Single-core vs Multi-core Performance and Efficiency

[@RogueHetero]

Gave thumbs up specifically because of the thumbnail. lol

[@abrarshaikh2254]

2:41 Why P= fCv^2 ? E = .5*Cv^2 P =∆E/∆t Where ∆t = T/2... T=1/f, and only take positive half cycle in account of energy dissipation. (Since clock has 50% duty cycle) Hence P= fCv^2.

[@arjaydee6379]

Seems like that 8 cores cpu is the sweet spot fot Android for now. As the 10 core cpus of Mediatek did fail miserably on thermals & power draw.

[@AkshatSharma1505]

But why did they fail? If multi-core is supposed to be power efficient?

[@rafizahabdjamil2111]

@@AkshatSharma1505 It’s not straight forward as a race track , imagine it as a highway traffic not all vehicles are the same size and the driver too, that’s how it work

[@johnj8639]

Most programs don’t scale linearly with the number of cores added. Finding prime numbers is more easily able to be divided among cores, but for most other programs there are only so many cores that you can use and each core won’t be used equally. If you’ve used a processor with more than 6 cores in gaming for example - you can have scenarios where where the game is bottlenecked by the processor even though not all cores are being used, and this is more evident in many other programs when you look at AMD’s 32 core processor, and why you’ll see people refer to AMD’s expected 64 core processor as having too many cores. Adding more cores for more performance is not the answer, having good single core performance is important too, because there are only so many ways you can logically divide a program up, and for any time sensitive program you likely wouldn’t want to divide up the load because you’ll have to wait for the other core to finish its side before the program should proceed - or same with some calculations in some sense, you can’t have one side of a calculation finish before the other and move on because the end result could be incorrect, so just because some programs don’t use all cores does not mean it’s unoptimizabled it’s possible they divided up the program as much as they could. So having multiple cores is important but there is a huge amount of diminishing returns, having faster cores is now becoming more important as we approach more than 4-6 cores.

[@stevenhe3462]

Yes, dividing tasks is difficult. Theoretically one can always divide the tasks nearly evenly through some work. But irl programming multi-core programs are much more of a headache than that of single-core programs and many programmers just can't border that unlike the developers of Linux. High single-core performance is definitely a convenience.

[@johnj8639]

​@@stevenhe3462 Most time sensitive programs would receive no benefit from being split into multiple threads, if anything it would be more inefficient than a single thread, the same goes for any thread that is dependent on a response from another - just because you can divide it doesn't mean you should. Plus for example many a-sync programming languages do not allow multithreading, like JavaScript for example, this is intentional and by design, hence why its a-sync.

[@sabirjee5313]

Fantastic explanation but I want to add something more to make these nice lectures more understandable that if these lectures are included with animations than hope they will provide more and more good information for all kinds of viewers.

[@bryantallen703]

I remember getting the Samsung Galaxy Note 3 back in 2012-2013. All because it was the 1st North American smartphone with a 1080p Super AMOLED and 3GB's of ram. That with a Snapdragon 800 (28nm/HPM) 2.23GHz with the Adreno 330. A Quad core running at that speed was unheard of. It even beat out the nVIDIA Tegra 4 in gaming. In 2015 when the mod and root scene was popular i overclocked the Snapdragon 800 to 3.1GHz on the Note 3. I kept it running long enough to finish GeekBench 3 and set a GB record with a SD800. It should still stand today on HWbot.

