1. Do you need support for Assetto Corsa Competizione? Please use the proper forum below and ALWAYS zip and attach the WHOLE "Logs" folder in your c:\users\*youruser*\AppData\Local\AC2\Saved. The "AppData" folder is hidden by default, check "Hidden items" in your Windows view properties. If you report a crash, ALWAYS zip and attach the WHOLE "Crashes" folder in the same directory.
  2. FOR ASSETTO CORSA COMPETIZIONE: If you report issues with saved games, please always zip and attach your entire User/Documents/Assetto Corsa Competizione/Savegame folder!
  3. If your game executable is missing, please add your entire Steam directory to the exceptions in your antivirus software, then run a Steam integrity check or reinstall the game altogether.

The Impact of RAM Size on ACC FPS Performance

Discussion in 'ACC Hardware Discussions' started by LeDude83, Jan 15, 2019.

  1. Jayk12

    Jayk12 Rookie

    Bummer. Oh I guess I'll have a look at that. At the moment I am quite happy with the performance from overclocking and ram upgrade, I think the RAM upgrade made the biggest difference.

    I need to see if i can push my ram past 3200, its 3600 ram, but AC kept crashing when running it at 3600 with 1.35v. Should i set the times from default to manufacturer timings?
  2. seb 6th gear

    seb 6th gear Racer

    If you keep same timings, you'll need to bump up the RAM voltage reminder up to 1.48 is safe. If you want to keep voltages closer to default manufacturer recommandation, you'll have to loosen the timings, but I don't know what values should you use on Intel plateform. Just so you get an idea 3200cl14 is very close to 3600c16 in terms of true latency, and that's what matters when evaluating the voltage you'll run them at.

    to give you a rought estimate

    the bold squared boxes are the standard low latency kits you can find on the market, the red boxes are the kits you will want to avoid like the plague, greener better.
    Please also remind that some games are starving for bandwidth, some on cycles (lower latency), it seems ACC is starving for both !
  3. Neilski

    Neilski Hardcore Simmer

    Just a small reality check here: the low-latency kits can be pretty expensive.
    I recently bought a 3200CL16 2x8GB kit for a little under £70 (haven't attempted to overclock it yet and tbh I probably won't bother because the gains are unlikely to outweigh the extra hassle).
    The cheapest 3200CL14 kits I can find with a short google session this morning cost upwards of twice as much as my kit. I'd imagine that most people building a new machine could better spend that cash on the CPU or GPU, given the really pretty marginal improvement that the faster RAM will bring to most applications (even ACC).
    Even as an upgrade to an existing machine, I'd argue that most people would get a larger "quality of life" benefit (across all things they do with their PC) from buying 16GB of decent RAM instead of 8GB of the fastest RAM their board will support. I'm certain that was the case for me, at any rate, having upgraded from 8GB of 2133 memory.
  4. seb 6th gear

    seb 6th gear Racer

    That is an undeniable fact.

    That unfortunately can't be less true on Ryzen, benefits still exists on Intel as shown here:

    It's not the first budget of your rig, that is for sure, but if you double the price of your RAM it's a 80€ more on top of 500€ GPU + 400€ CPU + 150€ Mobo we talk about +7.5% overall on an upgrade budget for mid range, 700€ GPU + 500€ CPU + 200€ Mobo it is only +5.7% on high end, even less if you consider a full new PC, and on another forum we pushed our configs to try and figure out fps limiting factors and you can have massive avg fps difference.
    If you run mid range CPUs (200€ ballpark) and entry level gaming GPUs (200/250€ ballpark) it actually doesn't matter much indeed to get low latency kits just go even 3000c15 if cheaper 3200c16 is indeed better as it is probably already samsung B-Die so you have OC potential (3200c14 maybe pushing the voltages).
    For VR users, it's a no brainer multiple examples of guys talking about a 10% improvement, that's like a next gen CPU upgrade just for the RAM sticks...
    In these examples guys had top of the line GPUs 2080Ti that is... And high end CPUs 3700x/3800x, performance beforehand was atrocious, for a very small budget difference in the end...

    Sure 16Gb will always do better in intensive games as your system wont be pagging so limiting massive stuters, but the recommandation for fast RAM kits doen't apply to guys who can't afford mid range machines.
  5. Neilski

    Neilski Hardcore Simmer

    I'm familiar with that thread, but it (or at least the OP) doesn't compare low-latency vs. "normal"-latency RAM; it just shows 2666CL14, 3200CL14, 4000CL16.

    If someone's choosing between the kits I mention above (3200CL14/3200CL16) they would be sensible to infer from that thread that the improvement they can expect from the more expensive one is really tiny (on Intel) because even the improvement going from 2666 to 3200 (both CL14) is pretty meh in my books, at a mere 6% or so. I would hazard a wild guess that the variation from CL16 to CL14 at 3200 is well under 2% but I'm not in a position to put that to the test.

