Groove_C

@Taxxor90 I have corrected my previous post to: It's may be obvious that the .exe is only visible only once it's launched, but it's really confusing for average user who first launches your program and can't see/select needed process (which is normal). Also it's annoying that one can only compare FPS of 2 runs but not framatimes (at least I couldn't find it). Frametimes comparison would have been much better than simply FPS.

Updated our server and everything is just fine - people can connect. The problem was just that the server was not updated to the hotfix.

@clawhammer 3866 CL18 is very bad for Ryzen, because Infinity Fabric can't operate at 1933 MHz (3866 MHz) and it switches instead to only half of the speed - 967 MHz, to guaranty 3866 MHz operation of RAM. And also 18-18-18-38 is really high for 3866 MHz. What you can do, is to manually switch your RAM in the BIOS to 3600 MHz, so Infinity Fabric can run at 1800 MHz instead of only 967 MHz and without changing RAM voltage or anything else RAM related, put your RAM to operate at 16-16-16-36 instead of 18-18-18-38. Because your RAM operates at 1.35 V, which is same voltage as for 3600 CL16 kits. You will see your FPS will improve and latencies will go down.

Not as expensive as 3800 14-16-16-36 🤣

Not as expensive as 3600 14-15-15-35.

RX 590 is much better than RX 570 to extract almost max possible FPS in ArmA @ 1080p ultra. RX 5700 would cost to much compared to RX 590 for a few FPS more @ 1080p ultra. G.Skill Trident Z Neo 3600 MHz 16-16-16-36 will be very very good.

@oldbear Would be nice to see screenshots in full screen window mode. Because it's very difficult to analyze small important peaks/details on screenshots with reduced window mode. May be if you find the original image resolution to be to big in resolution to display here - select the full screen window image link from imgur and press on the black eye in the post editor to put it under a spoiler.

@oldbear Frametimes graph tells you, how stable/constant the delivery of frames is, one after the other. In your case it takes from like 10 ms to sometimes more than 30 ms, which can feel/look stuttery when the difference in ms is very high, despite good or very good FPS. For example if you have 10 ms between 2 frames, then 35 ms between 2 frames and 20 ms beween 2 frames, it feels/looks stuttery. On the other hand, if between 2 frames you have 10 ms, 15 ms and 13 ms, it looks/feels much smoother (not faster, because it's not FPS). You can also have much higher time between 2 frames, like 23 ms, 30 ms and 27 ms, but it will still feel/look much smoother than 10 ms, 35 ms and 20 ms. So imagine a straight line and the graph should trying to stay as close as possible to it, to look as flat as possible. The flatter - the smoother game experience, whatever the FPS, as long as FPS is more than just playable. Like here. On the left image you see less FPS, but the graph is more stable/constant/calm, resulting in better game experience, despite lower FPS. On the right image you see jumps to more FPS, but the graph is less stable/constant/calm, resulting in worse game experience, despite higher FPS. You need to look how much time it takes to deliver each frame (lower=better), how big is the difference and how often the difference is big or very big. What is good is: less ms for each frame delivery, less ms difference between each delivery and this difference as rare as possible.

@oldbear look at my right screenshot. There you can see that YAAB takes always exactly 139 seconds from start to finish. With less FPS, it processes less frames, but it always takes exactly 139 seconds from start to finish. You must start to count 139 seconds exactly from the moment YAAB says "BENCHMARING STARTED". Not from when you seen the proposition on the screen to press "S".

@oldbear You need to press F12 in CapFrameX when in YAAB it says "BENCHMARKING STARTED" and need to press F12 again when in YAAB it says "BENCHMARKING FINISHED". The easiest way is to set in CapFrameX "Capture time in seconds: 139". And it will finish capturing frames itself after this time i nseconds. No need to press F12 again. Only first time when YAAB says "BENCHMARKING STARTED". I will then add 1080p ultra result, but you have to understand that there's still a lot of people, that can think, that FPS is lower mostly because of AMD CPU and not because of the RAM and GPU limit for given resolution, which is not the case anymore with Zen 2 (like it was with FX/Athlon/Phenom). Otherwise, the statement, that overall FPS is lower because of complete system (CPU+RAM+GPU) would be correct.

Updated your rig specs in YAAB results list. Not included ultra results, because you're in GPU limit with RX 570, which wouldn't reflect correctly possible FPS with this CPU and RAM at this graphics preset.

https://www.computerbase.de/downloads/benchmarks/capframex-tool/ Frametimes/FPS captured (by pressing F12) with CapFrameX free program in YAAB 1080p standard for 139 secs (starting from the moment when in the left bottom corner in YAAB it says "BENCHMARKING STARTED"). i7-4790K 4.8/4.4 GHz core/cache | 4x8 GB DDR3 2400 MHz 10-11-12-18-204-1T | GTX 1070 Ti arma3_x64.exe will be visible in the list only once it's launched. arma3_x64.exe must be selected in CapFrameX program in CAPTURE->Running processes list.

