Mastering Efficiency: The Distinction Between Multitasking and Multithreading - Byte Size Tech

[@daddski1]

One of the reasons I keep coming back to your videos is the simple fact that unlike all the others you mention plus others, YOU talk TO us, not down AT us. I appreciate that very much....

[@dantheman7357]

I can't tell you how much I enjoyed this video. You spoke of complex matters with a clear and easily understandable explanation. I wish it was 30 minutes longer. Well done, as usual.

[@TSUNAMI17]

I've particularly appreciated your real world use perspective. It's easy to get distracted by a chart that shows you getting a few more FPS that your eyeballs can't actually notice.

[@MaculaPravus]

I'll take my 3 stars thank you. I was unable to get a 5900x when I was upgrading my gaming PC. And my wife encouraged me to get the 5950x. I have never regretted that upgrade.

[@weltsiebenhundert]

How to cool that monster?

[@daddski1]

@@weltsiebenhundert never had a problem cooling mine. Just a simple 280 mm qater cooler and it runs 24,7 with zero issues

[@L.Scott_Music]

I explain multi-threading and multi-tasking in this way. Multi-tasking is the guys picking up three boxes. Light boxes get moved quickly; heavy boxes move more slowly. Multi-threading is two or three guys working together to moving one box. Light boxes are a waste of resources but if the box is very heavy the two or three guys will move it much faster together. The OS is the manager lady telling the guys what boxes to move, when and how.

[@Segrey]

Not a great analogy for a few reasons. Multithreading isn't "2 or 3 guys" and it's not necessarily one box (because multithreading helps precisely when there are different boxes to focus on), it's rather how many hands(=threads) each guy(=core) is using to pick up a bag - one or two. Multitasking in this analogy would be how many separate conveyor belts(=workloads) are feeding those bags to the guys. Here, the frequency of each core can be regarded as each guy's strength.

[@ByteSizeTech]

While all such examples are never exact, that's a pretty decent one. :)

[@ironfist7789]

and an e-core is like one of those slacker employees that doesn't work too hard, but they don't get paid as much?

[@ejosephsimon]

i 100% agree with better 1% lows is more important than supper high FPS.. Especially if you playing Online shooters.. great job.

[@christopherzaf5768]

I've played Fallout 4 before and after an upgrade, and I can confirm dips are noticeable and trash your gaming experience while super high fps doesn't change your life (unless you play competitive games?

[@jakeonly]

Very few reviewers talk about 0.1% lows. One CPU will have 68 FPS 0.1% lows and another 98 FPS 0.1% lows, but the one with 68 has 10 more average FPS and is considered the "winner". It's often not even mentioned unless a game dips so far it becomes stop motion.

[@darudesandstorm7002]

it can be really volatile though. my 0.1% lows can vary *wildly* running the same benchmark back to back over multiple passes.

[@rcvillapando]

What’s the best gaming CPU that has the best 0.1% low?

[@benprice1984]

I totally agree the 0.1% lows are the most important number to look at because u can feel the stutter if they get too low. Chris from the good old gamer channel is the only person i know of who specifically benchmark's and tunes system's around the 1% and 0.1% lows

[@richard.c.guitars]

Yours and hardware unboxed are my fave tech channels, I admire your humbleness and honesty. Great work!

[@kurthectic2902]

if you spend 1 or 2 hours talking about computers, i don't care the topic, i will watch EVERYTHING.

[@GamingForLocos]

Have never bought top of the line, always got the Ryzen 5 cause gaming, now have the 13600KF cant go back to less cores, the smoothness of doing several things at one is there.

[@908967]

Back in early 2021, I bought a Ryzen 5900x over a 5800x or any intel 10th/11th gen options at that time. Before the upgrade I was using an i7 8700 since November 2017, I was noticing bogging in general multitasking while having multiple game launchers, and programs open while gaming. So I had to close some those down to free up resources while gaming. And then I noticed in newer titles that used multithreading more efficiently was pegging that i7 8700 at times. I wanted a proper upgrade from that i7 8700, and didnt want to worry about managing background tasks and have enough cpu resources looking forward.

