M3 MacBook Pro and the trouble under load

Python speed and usability tests on the M3, M3 Pro, and M3 Max MacBook Pro. You might want to upgrade, but only in certain conditions.
Temperature/fan on your Mac: www.tunabellys... (affiliate link)
Run Windows on a Mac: prf.hn/click/c... (affiliate)
Use COUPON: ZISKIND10
🛒 Gear Links 🛒
* 🍏💥 New MacBook Air M1 Deal: amzn.to/3S59ID8
* 💻🔄 Renewed MacBook Air M1 Deal: amzn.to/45K1Gmk
* 🎧⚡ Great 40Gbps T4 enclosure: amzn.to/3JNwBGW
* 🛠️🚀 My nvme ssd: amzn.to/3YLEySo
* 📦🎮 My gear: www.amazon.com...
🎥 Related Videos 🎥
* 🌗 RAM torture test on Mac - • TRUTH about RAM vs SSD...
* 🛠️ Set up Conda on Mac - • python environment set...
* 👨‍💻 15" MacBook Air | developer's dream - • 15" MacBook Air | deve...
* 🤖 INSANE Machine Learning on Neural Engine - • INSANE Machine Learnin...
* 💻 M2 MacBook Air and temps - • Why SILVER is FASTER
* 💰 This is what spending more on a MacBook Pro gets you - • Spend MORE on a MacBoo...
* 🛠️ Developer productivity Playlist - • Developer Productivity
👨‍💻 Matrix multi code - github.com/ale...
👨‍💻 Primes - github.com/Plu...
👨‍💻 Mandelbrot - benchmarksgame...
💻 MacBooks in this video
M3 Max (16/40) 16" MacBook Pro 64GB/2TB
M3 Pro (11/14) 14" MacBook Pro 18GB/512GB
M3 (8/10) 14" MacBook Pro 8GB/512GB
- - - - - - - - -
❤️ SUBSCRIBE TO MY KZbin CHANNEL 📺
Click here to subscribe: www.youtube.co...
- - - - - - - - -
📱LET'S CONNECT ON SOCIAL MEDIA
ALEX ON TWITTER: / digitalix
- - - - - - - - -
#m3 #macbook #macbookpro

[@AZisk]

JOIN: youtube.com/@azisk/join

[@camsand6109]

All your mac benchmarking videos have been great. Glad you're filling this lane of testing from the pov of an engineer as not too many people are providing this valuable information (along with good editing and actual in depth testing).

[@peterbreis5407]

I love your tests, it's good to see real world comparisons all the rest don't mean much.

[@radityadanu]

Thank you Alex for your in depth reviews of these MacBooks for engineers and devs!! These videos help me a lot to decide based on my actual needs, instead of thousands of finalcutpro tests videos

[@user-gj4ec5mq3v]

That matrix multiplication example is memory bound on the base M3. I tested the same code on my M3 and it uses up all available memory (8GB) and takes 102 seconds to finish.

[@AZisk]

nice! thanks!

[@Vilverin]

Nice comparison and I appreciate the inclusion of fan speeds. I think the faster battery drain on the base M3 can be explained quite easily: it takes longer to complete a given task, so the CPU spent less time in idle and more time under load. What's maybe more surprising is that the 14" M3 Pro kept up with the 16" M3 Max in terms of battery life (despite taking longer for most tasks and despite having a smaller battery). It seems the 6 efficiency cores are really showing here.

[@LordBagdanoff]

M3 Pro looking a great buy after all

[@PassiveSmoking]

Should be noted that the Max is in the 16 inch chassis, which includes (among other things) a bigger battery.

[@LordBagdanoff]

@@PassiveSmoking nahhh overkill noisy fan and battery drainer

[@rpaiz87]

Thank you, As a data scientist I was curious about python performance on these models. Much appreciated!

[@IonutNegru87]

Found this after I ordered MacBook Pro 16" w/ M3 Max 14CPU, 30GPU... really glad i did as the performance seems awesome. Thanks Alex.

[@Guer778]

Thanks for your videos, I have been following them since I am between an M3 pro and an M3 max, currently having an Intel i9, I am a software developer and the trend leans more toward the M3 max.

[@johningram420]

What did it for me was the two external monitor limit on the pro. I can't guarantee that I won't need 3 external monitor for as long as I have this laptop.