[@iflnr978]

one problem with most phones including all Samsung phones is they don't have developer additions like the one plus one and the one plus6 are, Samsung phones, can only run locked down closed source proprietary Android OS (i know it is based on open sourced Linux) which most likely reports on your activity to Samsung or Google and the NSA (else why keep it secret?). perhaps it's ok if you are not doing anything illegal or immoral, but if you paid for your phone and supposedly own it, would you choose to send your data off to them if it were up to you? if you had kids, do you approve of Google and the NSA having file on their activity cause one day they'll nominate them for sainthood? you could get a $150 Pine Phone that ships without an OS (new but uses 3 or 4 year old technology) and if you can get it running with Ubuntu Touch you can have better privacy, which for some people is important. however faster phone and technical advances are needed. so thanks for pointing out how great the Notes are. I didn't think they could be overclocked. the recent announcement of Samsung partnering with Microsoft may produce better phones if they go down that path. while Microsoft has done some pretty bad stuff to open source projects in the past, in some ways they have learned their lesson and not only contribute money to almost every linux convention that takes place, they developed WSL, the Windows Subsystem for Linux and open sourced many projects (Ex Fat and PowerShell) so they can be audited and used by privacy caring individuals. I got a One PlusOne and got Ubuntu Touch to open on it, but it lacks important functionality. I cannot get it to function as Rob Braxman tech, KZbin channel has done. thanks for posting

[@logicalfundy]

Interesting video. Although I should note that in practice not all code can be made parallel. In particular, if one operation requires the result of another operation to complete then it'll have to wait regardless, and more cores won't speed it up. I also play my share of older games that weren't developed with many cores in mind, so I did buy a CPU that focused on single core performance for those games. That said, it's still a 6 core / 12 "thread" CPU (i7-8086K), so it'll work well for multithreaded tasks.

[@megamind5774]

Happy New year Professor...😎... And a very happy new year to all of you viewers...👍🏻👍🏻👍🏻

[@1MarkKeller]

*RAJDEEP PANDEY!*

[@_syedmx86]

3 Garys in the thumbnail. Nice!

[@neuroronan]

Gary, you are great! A true professor! I'm in medicine, a completely different background, but still you manage to make me understand most of what you say, and make it interesting, with good depth and applicability! I always rush to see your stuff, even if it doesn't mean much to me at first!

[@ObsidianMercian]

Thank you for such an informative interesting video! Always great to find well presented educational content on KZbin. You have yourself a new subscriber!

[@ScoopDogg]

Nobody says why we have more cores, because of getting smaller and more powerful frequency we have made the electrons travel over a longer path, this leads to a longer time to travel so we make more cores of lower freq to make the 1 core distance freq. Multiple apps running is good for more cores, but bottlenecks occur. Moores law is here we are reaching limits 5. Ghz 1 core is always expensive than multi less power cores for a reason.

[@erickbaka]

More cores is a bit faster because the whole CPU has more L1 & L2 cache available than for a single core. If you had a Raspberry Pi CPU that had double the normal L1&L2 cache per core, it should theoretically be faster than 2 half-speed cores.

[@bnadem.panormal]

Exactly the question I was googling minutes ago an then got this video notification!!

[@anilmhatre1069]

Maybe 2 cores with 3GHz

[@janeric1810]

That would be core 2 duo e8400

[@PrathamPatel18]

Amazing as always :)

[@entertainmentgyan202]

Happy mew year Gary. Multiple cores r much better then single core. I agree with uuuu

[@MohdAkmalZakiIO]

That's why its exist, even for consumers.

[@TA-eo2ww]

Gary you should have compared the power consumption in each test. Would have been even more interesting. The future is multi core, and it's going to be up the programmers and processor designers to make the advantages of multi core more effective for users. They have to collaborate with each other closely. Like your programmes very much!

[@GaryExplains]

I do compare the power consumption, there is a whole segment about that at the beginning of the video. What I don't do is measure the actual power consumption during those tests.

[@Raven-lg7td]

damn that was amazing, I learned more than in 11 years at school. Just 1 question - some code cant be multi-threaded by nature, so single-core perf will always matter, right? For example games, can they really digest all 16 threads in future ps5? Very unlikely

[@EmilBozhilov]

dx12 is mainly made to make games and developers use multi threaded cpus

[@U_H89]

With multithreading it is difficult to use all cores identical, so usually one cannot run all of them at 100 percent, especially in gaming. For gaming you want to prevent frame drops, even in case the cpu is already starting throtteling... so the game tries to get cpu power without using it/at least without causing a lot of heat...

