I can explain this... There's several factors in play here: Sometimes thermals can play a part, but frequency is more importantly determined by active cores on modern CPUs... if you force park 2 cores, the CPU frequency lookup will just boost higher (Intel is easier to demonstrate because it's definitively set, but AMD works pretty much the same but is more floating based on thermals too, however in gaming thermals are seldom the issue; for example, i7 12700k, 1 core 5.1ghz, 2 core 5ghz, 3&4 cores 4.9ghz, 5&6 cores 4.8ghz, 7&8 cores 4.7ghz (or something like that))... when playing games, even if they only use 4 cores max in the actual game engine, Windows will run background tasks in one or two extra cores so they never actually park, so the CPU downclocked slightly for those extra cores running background tasks... when you limit the number of cores in windows, then those tasks are moved to logical threads of your active cores, and the extra cores are actually parked, therefore the CPU will boost slightly higher... this frequency boost largely accounts for your improved average FPS. The fact that the background threads have now been moved to hyperthreading/SMT logical cores also accounts for the typical slight REDUCTION in 1% lows/frametiming, because the game threads can now occasionally end up waiting on a background task to complete on the logical core corresponding to that game thread's physical core. Many games still only use 4 cores, max, so for those, 4 active cores is the sweet spot, extra cores causing a CPU downclock, and fewer cores hindering the game engine. Further to this, game engines which are correctly programmed will use physical cores first, then use logical cores afterwards if they can utilise more... for example if a game CAN use 8 cores, but you have a 6 core CPU with HT/T, then it will use 6 physical cores and 2 logical cores... poorly optimised game engines can get this wrong, famously for example, GTA5 on release was incorrectly optimised and would just use the next available thread, so turning off hyperthreading in bios could give you a good 30-50% improvement in framerate... this has of course been patched now, but it was a relevant tweak on release. As I mentioned earlier, most games still use 4 cores or less, with only a few exceptions off the top of my head, notably CP2077 with the newest patch (but not the earlier version, and this is why you're getting better in game performance now, compated to previously), and also RDR and RDR2 can use 8 cores plus (although there are severely diminishing returns above 8 cores)... What is interesting is that even those games that do use 8 cores or more, do all have very diminished returns above 8 used corrs/threads. That is why I actually semi-cripple my 8p-core CPU for daily use/gaming, by turning off hyperthreading in BIOS but also overclocking it (I do have a second profile saved in BIOS with HT enabled, for when I want to do some workstation workloads)... if im playing a 1-4 core game, I'm definitely only using physical cores, and without the strain of running logical cores, the power consumption and temperature are reduced, which means I can run a higher overclock on more cores, which means that for 4 core or less games, its not even downclocking... for 8 core plus games, with my overclock it is running 4.9ghz, which is about a 4% increase over stock, and offsets any advantage that enabling HT but reducing clocks might give. PS. As well as being overclocked, it's also slightly undervolted, and in this configuration it uses about 3W at idle, 5W watching UKposts, and 145W under full load in R23 (compared with 7W, 15W and 190W at stock)... as a result, it runs exceptionally quiet as I have targeted fan speed for 67C operation (29C at idle)

I run a Ryzen 9 3950X with 16 cores. RTX2080 super and 64GB RAM. I use it for CAD work, but occasionally run a couple games. When Flight Simulator came out Task Manager showed only 8 threads being used, but now after many updates it shows 28 to 30 being used and my GPU is down from 100% to 70% being used at 3840x1600 resolution. I had a similar situation with Cyberpunk 2077, but it's never used as many threads as Flight Sim. Ratchet and Clank seems to use about 15 threads and 80% GPU. Granted this is very half assed, but I don't do much with games. I think it is about how the game or program is optimized for the specific hardware. I don't think all systems will run best in the exact same configuration. I don't think that is possible with so many different options now days. Good video as usual. Thanks.

I have seen cores turned off in the bios before, but never through the OS. A great video.

Pretty sure some overclockers will disable cores to maximize stable speeds on one core. Pretty cool stuff. I've done this in Linux to save power when the battery is running close to dry, but I never knew Windows could too.

