CUDA Simply Explained - GPU vs CPU Parallel Computing for Beginners

[@PythonSimplified]

How did your speed test look like?? 🤩🤩🤩 ************ CPU SPEED *************** 7.952215671539307 verify device: cpu ************ GPU SPEED *************** 0.6196999549865723 verify device: cuda:0 ************ GPU SPEED *************** 0.24684429168701172 verify device: cuda:0 ************ GPU SPEED *************** 0.24314093589782715 verify device: cuda:0 my CPU: Intel's i9-12900K 12th-Gen Alder Lake Processor (+5200mhz DDR5 memory) my GPU: Nvidia GeForce RTX 3090 matrix_size: 32*512 OS: Linux Ubuntu 20.04 (I've upgraded from 18.04 mid-video hahahaha)

[@Yachid]

Merry xMas!

[@FinickyHorseGameStudios]

#PythonSimplified Thank you for this video! Kindly share more videos about this CUDA tutorials with real time examples..

[@reconchrist]

***** CPU SPEED ***** 19.829071521759033 Verify device: cpu ***** GPU SPEED ***** 2.2739994525909424 Verify device: cuda:0 ***** GPU SPEED ***** 2.0090506076812744 Verify device: cuda:0 ***** GPU SPEED ***** 1.9959523677825928 Verify device: cuda:0 my CPU: Intel i7-9750H @ 2.60Ghz my GPU: Nvidia GeForce RTX 2060 matrix_size: 32*512 OS: Windows 10

[@Gyanateet2077]

*****CPU SPEED**** 9.473852396011353 verify device: cpu *****GPU SPEED**** 1.2871942520141602 verify device: cuda:0 *****GPU SPEED**** 0.8036797046661377 verify device: cuda:0 *****GPU SPEED**** 0.7862038612365723 verify device: cuda:0 SPECS: CPU: Intel's i9 7900xe 10 cores 20 threads (Oc'd to 4.5ghz)(3200mhz DDR4 memory) GPU: NVIDIA GEFORCE GTX 1080ti matrix_size: 32*512 OS: Linux Ubuntu 20.04

[@djoezi3662]

My god you are so beautiful and have good sense of humor 😘

[@muhammedaneesk.a4848]

Please don't abandoned this series. I'm really looking forward to it.

[@PythonSimplified]

Thank you Muhammed, I'm happy to hear! 😁 Don't worry, I won't rest until all of us become CUDA ninjas!! 🥷🥷🥷

[@RiverReeves23]

Being a developer of 15+ years. You're doing a great job at explaining an incredibly complex topic in a very easy way. Excellent work. Really love your channel.

[@ledkicker2392]

Didn't expect such a well presented tutorial. Your explanation is so structured and without distractions, and your manner of presenting is very eloquent 👌

[@PythonSimplified]

Thank you so much for the incredible feedback! 😀

[@benrav924]

@@PythonSimplified you're the best!

[@Spanu96]

Very, very, very well documented this video. Keep up the good work.

[@PythonSimplified]

Thank you so much!! Will do! 😃😃😃

[@MrRobot222]

I've just finished my Google Cloud Big Data and ML course. So this was great to understand the Cuda element. Thank you! Will post my benchmarks and list them below, but no 3090 here for me! 😂

[@PythonSimplified]

Congrats on finishing the course!! 🥳🥳🥳 I hope you had lots of fun!! The one AI course I took in 2018 (where I discovered Python actually 🤪) didn't explain CUDA at all! We've used it - but I had no idea what it actually was! I only recently found out that you can get it directly from your GPU rather than having to use some kind of online, often paid, third party service!! 😅 It was definitley a pleasant surprise! But I wish somebody bothered to explain it to me when I started my ML journey... it would save me so much time and frustration with Google Colab disconnecting my runtime with no reason!! Anyhow, I'm glad this video came at the right timing for you! 😉

[@szilike_10]

I just found this channel. I think it's amazing and it is everything someone that wants to learn the basics ever needs. I am a true believer that the most important thing is to get a grasp of the intuition and then slowly try to dive deeper into any topic. And of course you have A LOT of question when trying to learn something new and I love the way you approached it from the newbie's point of view, and focusing on what needs to be cleared first. Unfortunately, sometimes it's really hard to find any tutorials like that. Especially that at uni, it doesn't work like that at all :))) So I'm glad I've found you, and hope you keep posting. Wish you a great day!

