Need PC recommendations for data analytics and ML

[saksham77]

Hey everyone,

I'm a data analytics student and my current laptop has a ryzen 5 3500u with 8gb ram(2gb vram + 6gb ram) its quite slow for my current low workload as a result I'm looking for a serious upgrade with a ryzen 5 7600x and 7700xt with 32gb 5200mhz ram is it enough for the next 5 years with the intensive tasks of data analytics and machine learning.

I will wait till September at least for the 7700xt prices to drop.
and will a 7750xt launch in the next few months or year so the 7700xt will get more cheaper.

Thanks!!

[jeyakumar16]

If you're gonna run all your ML/DL programs locally on your machine,  then you might wanna consider getting a Nvidia GPU as it has CUDA which can provide gpu acceleration for deep learning tasks, most of the libraries and frameworks are optimized for Nvidia GPUs 

 

There are dedicated Nvidia cards for Deep Learning and other AI tasks like the Titan series.., most industries use Nvidia GPUs for their AI workloads

 

*My recommendation*

If you dont want to spend much on a new PC just for AI,  i will definitely recommend you to use Google Colab, as it can even provide better training speeds and acceleration than some GTX cards and is a great platform for beginners even to some rough extent.

I personally love Google Colab, it has a ton of features, all of this for free!!!

 

I will also suggest you to go for a i5 13600k over R5 7600x, as it is a much better cpu and  i guess you cant go wrong with a rtx 4060ti or even 4070
now you can get ddr5 rams at 6000 MT/sec
This wont bottleneck as well, you can game at 4k resolution with no issues

 

I dont have much experience, but i hope these parts go well 

 

i hope this helps!!!

[saksham77]

11 minutes ago, jeyakumar16 said:

If you're gonna run all your ML/DL programs locally on your machine,  then you might wanna consider getting a Nvidia GPU as it has CUDA which can provide gpu acceleration for deep learning tasks, most of the libraries and frameworks are optimized for Nvidia GPUs 

 

There are dedicated Nvidia cards for Deep Learning and other AI tasks like the Titan series.., most industries use Nvidia GPUs for their AI workloads

 

*My recommendation*

If you dont want to spend much on a new PC just for AI,  i will definitely recommend you to use Google Colab, as it can even provide better training speeds and acceleration than some GTX cards and is a great platform for beginners even to some rough extent.

I personally love Google Colab, it has a ton of features, all of this for free!!!

 

I will also suggest you to go for a i5 13600k over R5 7600x, as it is a much better cpu and  i guess you cant go wrong with a rtx 4060ti or even 4070
now you can get ddr5 rams at 6000 MT/sec
This wont bottleneck as well, you can game at 4k resolution with no issues

 

I dont have much experience, but i hope these parts go well 

 

i hope this helps!!!

I cant get the 4070 cause its 50k inr you can sure relate Im stuck at 35k for graphic card and the 13600k is still out of budget hence thought amd's price to performance

[jeyakumar16]

Ohhh alright, im not much experienced in picking pc parts

 

In the context of running AI workloads, you can go for Nvidia !!!

But still running in colab is a great choice !!!

 

 

[igormp]

7 hours ago, saksham77 said:

Hey everyone,

I'm a data analytics student and my current laptop has a ryzen 5 3500u with 8gb ram(2gb vram + 6gb ram) its quite slow for my current low workload as a result I'm looking for a serious upgrade with a ryzen 5 7600x and 7700xt with 32gb 5200mhz ram is it enough for the next 5 years with the intensive tasks of data analytics and machine learning.

I will wait till September at least for the 7700xt prices to drop.
and will a 7750xt launch in the next few months or year so the 7700xt will get more cheaper.

Thanks!!

The 7700xt is going to be useless for ML, you would be better with no GPU instead. Go for an nvidia one.

The 7600x is a reasonable pick.

 

I can't say how long it'll last you without knowing your specific workloads, but it also depends if you'll keep doing the same thing over and over again for those 5 years or not.