It should also be noted that coding techniques can also allow specific parameters to take advantage of CPU or GPU power and capabilities. But these tests do ask some questions that should open up more user testing to see if they can get better gameplay from using reduced cores, which may also mean using less energy to achieve better things too.

Great video ty

thanks for test sir, so far SMT off/HT OFF usually more performance, if 8 thread better, i think 8 core cpu, SMT/HT off, would be sweet.

Honestly, It's stupidly funny in retrospect. The magic number is 4 logical cores (Be 2 HT cores, or 4 ST cores). Mostly because AAA gaming had been stuck on that montra for over a decade, and literally only recently within like 5-6 years have developers been attempting to use more core counts in the system. We literally have about 15+ years worth of AAA games that were soft optimized for 4 threads overall, and that kinda ignores the increase in power overall with processors. Notice how only the most recent game CS2 had expected scaling? Ya.. that's an entirely brand new engine built for a age with your average low thread count being 8 with the normal being 12-16. The big boy problem is ... we've still got to wait for game engines & games to really take advantage of it... and only the newer games (1-2 years) really can because their game engines now support it. That's also to point a reminder that ... there is less optimization being done for higher core count because of the insane growth in power for CPUs in general... which means the soft magical number is 6 or 8, only because of diminishing returns overall. The secondary problem is ... programs (be it games or whatever software) don't get optimized for more power & processor count unless they kind of need it. Sure uber FPS is great... but if a dev is AAA dev is seeing 120+ fps on say 6 threads, sure it might scale a bit more with 8 or higher but they aren't going to be any resources given to that from the publisher or overhead. -- Games aren't like video editors, where the better the usage of the CPU/GPU... the better the product. Games have this 'good enough' point where any more dev time or resources used on it is effectively wasted money for them.

most game engines are coded to use no more than 2 threads, that's why. so you need 1 core for Windows and background tasks, and 1-3 cores for the gaming app. if game only used 1-2 cores it's faster with less extra cores because the system doesn't need to do the overhead work of switching operating cores.

Did some Tests with Ashes of the Singularity. (It´s only a small sample but still interesting I think) My System: MSI MEG x570 ACE Ryzen 5900X ASUS TUF RTX 4090 32GB RAM 3600 CL16 To make sure the CPU is at 100% the bottleneck, I also set the graphics to the low preset. But I only have the AVG. FPS (The built in Benchmark unfortunately does not show 1% Low) I didn´t bother going below 6 Threads though, since the game really uses 24 Threads and scales well up to 12. For these results I set the core limit in Windows like you did in your video: 24 Threads: 41.34 FPS AVG 12 Threads: 40.25 FPS AVG 6 Threads: 29.55 FPS AVG But here comes the interesting part. For these I additionally deactivated SMT in the BIOS : 12 Threads SMT off: 44.84 FPS AVG 6 Threads SMT off: 38.33 FPS AVG For this one I deactivated one of the CCDs but turned SMT on: 12 Threads SMT ON: 40.00 FPS AVG And for this one I deactivated one CCD and turned SMT off: 6 Threads SMT OFF: 37.23 FPS AVG I think these results confirm my theory with the temperature and the clocks. SMT uses unused functions of a CPU core in a separate thread. That additionally heats up the core. Deactivating a CCD means the load isn´t spread across two chips anymore, but concentrated in one. Less surface to dissipate the heat. With that the fastest configuration is 2 CCDs and no SMT.

I'm not a gamer but I am a consultant for business networks. I wonder if the results you got is related to the thing we have seen on VMware servers. Sometimes if you give a VM too many processors or cores it will slow the VM down because it is spending so much time switching between cores. We have noticed this a lot with SQL servers because SQL will try to use all the cores available to it. I'm not sure if this is right but it feels similar.

From your testing any games before 2020, and some after are most likely programmed for a max of 4 cores. I was getting bite by the computer building bug, and found enough parts to build a i3 3rd Gen Intel with a GT1060, saved my money and plays Dirt Rally super well.

i need help about windows 10 and 11 drivers update , i have and old dell laptop (dell inspiron N11510) when i put any windows of them all drivers gets updated and installed by windows update easily and if go into the (device manger) and tried to update any of them its all good and no need for those driver to get updated and Microsoft downloaded the good drivers for this laptop , but when i use any (third party) app its says there are more new updates and needed to be downloaded . the question is : 1- do i download those updates or not on this type of old laptop or not ? 2- if yes which is the best third party app do you advice to use to download them so i don't get any malwares or viruses ? sorry for talking so much 😅😅

On my Threadripper Pro 3975WX with 32C/64T, in-game benchmarks significantly decreased when set for 4C/8T in FarCry6, Hitman3 & CP2077. Certainly not a mainstram CPU, but an interesting datapoint nonetheless.

Very interesting, but my questions is, what about cpu temp, any effect?

you partly right about master core. Also Windows Manage Cores to be best used from XP and up. To get a more reliable result is to run the game, go to task manager and mark what cres the game can use without windows determining which ones to optimize. I can't get this to work on windows 11 and I am not sure its on windows 10 but it is possible on windows 8.1 and below.

I agree with that assumption that core 3 became more dominant with running threads but I am also thinking that on top of that less cpu cores active makes them have more L1 and L2 cache Ram between themselves and this could be another efficiency improvement plus less thermal or power throttling allowing cores to run slightly at higher average clocks all the time, I have just disabled 1/2 cores on my potatoe laptop that has 2 cores 5 threads (intel i5-7200u) 😀and it`s just running fine on 1 core 1 thread watching youtube and doesn`t feel sluggish at all and is super cool and quiet and I just love it knowing that if I ever need long battery life or especially now in hot Aussie summer I can just disable 1/2 cores and it`s all going to run just fine.

Nice review. Yes they always have.This is well known especially to old players who've played with tech since Win98. Agreed, thermals and threads do play an important part, but its mostly a software thing, how the cpu relays information between RAM and GPU and the actual coding itself. That's why a too slow cpu can cause a bottleneck for games too. Its all a balancing act. It was a huge hype when Quad cores came out after i2 Dual core cpu's, which improved games massively. After that release, game developers stopped focusing on cpu and put all their effort into GPU cards and shading. That's why XEONs or dual cpu's on MOBOs aren't better either. When the i7 hex cores came out, people found out quickly then that game developers basically but not entirely, stopped writing code to support more than 1 single cpu with 4 cores for games and any excess cores were utilized then more for networking, sound etc to disperse load. Amazing fast and multi cored hardware is not all the magic either. Software, as simple such as DirectX which is built into the OS can make games look and run amazing. i.e. if you run an old game on DX6 which has an ocean (the old win98 graphics platform) the water will look square & choppy sort of like minecraft. If you run that same old game on DX12 (the new win11 platform) it'll look fluent like real water and a million bucks. Essentially, if you want better looking (and sometimes) faster depending how the game was written, responsive game play (to a point), just upgrade your OS. Random UKposts example -> ukposts.info/have/v-deo/epeViGmNnq2i23U.html *edit: To see what version you're running: Click start. Type: dxdiag and run the builtin tool

You could also disable Hyper Threading from the BIOS. You will notice improvements in games. Some game engines just don't know what to do with so many cores & threads. I just upgraded my PC with a 4080 RTX 16Gb & a Xeon 2696 v3 18 cores with Turbo Boost upto 3.Ghz. I disabled Hyper Threading and my Port Royal score improved. Also I wont to experiment with turning off some cores because 18 cores is way to much for most games. Also my CPU is 145w TDP bound. So it will be hard for it to run 3.8Ghz all cores with Turbo Unlock. Also I won't to the "resize BAR" mode. I also unvervolted my CPU to help with its 145w TDP. Also I remember you saying why people turn on all the CPU cores & threads in MSI Afterburner this is exactly why you can analyse how each game engines uses the CPU. You will find games that utilise well the core counts or you will see only 1 core use in really old games. PS: also I can explain why less core/threads gave you in some titles better results its just that the game engine doesnt know what to do with nay cores & threads so wheb you give it less cores & threads it just utilizes it more. In RDR2 the game engine knows that you have a good CPU & it knows how to use all the threads & cores. Although I tell you try disabling the Threads on your CPU.

HUB tested various Intel CPUs with cores disabled and showed performance was affected more by cache size than extra cores. Chips and Cheese found games were experiencing low IPC below 1 on 5600x which is 6 wide, so perhaps having SMT threads isn't a real performance hit in these games. A lower active core count may have allowed higher boost or working sets to fit within L3

Rich, testing MSFS 2020 would have been interesting

I was hoping that you would test voxel based games like Minecraft and 7days to die ( vanilla and modded versions ) or maybe Vintage Story that have procedural generated maps. I always see the same games being tested and very few testing of voxel games. otherwise very interesting results.

Maybe there is also a point where the process threads are filled, while reduced core to manage that efficiently regardless of IDLE state

Cores seems to be just holding some points in the gamea computation I think, but it is becoming much more evident that cores are mattered as these games have taken the advantage of it. Cpu clock also matters in the recently demanding games, but as the newer game assets came into the point, the picture again changes. I assume how this is progressing really.

one explanation is that less cores can boost higher and if the game is ipc depended then you see a difference , the other is the cache and how is utilized if 4 cores have increased cache vs 6 cores then cache misses are lower and thus games run better

I knew about this before ryzen existed as a whole when ryzen came about older games on steam had trouble with the core count and you needed to disable some cores to play specific games

Another thing to think about is power delivery. More cores in sleep means less power to the CPU, which means lower temperatures and higher clocks for a longer time. Also same amount of L3 cache for less cores. I have a 5900X. I wanted to test something like that for quite some time now. Especially with Ashes of the Singularity, which claims to use 16 cores or 24 threads.

Might be an idea to turn off Hyper-threading and run the tests again. HT threads are not as powerful as Core threads.

Test 2 cores 4 threads against 4 cores 4 threads! This means you would have to disable the hyperthreading in the CMOS setup. The difference between a virtual core and a real core is that the virtual HT cores run on a real core. The idea behind Hyper Threading is that normal programs are not running an optimal instruction mix to keep all the resources of the core active at all timmes. So by allowing two threads to sun at the same time on the same core you could have, for instance one thread executing a floating point instruction at the same time as the other thread is doing integer math. Fetch instructions from ram can be run at the same time as the instruction decoder handles a instruction from the other thread and so on. To enable this the CPU registers are doubled up and some other parts that have a very high utility rate are also doubled, but a lot less is added to the silicon so it's no where near as "expensive" in area added as adding an entire core. Another effect is that running two distinct threads on a single core using HT does not double performance. There will still be occasions when both threads need to access the same resource in the core at the same time. When that happens one thread will simply have to wait. However adding real cores is almost linear. Adding a second core for a second thread there will be no sharing of the core resources. No waiting on the FPU to become available or anything like that. Memory access can still collide, but it's still faster than when you have two threads on the same core. So it would be interesting to test both 4 cores 4 threads and 2 cores 2 threads to see how much of a difference Hyper Threading makes. Heck if you do this why not give 3 cores 6 threads and 6 cores 6 threads a go? More data is fun!

Are those windows keys time unlimited, or they expire after some time?

Depends on the type of work being done

Do u think that 4 cores is the general sweet spot or its relative depending on the number of cores you have

test planetside 2, pretty cpu intensive game in the large battles. also, your 3440x1440 might be a little too high, leaving most the work to be done by the gpu?

I wish you could use all cores in gaming....then those Threadripper's would be the coolest 😎

You can disable cores for individual programs in the task manager chief

I think this was already covered on Gamers Nexus I think due to windows Scheduler insonsistencies on workload distribution and CCD load on the ryzen if memory serves me right.

Just shows that PC games today play on older machines with less CPU cores. This is because they have to sell as many copies as possible, and making the game accessible to 4-6 year old PC's help them do that!

how about to test the same but w/o hyperthreading

I changed mine now my computer is completely not running good at all. What did I do and how to I fix it

I wonder if there's a way to selectively disable cores, like disable only performance cores and use the system only on efficiency cores on intel for maximum battery when on the go, intel laptop's are known to be worse than AMD when it comes to battery performance.

Dont forget more cores can make more stable and smoother performance. As the other cores are working your background tasks and some games now can use 6 cores. Like baytleifled 2042

mine showed 16 processors so i put it at 8 and saw a huge jump in fps 120 to 180..windows 11 msi rtx 3070 32gb i7-10870h laptop ..is there any danger in doing this?

I have a RX 6800 and a 5800X3D...I wonder if me enabling all 16 threads will improve overall system performance? What about, changing core counts to individual running games rather than a global system setting?

I still run a 7th gen board with a pentium cpu. Still waiting for the price of secondhand parts to go down before upgrading. I'd be happy just to get 10th gen stuff.

i just had an interesting result when i tried this out. i have a three year old MSI GS75 Stealth, and when i set the cores to 12 (it was at 1) my framerate jumped by 20 in Starfield. in New Atlantis i usually averaged between 30 and 40 FPS. now i'm getting between 50 and 60. i'm going to try this out in Forza Horizon 5, and i think i will be getting the same result.

These results make perfect sense to me, you're testing older games/engines that at the time of their developement, were built and tested using cpus with lower core/thread count than modern processors, thus causing them to run better. Older games were optimized for duo/quad core cpus since thats what was available at the time. As you tested more new/modern performance intensive AAA games (eg: rdr2) you start to get opposite results since more modern games like rdr2 were built and tested on modern hardware (cpus with more cores/threads, multi/hyperthreading) Also could be amds cool n quiet feature that automatically and dynamically boosts the clock frequencies of only cores that are under load. A feature that helps with older games that lack multicore/multithreading capabilities, or basically only utilize a few cores. This can improve cpu bottlenecking and could also be another possibilty to explain your test results. If an older game utilizes only 2 cores and 1 thread per core and your cpu has 6 cores and 12 threads (2 threads per core/multithreaded) then cool n quiet will boost only those 2 cores past the frequency you would get on each core if you had all 6 of them being utilized.

do a video with nvlddmkm error fix

I would think that performance with background tasks active would be worse with fewer cores/threads

I don't think this was a very good selection of games to produce a result that is valid in 2023. Cyberpunk is moderately demanding on a CPU but really more as a function of needing a high single thread score, as opposed to needing more than 4 threads. Everything else you tested is light on the CPU. When testing Wonderlands recently on an older i5, I pretty much had to lock it to 30 fps to avoid stutter at 1080p. Maybe a very recent CPU could cope because of the single thread score being way higher. But it really isn't a valid argument because if someone is trying to game on a 4 thread CPU in 2023, it is going to be a few generations old. The question you are really asking here is 'how many CPU threads does a gaming PC need in 2023'. This is a common question from people looking to build a PC. This is very important when building a budget PC and asking someone who tested the premise on a current gen CPU by disabling cores really isn't going to give a helpful answer. The simple answer is that you need a 4c8t CPU and a 6c12t would be advisable if you want it to play new releases for more than a year. Even the very first 4c8t i7 can still game as long as you don't need AVX2. But almost any 4 thread CPU is going to run into problems often. Even a non-HT 6 core is a bad idea. Stutter is a very real possibility when running unlocked. If you just want a 30fps experience, it will work but that's all I would expect because anything more will be a bonus. One thing I can be fairly sure of is that it will be awhile before you really need an 8c CPU for gaming. All of the focus seems to be on improving graphics right now. Part of why the new releases all run like crap is that they are just dumping massive textures into ancient game engines. Another thing to consider which I didn't hear you mention is frame pacing. The fewer the threads and the weaker those threads get, the harder it will be for the CPU to maintain consistent frame pacing. This can lead to microstutters and eventually full on frame drops. You also have latency go up. Low latency and consistent frame pacing are what produce a good gaming experience, not absolute framerate.

super

Erm yes it does, even the FX 8350 does better now compared to the competing i5/i7s of the time

Less cores may mean less heat. Ryzen cpu will boost higher with less heat. Forcing off the 2 cores making it use core 3 more often at higher boost may be why performance goes up. 1% lows go down as back get task now are being run in cores running the games thread.

Did your wattage drop?

The results would be better with a better card installed as you are hitting a gpu bottleneck which has a massive effect on results.

