How Good is RTX 3060 for ML AI Deep Learning Tasks and Comparison With GTX 1050 Ti and i7 10700F CPU

[FurkanGozukara]

How Good is RTX 3060 for ML AI Deep Learning Tasks and Comparison With GTX 1050 Ti and i7 10700F CPU

Full tutorial: https://www.youtube.com/watch?v=q8Q8CCDdSKo

If you are wondering which Graphic to purchase to run recent Artificial Intelligence (#AI), Machine Learning (#ML), and Deep Learning models on your GPU with CUDA, then this is the right video for you.

I have purchased the cheapest and yet largest VRAM having GPU #RTX3060.

In this video I am going to compare the performance of Gainward RTX 3060 Ghost 12 GB GPU with MSI GTX 1050 Ti OC 4 GB GPU model and with my CPU which is Core i7 10700F running at 4.59 GHz.

For performance tests, I will use OpenAI’s very newest AI model release Whisper.

So this is a video of GTX 1050 Ti vs Core i7 10700F vs RTX 3060 in terms of Machine Learning applications performance.

Whisper is used for transcribing speech into text in 99 languages.

You can check out my tutorial educational video regarding Whisper here: https://youtu.be/msj3wuYf3d8

Also, in this video, I am doing a box opening of Gainward RTX 3060 Ghost. Moreover, I do a physical comparison of RTX 3060 with GTX 1050 Ti.

Furthermore, I use an AC power meter plug (digital wattmeter - watt energy meter) to calculate GTX 1050 Ti, RTX 3060, and Core i7 10700F power consumption.

I am very satisfied with the performance of RTX 3060. Moreover, it is even able to run the large model of the Whisper which is the best-released model.

Please join Our Discord server for asking questions and discussions: https://discord.gg/rfttctFewW

Please follow us on Twitter: https://twitter.com/SECourses

Please follow us on Facebook: https://www.facebook.com/OfficialSECourses

If you are interested in programming our playlists will teach you how to program and code from scratch: https://www.youtube.com/c/SECourses/playlists

[1] Introduction to Programming Full Course with C# playlist

[2] Advanced Programming with C# Full Course Playlist

[3] Object Oriented Programming Full Course with C# playlist

[4] Asp.NET Core V5 - MVC Pattern - Bootstrap V5 - Responsive Web Programming with C# Full Course Playlist

[5] Artificial Intelligence (AI) and Machine Learning (ML) Full Course with C# Examples playlist

[6] Software Engineering Full Course playlist

[7] Security of Information Systems Full Course playlist

Thumbnail source : https://www.freepik.com/free-vector/isometric-computer-hardware-parts-set-with-monitor-system-unit-electronic-components-details-isolated_9647137.htm

Video Transcription

• 00:00:00 Hello everyone, welcome to the Software Engineering Courses as known as SECourses.

• 00:00:06 This is Dr. Furkan Gözükara.

• 00:00:08 If you have watched my previous How to do Free Speech to Text Transcription video, you already

• 00:00:13 know that I have purchased a 12GB VRAM having RTX 3060 Gainward Ghost Graphics Card.

• 00:00:24 This is the video that I mean, let me show you.

• 00:00:31 How to do Free Speech to Text Transcription.

• 00:00:34 The graphic card has arrived today.

• 00:00:37 So in this video, I will do Artificial Intelligence / Machine Learning / Deep Learning

• 00:00:43 processing power comparison tests between Core i7 10700F CPU, which is my CPU and here

• 00:00:55 is its CPU Z and that is running at 4.59 GHz versus MSI GTX 1050 Ti, this is my graphic

• 00:01:12 card MSI versus RTX 3060.

• 00:01:19 I haven't installed RTX 3060 yet, but I will do.

• 00:01:24 I will also test their power consumption by using a digital wattmeter plug.

• 00:01:30 Okay, here let me show it to you.

• 00:01:36 This is the digital wattmeter plug that I am going to use to calculate power consumption

• 00:01:43 of the CPU and my graphic cards.

• 00:01:50 This is how it works.

• 00:01:53 Okay.

• 00:01:56 So let's start with checking out the power consumption of my PC when it is idle.

• 00:02:01 We can say that it is currently idle.

• 00:02:05 Right now, I am going to look at the wattmeter and tell you how much power it is consuming.

• 00:02:12 Okay, I am seeing 92, 93 watts per hour.

• 00:02:20 Now I will record its video and show it to you.