[@logicalfundy]

It's definitely true that not all code can be made multi-core - and many games can't take advantage of multiple cores. I bought a 5 GHz CPU because many of my indie games are basically single threaded (Kerbal Space Program used a single thread for physics, but it's been a while since I've played so I don't know if it's still the case). That said: Actually, there are things games do that can be multi-threaded. Physics and some types of AI can take advantage of multiple cores. The game loop itself can be put into a different thread than other tasks, as can networking and file access. Doing so can increase performance because if they're in different threads they're not blocking each other. I should note that modern AAA games do generally take advantage of multiple cores - so if you're into the big blockbuster releases, then more cores makes sense. Rendering the graphics can theoretically be done on multiple cores as well, but that's offloaded to the GPU these days (which is actually massively parallel, more than any CPU - high end GPUs can have thousands of "cores").

[@Phynellius]

NPC Ai should be able to leverage some of that extra computing capability. I can imagine there being some limiting factors when it comes to multi-threading single tasks such as part of the code requiring a result produced by the earlier portions of the code requiring the instructions to be sequential rather than parallel

[@mas921]

16 is nothing for games, GPUs have THOUSANDS of threads on the fly for rendering.

[@carmiladewi7618]

Besf of the Best explaination in all youtube channel, you talk about all the sistem integrated when other people only think about multi core & frequency

[@hugbearsx4]

Small nitpicking: actually *power* is of less concern than *energy* (which integrates various levels of power over time).

[@Archer_Legend]

Great video as always! Make a video on gpu performance! Maybe about 4cu of a12 vs 7 cu of a12x😉

[@redocious8741]

A D not you again. Like i said the A12X is 2.5x the power of A12 and thats on period

[@nintichance]

Woah, you know so many things. How do you keep it all ordered? How did you learn all this? Thanks for the very well explained video! :)

[@MohdAkmalZakiIO]

Explain to us about Thread and Multi-Thread as well.

[@davejordan87]

What you mentioned in the end "IF you're programming it right". That's a big if right there. Unfortunately, a lot of software out there is still single threaded. So, unless I'm on a phone and I care about those extra watts, I would go for a single core with higher clock

[@عبدالرحمنالهاشمي-د8ت]

I would love to see 2+4+2 on android 2 big cores, 4 mid cores, and 2 small cores. Like 2 X4 each at 2.0 ghz, 4 A720 each one on 0.75 ghz, and 2 A520 each at 1.0 ghz, or just any cpu in market currently and apply the video’s idea on it something like that, I just would love to see how it would perform against the competition, efficiency and speed.

[@chacmool2581]

What we did not learn is what types of tasks are inherently single-core and what tasks can be multi-cored.

[@voivittupeccaehet]

So this only applies if the program you're running is optimised to use multiple cores?

[@1MarkKeller]

*GARY!!!* *Good Afternoon Professor!* *Good Afternoon Fellow Classmates!*

[@GaryExplains]

MARK!!!

[@jordanwarne911]

Good evening Mark!

[@1MarkKeller]

@@jordanwarne911 *JORDAN WARNE!*

[@sonnyyyaji16]

i love reading your comment in every single garry's video haha

[@1MarkKeller]

@@sonnyyyaji16 *SONY AJISUKMA!* Thanks Fellow Classmate!

[@SamsQuarters]

Thanks for clearing that up, Gary!

[@tomcombe4813]

Should probably mention that some types of program can't really be split up to run across multiple cores. So for these it's best to run them on a single, powerful core.

[@anilmhatre1069]

But Snapdragon 625 has all same cores at cortex 2016MHz

[@GaryExplains]

True. And they are all efficiency cores (A53) as well, no performance cores.

[@its_rns]

That was a really nice chip (for that time)

[@wisdomboyzzz3334]

We need 4 high performance gary and 4 high efficiency gary

[@СергейД-ч4ь]

Enjoyed this video very much! Fantasticly interesting!!!!Thanks a lot!!!