[@thaZeref]

Curious how m3 max performs in Unity and UE5

[@purplecorpse4230]

Yesss

[@MarekG438]

I will test m3 max with 36GB 14”, but i have to wait one month for it because default keyboard is ISO and i like ANSI so i have special order. I can answer here and compare it with i7 13700k 64GB ram if you like. With production game not just one asset in project. Pls remind me if I forget

[@purplecorpse4230]

@@MarekG438 sounds good

[@theravigangwar]

@@MarekG438 when u r going to test??

[@MarekG438]

@@theravigangwari need to w8 for my macbook 2 weeks from now

[@jameshancock]

Please do tests against local LLMs including some 7b, 13b, and 33b models with 4,8,16 and 32k context lengths with and without metal acceleration. Specifically compare them to nvidia stuff and intel cpu stuff.

[@TheWallReports]

Excellent suggestion & request. I do a a lot local LLMs work on my machine and plan on upgrading to a M3 Max 128GB machine so I can work w/the larger models.

[@AZisk]

which LLM specifically?

[@d.paradyss8791]

Sería genial un test usando algún llm que tú elijas y un modelo de generación de imágenes como stable difussion

[@dbbbbbbb1952]

@@AZiskprobably llama they match those parameter numbers

[@gerald1964]

This is really valuable info. Thanks for performing these tests. It seems that M3 Pro MB Pro 14 inch will still experience noteable thermal warming and fan noise under full CPU load. Certainly, the base M3 Max in the same form factor will be off the charts in this regard based on another test that I have seen.

[@sreeharisreelal]

Can you make a video comparing the 18GB M3 Pro MacBook Pro to the 36GB M3 Pro MacBook Pro with regards to developers' needs, including VM?

[@SadamBober]

People already mentioned LLMs, in addition to that some Stable Diffusion tests would be nice too

[@rpaulseymour]

Yes yes yes. I found this channel because as a creative, these chips aren’t just for Adobe products anymore and average “rendering” benchmarks don’t tell everything. On my old Intel MPB, wouldn’t dare trying any Stable Diffusion, but I’m very curious about how dynamic caching and the higher GPU transistor counts in the M3 (especially mid range Pro) would do, as well as the newer NPUs. I’m a graphic designer, and the limitations of commercial implementations don’t work for what I need, subject-wise. Now is when design and creative coding might really start to merge or overlap.

[@sweealamak628]

Awesome video with the single-CPU core test I really wanted to see. As "slow" as python is, the speed bump would be nice to experience, especially on the M3 Pro which seems to be the sweet spot for me.

[@robertleal7455]

This was the most useful video in helping me decided which model to go with

[@AZisk]

glad it helped you

[@alunelu111]

I would love to see a 16" M3 Pro vs 16" M3 Max test. That would provide a more realistic comparison between the two. The 14" form factor and fans don't seem to help with the heat, and that's clear.

[@phxntomplanet]

i wouldve rather have seen all in the 14 inch beause we already know the 16 is good at heat dissipation

[@TheHardcard]

The tests I’ve seen show the 14” does very well with the heat. Not enough for the Max, but if you watch comparisons of the two sizes for the Max, the 14” can keep up under some heavy loads, often a little slower. There have been very few scenarios where it is a lot slower, when it happens is when performance is way out of the Pro’s league. I don’t know if anyone will do 14” Pro vs. 16” Pro, but I predict essentially the same speed under all conditions.

[@tejasnipankar2185]

Purchased the M3 Pro MacBook Pro with 18GB RAM and a 512GB SSD for personal use, upgrading from my employer-provided Lenovo ThinkBook with an i7 11th gen processor and 16GB RAM. Initially considered the M2 Pro but opted for the latest release. Now, I'm questioning if it'll efficiently handle my personal data analytics and mild machine learning tasks without overheating. Invested $2k and seeking insights on its performance. I wish to run docker and other data engineering tools on it as well

[@woolfel]

even for light ML work, you always want as much memory as your can afford. The macbook pro will definitely be better than Lenovo thinkbook, but 18G will limit you. I have M1Max with 64G and M2Max with 96G. I can train with batch sizes up to 1024 using images that are 300x300 on M2Max. On RTX 2060 6G max batch size for 300x300 images is 16.

[@motoshan]

Hi Alex 👋 I believe (at least part of) the reason for the battery drain difference comes down to how long the CPU is under load. Generally you get better battery life by getting the CPU back into idle state more quickly. This means even if you have to use more cores to get the task done, the fact that the CPUs complete the work much more quickly allows them to go back to idle and save the battery!