[@YoloVib3s]

I completely agree … that’s honestly why I have yet to need or want to upgrade my 3950X. I can game, listen to music or watch YT & render a video not skip a beat!

[@YoloVib3s]

@stephan4923 lol too late my motherboard died so I upgraded to a 7900X/Asus B650E-F

[@mtaufiqnmtn]

​@@YoloVib3shey that's a nice upgrade, hope it will last longer

[@TheRockMoved]

At first, I clicked on this video under the impression that it would be a motivational or self-help video. (I only read the first few words) Regardless, I decided to continue watching due to my genuine interest in the information being shared.

[@Ladioz]

I never cared about the cpu until I watched this man. I bought the 5800X3D in Feb 2023, for 298$ and never looked back

[@cluckerbeatz898]

This is the best explanation I've seen on this issue yet! Thankyou

[@SpoonHurler]

One of the interesting things I've noticed on this subject is how many modern games are responding to having e-cores on and off. People were saying to turn them off but I've noticed higher (5-10%) fps, particularly the 1% lows. I suspect that having extra L1 cache to hold stuff and processing all the low priority threads is what's happening. And those have been tests on mostly clean set ups on my 13600k.

[@rpospeedwagon]

Keep them on. For sure.

[@SpoonHurler]

​@@rpospeedwagonthere was one game, Riftbreaker I think, that was a little better off (but I never play it anyway 😂) but almost all the other games where better

[@rotmistrzjanm8776]

Rainbow Six Siege used to run on I71200k in 600 FPS with E cores enabled and 700 so as you can see it's astonishing diffrence rendering game unplayable

[@PamirS-jh9oh]

There is common understanding in the industry that having superfast cores is a boon to performance, like the old PS4 having lightweight Jaguar class crappy but ultrafast cores but crappy none the less. A bunch of feral wild cats eating away unlike the cheetahs of early Ryzen.

[@shanent5793]

More cores means less context switching. One thread could draw 90% of a frame's geometry, get interrupted, and then only resume with empty caches after the frame is already too late. That's less likely to happen with more cores. With more cores it should become possible to dedicate cores to tasks and guarantee the lowest latency. They do this on servers. Instead of having every network packet interrupt a task, switch into the kernel, copy the data from the kernel to application memory, then resume the application, it's possible to have a core that runs at 100% utilization just spinning around and polling the network card, and telling the card to put the data directly into application memory evey time it hits. Doing that can require a lot of manual tuning, and a general solution that pleases everyone may literally be impossible. One advantage of consoles is that the tuning effort only has to be done once and is amortized over many identical systems. I think doing that kind of adaptive tuning on a PC that would be a great application of AI or ML

[@derenbong]

Bought 10900k with Samsung b die and 2080ti when 30 series came out second hand, have no need to upgrade yet, good deal indeed, if you’re willing and smart, the 2nd hand market is well worth it!

[@gulden_lover]

Thank you for that explanation. It was both informative and entertaining!

[@Jarmanreich]

I used to have R5 3600 (and 2600X before that) with 2 monitors and a lot of background activity, never again. All the stutter was gone with R9 5900X. Thanks for the advice back in 2020! :)

[@turtletech22]

Yea! 3 gold stars!!!

[@walter274]

Your explanation makes a lot of sense.

[@fmatax]

Thank you! I loved the video, you bring such a fresh down to earth perspective that i don't usually see anywhere else. Your Ryzen 7 7700X PC Build Guide was truly an eye opener and make me consider thing in a way that i have not before. (Still waiting for part 2 XD)

[@canonwright8397]

I always buy a CPU with a thought about what I'm gonna do with it after it's done for gaming. For instance: more cores are nice when you know it will make a good home lab later on. (you know... for fun!) Have a great day. 😎

[@fidelgames]

I enjoyed the explanation ngl it was very informative

[@jouniosmala9921]

The benchmarks can also lie to other direction in terms of importance for multitasking performance vs single threaded performance. Code compilation benchmarks are multitasking benchmarks that run multiple copies of a compiler to compile lots of files in parallel. In normal compilation the results of each compilation is stored on a disk, and only the changes from previous compilation gets its own single threaded tasks. And from testing perspective its better to test smaller changes frequently than wait for a large change. So the number of processes is way lower than in artificial benchmarks.