[@RCmaniac667]

Remember all that marketing when the Galaxy S2 was announced? Especially two tunnels advertisement?

[@williamhart4896]

Standard arm architecture fastest cores do the heaviest work , the most efficient ones do everything else ,ie all the background noise apps .

[@PrivateSi]

The Xeon Phi series was the closest the world got to one of my ideal architectures but I've thought up a design from a rigid premise... a 2D, single layer architecture.... I know CPUs and PCBs have many layers but in 2D Flatland the maximum number of cores that can directly link to every other core without crossing lanes is 4. I think a triangle is the most effecient packing of cores and connecting lanes.. You can have multiple lanes connecting each core in this each-to-each network, but 4 cores max.. I mused with the idea of a fractral-like configuration with 4 joined cores forming a higher level core, with 4 of these then joined to form the next level core, and so on.... this is all in 2D but you can have multiple 2D layers. The idea is you have far fewer vertical connections, errors in which reduce chip yields. You can go as big or as small as you like, with as much redundancy as you like (if a level 0 core fails there are plenty left)... 4 cores can share each other's on-chip memory and work together in a number of modes. Level 0 cores could be very simple, or a full on X86 core.. I'd start with simple ALUs, probably a multiplier or 2 each. perhaps floating point at level 0, perhaps level 1.. They can work together to form a GPU, or multicore CPU, or both, on the fly... Having said all that, it's probably not a great design in this ever more complicated, 3D chip world, but you never know.. it is based on logical principles.

[@joechang8696]

The basis of the pre-2003 concept of Moore's Law was each process generation, a 0.71 (1 over square root of 2) linear shrink doubled the transistor density at a given die area, supporting a 1.4X increase in performance per cycle. There 0.71X linear shrink also made transistors 1.4X faster for a combined 2X gain per generation. Suppose we have hypothetical 3GHz core on a specific process and die area. Then in theory, we could design smaller core of half the area of the first core, and this core would be 0.71X lower performance, which we could imagine to be 2.1GHz. Hence the correct equality is one 3GHz core should be about the same area of 2 cores designed to achieve comparable performance of 2.1GHz of the baseline, or 4 cores of 1.5GHz So really, its about how much performance you want in a single core versus having confidence to scale over multiple cores.

[@codyrap95]

Thanks so much for this vid Gary. It all makes sense

[@Aleortgu]

Great video, Thx

[@kuriang1802]

Bravo, that was pretty informative Sir...

[@Immaustine]

Great teacher there👏👏

[@_Liam__]

Can someone please help me answer this question? What is more important for high graphics 3D mobile game? Single-core power or Multi-core power? Thank you very much!!!

[@mrjafs4616]

I wanna know too :)

[@mrjafs4616]

Did you ever figure that out?

[@SouravTechLabs]

I think it's better to run single core on GIL programming languages like Ruby, Python, Some JS Engine (Spidermonkey for example)

[@rondamon4408]

Excellent video, but you have shown different software scenarios, like real time operating systems, actor model, etc.

[@nobeltackful1]

Great video Gary you answered the question regarding Multi-core vs Single-core or Apple designed processors vs ARM designed processors nicely: "it depends." One can't simply say that A13 is better than SD865 and vice versa. The comparison you did with Raspberry Pi is repensentative for multi vs single-core, but as you stated not between Apple vs ARM designed cores. Android favors Multithreading more than iOS, since the ideas behind the design of SoCs are quite different. Apple SoCs behave quite like desktop CPUs by having fast big core to do the tasks as quick as posible then return to idle (Race to idle/sleep). Snapdragon SoCs for example have more performance cores while use less power. But less power doesn't always translate into less ENERGY or better ENERGY EFFICIENCY, as it can take longer to do the same tasks. As seen Apple A12 and A13 use newer ARM instruction set, have better memory subsystem and better IPC according to SPECInt comparing to SD855, SD865. All in all, there isn't one that completely better than the other. It depends on the usage and how well writen the softwares for respective platform are. In the future videos I would love to see more of your views on IPC and energy efficiency.