[@AZisk]

true

[@scottldn]

Great to see some usage from the perspective of developers. Most videos seem to resolve around video rendering and gaming 👍 new sub here

[@xingzhexin8843]

Alex I just have to say that your thumbnail is awesome and so is your content.

[@dsimon3004]

Great video !! Would be amazing to see a comparison between the M1 Max, M2 max and M3 max models

[@Chewbacca0678]

It’s been noted before that the 14 inch models are throttling because of overheating. This doesn’t happen with the 16 inch models.

[@takomayowasabi6491]

very useful information before black Friday!

[@mikhailtukhachevsky3277]

Thanks! Your test is so helpful! Have made up my mind for a 14’ m3 max

[@JohnSmith-sh1sy]

wish it was m3 pro 16 inch to compare temperatures with max 16

[@georgi.georgiev]

Love your tests, but 14" vs 16" is unfair comparison, though. Can't remember where I saw this, but the 16" with bigger fans and bigger thermal absorption capacity of the body gives you about 6-10% better performance under heavy loads under equal other conditions (aaand that's why a 14" is a no go for me 🙂).

[@AZisk]

nothing can really be fair

[@AnggaTirtaFL]

Just bought Macbook Air M2 and surprisingly good for my Flutter environment demand, my last Macbook Pro 16 2019 certainly better at any aspect but the price to performace on this puppy M2 is no joke. Love it and not gonna upgrade to M3 any sooner knowing they will release another next year like they always do. I'll just stick to it for another 3 or 4 year.

[@Tvj_films8452]

Dang, we have a python data scientist that does well thought out and edited videos. Too good.

[@AZisk]

i think im far from either of those, but have aspirations

[@davidpsp89]

I am waiting for MATLAB tests

[@leonplis9926]

I did run the matrix multiplication test on my M1 Pro with 16GB, with safari +30 tabs, one Xcode and 2 vscode instances, iterm with 4 terminals, one Debian VM with parallels with 4GB RAM with a firefox instance with 10 tabs, apple music, discord and wp web also opened, and it only took 26,2 seconds. the power of apple silicon is so insane that I think M1 users don't need to upgrade unless they're video editors or smth.

[@Joe_Brig]

We need some local LLM tests. How much does the 150, 300, 400 memory bandwidth make a difference?

[@AZisk]

what llm specifically?

[@Joe_Brig]

@@AZisk would like to know startup time and response times for codellama:34b. using Ollama.

[@synen]

Thank you Alex for this informative, excellent and funny video.

[@AZisk]

Glad you enjoyed it!

[@arnav_patange]

Pls tell which laptop is best for ai/ml . Also make comparison video for these topics for the newer generations like you did with m2 series.

[@oscarhealey]

Alex sounds like Robin Williams trying to do an accent and I'm loving it!

[@AZisk]

if i had half the talent robin williams had in his pinky, i’d be happy to sound like robin williams farting.

[@-Burb]

Wow, I ran most of these tests on my Windows desktop (Ryzen 7 7700X w/ 32GB DDR5 ram @ 6000MHz) and the M3 Pro seems to keep up extremely well. The Max is even faster in most tests! My desktop results just to have fun comparing with these beasts of laptops: Pyperformance: python_startup: 4.71ms regex_compile: 77.9ms json_dumps: 6.73ms logging: 4.23us Matrix Multiplication (16000x16000): 13.01s Mandelbrot: (16000): 26.025s (25000): 1m 3.355s I did run these in WSL rather than “native” windows or Linux, but it shouldn’t affect the times by very much if at all. Very excited to get an M3 Pro MBP soon! Edit - Is the M3 Pro in the video the binned or unbinned model?

[@crearg8259]

Ok, M1 Max 14 inch (32 core 64GB version) here, matrix multiplication took 28.2 seconds with other background tasks. Basically same level of M3 Pro, no need to upgrade.

[@enopticon]

Thank you for your videos. Very interesting to watch. I was hoping to hear your thoughts of the M3 max on ML workloads with pytorch/tensor flow. Is it worth it to max the memory to 128GB for ml model training?

[@user-yg3vv3zp1l]

Very helpful content. Thank you. ❤

[@AZisk]

So glad!

[@akin242002]

Great video! Oh boy! The M3 MacBook Air is going to be very popular with this new M3 CPU for automation programmers. With a Geekbench 6 multi-core score of 11915 and its amazing battery life, it will be even more popular than the M2 Air 13.5-inch (GB6 multi-core score of 9800). Including 16GB of RAM of course.

[@DojoOfCool]

The numbers of performance cores vs efficiency core make a big difference because some app's don't use the efficiency cores. That's why the Max chips do so well because they have the most number of performance vs efficiency cores.