[@mibrahim4245]

I wrote you somewhere else that this tutorial is great .. but I wasn't at my best concentration .. now that I again watched it... I have no words to say more than FANTASTIC .. clarity, knowledge, and everything else .. so thank you beast ,, thank you M .... I'm following what you present .. as always..

[@PythonSimplified]

Yeeeeeey!! I'm so happy to hear that!! 😁😁😁 Thank you so much for the incredible comment dear!! I worked extra hard on this video so your feedback really made my day!! 😃 Happy new year and see you in 2022!!! 🥳🎆❄

[@mibrahim4245]

@@PythonSimplified happy new year M ❤❤ .. You should've got used to my comments cuz I don't pass a video without a comment, or anywhere else 🤩🌹

[@italoaguiar]

Simply one of the best tutorials I've ever seen on this specific topic!! Congrats!!

[@PythonSimplified]

Thank you so much!!! super happy you liked it! 😀

[@grjesus9979]

Very nice introduction. Using additional software like GPU-Z while processing data (e.g: training neural network) you can check your GPU load and temperature. When processing big chunks of data for long time ( days) check that you GPU doesnt pass the maximum temperature joint parameter (in my case it was 100°C). Another thing is that for PCs it is really important to have good coolers and a notable spatial separation between CPU and GPU (GPU support better high temperatures than CPUs).

[@PythonSimplified]

Thank you so much for the incredible tips grjesus99! 😀 I haven't had a chance to use GPU-Z before, I usually use the GeForce performance overlay with Alt+R and it shows me all the important stats 😊 I absolutely agree about the cooling! it's a key component, especially in overclocked systems. a good air circulation will ensure your hardware lasts for much longer!

[@silvermica]

I use CUDA with HFSS - a numerically intensive electromagnetic solver (solves/satisfies Maxwell’s equations in 3D space).

[@satviksharma3722]

How???

[@silvermica]

@@satviksharma3722 The software works with CUDA. I think their competitor that makes CST also uses video card memory for ultra fast calculations of insanely large matrices (inverting the matrix) .

[@satviksharma3722]

@@silvermica I am a user of hfss too do i indeed to enable cuda from somewhere or is it supposed to detect that automatically?

[@silvermica]

@@satviksharma3722 - You need to contact your licensing manager and they will set you up - after you pay a lot of $$$. I've seen problems solved in 10 minutes that take over 24 hours without using cuda.

[@satviksharma3722]

@@silvermica thanks a lot for that. On my way to mortgage my house rn xDD

[@NassimEssaidi]

I'm in love now with your Computer Specs, I9-12900K with RTX 3090, damn, that's absolutely a beast PC.

[@jaywalker.]

This was a wonderful explanation. I've only ever had a vague idea of what CUDA was and what a CPU actually does.

[@demandelz]

Cool!! Thank you for the simple, clean, well presented video. You are inspiring me to try writing code using my GPU's. I'm sure I will be watching some of your other videos.

[@slcooIj]

KZbin brought me here. Although I am not a big fan of python, I understand the intension of your examples. Thx for that. I also like your gestures and style of explanation. Keep it up

[@mdibrahimshariff3386]

This is one of the best explanations I have watched

[@Luredreier]

3:13 Well... You're right that with x86 we're looking at 1 or 2 threads pr core. But the CPU can do multiple tasks pr thread at once. So it might do task 1 from thread A and task 1 from thread B, but then there might be some spare execution resources allowing task 3 from thread A and task 7 from thread B and perhaps if you're *really* lucky perhaps even task 13 from thread A to be executed all at the same time. The OS only assigns two threads to the CPU core to run at once till the next interrupt, so it's not doing more at once in that regard. But multiple tasks within each of those threads will be executed at once if possible (not always the case). Of course in normal spagetti code you can't really expect the CPU to be able to do particularly many of those instructions at once. And more advanced instructions sets helps the CPU understand exactly what you're doing and achieve more instruction level parallelisation since those advanced instructions in essence combine multiple instructions into one so even thoough they're multiple instructions *inside* the CPU they're just a few on the outside, meaning that the CPU knows how those internal instructions relate to eachother more and can do things better as a result.

[@kecoma]

That is not a computer, is a god-level beast. 3090 + i9 crazy!! Great video

[@PythonSimplified]

Hahahaha thank you so much Kevin! 😁 I've hustled a lot to get all these parts (well... besides arguably the most important part which Nvidia sent me for an upcoming TensorRT tutorial 😅 hahahaha) But the DDR5 for the processor was super hard to find!! I almost gave up and got a Rayzen 11th gen + DDR4 built instead... but thank God a bunch of Christmas miracles happened and everything worked out!! 💪💪💪

[@Vaibik]

This was quite nicely delivered. CUDA for beginners. Even a 1500$ MIT course doesn't explain it so nicely, because they expect us to know things already, which is not always true. Спасибо!!!