    I'd be interested to read something about the bigger impact of lower-latency DIMMs on Ryzens, as you suggest (a quick bit of googling didn't throw up anything very meaningful for me). Links gratefully received.

    I guess I'd just move the decision point a little compared to you and recommend the fastest kits only for those who can afford high-end boxes. (But my definition of mid-range probably differs from yours because a mid-range GPU for me is no more than £300 ;))
  6. seb 6th gear

    seb 6th gear Racer

    I have bits of stuff that I can share for independant information, but otherwise it's my own experience on friends system or my owns.
    One video I can higly recommend you is from Mindblank Tech it was on 2700X vs 7700K but that is fresh enough to be valuable (as 7700K @5.0 is soo close to 9900K @ 5.0 on single core perf)

    On R3E forum we did a CPU benchmark survey you can find it here

    The aim of that survey was to evaluate CPU/RAM bottlenecking with a fixed scenario Heaven benchmark in DX9 (R3E is DX9 and DX9 is poorly CPU optmized it seems), 640x360 0xAA windowed, low (to remove completely GPU bottlenecking)

    To sum it up on two pairs of very similar configs where almost only RAM timings differ we got the following

    So aside from 100Mhz between the two 9700K on one hand and 1080ti vs 2070 Super on the other hand it was only about memory timings and we got in the ballpark of 10% better CPU performance.

    So in a CPU bound scenario, which ACC represents so well, there is indeed a significant difference. At least for VR users, as I already linked in previous posts above, the almost 10% uplift works at full tilt.

    And yeah you indeed can do a good mi range config with a 300£ GPU, RAM upcost will still represent maybe 8% more but if you're not GPU bound, it will be worth it (as long as you do need the added fps whatever your reason maybe).
  7. Rogue Leader

    Rogue Leader Racer

    100% Correct Z170 and Z270 will do 6th and 7th gen chips, for 8th and 9th gen you need Z370 (note only talking about high end here, yes I realize there are B and H boards too).

    As a general warning that game-debate website is pure trash, as is "the bottleneckr", userbenchmark is slightly better but still junk too. But for specific specs always go to the manufacturer.
    Neilski likes this.
  8. Neilski

    Neilski Hardcore Simmer

    OK, those are indeed astonishingly large performance differences. They are so big as to make me really suspect that something else was going on, because even in the most extreme synthetic benchmark I'd have thought you'd struggle to convert the 10% latency improvement into a 10% fps improvement.

    But also do consider, with respect to ACC, that your figures are not reconcilable with the data from that post you linked.
    In that post, the fps improvements seen between 2666CL14 and 3200CL14 were about 6%. That's with a 20% difference in the latencies (and that's totally ignoring the bandwidth increase that comes with the latency improvement in this case). Even CPU-bound tasks tend to spend comparatively little time hitting the RAM latency bottleneck.

    Edit: the really simple apples vs. apples way to address the question might be for someone who owns a low-latency kit to tweak the timings and compare ;)
  9. seb 6th gear

    seb 6th gear Racer

    I do agree it isn't either scientific, neither realistic as the GPU load was lowered to an unrealitic level, but it was done the same way for every individual, and results were registered only when CPU-Z screen shots were taken. So the test is unrealistic in the sense of how it is applicable to real resolution/LOD but these are real results anyway. I can confirm as a VR user and being part of the DX9 test that the gain was full on Ryzen.
    Now, even on that test there is no linear relationship between memory latency and fps/frame time, and the gains aren't the same between Intel and AMD platforms.
    So in my first post I linked tests that were ran on ACC and R3E but in 2D, in VR the CPU bottleneck is way worse so gains are higher, closer to the DX9 test.
    Also mind that the ACC test linked in first post is based on replays, and replays are a much lighter load on CPU as there is no physics nor AI computation, so it is more a pure GPU test, and even then, gains were to be seen.
    That means that even in 2D when you have AI, the physics, dynamic weather and so on, you can be CPU limited rather quickly, lots of users reported "High CPU usage warnings", in these case lower latency kits bring full advantage.
    Neilski likes this.
  10. Ästhetiker

    Ästhetiker Racer

    i got 2x 8gb Corsair Vengeance LPX 3000MHz in my system.. Its allready running 3000Mhz 1.35v with the XMP profil... I want to overclock the RAMs even a little bit more to see if the performance still stable... could you give me a guideline? i dont want to crank up the Mhz to 3600 and raise the Volts without any knowledge of timings. The Timings of the LPX are 15-17-17-35.. These are the stock timings, even though i use the XMP profil...
    Last edited: Feb 5, 2020
  11. Jayk12

    Jayk12 Rookie

    I have set my RAM to 3200 mhz running 1.35v which is pretty stock according to the vendor. I left my timings as default and it was stable.
  12. Ästhetiker

    Ästhetiker Racer

    okay, i will try this... and what im going to do, when it doesnt work? Change the timings or raise the volts first?
  13. Hypersonix

    Hypersonix Rookie

    FB2F4051-8963-4F1E-8699-F32939826AEF.jpeg 1D324BD4-DEEF-4632-870A-DDFF4803E9D6.jpeg 0F24BD29-A5A1-46AF-B10B-A561F6489F82.jpeg Could someone take a look at my ram details and see if it’s correct please?