G.Skill Tridentz Z Neo 3800 MHz 14-16-16-36 (2x8/4x8 GB)

@Topmin2 I can't tell you for sure. May be @oldbear can tell you better here.

It's because of your external HDD. ArmA needs to be on an internal SSD.

@Topmin2 Please, continue the discussion here

YAAB results comparison updated with R5 3600 (X) and R9 3900X.

YAAB results comparison updated with R5 3600 (X) and R9 3900X.

*results sorted by CPU models + FPS YAAB v.1.00 | patch 1.90 i9-9900K (8C/16T) @ 4.7/4.3 GHz core/cache | 16 GB DDR4 3600 MHz 16-16-16-36-631-2T (single rank) | RTX 2080 Ti (no info on AI LEVEL) 66.9 FPS 1080p standard 52.9 FPS 1080p ultra 66.7 FPS 1440p standard 50.2 FPS 1440p ultra YAAB v.0.98 | patch 1.84 i7-8700K (6C/12T) @ 5.0 GHz | 16 GB DDR4 4200 MHz CL16 (single rank) (most certainly not killed unnecessary background tasks / no info on AI LEVEL) 65.1 FPS 1080p standard 57.1 FPS 1080p ultra YAAB v.1.00 | patch 1.90 i5-8600K (6C/6T) @ 4.8/4.5 GHz core/cache | 16 GB DDR4 3200 MHz 15-17-17-35-540-2T (single rank) | RTX 2070 58.2 FPS 1080p standard 45.5 FPS 1080p ultra 57.4 FPS 1440p standard 44.2 FPS 1440p ultra YAAB v.0.98 | patch 1.68 i7-7700K (4C/8T) @ 5.0/4.4 GHz core/cache | 16 GB DDR4 4200 MHz 17-17-17-36-260-1T (single rank) (most certainly not killed unnecessary background tasks / no info on AI LEVEL) 66.0 FPS 1080p standard 52.3 FPS 1080p ultra YAAB v.1.00 | patch 1.90 i7-6700k (4C/8T) @ 4.4/4.1 GHz core/cache | 16 GB DDR4 3200 MHz 14-14-14-34-560-2T (single rank) | GTX 1080 Ti 54.5 FPS 1080p standard 42.2 FPS 1080p ultra YAAB v.1.00 | patch 1.90 i7-5775C (4C/8T) @ 4.2/3.8/2.2 GHz core/cache/eDRAM | 8 GB DDR3 2400 MHz 9-11-11-31-1T (dual rank) | GTX 1080 Ti 65.8 FPS 1080p standard 50.2 FPS 1080p ultra 63.2 FPS 1440p standard 49.9 FPS 1440p ultra YAAB v.1.00 | patch 1.94 i7-4790K (4C/8T) @ 4.8/4.4 GHz core/cache | 32 GB DDR3 2400 MHz 10-11-12-18-204-1T (dual rank) | GTX 1070 Ti OC (Windows 10 1903) 66.1 FPS 1080p standard 47.3 FPS 1080p ultra 63.2 FPS 1440p standard 44.8 FPS 1440p ultra YAAB v.1.00 | patch 1.90 i5-4690K (4C/4T) @ 4.5/4.3 GHz core/cache | 16 GB DDR3 2133 MHz 11-12-11-30-278-2T (dual rank) | R9 270 (Windows 7 x64) 48.2 FPS 1080p standard YAAB v.1.00 | patch 1.92 R9 3900X (12C/24T) @ 4.225/1.867 GHz core/infinity | 32 GB DDR4 3733 MHz 14-15-14-32-262-1T (dual rank) | GTX 1070 Ti (Windows 10 1903) 62.3 FPS 1080p standard 58.7 FPS 1080p high YAAB v.1.00 | patch 1.94 R5 3600X (6C/12T) @ 4.250/1.6 GHz core/infinity | 16 GB DDR4 3200 MHz 16-18-18-36-1T (single rank) | RX 570 (Windows 10 1903) 53.1 FPS 1080p standard 49.9 FPS 1080p ultra YAAB v.1.00 | patch 1.92 R5 3600 (6C/12T) @ 4.1/1.6 GHz core/infinity | 16 GB DDR4 3800 MHz 14-15-14-32-266-1T (single rank) | GTX 1070 Ti (Windows 10 1903) 63.8 FPS 1080p standard 59.7 FPS 1080p high 48.4 FPS 1080p ultra YAAB v.1.00 | patch 1.92 R5 3600 (6C/12T) @ 4.1/1.6 GHz core/infinity | 16 GB DDR4 3200 MHz 14-14-14-34-560-1T (single rank) | GTX 1070 Ti (Windows 10 1903) 57.3 FPS 1080p standard 51.8 FPS 1080p high YAAB v.0.98 | patch 1.88 R5 2600X (6C/12T) @ 4.2 GHz | 16 GB DDR4 3600 MHz 15-15-14-28-160-1T (single rank) | RX 470 (most certainly not killed unnecessary background tasks / no info on AI LEVEL) 58.3 FPS 1080p standard *results sorted by CPU models + FPS