• 00:02:24 Okay.

• 00:02:25 Okay, right now I see 93, 92, 94.

• 00:02:32 Okay, so I am going to note this 94 watts per hour into a file so that we can compare

• 00:02:39 all the values at the end.

• 00:02:41 Here I have made a text file already, and I have noted it as 94 watts idle.

• 00:02:50 I have counted it as idle, but OBS Studio was recording the screen, so it is worth mentioning.

• 00:02:57 So okay, for testing purposes, I will use OpenAI's Whisper model to transcribe first

• 00:03:03 10 minutes of my latest video.

• 00:03:07 My latest video is this one, and I already have prepared its first 10 minutes audio as

• 00:03:17 can be seen here.

• 00:03:20 Since my MSI GTX 1050 Ti graphic card has only 4 GB VRAM, I am only able to use small

• 00:03:29 model of Whisper when I am using CUDA with my graphic card to do the speech transcription.

• 00:03:44 So all tests will be made on small model for equal comparison because my GTX 1050 Ti is

• 00:03:53 only able to process small model.

• 00:03:58 Let me show you how much VRAM each model requires.

• 00:04:04 When we open the model card, we see that small model. Not this one actually. Small model requires about 2 GB VRAM and medium

• 00:04:26 model requires 5 GB VRAM and large model requires 10 GB VRAM.

• 00:04:30 Therefore I will do the experiments, the testing on 2 GB requiring model which is small model.

• 00:04:39 So if you wonder also how to use Whisper model, then you should watch my previous video.

• 00:04:47 You can find it on our channel with clicking videos and it is here.

• 00:04:54 So let's start testing with CPU.

• 00:04:59 I already have the commands in here which I explained in the previous video.

• 00:05:06 So I will just change the input parameter as this one, language device will be CPU,

• 00:05:19 model will be small and task will be transcribe.

• 00:05:39 Okay so it has started working or it is currently initializing because the CPU didn't increase

• 00:05:49 it or it is downloading the model.

• 00:05:51 It seems like I didn't download this model previously.

• 00:05:56 Okay so the download has completed and the processing has started.

• 00:06:00 You see the CPU usage is increasing.

• 00:06:08 Okay and let me check the power meter.

• 00:06:16 Okay right now I'm checking it, okay it is using about 200 watts, 210 watts, 215 watts.

• 00:06:29 So I will show it to you with my camera.

• 00:06:34 Okay you see that the wattmeter is displaying about 188, 190, about 200 watts.

• 00:06:43 The CPU usage is about 85 percent because I have stopped the OBS studio recording.

• 00:06:52 So I will say that it is using about 190 watts.

• 00:06:58 Okay as you have also as you have seen in the video it is using about 190 watts and

• 00:07:05 since I have started the video recording again it has jumped in CPU usage as well a little

• 00:07:11 bit and the transcribing is already started and it is displaying the output.

• 00:07:24 All right let's see the GPU-Z is displaying nothing as supposed to be but my graphic card

• 00:07:38 is also in being used probably due to OBS studio recording.

• 00:07:46 Okay so since recording also using CPU I will now stop video restart the transcribing and

• 00:07:59 I will record its time so that we will have perfect time calculation for CPU.

• 00:08:11 Okay so the processing has just finished it took 22 minutes 10 seconds and let's take

• 00:08:20 note of this value 10 minutes 22 minutes 10 seconds and 195 watts per hour actually 190

• 00:08:42 watts yeah.

• 00:08:46 All right so now it is time to test our graphic card I will start test and then restart test

• 00:08:54 for to calculate its maximum effectiveness because when we are recording with OBS studio

• 00:09:04 you see graphic card is being used a lot.

• 00:09:09 All right so let's modify our command here with device as CUDA okay and let's start.

• 00:09:26 Now we will see that our graphic card usage will get maximized.

• 00:09:35 I will restart this test with closing OBS studio as same as previously.

• 00:09:42 Okay now the graphic card is being used almost maximum.

• 00:09:48 Once the transcription starts here okay it started let's check out the power usage

• 00:09:58 I will show with my camera again.

• 00:10:06 Okay so now the power usage is 150 watts I have paused the OBS studio recording right now

• 00:10:13 so it is pretty decent it is lower than CPU usage actually I am amazed with this 150

• 00:10:24 watts.