[@geovanesoares]

Hey, i'd love to see a comparison between the M1 Max and M3 pro, because I didn't see any comparison like that, and I'm not sure witch machine should I buy

[@slawek0latka]

I like your vids. Can you also do price / performace benchmarks? PC vs Mac?

[@jack_mc8]

Thank you for reviews !

[@AZisk]

Glad you like them!

[@jigyansunanda]

I am at m1 right now. but man as I move into my masters in computational data science, I am looking forward to a 16 inch m3 pro / max

[@UFOGoldorak]

Anyway we can get some render comparisons between the newly released blender v 4.0.1 against the older 3.x especially on that M3 Max machine? Version 4 of blender supposedly has M3 specific improvements.

[@AliZamin]

Do test Android Studio as well

[@tylerwmbass]

Would love to see these comparisons with an AMD and Intel desktop running Linux. I have a few machines I might try some of these tasks with, a lot of the engineers I know including myself often are using Macbooks to just SSH into other hardware that's going to do most of the crunching anyways.

[@DigiDriftZone]

You should run the tests in a loop to see number of iteration per second for more meaningful results.

[@dudleygb45]

Would be awesome to see how this goes on the airs?

[@andrewfong2734]

Terrific assessment. I wonder how much the bigger 16" chassis helps with handling thermal load? The M3 Pro in the 14" chassis may be throttling more with the confined space.

[@lightenedstardust1883]

we need a review on linux os (zorin or pop os) from you. It seems to have potential giving a really good developer experience

[@justinyadao744]

how do the ray tracing hardware perform in unity or unreal?

[@DiabloMatrix968]

I'm curious about Java and Android development , what kind of tests can you run to see what is the best ?

[@alptaken_]

You remind me of II. Mehmed (Mehmed the Conqueror)

[@AZisk]

how do you know?

[@alptaken_]

@@AZisk 'Portrait of Mehmed II. by Gentile Bellini'

[@jeffr_ac]

Great vídeo

[@AZisk]

thx!

[@h1ghpower]

I’m so confused as I want to change to Mac…I mostly do QA..web testing and lots of multi tabs and multi tasking ..maybe some automation and light programming soon…not sure what do buy to really be proof for 5 years …..I’m now on a powerful i7 8 core and 16gb ram 17.2 inch laptop…..…what you guys recommend (16 inch plus only)

[@_Chris390]

You don’t mention at the start of the video which of the laptops is the Max - the 16 or one of the two 14s. You also do not specify the full specs of each machine and if they are the higher tier or binned models, how much RAM they each have, etc. Pretty vital info for a comparison like this.

[@josecasillas4046]

Excellent videos

[@user-fr1ux1de4t]

Can you setup a TypeScript compiler performance test? Pretty important for frontend and nodejs devs

[@joel9909]

Amazing videos

[@AZisk]

Glad you like them!

[@andrawssantos3168]

Hey, your videos are always very good and informative! I need a little help! I have a 2015 MacBook with 16GB of RAM, an i7 processor with dual graphics. I'm a data scientist and usually work with Python to create and train machine learning models, including neural networks. I've been working with databases of at least 5 million entries, and in many cases, the models take more than 8 hours to train. I'm eyeing an M3 Max, but I have doubts about what to focus on: GPU cores, RAM, CPU cores, SSD? What do you suggest, considering that I work with large databases and various machine learning models? If you can recommend an option with no budget limit and two or three options below 3500 euros, I would greatly appreciate it. Thank you.

[@AZisk]

you’ll want to get as much ram as possible, but the m3 max is a good one for that. keep in mind that whatever you’re training needs to be apple compatible

[@RomPereira]

"JSON Dumps", 3 seconds later almost LOL... It is like farting in front of kids, you never fail to crack them up.

[@juniorg1024]

Thank you for this video! I have just started taking courses to learn python and will follow up with SQL. I have purchased the M3 Macbook pro 14' 18gb 1tb and was happy with my choice but after reading around, I see programmers stating that 36gb should be the min for programming. I just want know is the model that I purchase good enough to get me through the learning stages of programming? I love how my mac is working and for a few of my small projects I have no issues. But again I am just starting to learn so not sure if this will effect me much.

[@ishanpanchal2183]

Hey there, I am in the same situation as you, I have tried to ask around but have not gotten an answer yet, have you gotten one? I was also trying to see wether I need more ram or if 18 Is good enough

[@mittamoa]

18gb is enough for 'learning stages' and even more. Invest spare money in other areas for example if you need to learn cloud services, software licenses or similar.