[@MiguelAngel-rw7kn]

"One can't simply say that A13 is better than SD865 and vice versa." In this case, yes you can. A13 is faster at both single and multicore.

[@nobeltackful1]

​@@MiguelAngel-rw7kn If you meant better = faster then yes. Even the Cortex A77 in new SD865 has just caught up to Apple A11 's Monsoon cores. Multicores ? I think they are almost the same (maybe wrong though, i dont have any info on that) Better Power efficiency? No. Better Energy efficiency ? not any more with the new SD865, they are pretty the at the same ballpark now. (we saw no increase in EE between A12 and A13). Faster NPU ? Nope, with A14 maybe :). Faster GPU ? Yes, the Adreno 650 in SD865 trails behind A12 's GPU in term of speed and efficiency. Like i said it depends on your usage. If you game alot then yes A13 is a better pick for you.

[@MiguelAngel-rw7kn]

John MacTavish I meant in the topic of the video; the A13 is faster at both metrics: single and multicore even if the later is just for a little bit. Everything else will depend on the task, and to be honest, we don't have any metric on how much energy every soc consumes when all cores are fully active. As you said, the lighting cores on the a13 are less efficient than the cortex a77, but the thunder cores are much much better than the little a55 ones, so it can compensate when fully load. But we don't really know.

[@mas921]

Thanks a lot Professor for the explanation! Will use this video to "educate" all those Apple fanboys bashing android SoC's 😈

[@1MarkKeller]

*MAS921!*

[@Nossieuk]

I hate to say it - but if you are apple and you know exactly what hardware your running software on - it's ALWAYS going to be better than any random generic android soc. Diversity is both Androids strength and weakness, especially in todays trend of race to the bottom.

[@manulovesjesus]

Stop always making it about android vs apple. Also: The video compared cores with similar architectures. Dont see how you can use it to lecture anyone.

[@mas921]

@@manulovesjesus What is "Geekbench scores?" ....I😁

[@mas921]

@@Nossieuk have you developed an android app? You won't say so if you really get into how android software stack runs.

[@Nossieuk]

The simplification in this video is .... mind blowing - without taking everything else you have mentioned and said and pointed to in other videos into account this really says NOTHING. You might as well also say that it depends on how much cache per core? is it shared over multiple cores? is it dedicated to one each? Water is wet, unless it is dry.

[@GaryExplains]

LOL, I think the problem is that you didn't understand it.

[@Nossieuk]

Your efficiency does not scale and outside of specific symmetrical mathematical computes does not make one squat of difference. Performance will scale less than linearly with the number of cores put in place. In fact, the incremental value of adding another core approaches zero as the high degree of intercore communication eventually overwhelms the gain adding the core.

[@the1barbarian781]

All cores are equal, but some cores are more equal than others

[@アニメの歌]

Which is the best? Single core fast or multi core fast? Antutu result Redmi note 3pro SD650 ADRENO 510 6x 1.8GHz Single core-1465 Multi core-2766 OR Antutu result Redmi note 4x SD625 ADRENO 506 8 x 2 GHz Single core-831 Multi core-2957 WHICH IS THE BEST PHONE FOR GAMING? PLEASE ANSWER ME

[@patdbean]

Were there not single core, but hyper-threaded CPUs between the single and multi-core CPUs? I am sure I remember having a 3ghz P4 with hyper-threading in about 2003/4 before getting my first true multi-core AMD phenom II x4 @ 3.4ghz in about 2010.

[@yuvrajchandra2136]

isnt the power should be half of what he mentioned ...

[@OroborOSX11]

Well it’s about time Android finally figured out how to separate the display compositing process from the app threads... Core Animation has been doing this since iPhone OS 1.0.

[@GaryExplains]

Eh? It has, that is what the SurfaceFlinger does.

[@ScottJPowers]

Gary Sims, eh? Are you saying you're a Sim on a computer?