5 hours ago, jeyakumar16 said:

I will also suggest you to go for a i5 13600k over R5 7600x, as it is a much better cpu

Not for ML, since current Intel CPUs lack AVX512: https://www.phoronix.com/review/ryzen-7600-7700-7900-linux/12

5 hours ago, saksham77 said:

I cant get the 4070 cause its 50k inr you can sure relate Im stuck at 35k for graphic card and the 13600k is still out of budget hence thought amd's price to performance

What about a 4060ti 16gb? It'd be the perfect budget GPU for ML given its vram. Otherwise, go for a 3060 for its 12gb of vram.

[saksham77]

2 hours ago, igormp said:

The 7700xt is going to be useless for ML, you would be better with no GPU instead. Go for an nvidia one.

The 7600x is a reasonable pick.

 

I can't say how long it'll last you without knowing your specific workloads, but it also depends if you'll keep doing the same thing over and over again for those 5 years or not.

Not for ML, since current Intel CPUs lack AVX512: https://www.phoronix.com/review/ryzen-7600-7700-7900-linux/12

What about a 4060ti 16gb? It'd be the perfect budget GPU for ML given its vram. Otherwise, go for a 3060 for its 12gb of vram.

Hey I checked some videos on the 4060ti it dosent perform as good as the 7700xt, and also checked that amd is introducing rocm to rdna 3 gpu's hoping it comes to windows as well

[igormp]

9 minutes ago, saksham77 said:

Hey I checked some videos on the 4060ti it dosent perform as good as the 7700xt

Are you going to use it for games or for ML? For games the 7700xt is surely better, but it's really bad at ML.

9 minutes ago, saksham77 said:

and also checked that amd is introducing rocm to rdna 3 gpu's hoping it comes to windows as well

1- Doing ML in Windows is way worse than on linux.

2- ROCm is awful even on linux, don't get your hopes up for windows support.

 

FWIW, a 7800xt is slower than a 4060ti 8gb when it comes to ML.

[saksham77]

16 minutes ago, igormp said:

Are you going to use it for games or for ML? For games the 7700xt is surely better, but it's really bad at ML.

1- Doing ML in Windows is way worse than on linux.

2- ROCm is awful even on linux, don't get your hopes up for windows support.

 

FWIW, a 7800xt is slower than a 4060ti 8gb when it comes to ML.

So do you have any idea if machine learning is used way too much in data analytics, that I would seriously need a nvidia gpu becasue in my country they are beyond overpriced. For example a 4060 ti 16gb is priced same as a 4070 and beyond my budget for sure like  a 7700xt will get to 35k in few months because amd has near to no value in India and an 4060 ti 8gb will still be at 37k. So should I opt for the 4060 in such scenario as I'm a data analysis student and ML wont be my main priority IG

[igormp]

5 minutes ago, saksham77 said:

So do you have any idea if machine learning is used way too much in data analytics, that I would seriously need a nvidia gpu becasue in my country they are beyond overpriced. For example a 4060 ti 16gb is priced same as a 4070 and beyond my budget for sure like  a 7700xt will get to 35k in few months because amd has near to no value in India and an 4060 ti 8gb will still be at 37k. So should I opt for the 4060 in such scenario as I'm a data analysis student and ML wont be my main priority IG

You did mention machine learning in your OP, hence why I mentioned that. If all you're going to do is pure analysis, such as transforming data with excel/powerBI/pandas and building charts out of that, then you won't need a GPU at all.

You might need more RAM tho, but it really depends on the size of the data you're working with.

[saksham77]

4 minutes ago, igormp said:

You did mention machine learning in your OP, hence why I mentioned that. If all you're going to do is pure analysis, such as transforming data with excel/powerBI/pandas and building charts out of that, then you won't need a GPU at all.

You might need more RAM tho, but it really depends on the size of the data you're working with.

Thanks mate, I searched everywhere but couldnt understand this part really thank you!!!