[@latrogeniwile58]

Great explanation. I followed it just fine.

[@MaximanLP]

3:28 He actually does this basically every time Rainbow 6 Siege shows up in benchmarks. He sarcastically talks about how unplayable the 7950x3d is at 841 fps compared to the 858 fps of the 13900k.

[@richardlittlejohn8184]

100% agree! I went from a 5600x to a 5900x a while ago, and whilst I wouldn't say the 5900x is quicker as such, it doesn't slow down so much. It's as if they both hit the same top speed, but the 5900x is relentless, it just doesn't slow down. I am unable to do enough to slow it down!! lol It's like having a Toyota Corolla and a freight train. They'll both hit 70mph, but with 5 people in each, the Corolla will slow down, the train just wont care! Great video tech :), but may I ask without twitter, and now we no longer have the discord, how do I go about asking a question :)

[@weltsiebenhundert]

WTF same platform massife cooling difference. Hopefully for cheap.

[@jackeyniraula]

What I am learning from these videos is that the newer is not always the better and sometimes old is gold.

[@NYI2010]

2:47 interesting how this aged lol

[@baldeagle1171]

Got the 13900k and I don't regret it. I can do is all with it

[@Shobu]

I'm very excited to hear this properly explained! //EDIT after watching the clip. Now I have an idea related to memory bandwidth. To my knowledge, ryzen 5000 series with 2 CCD dies, also come with 2 IMCs. I don't own a ryzen 7000 chip so I can''t comment about that one but... While multitasking I get the feeling that even this small detail speeds up my system so much, I never want to go back to 1 CCD unless there are huge architecture changes. (only talking about AMD here from a multitasking perspective, Intel is fine too)

[@ByteSizeTech]

The memory controller is not on the CCD, it's on the separate I/O die. Or, to be more specific, all RAM ACCESS is via the IO die. Yes, the CCDs have SOME memory management on them of course. I don't know that it makes much difference from that point of view, but a pair of CCDs DO have double the L3 cache.

[@Shobu]

@@ByteSizeTech thanks a lot for the explanation. I saw double the performance of ram read/write throughput (measured with aida) from 5800x to 5900x. 5900x had the same performance as 5950x in aida. This is why I guessed that even on the I/O die, there must be double IMC or something to do with the CCDs that double the throughput. I am no expert but these are my findings.

[@hexacarbide268]

I will admit my 3600x build is super solid. Not the best yes but super reliable and zippy. Given it was what I could afford at the time. Getting ready to build my 5900x with the sweet deals! Can't wait!

[@link3rd]

SHOTS FIRED!

[@dianaalyssa8726]

Great examples here.

[@jameswubbolt7787]

Great show.Thanks

[@andregomes3317]

