Are these specifications decent enough if one needs to work on hardcore deep learning amd ML projects?
-
Hey there! Welcome to TFC! View fewer ads on the website just by signing up on TF Community.
Need help with PC specifications
- Thread starter Walter White
- Start date
- Replies 12
- Views 558
Graphics ram 8gb*
"Ryzen 7 5800u' not great for hardcore*
Why don't you build a PC?
At least take a 6gb Vram graphics card and 16gb Ram if you plan to keep it for a longer time.
taskmaster
TF Premier
No you need a better graphics card. Even if you wanna go software you need atleast a 3060 which has 12gb vram too
Wait for intel to launch AI gen chips .
I won't recommend a laptop for hardcore ML stuff, maybe consider building a PC?
If you have to go with a laptop, then I'll say check if you can bump it up to ryzen 9/ 8gb graphics, also, no graphic cards with Ti versions, they are built for portability, not performance.
If you have to go with a laptop, then I'll say check if you can bump it up to ryzen 9/ 8gb graphics, also, no graphic cards with Ti versions, they are built for portability, not performance.
when it comes to computing power, desktops often outperform laptops with similar configurations. If budget allows, consider building a desktop computer and complement it with a decent laptop. This setup allows you to remotely access your desktop while traveling, providing the best of both worlds.
Cc_creditguy
TF Premier
Your budget bro
In this case- clear no.
5800U is for ultrabook. You need H series. U series Targets battery life. Tdp is low. And oem limits it even lower. Because they can't fit the required cooling system in an ultrabook.
Ti series is step up from base card. But both are pretty old. 5800u wasn't good for heavy stuff then and it definitely isn't now.
If you are looking at hard core use as a primary device, I would never recommend a laptop. It won't handle the power /cooling requirements. Plus it might get outdated in a short time.
Build a PC if you can. There are many reputed assembler who do a neat job. They do charge a premium for their service. Some of them also provide consultancy to choose best specs according your need/budget. Which you can use to order from them or local assembler.
You can also go with local assembler but you need to haggle hard, and would require a help of friend who have decent idea about this. Sometimes you can get good deal. Meaning the extra money you pay/save more than justifies the performance difference.
I have seen many local assemblers recommending parts which aren't furureproof/ compatible or don't give optimum performance.
But know that all this would require a time commitment from your end.
Intel GPUs lack behind NVidia. Most of the AI/ML frameworks developed around CUDA for NVIDIA GPUs.
OP can better go for desktop if he is seriously into AI/ML training.
@Walter White get some Ubuntu Desktop with Nvidia GPU and use it anywhere with Jupiter notebook from laptop.
P h y t h o n
Yes. Python just glues things. Under the hood AI/ML frameworks use CUDA(C++) to talk to GPU.
Python is syntactic sugar to avoid brain freeze of C++ daunting syntax.
Do you want to build a workstation?
If not, just buy a predator 🤷♀️
If not, just buy a predator 🤷♀️