[@RomPereira]

Sweet video! You are taking your sweet time doing them, I did not buy my computer yet. By the way you could do a live about this black friday thing. There are plentiful discounts on the new M3s, from 200 dollars and up. Help us out! I am between a 14 M1 Max at BH Photo (2499 M1Max/64/2TB), 14 M3 Max base, 2999 at Adorama with the "apinsider" coupon discount and for those luck enough to have a Microcenter close by, they have a 10% discount on some M3 Max models, yes, 10%, 2879 for the base model... Come on Alex!!!!! Live Black Friday Buy!!!!!!!

[@AZisk]

yes, i have a big dilemma, since editing these video myself mostly, and it’s taking a bit longer to get them out. but hopefully the videos that are out are better and more complete than previous ones.

[@Br0adCastYourS3lf]

Is it safe to assume that hypothetical **M3 Mac Mini** (M2 Mac Minis announced January 17, 2023) and **M3 Mac Studio* (M2 Mac Studios announced June 5, 2023) will have the exact same core counts?

[@ManojVerma-rv9oq]

Hi Alex - I have seen lot of KZbin channels but your approach to analyze multiple configurations is awesome. My needs are pretty simple - I do financial modeling using excel with data tables, macros and formulas which are volatile like - offset, lookups etc. I don't multitask..just safari with 2-3 tabs, Apple Music and teams for calls and pdf. I intend to buy MBP-M2Pro-12C (8C+4E) with 16GB or MBP-M3Pro-12C (6C+6E). Not sure which one is good..I think having more P-cores will help. Can you please advise ? Thanks a lot. Manoj

[@akasharakharao3091]

Hello, I'm thinking of buying macbook M3 air with 8GB Ram 256SSD is it Sufficient for machine learning and deep learning as I will be using this laptop just for my own practice and to develop portfolios or should i go with the gaming laptop's with dedicated Graphic cards like Nvidia?

[@izzathnisfer5040]

Hi Alex I’m waiting for the m2 air 15 vs m3 (1599) I’m confused to choose Upload asap

[@watchchat]

Am I missing the startup config link?

[@Macintoshtcftrus]

Does the matrix multiplication test on 3:07 use the Neural Engine on the silicon? Or one would have to manually go and utilize that resource?

[@terror403]

Why did not you write on your terminal in the prompt (PS1) the model it is ?

[@edmondj.]

Thanks

[@edmondj.]

If you can train huggingface's Mistral with mps I pay you 10 ;)

[@AZisk]

🤩 thanks a lot!

[@AZisk]

i have a few ml tests in the rep, but curious about training. please send a link to my email: [e-mail]. Is there already a mps implementation?

[@caucasia_lucas]

Alex, I guess the higher temps causing the less percentage on the bat and not fan time :).

[@AZisk]

perhaps

[@stuff7664]

Never seen laptop temps go up this high

[@daan3298]

How many tokens per second for LLM inferencing? Llama.cpp models. + How long to describe image with Llama.cpp Llava? That is what's interesting for these large RAM M3 models.

[@S3basi1]

Great video!!

[@AZisk]

Glad you enjoyed it

[@user-dp8dr4wl8z]

Mandelbrot test shows what would happens if Apple would let 8 P-cores on Pro. It would beet M2 Max. They didn't want that.

[@InternetListener]

Can you test numeric keypad input speed? How can you tell one MacBook from another since 2017 if they are turned off? Those could be M1 or M2...will M3 16 gb be faster than 8Gb?

[@yashj8238]

Please tell me this about m3 max, if I am not throwing intensive tasks at the machine, just doing simple coding on pycharm, will the fan kick in?I want my machine to be dead silent that's why it's crucial to me. I can withstand fan noise under heavy loads

[@imopn]

Im done with Sena untill they support open coms standard

[@alexh2665]

Glad I know I made the right choice with m3 pro coming from M2 MBA. I was worried M3 would have better single core performance than M3 pro due to some benchmarks and cuz the base M2 air last year was beating out M2 pro in a lot of single core tasks but seems like in real world usage that isn’t the case with M3 and all the M3s beat out M2 max in single core which is also a big plus

[@TheHardcard]

The differences in single core performance is just the silicon lottery. They have identical performance targets, but each individual core varies slightly in actual production. The M2 Air just happened to get a faster core variantion. The only way to affect that is to have the money to buy 10 or so machines and test them right away. Almost guaranteed one will have top single core variation. Then return the other nine or so.