I agree that 1% lows / 0.1% /frametimes are more important for the "game feel". Average FPS on its own is meaningless, well that´s my opinion anyway. If it dips hard and often, or if it crawls the minute you open a few extra background stuff, it doesnt matter if the average fps is 200 or more, the game experience will still suck. However, I´m not sold on the idea that a stronger 6 corer will have worse 1%/frametimes than a weaker/older 8-16 cores. I´m not considering 4 core CPU´s because AMD doesn´t sell any of those for a long time anyway, may be for OEM but that´s about it. The reason being is that those typical background tasks (discord, spotify, youtube, windows stuff, etc) resource utilization are very modest for modern systems, so I don´t think those small queue hiccups would be able to surpass the reduced single core performance dunk you´d get with an older gen CPU. Many games , even recent ones, are still limited by their "big main thread", even if they can parallelize the workload over many cores. However, that´s definitely possible if the game is using a shit ton of compute power. I would like to see that possibility explored, and I don´t think mainstream tech media explore that possibility very well. To explore it, I would select a bunch of games that are super hard on the CPU and can effectively use a lot cores(HUB selections feel outdated sometimes), then select a realistic but somewhat heavy panoplia of background apps , even multi-monitor stuff, and disclose it in the test configuration. And then "ignore" average fps altogether and only compare frametimes, 0.1 and 1% lows, since we´ve established those are much better metrics to evaluate the game experience. If you can prove a 5900x outperforms a 7600x in those newer heavy games (using those metrics, not average fps), that will change the way I view modern hardware. After all, a 5900x system costs about the same money as a 7600x one, when you consider the RAM and board. This would be an example of what I propose comparing, with current prices in mind (this is just an example, there are other reasonable configurations ) 7600x system (6 cores) - pcpartpicker.com/list/zPzRjZ About 480$ foir the platform and cooling 5900x system (12 cores) - pcpartpicker.com/list/G8ZDpH About 500$ for the platform and cooling, not exactly equal but I´d say close enough to make a fair comparison. The challenge is made! EDIT: One challenge with this test is that 0.1% / 1% lows can be very volatile and have a lot of run to run variance. I don´t have a solution for that, perhaps make longer /multiple tests to reduce it as much as possible and keep it scientific. I don´t exactly know, I´m not a benchmarker, I´m just tired of this average fps nonsense, I think it´s a terrible metric for us gamers. Perhaps single core performance is still "all that matters", perhaps not. I have an open mind.

[@pauldufresne3650]

A thread is an instruction pointer (and a stack) inside a single program. Some cores have 1 instruction pointer, some have 2 (multithreading). But the program must be designed to execute at more than one point at a time to use it. Each core can execute one program (different than the other cores) at a time. So when there was only one core, the OS was giving the illusion of executing more than one program by switching which one is executing many times by second. When there is N cores, there can be N programs really executing at a time.

[@AHOY_MATEY]

Brilliant.

[@The1commenterguy]

the analogy I always thought about in this situation is is like that one bodybuilder multi-armed pokemon. Is having 6 very strong arms that can curl "X" pounds while another being having 8 weaker albeit still strong arms than can curl "Y" pounds (X>Y). When left to a task of only curling the most weight with only 6 arms the first being will curl more weight but vs the other 8 armed being but that being will still have arms leftover to do other tasks.

[@mateiberatco500]

Well, my conclusion of your speech is: cache (2700X 2x4x512K vs 5600X 1x6x512K). The caches are there to keep the cores fed as much as possible. So the only agreement with you is that any single task will be penalized just by the cache splitting with a new task (I remember seeing the cache splitting slides on P4 HT, which might be a good reason why 1-core HT was slower than 2 cores; or any x vs 2x since. Yes, I know about ALU limitations). Therefore a slower CPU with more cache can be helpful on 100-tab users. Also there would be less thread inter-core movement (which means less inter-cache movement) when the system is close to 100%.

[@ByteSizeTech]

The 2700X to 5600X might have been a bit of a stretch. A better example would be 3700X to 5600X. Such is life.

[@C7TJFVB7CAZ_TyonorShap]

@@ByteSizeTech back then when i bought the 3800X i did a smart moove and i didnt feld bad but ppl told me sonstantly that i should have gone with the 5600X then i said, yeah okay but only if you do my work on it then, cuz i dont wanna use that shit! xD

[@Road2Rally]

This was really helpful, thanks o7

[@MyHhobbies]

for better or worst you are right

[@Ruby_Mochii]

I think it was the 5800x for 450 and 5900x for only 100 more. There's probably other examples too. Jumping from 6 core to 6 core by 1 gen or 2 is more of a sidegrade than upgrade, might as well get extra and there's larger difference.

[@rcald-gz5jd]

Very interesting.

[@kevinnel9275]

Does this mean - " silly example " - lesser cores at 'higher speeds' would be like funneling sand through a funnel with one large size draining hole, draining sand at high speed through the large outlet, whereas having more cores at a 'slightly lower speed', would be like draining the funnel through two medium-sized outlets, resulting in more sand/data to pass through the drainage holes . . . " more effective transfer of sand, even though the speed it travels is lower, the volume of sand at a set time is more/higher."