YAAB results comparison is meant to help people to decide if they simply can/need to upgrade only specific parts of their existing hardware configuration or if it's better to build a completely new PC (possibly keeping some parts from the previous build, like case/PSU/cooler/GPU). Here you can discuss ArmA hardware. Your YAAB results (submitted to me via PM) will be published here, by updating already existing list with new results, to keep everything as compact/clean as possible. Please, don't post your results here yourself! This way, I will make sure, that you've "properly" set up everything, so we all can compare results that are as consistent/accurate as possible. Other YAAB results, that I've found so far all over the net (including this forum), can't be used for consistent/accurate comparison, because of custom video settings, people not indicating their monitors' resolutions, CPU's cache speed, RAM's timings etc. Results published will be mainly limited to standard and/or ultra video options presets and to 1920x1080 and/or 2560x1440 (16:9) resolutions, for consistency/accuracy. If you wish to submit your non-standard monitor resolution 21:9/32:9 or High/Very High video options presets results, you still can submit results at these resolutions/video options presets, but you will still have to submit 16:9 and Standard/Ultra video options presets results as well. Results at "Standard" video options preset cover pretty well the FPS difference for low end/midrange CPUs and GPUs. Results at "Ultra" video options preset cover pretty well the FPS difference for powerful CPUs and GPUs. Here you can download the latest YAAB version from Steam Workshop by subscribing to it. Big thanks to @Greenfist for this amazing benchmark! P.S. - I strongly recommend you to go for more than 3 YAAB runs, if FPS differs to much between runs, for your liking. This way you can drop run(s) with FPS that differed to much from other runs. - close all possible background programs prior to launching/running ArmA/YAAB (except sound and mouse software), to minimize "false" results due to background programs interference - disable all mods in the ArmA launcher - launch ArmA - under "OPTIONS -> VIDEO -> DISPLAY -> VSYNC" select Disabled - under "OPTIONS -> VIDEO -> OVERALL QUALITY" select Standard - under "OPTIONS -> GAME -> DIFFICULTY -> AI LEVEL" select Normal (regular) - restart ArmA - find "Yet Another ArmA Benchmark" under "SINGLEPLAYER -> SCENARIOS -> Steam subscribed content" - run YAAB 3 times and submit minimum and average FPS from each of 3 runs (to me via PM), so people with low end/midrange CPUs and GPUs can compare - under "OPTIONS -> VIDEO -> OVERALL QUALITY" select Ultra and restart the game - find "Yet Another ArmA Benchmark" under "SINGLEPLAYER -> SCENARIOS -> Steam subscribed content" - don't press "S" at the start of YAAB, otherwise it will modify selected video options preset by lowering landscape and objects view distance from ultra (3800/2300 m) to standard (1600/1300 m) - run YAAB 3 times and submit minimum and average FPS from each of 3 runs (to me via PM), so people with powerful CPUs and GPUs can compare. CPU: indicate the model of your CPU (and its frequency, if overclocked). Also indicate CPU cache frequency (i.e. i7-8700K 5.0/4.7 GHz core/cache). CPU cache frequency can be checked using CPU-Z -> Memory -> NB frequency (north bridge). CPU cache frequency is very important! Higher than default CPU's cache frequency, additionally increases FPS, increases RAM and CPU cache read/write/copy bandwidth and decreases RAM and CPU cache latencies. RAM: indicate the model of your RAM (and its frequency and primary timings and indicate if it's overclocked, i.e. 3200 MHz 14-14-14-34) and if 1, 2, 3 or 4 sticks installed GPU: indicate the model of your GPU (and if its with double vRAM capacity, i.e. RX 470 8 GB) and whether it's overclocked or not. Hard drive: indicate where Windows and ArmA are installed, whether it's a HDD or a SSD Windows: indicate Windows version and update version (i.e. Win 10 1903) Monitor: resolution (for consistency/accuracy): If you have a 1680x1050 (16:10) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1600x900 (16:9) and sampling (OPTIONS -> VIDEO -> GENERAL) to 120% to match 1920x1080 (16:9) resolution. If you have a 2560x1440 (16:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080 (16:9). If you have a 3840x2160 (16:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080/2560x1440 (16:9). If you have a 2560x1080 (21:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080 (16:9) If you have a 3440x1440 (21:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080/2560x1440 (16:9) If you have a 5120x2160 (21:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080/2560x1440 (16:9) If you have a 3840x1080 (32:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080 (16:9) If you have a 5120x1440 (32:9) monitor, change its resolution (OPTIONS -> VIDEO -> DISPLAY) to 1920x1080/2560x1440 (16:9) @1080p RTX 2080 Ti = RTX 2080 = GTX 1080 Ti = Radeon VII @1080p +~1-2 FPS vs. RTX 2070 @1080p +~3-4 FPS vs. RTX 2060 / GTX 1080 / Vega 64 / GTX 1070 Ti / GTX 1070 / GTX 1660 Ti @1080p +~__6 FPS vs. Vega 56 / GTX 1060 6GB / RX 590/580 @1440p RTX 2080 Ti = RTX 2080 = GTX 1080 Ti = Radeon VII = RTX 2070 @1440p +~_5 FPS vs. RTX 2060 / GTX 1080 @1440p +~_7 FPS vs. GTX 1070 Ti / RX Vega 64 @1440p +~15 FPS vs. GTX 1660 Ti / GTX 1070 / RX Vega 56 @1440p +~32 FPS vs. RX 590/580 / GTX 1060 6 GB

