltage or a lower speed to be stable.
    post edited by arestavo - 2018/10/29 21:57:15
    #3
    batboy88
    Superclocked Member
    • Total Posts : 234
    • Reward points : 0
    • Joined: 2018/03/07 16:01:58
    • Status: offline
    • Ribbons : 0
    Re: Frametime and memory speed correlation? 2018/10/31 12:31:04 (permalink)
    It's small man like 2-5% and pretty redundant after 3200mhz. Maybe with better capable timings, most of it's pretty slacked past that.

    7700k/5.2ghz-1.34v/ FTW K/ custom Loop/ Crucial/Micron Ballistix- 3200mhz/ Vega 56 Reference/ Cooler Master GXII Pro 750w
     
    2600x/MSI X470 4.3ghz
    #4
    arestavo
    CLASSIFIED ULTRA Member
    • Total Posts : 6916
    • Reward points : 0
    • Joined: 2008/02/06 06:58:57
    • Location: Through the Scary Door
    • Status: offline
    • Ribbons : 76
    Re: Frametime and memory speed correlation? 2018/10/31 19:54:26 (permalink)
    batboy88
    It's small man like 2-5% and pretty redundant after 3200mhz. Maybe with better capable timings, most of it's pretty slacked past that.




    You're likely not going to get 3200MHz working with DDR3!
     
    And it was NOT 2-5% for dual channel systems (DPS dips were much more pronounced for dual channel) - it was ~2-5% for my quad channel system. That was for DDR3 systems and those processors, and newer processors with faster dual channel RAM aren't going to be as affected in CPU bound games (for now).
    #5
    Jump to:
  • Back to Mobile