[@ben433]

I’d enjoy seeing benchmarks to show this comparison in action!

[@ben433]

And yes, I realize that it’s tough to benchmark a dirty system. But that’s why other reviewers struggle with these concepts!

[@skarletice7768]

I've got a asrock z790 pg riptide and a 12700 and it's really nice but I've been using it and seeing some of my friends enjoy their 13700k's and I've decided to just get a 13900k sell the 12700 and be done with it for a while.

[@bradt2656]

huh good info man, glad i choose the 3700x over the 3600x

[@dwardoggo]

cage match !!

[@BOXabaca]

Working with ESXI I just let the hypervisor abstract the CPU.. One ultra-fast quad core is as good as one slow as balls 8 core. Linux on metal is pretty good about this as well. It's really only windows that chokes to death trying to thread so many competing programs.

[@RyanMah34]

Bang on regarding what the other KZbinrs are good at and not good at. I’ve been telling people for years that some channels is not the go to for all advice. #itdepends 😂

[@mrhappy8966]

Been gone for awhile where is the misses tech, hopefully she's well??

[@Dogex670]

What do you think of a 5950x, 4090 and 64 gb ram 3600.. is it time for a cpu upgrade? I play at 1440 165hz

[@q3mago]

amd have dips is so weird cos there so much cache awesome video very informative ty mb it windows, Win32PrioritySeparation and such

[@vin.k.k]

Get those cores.

[@dusctilldawnrttc2308]

Ok 4 gold stars! My question is whether I would get better gaming performance using 2 or 4 of the fastest cores versus all 12 cores? My spec.s: Ryzen 9 5900x cpu Radeon 6800xt gpu 64 gigs ddr4 3600 trident z royal silver Samsung 970 1 tb mvne 2x crucial 1 tb ssd.s On a Gigibyte 570 motherboard

[@KinGDGaminG63]

No upgrade Path my ass. LoL I'm going from i712700k to i914900k in Oct. I know it's a one off. But I'm still doing it lol

[@ByteSizeTech]

My understanding is the 14900K is just a slightly faster 13900K Which is itself a nice update over the 12700K, but single gen upgrades are rarely worth it.

[@SpoonHurler]

Or it will be a 13850KS Ultra 😂

[@MrRmeadows]

I got there for 3 gold. So I will comment. If you are going for 2 CCX might as well get 16 cores.

[@trentginns1477]

i used to have an and ryzen 5 2600xwith cooler master hyper 212 turbo cpu cooler, crucial ballistix 16gb 3000mt, evga gtx 980 ti sc, samsung 970 evo plus 250gb m.2 os/2tb wd 5400rpm 64mb cach, 24" ips 75hz 1080p monitor. and i upgraded to an intel i7 12700k with a corsair h150i lcd cooler, corsair vengeance pro sl 32gb4x8gb 32mt, msi rtx 3080 10gb gaming z trio, samsung 970 evo plus 250gb os m.2, samsung 980 pro 1tb m.2, corsair hx1000w psu and for me that was a weeee upgrade multi tasking/threaded workloads basicly everything was way more snappyier this was all to play anything and everything i dont play a handful of games i play allsorts.

[@trentginns1477]

i got a lg ultragear 27" nano ips gtg 165hz 1440p monitor

[@rangersmith4652]

If total compute performance was the real measure, wouldn't it be easier for CPU makers to build a single-core CPU that runs really fast? The fact is, it would not. At a given clock speed and IPC, and with the rest of the system also being the same, more cores will always be better. Not in every application or game, but in enough of them to matter. And especially if you're running a lot of things (multitasking) at once.

[@LiteOS]

mb it all because low cache size ? intel old cpu was low on cache so with 10400f and i had to use older windows 10 and tweaked it to make smoother now with 12700k i dont feel any diff between bloaty win11 and tweaked win os and even i use most of the ram cpu pc still running without stutters

[@christopherzaf5768]

Does that mean betting a 13600k was a good future-proof cpu wasn't a bad idea! Knowing 1440p would soon become mainstream and 4k wasn't that far with my 1440p + 2 side 1080p. Time I'll tell...