@Dwarden 16-18-18-36 R5 2600X 16 GB 3600 MHz 15-15-14-28-160-1T + RX 470 = 58.3 FPS 1080p standard vs. R5 3600X 16 GB 3200 MHz 16-18-18-36-xxx-1T + RX 570 = 52.8 FPS 1080p standard

Very quick YAAB test - 1x standard + 1x ultra @1080p Best case scenario for R5 3600(X) because of 3800 MHz CL14 RAM (1:1 mode) v-sync off | AI difficulty level normal/regular | all possible background tasks like VoiP, recording and browsers off R5 3600 (SMT on | Win 1903) + 2x8 GB 1900 MHz (3800 MHz) 14-15-14-32-266-1T 1.5 V (G.Skill Flare X 3200 CL14) (1:1 mode) + GTX 1070 Ti Now I'm at least 100% sure that I can order i7-5775C with 128 MB L4 cache. Because @ 1080p @ 4.2/3.8/2.2 GHz core/cache/eDRAM and with only 8 GB DDR3 2400 MHz 9-11-11-31-1T (XMP) it's still 2 FPS faster, @ standard as well as @ ultra. OC to 4.3/4.0 core/cache + my 32 GB 2400 10-11-12-18-1T (secondary+tertiary timings max optimized as well). This way I will most probably additionally gain around 3-5 FPS more. So I can wait for Zen 3/ Ice Lake/ ArmA 4. Ryzen 3X00(X) is only 1-2 FPS better (@ 1080p) than my i7-4790K 4.8/4.4 GHz core/cache with 2400 MHz CL10 DDR3 ))) 3-5 FPS slower with best possible/acceptable RAM vs. stock 9900K 4.7/4.3 GHz core/cache with not expensive 3600 CL16 RAM (XMP). No - thanks ))) R5 3600(X) @ 3200 MHz 14-14-14-34 (1.35 V) vs. 3800 MHz 14-15-14-32 (1.5) is 7 FPS slower @ 1080p standard (YAAB). R9 3900X in YAAB same FPS as R5 3600, but was not able to go to 1900 MHz infinity - no 3800 MHz CL14 RAM, only 3733 MHz CL14 RAM was possible. With 2X16 GB (dual rank 3733 CL14) a little bit better FPS than with 2x8 GB (single rank 3733 CL14). Intel "K" still king in ArmA @ stock with average RAM frequency/timings vs fully OC'ed AMD with best possible RAM for it, if you don't consider Intel CPU prices. + possible to OC to 5.0-5.2 GHz with 4000+ RAM.

@Smart Games RX 570 8GB OC is ok for not more than high video settings with normal FPS @ 1080p. Yes, big difference. Buy GTX 1080 only if very very good price. Otherwise wait for custom RX 5700 (XT) models.

@Smart Games here is a very good board for Ryzen 5 3600 Gigabyte X570 Aorus Elite

@Smart Games 9600K is just slightly better in A3, yes. But one never knows how well multi-threaded next ArmA will be. R5 with 12 threads is more future proof. It's behind 9600K not even 10%, I think. And this is in single thread performance. Before new Ryzen CPUs I wouldn't even have hesitated. But now I'm not that sure about 9600K.