[@GaryExplains]

No, I don't think I am... but even if I was I wouldn't tell anyone because there are lots of nation states that would like to pick me apart! 🤖

[@ScottJPowers]

@@GaryExplains hehe

[@1MarkKeller]

@@ScottJPowers The Professor is a Time Lord. That much is fact.

[@kivsa85]

Thank you I'm getting the 2x 4 cores instead of 8 cores

[@itsmuhammad2305]

So... Should we start wanting 10 core mobile proccesors?

[@GaryExplains]

Striking the balance is always key, but don't forget the MediaTek Helio X20 was a deca-core processor.

[@itsmuhammad2305]

@@GaryExplains I know, they will eventually up the core count. Maybe.

[@itsmuhammad2305]

@@jamesjames7601 I believe you but I want to hope as well.

[@Sukhmani_randhawa_artistry]

Hello sir, First of all, amazing video. Helped me clear a lot of doubts. One thing that aught my attention in this video is "quantum tunneling". I googled the meaning, but how it limits frequency increase, I would really like to learn about it. Thanks

[@Edgar4T007]

Very well explained.

[@joeshmoe000]

They should make multicore CPUs have a firmware that handles all the stuff, so the OS and apps see it as just one much faster core. The reason is cause programming for multicore is much more complicated. I learned C# and it became clear that making apps that take advantage of multiple cores is complicated and annoying. It would seem more elegant to do it this way. I wonder if it's possible.

[@GaryExplains]

CPUs already do that. It is called Instruction Level Parallelism (ILP). That is why modern high perf cores use out-of-order pipelines and have multiple units (like ALUs, load/store, branch) in the execution core at the end of the pipeline. From a software point of view it is worth looking into OpenMP, it is a topic I have on my video TODO list.

[@metesnsr4544]

Thanks for great content

[@EyesOfByes]

This video was really on another level

[@Ureallydontknow]

like the difference between single gary and multi-gary

[@YouCantParkThereMate]

Hi Gary! Dont know if imissed this bit while having to look after a 3 year old monster but.... The load the program requires would have made for another comparison. An app that prefers single threaded, high frequency operations would never be faster on a power efficient cpu as the program isnt designed to split the load. I would call this a badly designed app but im using it as an example. Seperate from gpu and different hardware architectures. Only say this as i think theres some simple people who wont take "programs being written properly" to cover this point.

[@AK-ny5bz]

Did you really measure power consumption in the above experiment too?

[@josephgamal3135]

One big "Gary", Vs two small "Gary"s Nice thumbnail

[@U_H89]

I will prefer single core for single thread processing of course 😉. Sure with multi core you can save energy, but more cores need more space, more die, so will be more expensive ... Not easy to find _best_ compromise for all (people/situatitions) ...

[@GaryExplains]

But also remember that a single core with high single thread performance takes up a lot of space, making such a powerful core takes space. The transistor count of the hexa-core A13 is similar to the transistor could of a octa-core Snapdragon.

[@U_H89]

@@GaryExplains I aasumed we were talking about the _same_ cores, so f.e. A53 core. So I would prefer a single A53 core with 3 GHz instead of two A53 cores @ 1.5 GHz... (Although A53 might be a 'bad' example because of missing out-of-order feature...) And usually you cause an overhead when using multiple cores instead of one. You need (different) multiple caches, RAM management (read/write), consitency(?), you cannot easily predict/optimize codes (which core will run its thread faster, which will be first and read/write RAM or continue with next thread/task).

[@C0nanblah]

How does hyper threading play into power consumption and speed.

[@Niravbagada]

Does It applies to gpu accelerated tasks?

