What computer for machine learning in 2023

Рет қаралды 42,390

Jeff Heaton

Жыл бұрын

#gpu #machinelearning #deeplearning

Пікірлер: 40

@Virgil-Binny Жыл бұрын

Better video than most of the so called tech channel,short simple precise.

@kolakareem2441 Жыл бұрын

You are the best, Prof.

@HeatonResearch Жыл бұрын

Thanks!

@abhishekrohilla5371 Жыл бұрын

Is these specs for those who have to just start AI and DL...

@daddydope258 8 ай бұрын

@@HeatonResearchAcer Aspire 3 : 35,000 rs 1) intel core i3 11th gen processor 2) 4-8-12 GB ram (upgradable) 3) 256 GB NVMc SSD supports 2 TB HHD 4) 15.6" full HD LED backlit Display 5) intel UHD Graphics 6) upto to hrs hours battery life Ye DL engineering ke liye enough hai kya

@rahulsingh-yf8kc 8 ай бұрын

Seen best so far Straight to the point

@Ceilliau Жыл бұрын

Super helpful thanks. Did not know NV Link was on the way out!

@Danieljordan2 5 ай бұрын

This is EXACTLY what I was looking for.

@very_nice_777 10 ай бұрын

core i3 4030U with Intel Integrated Graphics. Peak performance is what I always get.

@unabonger777 8 ай бұрын

you mean your cpu is always maxed out

@redmi26635 8 ай бұрын

@@unabonger777 100%

@HelloWorlds__JTS 8 ай бұрын

Thanks for sharing your wisdom and experience with us plebs! Can you say a bit more about PCIe 5 replacing NVLink? I have yet to find a competitive [consumer] gpu using 5.

@luisff7030 8 ай бұрын

If we populate all the 8 ram slots, we will get more bandwidth if CPU can handle

@FielValeryRTS 9 ай бұрын

i5-12400 with RTX 3060 12gb... Ram only 8 so will have to upgrade... Sadly, I'm broke. 😢

@redmi26635 8 ай бұрын

you could have get f series cpu and save bucks to buy extra 8 stick

@nond8957 10 ай бұрын

What do you think about getting the intel arc a770 16gb?

@arindamdey1724 11 ай бұрын

Thanks

@theccieguy Жыл бұрын

Nice

@HeatonResearch Жыл бұрын

Thanks

@shivam_7208 6 ай бұрын

Is there a dedicated models for gpu for machine learning from nvidia ? Like they have for gamers graphic specific. I did my research but couldn't get something specific fo ml that is available to consumer

@LucasSouza-sw7yx 6 ай бұрын

Why PCIe 5 ports we don't even have PCIe 5 devices yet. The latest Nvidia card (RTX 4090) uses PCIe 4.0. If you're aiming to be safe for future upgrades then I have nothing to say but then there's the drawback of the processor socket. In case of AM5 AMD has plans for support up to 2024 (if I'm not mistaken) from that point onwards it's a hard guess. Might be aproblem if you're fictionated on having the latest technology aviable.

@karthikpai2537 11 ай бұрын

Need suggestions between 5700x and 13400f for my rig for machine learning tasks

@redmi26635 8 ай бұрын

What about Tesla K80 24GB gpu?

@daneillasso1358 11 ай бұрын

Hi, I'm wondering if is it better 5700x or 7600x on CPU. I mean, better 6 cores with a great speed or is it better more cores. Thank you very much.

@AntonioDev24 5 ай бұрын

I can barely run codellama:7b with a gen1 P15 23gbRAM, 4GB GPU. Im thinking in adding a eGPU (rtx360 12GB) any suggestions?

@ickorling7328 8 ай бұрын

How about mini pc iGPU AMD, because systems like these have no vram, and thr iGPU directly (and yes, without flaw because of integration with CPU) use the system ram for vram. The archetecture is then the only thing in question, and CUDA is different but not exclusively the only way to run AI. It seems that Nvidia is just more optimized. For comparable class iGPU, or GPU, the AMD is likely within 25% or pess speed difference. Who can confirm or debunk this theory?

@JReuben111 Жыл бұрын

what about optimal laptop for ML ? RTX 4090 enabled gamer machines ($$$) will be on the market in Q1-Q2 2023 - should have equivalent CUDA / Tensor cores to a desktop 3080

@HaunterButIhadNameGagWtf Жыл бұрын

No. You will destroy ANY laptop on ML... I tried to run some training on Dell Precision 7550, which is pretty well cooled and even this laptop didn't like it much... It's not build for max temperature for long time. Plus the mobile GPU is not as powerful as the desktop one... by a lot!

@bbnCRLB 7 ай бұрын

I have 11GB GPU and 64GB RAM with Intel Core i7 11th Gen with Linux. I'd like to upgrade my GPU to at least 16GB. But these things are very expensive. I wonder why AMD isn't trying to compete hard against Nvidia's CUDA.