[@alexh2665]

@@TheHardcard I figured that t was something like that, seems like this year having the actual higher CPU cores is actually seeming to make more of a difference

[@TheDanEdwards]

How does sitting the laptops on your table affect the temperature performance? What would the temps be a laptop was sitting on, say, a metal baking cooling rack (as found in your kitchen, used to sit bread out of the oven)?

[@AZisk]

haha. nice try. i almost fell for it

[@fnkcgxgkgcfu]

Imagine the difference of performance when running those tests on the bios terminal

[@qwertzuiop875]

To be honest the mbp with 8gb is more than fast enough for almost most tasks.

[@not_anton]

Are you sure the Numpy libraries are installed correctly? It does not implement matrix mult itself, calls system BLAS functions instead. The 86 sec runtime is really slow for M3 - I got 50 sec on M1, and 92 sec on an ancient Intel i7-6820H.

[@AZisk]

if you are asking whether i used the accelerate framework, then no, none of these machines used it to show base levels. but, yes, i should have specified in the video.

[@marianbencat6658]

Now... some benchmarks on some programming language / virtual machine WITHOUT potato performance...

[@Fycstion]

Can you re-run this test with 24 gb ram on the base m3?

[@AZisk]

i don’t have that configuration

[@pxnchx93]

What do you think about that new Black color on the MacBooks? Is it dark enough?

[@AZisk]

it’s fine.

[@tarunbaluni133]

Hi , I'm a student eager to learn ios development . I have learned Python, Java , Android development , beginner level of Datascience in a high spec Dell inspiron. laptop Now I have some interest in M1 Macbook pro for ios development or swift . I am from India and the price is in my range around 700$ so would you suggest me to buy Macbook pro M1 or MB Air m1 or Macbook air m2?? Tell me how it will be performance of MBP m1 as compared to MB air m1 in mobile development or should I remove the thought of buying a new Macbook Laptop should solve all the previous learnt languages which I have mentioned above .

[@andersontapetti8420]

Weird, my MBAir M2 with 8cores makes Matrix Multiplication at 38s. Faster than M3? (this mac has 16GB of ram)

[@navb3663]

Great video

[@AZisk]

Thanks!

[@user-es4oz2nf9b]

Schwarzenegger🔥

[@RomPereira]

[@aj7_gauss]

how you install python , like is under intel simulation or arm?

[@adityadarekar5497]

On My Mid-2015 MacBook Pro with Intel i7: Matrix Multiplication took 63.31s (I am shocked that it ran faster than base M3) and the Mandelbrot test with 16k parameters took 3:27 while 25k p took 8:58. The multi core score is def bad on my Mac but what happened to matrx multi? Is it single core?

[@not_anton]

Same thing. Probably Conda installs bad math libraries by default, or Numpy not linked to use proper ones. Would be interesting to compare with Numpy installed e.g. with Homebrew instead of Conda.

[@anjansingh3093]

@2:43 its not five point seventy nine, its pronounced as five point seven nine.

[@AZisk]

oh yeah?

[@anjansingh3093]

@@AZisk for your knowledge, small things make big impact. Google it.

[@LordBagdanoff]

So the M3 having a high battery is pointless since the single fan has to work harder and eventually draining more battery

[@AZisk]

you’d likely buy the base model for lighter work

[@LordBagdanoff]

@@AZisk I want the new colour though 😂

[@flat6croc]

Er, the reason why the M3's battery is lower is because the chip is under load for much longer than the others in many of the tests!

[@boraoku]

My M1 Pro (16”) within 10% of M3 Pro and much more cool.

[@boraoku]

Oh scratch that I was running the tests on low power mode (set to active automatically while on battery power). With low power off, M1 Pro scores marginally better than M3 Pro!

[@phoneaccount6907]

You have rich set of HW. Btw are you checking correctness of ops? Like linx do. Anyway , why python? Afaik you are JS guy, why not node.js?

[@AZisk]

good question. many of these llms tend to be better at python than js imo. so im giving them what they should be good at. i know it really depends on what they are trained on, but python is the most popular language in this area, so it makes sense that people working w ml would also know python

[@RobTaylor1966]

I’m curious with all this extreme heat what kind of issues is this causing this machine over a long period of time?

[@AZisk]

heat seems to be managed pretty well, unless you are constantly putting machine through the ringer like in the mandelbrot test. few people will ever do that consistently, and if they do, they should consider running their jobs in the cloud