• 00:10:26 Okay so as you have seen in the in my camera video it is using 150 watts so let's note

• 00:10:35 it and I will then restart the test with closing OBS studio okay 150 watts it takes 1050 ti

• 00:10:52 150 watts and now I will close OBS studio and restart the test.

• 00:11:00 Okay the processing has just finished it took six minutes and one seconds as you can see

• 00:11:05 the GPU usage is now decreasing the memory usage has already decreased it used 3d

• 00:11:13 the power consumption during the process was really low I am surprised with that so the

• 00:11:20 CUDA cores are not using are not using a lot of power probably.

• 00:11:28 Okay so let's note it as six minutes and one seconds so you see the GPU is three times

• 00:11:36 faster on which model the small model on the bigger models probably it is making more difference

• 00:11:49 so now I will open the box of RTX 3060 I will also show you box opening and physical comparison

• 00:12:03 of GTX 1050 and RTX 3060.

• 00:12:11 Okay now I will stop the video and my camera video will enter now.

• 00:12:17 Okay now I am going to do box opening of RTX 3060 and compare it with my MSI 1050 ti as

• 00:12:26 you can see here okay all right.

• 00:12:36 Okay I took it out of the box and simply there is nothing else just the graphic card itself

• 00:12:45 it is pretty empty basic if we compare the both cards we can clearly see that RTX 3060

• 00:12:56 is bigger in size and a little bit heavier as well the design is simple as this one as

• 00:13:12 you can see they both look pretty awesome if you ask me now time to install it in my

• 00:13:20 computer and do our test once again okay.

• 00:13:27 Okay so now I have installed RTX 3060 and I have installed the newest driver restarted

• 00:13:36 the computer everything is ready but before we run our test let's check out whether the

• 00:13:45 idle watt usage has increased or not are there any differences okay when I check I see that

• 00:13:53 100 95 100 watts okay so what was it before it was before about 95 watts and now I see

• 00:14:06 similar to that so not much idle increase then we can start our test with the same command

• 00:14:14 same parameters okay now let's wait until the GPU usage increases and transcription

• 00:14:26 starts okay the GPU has dedicated 12 GB memory wow it's going fast okay let me check the

• 00:14:36 power usage okay now I will turn into my camera video okay I have stopped the OBS

• 00:14:51 studio and when we check the power usage it is about 250 watts and it changes actually

• 00:14:59 it fluctuates 270 watts 250 watts 266 250 so I will say that it is about 260 watts it

• 00:15:12 is a lot of increase but of course the time speed will also increase okay so it is about

• 00:15:19 250 watts all right okay so the test has been completed it took only one minute and one

• 00:15:26 second the speed increase is magnificent and let's record that with RTX 3060 it it uses

• 00:15:41 250 watts actually between 250 and 270 watts it fluctuates but takes only one minute one

• 00:15:56 second so the speed difference is magnificent the RTX 3060 is 22 times faster than my CPU

• 00:16:11 and six times faster than GTX 1050 TI okay so let's make it like this okay I think everything

• 00:16:29 is very clear the CPU takes 22 minutes GT GTX 1050 TI takes six minutes RTX 3060 takes

• 00:16:41 one minute only the watt usage of course a lot more with RTX 3060 but still it is similar

• 00:16:52 to the CPU watt usage however the speed is magnificent okay this is all about the comparison

• 00:17:02 I hope you have enjoyed. Please subscribe to our channel like our videos if you like

• 00:17:09 it and please make comments I read your comments actually my next video will be one of our

• 00:17:18 followers request. Our followers Tokyo family has requested that how can we how can we automate

• 00:17:32 whisper transcribing so I will code a software from scratch and show you that how can you

• 00:17:43 use any third-party library / exe / application through command window command prompt so we

• 00:17:54 will automate command prompt to use any software exe practically that can be used with commands

• 00:18:06 so stay tuned also one final thing is that I have opened official Twitter and Facebook

• 00:18:19 page for our channel so if you are on those platforms please also follow us there you

• 00:18:27 can access our Twitter and Facebook through clicking about top in our channel or from

• 00:18:36 here then click Twitter link and Facebook link they will be open it you see I have just

• 00:18:44 opened them so we have currently zero followers and this is our Facebook page you can also

• 00:18:51 follow me on LinkedIn if you want and you can also check out the game that I developed

• 00:18:59 Monster MMORPG and you should also join our discord channel and contact me through there

• 00:19:07 ask your questions through there thank you very much hopefully see you later.