11/14/2023 0 Comments Process lasso tarkov![]() ![]() ![]() This of course depends on what you're doing, but trying to keep them fed with the data they need is often the toughest challenge. Most performance is not throttled by CPUs not being fast enough or cooled enough, but by not being able to access enough memory quickly enough. Remember, modern CPUs run much more quickly than RAM, which is why we have three layers of caching between them and RAM. This is true, but, moving a thread from one core to another is far from 'free', as you invalidate caches. This allows peak speeds, which are faster than the core can cool because it's for such a short time. This means each core will heat up 1/4 of the time and cool 3/4 of the time. Windows shuffling processes to different cores acts to flush out the available cache, and for a game as clunky as ONI in terms of resource usage - I think that may be the key. I have a 2700x so i'd use cores 0-3 for games, then leave the remainder for other processes. For the video I simply showed how to split the workload with SMT still enabled - however for best results i'd advise (firstly, do some detailed reading) disabling SMT, googling which are the physical cores for your processor - and using half of those for your games. Knowing which cores are physical versus doubled can help hugely. This guy seems to explain it rather clearly where I cannotĪlso, are you going to tell me that you would actually close each and every other program you have running before booting up a game? Nerd You can't close *every* process - you just can't, and those windows essential programs, or peripheral software, etc - they can bugger up core usage and cache allocation. There´s nothing to gain by this.īefore I go on - i'm clearly not an expert (nor am enthusiast) but from the reading I did I think the cache in use is one of the more important factors here. But it will not really help, if you already have only 2 or 4 cores. If you need all of this programs, for example, as a streamer, this software can help. Windows is doing nearly nothing on itself in background of a heavy application like a game or such. Processes not running, cant use CPU-time. The other way, that works perfectly fine is: Close the damn things, you dont need right now. In other words (no offense btw), if you´re running many things in the background (for whatever reason), this can help, because the clunky windows scheduler wont assign the game to cores, already in use by another task, and then again and again. Currently installing the mainboard in the case, so I cannot check. Of course, that win10 installation is from yesterday, and I may still be missing drivers and stuff. CPU load on the FX8350 on Win7 was more like 4 cores at 50%, so completely different scheduling. Not sure how hyperthreading goes into core numbering, but the performance monitor lists 16 cores for this 8 core CPU.įPS went from still completely playable 13FPS on my old FX8350 to also completely playable 20FPS on a complex end-game map on the 3600X. On my new 3600X, I have 100% load on core 4 and 12 (I think) when running ONI. It seems that Win10 uses the fastest two cores for ONI. This means in my case it's best to NOT use dedicated cores. This means if you dedicate a core to the 100% load process, then there is a risk that the CPU will thermal throttle because you took away the ability to spread out the heat. However if you follow the process, it will only experience peak speeds because it keeps getting a new cold core to heat up. ![]() It will consistently switch in this order: 0->1->2->3->0. ![]() I have an Intel 4790k and I noticed that if I have a single core task, which takes 100% of the CPU time for a while, then it will not stay on the same core. One word of warning though: it's not always a good idea to dedicate a CPU core to a task. Also know about which cores are on which cache if you have enough cores to not let all cores share the same cache. In general you would likely get the best results from knowing which cores are real and which are hyperthreaded ones. If you set ONI to only run on a single CCX and let nothing else run on that CCX, then not only will you avoid the performance hit from switching CCX, you will also ensure that ONI gains control of the entire 16 MB cache. Each CCX has dedicated cache and there is apparently a performance hit when a process is moved from one CCX to another. They consist of 2-4 CCX, each with 3 or 4 cores. It's also interesting if you use a Zen 2 CPU. If you overclock and end up not running the same speed on each core for whatever reason, you can make ONI only use the fast cores. This has a bunch of interesting use cases. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |