Hey bro, i did my research before disabling SMT. It's not about a small headroom for temps but L1/L2 cache hit efficiency. Sharing cores for the double amount of threads mean halving the L1/L2 cache size and cramming extra work to this halved cache space. The only workloads that will benefit from SMT are the ones that are more reliant on paralelism and not L1/L2 cache, like video rendering, everything else will suffer. It's a marketing feature: "two for the price of one". After disabling it, not only my system was running more stable (less stuttering) but it was also way more snappier. 8 high-ipc, well working (aka not under SMT) cores is plenty to do everything nowadays.
You should separately test 8 cores with the first 8 and the last 8 cores as one CCD has the extra 3D cache while the other has less which gives it lower latency and thus a clock speed advantage
the thing with smt is not that the temperatures are lower, for this purpose people do undervolting. People are turning off smt because some games don't really use multithreading properly, they need single-core performance and sometimes they're accidentally use virtual threads that are worse than physical.
Thanks for making this video ! I recently got into PC tweaking after refusing to pay someone on twitter. Calypsos guide recommends SMT off but this video is great and I'll leave it on
Best way to know if something is better is test it on your system. I just yesterday tested smt on a 7800x3d, and I had better results with SMT on, but I've also seen scenerios where SMT off can help.
Amazing work 💪💪 Would really be curious to see 13700k what it will be like and Happy Birthday🍻🍻 Your channel must grow because there are free things you learn from here, once again nice work🤝Toate cele bune si distractie astazi🍻
Der8auer showed in his video "Just a Waste of Time - Raptor Lake R̶e̶f̶r̶e̶s̶h̶ Renamed" that disabling E-Cores and pushing the P-Cores to 6Ghz (Disabling E-Cores meant the CPU was cooler and so P-Cores had more thermal headroom) gave a small but decent performance uplift.
went and searched a bit of info about HAGS, apparently affects latency and is generally better if you leave it on but no one really tested it, for anyone that wishes to disable it try to test it if you can and see if it changes anything other than the fps
The reason disabling smt helps is because of core contention. This becomes increasingly worse the more cores u have. I always try to get away with disabling smt if at all possible the issue is that windows scheduler isnt really optimized with smt disabled but it does lower your system and dpc latencies.
@@XMoldyMe but the fps isn't really the whole picture hyper threading even with higher fps can cause weird issues with stutter or higher system latencies because of core to core latency and cache.
Also I need you to benchmark CSGO because of most people there are curious about Nvidia control panel settings and some people said that they got better input lag and fps with it tuned.
You just gained a sub, I have been looking for a video just like this, since MW3 and MW2 are what I want to play at the best performance. May I ask, what in-game graphics settings are you benchmarking with?
thank you for the video! I would also appreciate a video about tweaks for Windows and its registry, for some of them there is not much information on the internet
HAGS is supposed to works like Reflex, not to improve frame. You also needed it for DLSS frame gen. SMT off is better on games with bad memory allocation and l2 is also cut in half due to simulated CPU cores, also most engines are optimised for 7 threads so SMT off only make sense on 8+ cores. Btw when is part 2?
Hi Beefy Tech! I love your content, I just subscribed! I have a question or rather I need advice, because here I have a configuration rather good yet (CPU 5900X, 6800XT, 32 gb 3200mhz) I've just bought myself a benq 360hz screen and I'd like to get the most out of it. I'm hesitating between making a new configuration under 7800x3D but still keeping my gpu, or just upgrading from a 5900X to a 5800x3D.... I'm having trouble making a decision because I don't think it's worth spending 2000 euros on a CPU, motherboard and DD5 kit, to gain only 20-30 fps. I'm not interested in the new generation of gpu at the moment, I'd rather wait for the next generation. Should I upgrade from 5900X to 5800x3D or install a new ZEN 4 CPU kit? Thanks for your advice :)
I personally recommend going for a 5800x3d, as it's your cheapest upgrade. I recently got a 7800x3d though, and my build costs was 880 usd for the Cpu, B650 Msi mobo, ddr5 6400mhz ram, and an air cooler. So for sub 900usd you can get a full platform upgrade with brand new parts. To sum up your performance gains, the 5800X3D will be 35% faster in cache and cpu sensitive games. The 7800x3d will be another 35-40% faster than the 5800x3d itself, but basically the 7800x3d would be way to fast for the 6800xt, and wouldn't be worth your money unless you plan to upgrade GPU also.
My 13700K is getting an abismal 332FPS Avg score for season 4 down from 374 in season2. I used to get a score of 420 FPS on my GPU as well and that tanked to 391 FPS for the current season. I'm still curious how much extra your rig would be if you put in an XFX RX 7900 XTX...whoa baby.
Basically, but what ull notice is that minimizing cpu usage and maximizing gpu usage in a game that tends to often be cpu bound ends up with funky results, even in the benchmark.
One game testing isnt enough to settle this, ive seen multiple benchmarks show high core count CPUs in general perform better with it disabled testing like 20 games with like 3 of 4 games perform better with it enabled, with the amount of heat that impossible to cool of 13th/14th gen intel cpus and amd 7000 id keep it disabled and get that 3% better fps, like i said one game isnt enough.
@Beefy-Tech hi! love your vids and channel - could you help - everyone keeps telling me that timer resolution is broken on windows 10 and that things like islc for timer do'nt work? could you clear this up so i can tell them straight and know for me too
@@Beefy-Tech Do you agree with me there are problems in nvidia drivers I rey amd gpu i get like 180-207fps in ashika and 100% usage But at my 3070ti i get like 130fps and 45% usage
@@11d3x I'm not certain there either, I don't think that'd be the nvidia drivers because I haven't had a single issue on any of my nvidia graphics cards. Could be your 3070ti specifically
Sure. On the 7950x3d Game Mode actually has an affect on the Affinity (which is how the 7950x3d knows to park cores in a gaming load) so I'll have to try it on the 7800x3d once that build is finally finished.
Wouldn't be helpful as tarkov would require testing the same exact path manually rather then running a benchmark which is easier as it's always the same exact benchmark.
@@Beefy-Tech downloaded your config but i get only 250 fps in GPU low 1st. with 7800x3d, 4090, ram 6000 cl30. Gpu clock is 2800. CPU score is also a bit lower. Did you change anything else?
@@papapepo4321I don't think there's anything else done, especially on the GPU side, as there's not much to do there. My 4090 clock is 3030mhz though, so it is overclocked.
You oubvisouly dont know how software is made, SMT OFF can help because each core wont have to switch between 2 threads half the time. It all depends on how the game is coded. SMT only creates fake cores. It doesnt give you more cores, time it halves the time if each real core into 2 logical cores. it so each core pretends to be doing 2 things at once when it is only ever doing one thing at once. I can show you benchmark that will prove you wrong ... another bullshiter on YT
You wrote a whole paragraph and somehow managed to say absolutely nothing relevant in the entire paragraph...what you said doesn't disprove any of my benchmarks, and different games and different systems will show different results... so no matter how much you want to show me your you know better cause your specific PC does it differently, it won't mean anything. Another bullshit comment, next.
8 месяцев назад
Smt is 2 waiters in restaurant. You can have exact same cook but with twice serving rate for dishes. Enable smt if your app has 32 customers.
Hey bro, i did my research before disabling SMT. It's not about a small headroom for temps but L1/L2 cache hit efficiency. Sharing cores for the double amount of threads mean halving the L1/L2 cache size and cramming extra work to this halved cache space. The only workloads that will benefit from SMT are the ones that are more reliant on paralelism and not L1/L2 cache, like video rendering, everything else will suffer. It's a marketing feature: "two for the price of one".
After disabling it, not only my system was running more stable (less stuttering) but it was also way more snappier. 8 high-ipc, well working (aka not under SMT) cores is plenty to do everything nowadays.
same I just disabled SMT and my games feel way smoother I thought it was placebo but not really it feels way better also im on a 5900x
@@vectorwolf9502 i've disabled it and never looked back
Wow, thanks for this! Not much info online for optimizing a 7950X3D with a RTX 4090. I didn't know about ISLC!
A belated happy birthday to the Beefiest Beefy Tech i know.
Fantastic work, thanks for all of the effort that you're putting into these, definitely shedding a lot of light on the topic!
You should separately test 8 cores with the first 8 and the last 8 cores as one CCD has the extra 3D cache while the other has less which gives it lower latency and thus a clock speed advantage
the thing with smt is not that the temperatures are lower, for this purpose people do undervolting. People are turning off smt because some games don't really use multithreading properly, they need single-core performance and sometimes they're accidentally use virtual threads that are worse than physical.
Thanks for making this video ! I recently got into PC tweaking after refusing to pay someone on twitter. Calypsos guide recommends SMT off but this video is great and I'll leave it on
Best way to know if something is better is test it on your system. I just yesterday tested smt on a 7800x3d, and I had better results with SMT on, but I've also seen scenerios where SMT off can help.
The thing i like about the MW2 bench is that it results within margin of error run after run until a setting is changed.
Amazing work 💪💪 Would really be curious to see 13700k what it will be like and Happy Birthday🍻🍻 Your channel must grow because there are free things you learn from here, once again nice work🤝Toate cele bune si distractie astazi🍻
Der8auer showed in his video "Just a Waste of Time - Raptor Lake R̶e̶f̶r̶e̶s̶h̶ Renamed" that disabling E-Cores and pushing the P-Cores to 6Ghz (Disabling E-Cores meant the CPU was cooler and so P-Cores had more thermal headroom) gave a small but decent performance uplift.
This is what im looking for, thanks man for the great content ❤
you definitely deserve more views and more likes sir, the amount of work you put in across all of your videos is insane. You have my like and my sub
thanks
Very good video... These small settings exploration does definitely improve system latency
Summed up a weeks worth of my research down to one video,
sub earned
went and searched a bit of info about HAGS, apparently affects latency and is generally better if you leave it on but no one really tested it, for anyone that wishes to disable it try to test it if you can and see if it changes anything other than the fps
A lot of work,thank you for that👍👍👍
Amazing work! Thank you!
13700k would be awesome since I suffer from this choice of CPU since I primary play warzone 2. Wish I found your channel earlier
The reason disabling smt helps is because of core contention. This becomes increasingly worse the more cores u have.
I always try to get away with disabling smt if at all possible the issue is that windows scheduler isnt really optimized with smt disabled but it does lower your system and dpc latencies.
Because, some games are optimized for HT and SMT and if you turn it off you'll lose FPS
@@XMoldyMe but the fps isn't really the whole picture hyper threading even with higher fps can cause weird issues with stutter or higher system latencies because of core to core latency and cache.
Also I need you to benchmark CSGO because of most people there are curious about Nvidia control panel settings and some people said that they got better input lag and fps with it tuned.
nice one. the only advice to post all results at the end as chart or something like that
You just gained a sub, I have been looking for a video just like this, since MW3 and MW2 are what I want to play at the best performance. May I ask, what in-game graphics settings are you benchmarking with?
can you try looking into Process Lasso if it helps with the game since it really is CPU heavy
thank you for the video! I would also appreciate a video about tweaks for Windows and its registry, for some of them there is not much information on the internet
Any chance you could do this test on lower end pc. All these tweaks were to help lower end pc parts reach higher all around performance.
HAGS is supposed to works like Reflex, not to improve frame. You also needed it for DLSS frame gen.
SMT off is better on games with bad memory allocation and l2 is also cut in half due to simulated CPU cores, also most engines are optimised for 7 threads so SMT off only make sense on 8+ cores.
Btw when is part 2?
great video!
amazing video! subscribed.
Happy birthday and thanks for video. Interested to see intel platform
Thanks fam!
This is insane! Im wondering if you could do with an intel 12th gen / 13th gen cpu a vid like this !
Hi Beefy Tech!
I love your content, I just subscribed!
I have a question or rather I need advice, because here I have a configuration rather good yet (CPU 5900X, 6800XT, 32 gb 3200mhz)
I've just bought myself a benq 360hz screen and I'd like to get the most out of it.
I'm hesitating between making a new configuration under 7800x3D but still keeping my gpu, or just upgrading from a 5900X to a 5800x3D.... I'm having trouble making a decision because I don't think it's worth spending 2000 euros on a CPU, motherboard and DD5 kit, to gain only 20-30 fps. I'm not interested in the new generation of gpu at the moment, I'd rather wait for the next generation.
Should I upgrade from 5900X to 5800x3D or install a new ZEN 4 CPU kit? Thanks for your advice :)
I personally recommend going for a 5800x3d, as it's your cheapest upgrade.
I recently got a 7800x3d though, and my build costs was 880 usd for the Cpu, B650 Msi mobo, ddr5 6400mhz ram, and an air cooler.
So for sub 900usd you can get a full platform upgrade with brand new parts.
To sum up your performance gains, the 5800X3D will be 35% faster in cache and cpu sensitive games.
The 7800x3d will be another 35-40% faster than the 5800x3d itself, but basically the 7800x3d would be way to fast for the 6800xt, and wouldn't be worth your money unless you plan to upgrade GPU also.
My 13700K is getting an abismal 332FPS Avg score for season 4 down from 374 in season2. I used to get a score of 420 FPS on my GPU as well and that tanked to 391 FPS for the current season. I'm still curious how much extra your rig would be if you put in an XFX RX 7900 XTX...whoa baby.
My 13700k is at 328 FPS avg in the benchmark, so I feel you🤣.
Would be nice if you did this with Tarkov? Perhaps a video on stiff that actually works?
Great work,. shouldve included windows game mode on vs off too
I actually thought that HAGS on minimized the CPU usage and maximized GPU usage?
Basically, but what ull notice is that minimizing cpu usage and maximizing gpu usage in a game that tends to often be cpu bound ends up with funky results, even in the benchmark.
One game testing isnt enough to settle this, ive seen multiple benchmarks show high core count CPUs in general perform better with it disabled testing like 20 games with like 3 of 4 games perform better with it enabled, with the amount of heat that impossible to cool of 13th/14th gen intel cpus and amd 7000 id keep it disabled and get that 3% better fps, like i said one game isnt enough.
Just subbed 😊
Thanks fam!
WOW. I'm gonna try that ISLC thing.
❤❤❤❤❤❤❤🎉thanks broooooooo
Great for mw2 and 7950x3d !
Anyone having issues with their Cpu usage at like 98% and gpu usage at 1-2% when using MW2 benchmark?
yeah, but this depends game by game. but give us a great idea in how to do the things
Cries in 13900
what are your results with the 7800x3d and 4090 bro ?
350 fps benchmark
So essentially there is no way to optimise this CPU further?
can i ask how are your +5v +12v and +3.3v?, how is siting at idle and running some games?.
@Beefy-Tech hi! love your vids and channel - could you help - everyone keeps telling me that timer resolution is broken on windows 10 and that things like islc for timer do'nt work? could you clear this up so i can tell them straight and know for me too
hardware ACGpu off
When i recording in obs was so laging why but when its on i have no problem
I'm not sure exactly.
@@Beefy-Tech Do you agree with me there are problems in nvidia drivers
I rey amd gpu i get like 180-207fps in ashika and 100% usage
But at my 3070ti i get like 130fps and 45% usage
@@11d3x I'm not certain there either, I don't think that'd be the nvidia drivers because I haven't had a single issue on any of my nvidia graphics cards. Could be your 3070ti specifically
thanks2 support the game devs too by buying games you like, it has benefit of always been played in updated state, which gives best fps and graphics
can you test game mode on vs off on some different games?
Sure. On the 7950x3d Game Mode actually has an affect on the Affinity (which is how the 7950x3d knows to park cores in a gaming load) so I'll have to try it on the 7800x3d once that build is finally finished.
I wish you did this for cs2 😭
!mouse
using cod as a benchmark? nah. u wanna use a heavy CPU game like tarkov.
Wouldn't be helpful as tarkov would require testing the same exact path manually rather then running a benchmark which is easier as it's always the same exact benchmark.
@@Beefy-Tech downloaded your config but i get only 250 fps in GPU low 1st.
with 7800x3d, 4090, ram 6000 cl30. Gpu clock is 2800. CPU score is also a bit lower.
Did you change anything else?
@@papapepo4321I don't think there's anything else done, especially on the GPU side, as there's not much to do there. My 4090 clock is 3030mhz though, so it is overclocked.
P r o m o s m
You oubvisouly dont know how software is made, SMT OFF can help because each core wont have to switch between 2 threads half the time. It all depends on how the game is coded. SMT only creates fake cores. It doesnt give you more cores, time it halves the time if each real core into 2 logical cores. it so each core pretends to be doing 2 things at once when it is only ever doing one thing at once. I can show you benchmark that will prove you wrong ... another bullshiter on YT
You wrote a whole paragraph and somehow managed to say absolutely nothing relevant in the entire paragraph...what you said doesn't disprove any of my benchmarks, and different games and different systems will show different results... so no matter how much you want to show me your you know better cause your specific PC does it differently, it won't mean anything. Another bullshit comment, next.
Smt is 2 waiters in restaurant. You can have exact same cook but with twice serving rate for dishes. Enable smt if your app has 32 customers.
@@Beefy-Tech wondered if you help with my question above?
Yes more like 1 waiters shadow.
just download process lasso to disable smt