From 6c-12t to 4c-8t all the tests were within margin of error because all those games came out in a time where 4c 8t was mainstream. Except for cyberpunk 2077, that game can use more cores. Now if the game can utilize more cores then it would be better to go with a higher core cpu. Yes clock speed/cache does matter too but it’s usually the bigger core count cpus that have higher clocks/cache. Unless you’re a good at overclocking 🤷🏻‍♂️.

AMD Ryzen 7 7800X3D or 7900? hmmm

by disableing cores you lower temps and voltage it could be its boosting a little higher due to more head room giving those results

Sir i have dell e7470 . I purchased refurbished from amazon i have bill... I wipe data from bios and my hard drive got locked. I never set passwords on hard drive ... I tried all file systems of usb my usb and hard drive not showing in uefi mode plz help I don't have enough money im stressed from last month plz help🙏 dell costomer service dont help me because my laptop dont have warranty

you could have dropped the cpu down to 3 cores to see if the lack of power used by the disabled cores is affecting the dominant core's boost performance translating into the game speed boost CyberCPU Tech to see if the performance loss is from loosing core 3 your dominant core by not trying 3 cores you don't know if the sweet spot was 4 cores or 3

TLDR; Yes, they absolutely matter if you care about playing modern, and soon to be released games. This isn't really new info? The majority of games don't really use more than four cores, but many do and slowly, but surely the average core count is rising for modern games. Cyberpunk is a great example as to why something like the 9600K and its six cores just doesn't cut it for the Phantom Liberty DLC. The performance hit is drastic going from the base game to the DLC if using a processor with less than eight cores. The games where cores really matter more than ever are those that are either simulation based, or are trying to use a lot of game AI. The comparison you use in this video is just odd. You're comparing two processors (or the same with cores disabled) with low core counts to one another instead of one with a lower core count to a more modern CPU. More powerful chips that have more than 4/6 cores blow them away. Cyberpunk is *heavily* dependent on CPU cores, and it's especially noticeable in Phantom Liberty. DLC that was designed solely around current-gen systems.

May the developers use all the cores😂

Counter strike 2 heavy on processor side I think😂

*affect

What is 1%low

That should be Affect not Effect. Effect is a noun. That could be: "Do CPU Cores have an Effect on Gaming" "Affect" is commonly used as a verb, meaning to influence or produce a change. "Effect" is primarily used as a noun, representing the result or consequence of an action. However, "effect" can also be used as a verb, meaning to bring about or accomplish something. Sorry, couldn't resist. NO! I am not an English major. 😮 Interesting video though.

What about emulations example ps3 emulators

how did my pc able to shout down my town internet how can a man walk through a wall he the son of satan anti christ you can turn waki tolki into a cb radio there a way you can pick up air planes on a radio you know i can you a foot stall to track people in my hall way and i can use my bathroom wall as way to know if there people in the building a nes game you can skip to end level in every game you can have as many life as you want also in all games use 24 cores for gaming it's over but try it anyway for fun can your pc play blackwake you can do a lot if you put your mind to it great video life is a test as well

Your windows buying site sets off my Antivirus.....

Damn pc you confusing

explanation might be quite easy cores boosting for longer and higher when you have spare power budget in situation your game don't need more cores and never uses tham anyway

An 8 core processor was better for this test.

na na disabling cores wouldnt perform as 4 core cpus

It seems to me you are GPU bound......run everything at 1080 medium settings and I would expect the CPU to matter more

Does this video prove that intel big little architecture is a joke for most mainstream gamers?

your test methodology is erroneous, you didnt mention system specs or games graphics settings and you are doing the tests at high res instead of low and that setting is crap, you disable cores from bios 🤦

Lol i knew the i3 cpus are the best gameing cpu u can have. You whats funny if the game does not uses more the 1 or 2 core i does matter how strong is you cpu, get owned

Can't imagine that you would have the right results as I'm on Intel and we all know that Intel is the gaming king. With Intel CPU you would get more performance out of the box regards games than AMD since you are running an older 5600x Also, you did not test or explain its single overclocked core or overclocked for all cores/threads. Your CPU is running at 4.4 I see? but my CPU is OC at 5.1ghz and so many new and older CPUs can run higher than 5.1 so is your 5600X Overclocked to 4.8? Then test results are different. Then it's the game and software, not all games might be made that well,