[@GavriGames]

I've watch to the end of the video. so that's mean intel doing good job with E-Core to hit AMD CPU. btw I'm doing gaming, OBS streaming and music mp3. so intel budget cpu serve me better than budget AMD with low cores count. thanks for the video

[@hexacarbide268]

I like you spicy! Keep it coming! 💯 ❤

[@max-lee]

this is why im getting r9 7900.

[@C7TJFVB7CAZ_TyonorShap]

damn good choice bro 😘

[@alexisJonius]

Stability > raw performance anyday

[@Rayer24]

I've had a 3700x for 3 years now, I mostly play low demanding/older games. But in the most demanding games even 8 cores doesn't offer the best experience from a smoothness perspective. Add a little multitasking in there and the microstutters become even more noticeable. I could upgrade to a 5900x for only €240 but honestly I don't want to put that chip on a b350 board plus the 3700x fulfills most of my needs and wants. I'll upgrade it in probably 1-2 years from now. Hopefully AMD will increase the core count at some point like Intel is doing. Can't wait til the time we'll have 40 core consumer grade CPUs, we'll laugh at these 6 vs 8 core discussions :)

[@HELL_HOUND13]

so what i can gather, it will be best to upgrade to a ryzen 9 5900x instead of a ryzen 7 5800x3D. ill be on am4 platform for about 3 more years

[@martintop2605]

yes i have 5900x since 2 years and its a beast , 5800x3d can be better in specific gaming scenarios but in general 5900x its a better cpu and will last you longer. With 5900x no mather what is happening on the system you aways get smooth framerates. Can confirm that utilizes well the 4090 at higher res

[@HELL_HOUND13]

@@martintop2605 thank you !

[@user_____M]

I have a i5-10400F, 1660ti, 16GB RAM, 1TB SSD, 600 psu bronze. The motherboard only supports 10th and 11th gen. I want to buy it a 4060 ti or 6750xt, 64GB RAM and 2TB Samsung 990 SSD (all of this costs about as my PC two years ago), if I get it a new psu and i9 10900k it will make the upgrade around as much as buying a new low-mid gaming PC. Am I dumb and should instead buy a new PC with a i7 13th gen, 4070, 32GB DDR4, 1TB SSD for almost twice the amount of my intended upgrades (*if I buy the new psu and i9 cpu then the upgrade price is almost the same as this new PC) ? I play games occasionally but I do want to play new games too.

[@ByteSizeTech]

Unless you get a really good price for that 10900K, don't bother, just get a 13th gen. The performance improvement is noticeable just in Windows, much less games. You can sell that existing CPU/MB/RAM for a reasonable price and use that money, plus what you would have spent to buy the 10900K, and get a really solid 13th gen system.

[@user_____M]

​@@ByteSizeTechthank you, unfortunately no good prices/deals better than that in my region (Romania, Eastern Europe). I will buy a new PC for 2000 USD or more. 👍

[@SylphidUndine]

i did have a constructive comment but can i just get my 3 stars now.

[@ATVProven]

Quit teasing us with you and Rogue thumbnails bring her back already please.

[@ilovehotdogs125790]

Can you show proof that a 2700x will have better 1% lows than a 5600x. I’m calling cap on that

[@UNr34]

You can't compare the 2700x to the 5600x in any capacity, the 5600x is a much better cpu in every single scenario. Even comparing it to the 3600x is quite the extraordinary claim and you need to provide extraordinary evidence.

[@weltsiebenhundert]

Guess my choise... 5800X 220€ 5800X3D 300€ 5900X 300€

[@weltsiebenhundert]

@stephan4923 5950x for 5900x prive with a AIR cooler is ok cooled

[@rotmistrzjanm8776]

OwO 3 gold stars for me

[@Rye042]

Welp... 5950x it is.

[@adzansyahputra2675]

i really dont enjoy watching Linus content. his explanation is very fast and the tone is very flat, almost feels like he's reading through Teleprompter and he himself doesnt understand what he's talking about.