When did you stop being a techuck and realize you did not need a dedicated gpu?

come on man, I hate Apple as much as anyone here, but the air was unironically one of their great ones

>GPUs are only used for games
Idiot

I need one to run my machine learning routines semi efficiently.

fucking retard

Never. Even though I'm using laptops exclusively. Gaymen laptops cost about the same as meme ultrabooks with the same CPUs while having way better cooling AND a discrete GPU AND better upgradability and repairability. Even if I don't use the discrete GPU I have a better laptop overall, simple. When I finally settle in one city and if I decide to build a desktop I'll get some HEDT CPU which won't have an iGPU so I'll still end up having a discrete GPU.

They are not, but it is what 99% of people have them for.

very based user

What do you use them for then?

the best games are non-gpu anyway

Attached: serveimage?ref=http%3A%2F%2Fwww.bay12forums.com%2Fsmf%2Findex.php%3Ftopic%3D1353.0&ua=TW96aWxsYS (800x663, 140K)

What do you need ML for? Literally only PhD researchers need ML, and they've got V100 clusters

>river going through a mountain
200% Comfy.

>stop vidya gaming
>move permanently to Linux (no more double booting)
>realize that Intel HD Graphics is pretty good

Attached: intelhd.jpg (800x600, 49K)

I used my 1070 to develop and train a neural network for object recognition (pre-mining craze). It's a lot cheaper to use consumer grade hardware than to pay Nvidia for overpriced "professional" hardware.

Yeah, and what do you use that neural net for? Object recognition is literally babby's first deep learning project

Finishing a college course and winning first prize in a national software competition (that's how I got my job too). I feel like it worked out ok. What about you?

They are very fast at processing matrix math, so stuff like machine learning and rendering.

Can't speak for OP, but consumer gfx are good for prototyping or lighter ML.

I do hobbyist robotics. I don't need a lot of power to train, but I do have to do it over and over as I tweak the training data. However you don't need the latest and greatest cards. One or two generations back is perfectly fine and reasonably priced for the performance.

That's not a good reason to have a low resolution monitor or a shitty gpu.
I don't play games, but a mid tier gpu (say 1050 ti is still useful)

Fuck I messed up with the parenthesis.
Sorry ¯\_(ツ)_/¯

What for? Are you one of those ML larpers?

I never found my GPU useful for anything but vidya

I am a quant, and I building ny own framework to calculate risk in complex financial products and to forecast possible price movements. My plan is eventually quitting the investment bank I work in and make my own highly specialised company.
I use ML to calibrate model parameters. A lot of them are basically almost impossible to calibrate with regular statistics.

My point is that nobody gets hired to work on ML without a PhD (except that guy who got a job because he won a competition, I guess)

A bit of ML, a bit of fun in general with other OpenCL/Cuda programs, and a lot of shaders / other video tricks on mpv or madvr

competition guy here, I work as a programmer / linux sys admin. I got offers from universities to do ML related courses but didn't feel like it. Nobody gets hired without a PhD but you've gotta start somewhere.

Yeah, I mean, even then, the path you took isn't really replicable to get a job period, and the guy who is a quant already has the job requiring ML. I don't think it's rational for somebody to decide some day to just get a GPU and learn ML.
CUDA and OpenCL are useless for even 99% of developers, even more so for the general public.
If one day your job requires you to learn ML, then you ask your employer for a good GPU, but having one for personal use doesn't really make sense except for vidya.

If you want to succeed it's perfectly rational to start learning it.
Someone who wants to become a programmer should absolutely buy a computer and start learning. Better to find out it's not for you before spending years in education.

Quant guy here.
I work in derivatives pricing for market risk, I don't use ML in my job.
I did study SL extensively at uni during my MSc, though. And I don't really need any exotic shit for the type of calibration I want. I just compute the error on the forecast error. The forecast is a function of the parameters calibrated by the ML algorithm, so I am not working in the feature space, but by optimizing a functional. Sounds complicated, but basically, this means that the algorithms that I am actually using are mostly already implemented hy someone else in some python lib.

it simply clicked that the only stuff I would ever use my gpu is vidyas and graphical editing software. then after being on linux I learnt about brute forcing and mining and now I want to build a separate server to ssh when needed.

Larping about ML while turning Jow Forums in to /v/.

I realised the opposite, actually.

>I use ML for calibrating the models I use to do my job
>but I don't use ML for my job
Huh? I understand it doesn't play a primary role, but sounds like you do use it to do your job

No, I use ML for my personal projects, because I want to make my own specialized (small) company, so that I can research whatever I want.
Models is big investment banks go through huge layered validation processes. I have literally zero freedom to implement what I want. That's one of the main reasons why everything is super outdated and everything "just works" until that actual dark day comes.

Attached: 1519783618052.jpg (1100x2050, 356K)

*models in

Less fans, less noise.

Being annoyed by background noises is one of the signs of autism so you might as well spend the saved money on some kind of treatment.

kek
Not gonna lie I once dreamed I helped Ted escape from prison

CAD autism

Do you post in /smg/?

No, you don't because you are just a random, you can't even imagine how many of our customers are using telsa gpu atm

laptops don't need dedicated gpu graphics just get a ghost s1 case its the size of a external GPU enclosure but can fit a entire computer. and then get a crappy laptop to go with it.

Attached: Ghost S1 DXj.jpg (1200x900, 239K)

dafucks that symbol?
looks like some satanic kike shit