    The ram is Corsair Vengeance LPX DDR4 3200Mhz C16


    Thank you.

    I should add the 2 of the 8gb sticks are double sided Samsung and 2 are single. Does this have any impact?
    Last edited: May 1, 2020
  14. seb 6th gear

    seb 6th gear Racer

    CPU-Z Memory tab indicates the current speed of your RAM kit set in BIOS and you are running JEDEC#7 profile that is pretty slow as this is almost default DDR4 specs.

    You need to go in BIOS to activate XMP profile that will enable RAM speed of 1600Mhz (3200 MT) instead of current 1066Mhz (2133 MT) !

    You currently leave a lot of performance on the table pal !

    That could be your issue, on SPD tab please do a second screen shot to show the specifications of the second RAM stick (you have shown only slot#1), there is a possibility your motherboard defaulted to 2133 if the second stick isn't as performant as the one you showed in the screen shot (lowest speed is always chosen for obvious reasons).
    Hypersonix likes this.
  15. unpierrot

    unpierrot Gamer

    If you've never tested your RAM modules, I can only suggest to do it at least one time. Bad surprise can happen. For my case, it was a faulty module right out of the box. But a faulty module can hide itself for a long time. If the faulty bit take place on a picture located in RAM, then out of a wrong color pixel, nothing else will appear. Now if the faulty bit is in a core code, things can start going severely wrong.
    And moreover, if you start to overclock (or use XMP feature), it's better to lose few hours testing RAM modules and to be 100% sure that everything is working fine than trying to understand where does come a random crash.
    I'm used to memtest which quite efficient. https://www.memtest86.com/
    This kind of test could be very long (depending of your amount of RAM) and it's better to let it make two full pass.

    edit : forget to say that free version of memtest is enough.
  16. Hypersonix

    Hypersonix Rookie

    Thank you so much for helping here, I’ve been scratching my head about this for a while!

    Here we go.
    I set the ram to 3200 in the bios and it now shows correctly but I wonder if the timings are ok? These were set automatically.

    Attached Files:

  17. Hypersonix

    Hypersonix Rookie

    I found the XMP setting in the BIOS, this is the new screenshot. XMP on.PNG
    seb 6th gear likes this.
  18. seb 6th gear

    seb 6th gear Racer

    You will definitely feel the difference now!
    Hypersonix likes this.
  19. Hypersonix

    Hypersonix Rookie

    Nice! No issues with the timings? Just leave it alone now?
  20. seb 6th gear

    seb 6th gear Racer

    You are are the RAM brand advertised specs, so basically now you benefit from what you bought in the first place. And 3200 XMP profile is a memory OC as it exceeds both Intel CPU RAM speed support and default DDR4 specs.
    As not all CPU/Mobo combo allows to run XMP sometimes it needs a bit of tweaking.
    Now I recommend that you do some RAM tests to be 100% sure RAM runs error free.
    That said if stable you can try to push further, but be aware it takes a ton of time and acquiring quite a bit of knowledge on the matter for a results that is not always worth the hassle/results.
    So for peace of mind if your not so much into OC, leave it there and enjoy safely the new gains. But do run a full mem test to check everything is 100% safe on XMP profile.
Similar Threads
Forum Title Date
PlayStation 4 Discussions Impact with other cars or barriers can cause frame-stuttering Sep 8, 2016
Graphics and Models/Tracks Bugs & Issues Time of day/lighting impact on framerate Nov 14, 2013
ACC Hardware Discussions Screen resolution impact on FPS? Jul 31, 2020
ACC PS4 - XB1 General Discussions Do Thrustmaster control panel settings impact console behaviour? Jun 29, 2020
ACC PS4 - XB1 General Discussions Boost mode impact? Jun 24, 2020
ACC Multiplayer Glitchiness in MP and impact to ratings Feb 16, 2020
Troubleshooting - Workarounds Huge FPS Impact w/ PP Filter ON Nov 15, 2019
ACC Hardware Discussions Biggest settings impact on GPU? Aug 4, 2019
ACC Troubleshooting What impact on performance does the settings have? Jul 14, 2019
Troubleshooting - Workarounds What impact on performance does the settings have? Jul 14, 2019
ACC Rating & Leaderboards SP without impacting statistics (RC,SA...) May 29, 2019
ACC General Discussions Impact of fuel weight on GT3 cars Sep 21, 2018
Troubleshooting - Workarounds assetto corsa wall collisions: impact too large Sep 15, 2018
ACC General Discussions Will Enabling HDR via Windows 10 Impact Performance in ACC? Aug 18, 2018
Physics Bugs & Issues no ffb information in crash impact Apr 22, 2018

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