[@muthaheerayasmeen3845]

The most simplified explanation for beginner learners. Thank you

[@nonomnismoriar9601]

Another great tutorial, really like your clear presentation style and depth of information. Made it very easy to setup my system, thank you!

[@PythonSimplified]

Thank you so much for the lovely comment! I'm super happy to help! 😁

[@Dygear]

This had LTT / LMG levels of production value with one of the best / clearest explanations for what CUDA is and why it matters.

[@kushalbasnet8751]

Dyamn .. i dont know how i came accross this channel/video .. but have to say excellent work. I could watch these types of videos all day long ..

[@EmaMazzi76]

Super clear explanation Maria, great video, thank you!

[@PythonSimplified]

Thank uou so much Emanuele, glad you liked it! 😁

[@gareththwaite5128]

Haven't long come across your channel, I think it's now my favourite python channel

[@Noorpak634]

First i would like to practice on your kivy simple application. It's my starting point. Kivy is easy to me. But i have a problem, the kivy application not opening, I'm not testing my code.

[@udbhav3760]

GPU and CPU both are good in there way! Merry Christmas ☃️! Video is really good you did well! :+)

[@PythonSimplified]

Thank you so much Udbhav! 😃😃😃

[@dbassett74]

Great video! Please continue with this series.

[@dironin2363]

Merry X-mas, Mari! Nice subscribers count))

[@PythonSimplified]

Merry Christmas di ronin!! 😁😁😁 KZbin has been very kind to me lately 😊

[@DavidMedinets]

CPU Speed - 27.37293028831482 GPU Speed 1 - 2.607041835784912 GPU Speed 2 - 2.7270538806915283 GPU Speed 3 - 2.6175949573516846 CUDA was run once so the initialization time is not invoked. The matrix was 32*512. CPU=i7@2.60GHz GPU=GeForce GTX 1660. Running Windows

[@cannatoshi]

as always absolutely top-notch explained, your way to convey a thing is unbeatable good!

[@davidgm2821]

Awesome explanation!! 👏🏼👏🏼👏🏼

[@PythonSimplified]

Thank you so much David! 😊

[@marufmazumder8294]

conda install -c pytorch pytorch ====> did not include CUDATOOLKIT, why?

[@Isagi__000]

Thats really great explination. Thanks

[@PythonSimplified]

Thank you! Glad you liked it! :)

[@Luredreier]

2:10 That's a sort of dated statement. These days a CPU has multiple pipelines and can do multiple tasks at the same time in the same clock cycle. Exactly how many varies depending on the arcitecture. But in highly optimized code you can actually get more then 4 different instructions done at once in some x86 processors. However to do this you kind of have to tailor your code for the CPU in question, as each arcitecture can only do certain instructions in parallell with certain *other* instructions. Combinging the wrong instructions will lead to other instructions either being slowed down or outright stopped from running till that instruction is finished. If your code has dependencies then of course that also means that a lot of code has to wait till those dependencies are handled before the processor can start working on anything past that dependency and so one and so forth (although there's of course some speculative execution at play too).

[@bushratarif6028]

I've really loved the way you have explained it!!!!Justttt Amazing.

[@mshparber]

Finally I got the answer to my question - what’s the difference between CPU and GPU. Thanks!

[@sehbanomer8151]

Great video! One thing I want to point out is that a CUDA core isn't really comparable to a CPU core. CUDA cores are just floating point units, not processors. A streaming multiprocessor on the other hand, can be thought of as a CPU core that is very good at parallel floating point math. Another tiny detail is that most modern CPU cores have some limited parallel processing capability (SIMD).

[@PythonSimplified]

Thank you! 😃 This video is a beginners level introduction so I'm using lots of abstraction to convey the concepts rather than focusing on small detail. It makes sense to talk about it in intermediate/advanced level tutorials, but if I use these words to explain CUDA to folks without much background in programming - they will get frustrated and leave. The number of cores was compaired in terms of running processes in parallel - each core runs a process, which I believe is a correct thing to say (I actually sent this video to Nvidia before posting it to make sure my explination is accurate, but I wouldn't be so sure if it was just me proofing it hahahaha 😅) Anyhow, thank you for your comment! I'll definitley talk about it in due time 😉 ...it's just too soon, it's only the first episode of this parallel computing series

[@user-wr4yl7tx3w]

this was really well organized and explained. best so far.

[@venom_snake1984]

Merry Christmas Mariya! Have a good one! :)

[@PythonSimplified]

Merry Christmas Ali!! You too! 😀😀😀

[@高板丽奈单推人]

So good, I just start to get contact with the GPU accelerated computing topics, I think this field is so promising and active.

[@gersonovphysics]

This video is awesome! Very instructive. Congrats!

[@Banefane]

I like your positive attitude :D! Very good explained and easy to understand!

[@reihanehmirjalili4676]

You are amazing! Please keep posting these wonderful videos !

[@Matthewsavant]

I’m literally just here to understand why applications are running on my GPU and if that’s the most efficient way to be doing things. I haven’t done anything with code since the Commodore 64 I had when I was a kid lol. I am just coming back to PC after 17 years on Mac and damn a lot has changed so I’ve been just kind giving myself a crash course. Very fascinating and exciting stuff. Really my main use and why I want to keep my computer as stable, efficient and fast as possible is for Sound design and Audio engineering and I’ve been noticing that several programs are actually running on my GPU (RTX 3080), and not just OS stuff but even my digital audio work station and that just didn’t make sense to me, especially since I picked the intel 13900 specifically because of its number of cores and overall speed, after watching the nvidia keynote speech and this video I’m understanding why. Also has me interested in diving into Python just for fun

[@ss-o]

You need to put some soft material under the keyboard or re-allocate the mic. The noise when you type is similar to good drum and bass music.

[@morousg]

Very well done video. Just, the title it’s a bit misleading. CUDA, besides hardware and runtime, it’s also a programming language based on C and C++. Saying CUDA and then talking about pytorch… it’s like saying “how car engines work” and talk about how to drive a car…

[@randahan215]

wow!! Your explanations and examples are perfect!!. Thank you so much

[@hopaso6554]

Very nice, but for me conda install -c conda-forge pytorch-gpu is only working, installing pytorch with cudatoolkit included.

[@Luredreier]

2:28 That's true with a modification. *Technically* if you have the source code you can hippify the cuda code and run it on a AMD GPU. The AMD GPUOpen standards relevant to CUDA are a bit... Let's just say that they're not really production quality quite yet as far as I can tell... Please *do* keep an eye on them though. AMD is trying to regain market share here, and they're working on their software stack.

[@1MinuteFlipDoc]

nice tutorial! excellent video and clear window-in-window effects to show what you are doing!! however, I had to turn my audio way up to hear you. boom mic? lapel mic? Solid video! :)

[@PythonSimplified]

Thank you!! More like 10 foot ceillings 😅 the mic is great but the echo is an ongoing problem I'm trying to fix in all kinds of creative ways (when I'm filming - there's pillows and paper towel rolls everywhere!!! Hahahaha 🤣🤣🤣) Thanks for letting me know! I'll boost the volume up in future tutorials, I think my echo-cancelling effects are causing this low volume situation

[@vladmaiuga2934]

Brains and beauty!!! Informative video, which I enjoyed due to your programming style! Out of curiosity how much time and effort did you put behind this video? I have my wild guesses, but I think it’ll show folks your capacity if you state it yourself. Anyway happy videoing, and hope to see many more.

[@sadeghkalami1822]

All I can say, U have a hell of system 3090 RTX + i9-12900. daaaaammmmmn :DDD

[@kavanspace1654]

Amazing tutorial from a lovely person. Its great to see simple and clear explanation. Thank you...I found a new resource to my addition. Question : Why can't we just replace CPU's with GPUs only.... ?

you are using linux, perfect. this is my favourite channel now

[@PythonSimplified]

Hahahaha using a combination of Linux and Windows with a dual boot system 😉

@@PythonSimplified thank you, my 8yo daughter is very happy with your channel, the graphics help, and representation, it matters.

[@PythonSimplified]

Thank you so much Francisco! I'm so happy she likes it! 😃 I'm planning to start a "Python for kids" series soon, there will be plenty of colourful graphics and fun projects for little pythoneers 😊

[@caizza3]

Awesome explainer video thank you!

[@l0ksh]

Great video, you explained very well.

[@PythonSimplified]

Thank you so much, glad you liked it!! 😃

[@fahadalkamli]

Thanks for the great video I just have a question , you seemed to miss threads for CPU cores ? Perhaps for simplicity's sake. As per my understanding CPU cores will usually be split into threads so for 4 cores usually have 8 threads.

[@honahwikeepa2115]

Thank you. I'm much smarter and excited now. Love tinkering with computers in my late years. I have a couple of Quadro gpu's.

[@k.ballajiaxe6403]

❤️ lots of information and useful ❤️

[@PythonSimplified]

Thank you K.Ballaji Axe! 😀 Have fun with CUDA!

[@hank7v]

Great presentation. Great teaching style. Subscribed to look at more python material

[@riittap9121]

This is a great video! I've got a 3070 and have been wondering what to do with it. It seems obvious, that I need to learn parallel programming.

[@moayyadarz2965]

your way of explanation is really amazing

[@soultribe9]

Merry Xmas M!! God bless you !!!

[@alexnickolaevich9536]

Лучший ролик по теме на ютубе!

[@kalyanheng5657]

your explanation is very clear!

[@August-hc9ke]

Well explained, I love this video!

[@MolecularArts]

consider using time.perf_counter() instead of time.time(). "The perf_counter() function always returns the float value of time in seconds. Return the value (in fractional seconds) of a performance counter, i.e. a clock with the highest available resolution to measure a short duration. It does include time elapsed during sleep and is system-wide. The reference point of the returned value is undefined, so that only the difference between the results of consecutive calls is valid."

[@smalirizvi8026]

Mariya! Show up soon 😍😍😊😊 Thou time flies, but in your case it feels like the last video u uploaded was *9 years* back. 😀 hope to see you really soooon

[@PythonSimplified]

Hi Ali! 😀 I was actually about to post a new video today but couldn't get myself to wake up! 😅😅 (finished editing it at 3am and needed to wake up at 6am to review it and post it... but I was a bit too optimistic!! hahaha) See you in a brand new video tomorrow though! 😉

[@smalirizvi8026]

@@PythonSimplified yayyy!! 😀

[@CULTURE_dz]

thanx...realy you are the best ... any tensorflow-gpu 2.7 cours

[@rudrakshya1]

It's simplified. Thank you.

[@badprogrammer1280]

Thanx, Maria! Very informative information ))) Could i vote on a theme? Using Python at the robotics (such as ROS) seems useful to dig into...

[@PythonSimplified]

You're absolutley welcome! 😃 I think I'll wait a bit with robotics as I'm quite overwhelmed with the amount current topics... and also I know nothing of robots 🤣🤣🤣 (that might be a better excuse! Hahahaha) I'm sure we'll get there eventually though! my dad is a mechanical engineer so we might do some kind of robotics project together along the way 😉

[@badprogrammer1280]

@@PythonSimplified Thank you for response!!! i am also a parvenu in robotics))) But i love opencv and PCL looks promising(i'd been working on pcl, but without the MS library) ROS guys use python a lot! So, take a rest and look at the theme please!!!

[@gfeie2]

No CUDA code was shown in this video.

[@starsun9822]

Excellent video! thank you Mariya 🌷 ❤️

[@PythonSimplified]

Thank you so much! Glad you liked it! 😃

[@jsuperp1371]

You are my style 😍😘

[@markippo]

In the great simplification, this is the way rasterization process work. The final outcome is a tensor resolution width x height and each element (representing pixel) is an array of 3 integers (for RGB). We're pc master race and we want that operation at least 60 times per second. And... Before it even begins all rendering, postprocessing need to be done. Huge amount of floating point linear algebra math. Your example shows how deadly slow Python is :)

[@PythonSimplified]

Thank you so much Sarat! 😊

[@caruccio]

Thanks, really easy to follow explanation.

[@amortalbeing]

Though this was going to talk bout CUDA programming in Python!

[@kopa1999]

Hey Mariya ! I just watched your video teaching flask and I love it. would you do a video about deploying flask apps with databases?

[@PythonSimplified]

Absolutley! 😃 I'm filming a new tutorial tomorrow where we connect an Sqlite database to a Flask grocery list application! (Will upload the complete code to Github shortly, I'm almost done! 😉) At first we will deploy it with Wayscript X (as we don't need Heroku anymore!Wayscript figured out how to do this directly from the IDE 😱) But I'll defonitley film an additional deployment tutorial for those who want to use Heroku instead! 😊

[@kopa1999]

@@PythonSimplified Thank so much! You're the best

[@Luredreier]

4:21 Regarding the number of cores in GPUs... Those numbers are a bit inflated due to the difference in what is actually called a "core" in CPUs and GPUs. Cuda cores for instance is in reality the same as pipelines in CPUs. Meaning that *really* high end CPUs actually can beat many GPUs in terms of the equivalent of cuda cores. But GPUs of course *still* ends up having more overall at any given price as well as when comparing the best vs the best. I've been moving so I haven't been able to keep up with the latest in hardware news as I used to. But I seem to recall that a high end threadripper was roughly on par with a 1060 in terms of number of pipelines. Basically the GPU will have more pipelines pr what in the CPU world is called a core. While the CPU will have things like prefetchers, speculative execution etc that you won't really find in GPUs. Meaning that if you have a lot of simple tasks you can do more of them in a GPU pr clock cycle then in a CPU. But if you have a lot of tasks that depend on eachother finishing in sequence or a complicated code with a lot of branching etc or if you want the hardware to work out what parts of the code that can be run in parallel for you then your code will run faster on a CPU then a GPU. As the CPU can guess what data the core will need next and get it, can work out what branch is the most likely and start working on what's going to happen there in the future already before the CPU knows for sure what branche will be used, and just in general work on multiple parts of the same thread at once speeding up the process of progressing through the threads code fast.

[@TheJackTheLion]

You literally just said what she said.

[@brians7100]

Great tutorial! 💪🏻

[@siddharthanrajasekaran8977]

It was like watching an acting audition. You'd definitely get the part

[@sarkisissa2601]

A super great video, thx for your effort ❤ Why do you freeze the CPU in the for-loop? So that the GPU-speed result can be displayed and not by the CPU? Thx 👋

[@navidntg6082]

Wow this girl is on fire! Nice job🤩

[@twjoshua]

It is my first to find how the GPU works. In 1996, I had to run the ARIMA / VAR with Rats on SINICA servers….From years ago, it could be done with Cuda with GPU. Thanks for your introduction! The basic knowledge linear algebra and forecastmodel is comment now?

[@lukkenny5269]

looking forward to the next video since I have 2 GPUS and dont know to make it parallel computing in my deep learning programme

[@jipe4153]

Nice presentation! A fair comparison is more like~ 16 cpu cores ~ 256 FP32 cores (AVX 512). So 256 vs 10,000, and also the bandwidth difference is huge aswell

[@JuanPabloMolinaMatute]

Just excellent!

[@PythonSimplified]

Thank you so much Juan! 😀 By the way, I'm premiering the next episode in this parallel computing series in 45 minutes - come say hi! 😁 here's a link: kzbin.info/www/bejne/n3ekdaaIea-beq8

[@wrcz]

Maybe worth mentioning that Anaconda starts in System32 by default if you're on Windows. So when you start Jupyter from it, you start making Jupyter Notebooks in your System32 folder. Not ideal. Learn from my mistakes :)

[@MosesMakuei-b5z]

I love your explanation. You are really good

[@samibelattar1847]

Thank you your explanation is very clear.

[@1UniverseGames]

Would you like to make a video on building or creating a Single node level task scheduling for deep learning based RLScheduler in spark cluster?

[@roeslib10]

Very good explanation, but when installing pytorch there was no cuda toolkit package to be installed. This is the list of packages to be installed: blas, filelock,gmp,gmpy2,intel-openmp,jinja2,markupsafe,mkl,mpc,mpfr,mpmath,networkks,pytorch,pytorch-mutex,sympy,tbb,typing_extensions.

[@sanyajain862]

Amazing Explanation

[@PythonSimplified]

Thank you so much Senya! 😀

[@FritsvanDoorn]

Very intetesting. But your sound is a bit low and difficult to hear. I do not know why

[@PythonSimplified]

Thanks for letting me know Frits! 😃 I have very high ceilings in my apartment so whenever I use audio effects to remove the echo - it probably affects the volume as well... I'll boost the volume in future videos 😉

[@ryanmckenna2047]

You explain it very well

[@PythonSimplified]

Thank you so much! 😃

[@goodsunny5041]

Thanks for the great video!💐💐💐

[@MatthewSuffidy]

I'm not an expert on it, but I think a lot of GPUs support openCL which is a general GPGPU framework that does a lot of the same things. I can't think of a lot of examples besides graphic overlay and capture for example, but it is possible to directly busmaster without a cpu involved between 2 devices.

[@Hof79905]

This is for beginners? W O W !!!!!! What preceeds beginners?

[@60pluscrazy]

Very well explained 👌