[@leonardpetty6029]

They making all these multiple core processors and Harley anyting is made. 2 benefit from them

[@Dominus_Potatus]

So the conclusion is In ideal environment, more cores are better. Bur in real life, you can't do work without doing prior work, hence doing it faster in single thread will make it faster. Unfortunately life has no straight line, so the combination of more cores and more frequencies will be great. I really wonder... a modular processor such as Ryzen should make use of different architectures. Why don't they combined 1 chiplet of Zen2 and 1 chiplet of Zen+, and so on. It will drive their processor price down since it will reuse old technologies but will increase the overall performance. Imagine you are playing with 8 cores (4 cores Zen2 and 4 cores Zen+) While gaming nowaday can't take every advantagr of cores, having 4 higher performance will crank the performance in game rather than 8 low performances. The overall multicores will be lower than you use all Zen2, but hey... it will still be cheaper. And as far as I can see, It will prolong the life cycle of Zen+ and Zen2. Extend the 2020 end of life of AM4. I mean... I'll be satisfied to have $200 8 cores processor. But it will run like big.little ARM, could it be infringe IP?

[@Dominus_Potatus]

@Vishal Belbase I have heared it. Have a very little knowledge on it. Maybe you or even Gary can explain it 😆

[@samvortex4136]

Would you rather have 5ghz 8 core or 3.5 ghz 16 core ( same arch ) ? Multi threading code(program) to run on more then 8 cores is beyond capability of our software devs

[@GaryExplains]

Sounds like "your" software devs aren't very good then.

[@hatihitamrosmerah1692]

What about music production?single core or multi core?

[@GaryExplains]

That would depend completely on the software being used.

[@yogendratala8577]

Thank you

[@jhayrez]

Well Explained 👏😊

[@ozabid]

I have subscribed specifically because of the animal farm reference! Will definitely be watching more videos

[@michaelhazellful]

The clock speed per core doesn't translate to the efficiency of the core. Sometimes a 2.4 GHz core can out perform a 3.0 GHz core. It just depends on the cores efficiency and stability. If we went just off core speed, the sd845 would be superior to sd855. Cortex core a76 is definitely more efficient at work loads than cortex core a75

[@GaryExplains]

LOL, you clearly didn't watch the other videos I suggested and IPC etc. Following that, the assumption is we are talking about the same microarchitecture at different clock speeds. 🙄

[@michaelhazellful]

@@GaryExplains tbf, you can't expect people to watch every video you've made after you make a headline. All I'm saying is measuring the clock speed per core is a poor way of analyzing how well it would do. IPC(instruction per clock cycle) indeed is important. But if someone were to go off, oh this is clocked higher, so it must be better mentality is terrible. Something can be clocked higher and still lag behind a lower clocked performance core and being a lot less competent.

[@GaryExplains]

It isn't a poor way when the CPU is the same, it is ONLY way! 🙄 Again, you need to watch the other videos, if you don't want to then I can't help, I ain't going to repeat what I say in the video here in the comments.

[@michaelhazellful]

@@GaryExplains I think it's still important to disclose if the CPU is the same with the same cores. Again, expecting people to be returning to your videos after watching previous videos is a bad way of informant. Could be a random person who never watched any of your videos, and watched this one then was misled that higher clock speed = better every time xD Every video should carry itself and not be dependent on a previous video

[@GaryExplains]

Well, I thought it is clear enough and no one else has raised the same questions.

[@westernshitt732]

That's great explanation professor

[@1MarkKeller]

*SHIVA AHUJA!*

[@westernshitt732]

@@1MarkKeller MARK!

[@Unknown-hu4gf]

So, what would be the best computer rig to use for Android Mobile App Development now a days? How many thread processor would you recommend for said development?

[@illegalquantity]

So which one is better for the KZbin?? I'm before buying a low consumption mini computer for everyday use. Internet browsing, watching videos on youtube in 1080p. I will use IGP which will be probably the Intel HD 530. I prefer Intel 6th generation because my need of Windows 7. ;) So i5-6500T or i3-6100?

[@aliuzel4211]

Could you analyze Microsemi Polarfire SOC in one of upcoming videos?

[@billyheng4824]

incorporate your testing into the video will be more interesting

[@M.N.922]

i will use the program windows server linux termux kalilinux which one do you prefer single or multi ?

[@onyxtraveler]

So what do you think of the intel 11th gens vs intel 10th gen?

[@hubcap4KHDRgameplay]

Why is my game not using the best 2 cores of my i9 11900K for the heavy lifting in gaming....( Not an all core usage game) ? ?

[@portman8909]

You can force the game to use the 2 cores you want if they aren't.

[@nathanhamman418]

1 core at 4ghz vs 2 at 2ghz vs 4 at 1ghz

[@AbdelrhmanKamal1]

Can you measure the power consumption for RaspberryBi tests?

[@mrjafs4616]

What about heavy online games? Like 4 to 6 hours of gaming? Which one should i look for? Single core or multi core? TIA

[@JamesSmith-cm7sg]

Multi core, single core gaming died a long time ago

[@portman8909]

Look for a CPU good at both. Some games are single threaded, some aren't. Why compromise. i7 13700 is a beast in both single and multi thread performance.

[@wanmaziahahmad151]

Great video gary... i think this video apple user can open their eyes concern more about single core performance rather than multi core performance Lol 🙄

[@marcuslex8654]

Anyone know why? A processor specs has 2 cores but when installed in a laptop declares 4 cpu's?

[@GaryExplains]

My guess is that it has two cores and four threads. I have a video on hyperthreading here: kzbin.info/www/bejne/o4S9oXd8Yt-Ipas

[@marcuslex8654]

Thank You.

[@dedelefoudu88]

You know you can use taskset to run on a given core

[@prakharposwal9]

Now we gave 64 core threadripper

[@TurboGoth]

Intel would certainly disagree. And they would make a good case that all workloads aren't parallelizable but power efficiency has never been their game. But if you want a super powerful hog with no concern for power and you want it to run software from JUST BEFORE multicore was available where NO EXISTING PROGRAMS were available that take advantage of multicore then your most easy path for selling new chips that offer improved speed on old software then cranking up those MHZ without concern for power is your goal. Then again, Prescott Pentiums D's extended out their pipeline some 31 stages in order to crank up their speeds to an absurd degree only to completely waste power and this proved a mistake for the server market. Less of one for the Desktop, their primary market, in that age just before the internet but once it was sort of rectified by AMD who shot for more reasonable designs, people began to expect better. And the race to 4GHZ was called off.

[@OneStepToday]

Suggest how is i3 10th Gen. 4 core 8 threads

[@nissyvivek87]

can somebody explain me why 12 gb ram andriod lags and 3gb ram iphone runs smoothly?

[@GaryExplains]

Why are you assuming the amount of RAM is the important factor?

[@ArunG273]

I take medium clocked multicore setup.

[@Isaacminer1]

Good video, Gary, i usually disagree with you a lot but i still enjoy your videos

[@abineemmanuelshorts2219]

Why are u not teaching in Stanford university U really know ur stuff sir

[@GaryExplains]

Stanford never offered me a job, plus I don't live in the USA.

[@abineemmanuelshorts2219]

@@GaryExplains oh it's alright You leave in the UK I presume🤔

[@mmawad100]

so isn't, there are some programs which can't make use of multiple cores or I am getting that wrong?

[@rrsharizam]

LibreOffice runs on single core. It runs slower on i7 than i5

[@logicalfundy]

Yup. Gary kinda explained poorly, and his benchmark basically prefers multiple cores because it's a task that can be easily split into many pieces. Not all tasks can be so easily split.

[@GaryExplains]

What tasks are you doing in an office suite that allows you to measure its performance across different processors?

[@tunkunrunk]

I remember a long time ago, the competition between Intel and AMD to break the 1ghz barrier, intel was always ahead, they released the Pentium III 1ghz. by the time AMD was struggling to have their 1ghz CPU, Intel released their socket 478 CPU and broke the 2ghz and 3ghz barrier with the Pentium 4. the turning point in this race was when Intel decided to introduce dual-core CPUs instead of trying to always get the highest speed

[@projectlij9777]

I have 2.1 ghz with 4 cores