Jump to content

Deep Learning Build in 2020

mahmoud.tabikh
By mahmoud.tabikh
in New Builds and Planning
 Share
Go to solution Solved by igormp,
9 hours ago, mahmoud.tabikh said:

Budget (including currency): 2000 Euros 

Country: Germany

Games, programs or workloads that it will be used for:  

 

Hi Guys,

I want to buy a PC to run my machine learning models on and i am really confused by how my choices would play out with running models and parsing data.
My budget is around 2000 euros max and i wonder how could my options affect my efficiency working with Deep Learning.
I cannot say for sure what models i will be working on, as i am still a novice but i will be running deep learning models like CNNs.

Here's what i am thinking about:

  • CPU: Ryzen 5800x vs i7-10700K
  • RAM: 16 gb vs 16 gb Ram 3200 MHz
  • MoBo: X570 vs B550 (+BT)
  • GPU: NVIDIA RTX 3070 vs RTX 3080
  • Storage: 1 TB M.2 SSD vs M.2 NVME
  • PSU: 750 W vs 850 W (Modular)

Is this system enough ?
What is you advice for choosing, changing or adding ?

Thanks in advance for your help!

 

-Mahmoud

Are you really sure about only 16gb of ram? I currently have 28gb and often end up using over 50gb for some reasonable models (like Resnet100+). I'd go with 2x32gb sticks, which would allow for a later upgrade to 128gb if you ever need.

 

I'd say to go with a cheaper CPU, but it's not like you'd be able to use the extra money to do anything else. Instead, I'd recommend a ryzen 3900x due to the extra cores.

 

Are you in a hurry about buying this system? It's rumored that nvidia will release a new revision of their cards with more VRAM, and more vram is always better for ML, a 3070 16gb rould do better than a 3080 10gb due to the increased batch sizes you'd be able to use.

 

I recommend you to read some posts on that topic, specially the ones from Tim Dettmers. (here's another one).

 

If you are just beginning, you could go with a 2060 Super or 2070 Super, they'd be able to handle most stuff a beginner would throw at them, and would be neat placeholders until you get the hang of things and can decide for yourself what you actually need (more power vs more vram, or even multiple GPUs for increased data parallelism).

 

I'd say to go with a 850W PSU just in case, since the newer cards do like to sip tons of power, or in case you want to run 2x older cards (like the 2080Ti).

 

Go for an nvme as your main boot drive and get a regular M.2 sata one for data.

 

9 hours ago, Radium_Angel said:

Intel and nVidia have the tools you need (software and hardware) that are much more mature than AMD's offerings.

True for GPU, irrelevant for CPU. He could do just fine with an i3 or a ryzen 3.

 

8 hours ago, Downkey said:

Zen 3 benefits from quad channel. Here's a great video on it 

Consumer platforms don't have quad channel support, you actually meant dual rank on each channel. That's achievable with high density sticks too, and it's also irrelevant for machine learning tasks.

Posted

Budget (including currency): 2000 Euros 

Country: Germany

Games, programs or workloads that it will be used for:  

 

Hi Guys,

I want to buy a PC to run my machine learning models on and i am really confused by how my choices would play out with running models and parsing data.
My budget is around 2000 euros max and i wonder how could my options affect my efficiency working with Deep Learning.
I cannot say for sure what models i will be working on, as i am still a novice but i will be running deep learning models like CNNs.

Here's what i am thinking about:

  • CPU: Ryzen 5800x vs i7-10700K
  • RAM: 16 gb vs 16 gb Ram 3200 MHz
  • MoBo: X570 vs B550 (+BT)
  • GPU: NVIDIA RTX 3070 vs RTX 3080
  • Storage: 1 TB M.2 SSD vs M.2 NVME
  • PSU: 750 W vs 850 W (Modular)

Is this system enough ?
What is you advice for choosing, changing or adding ?

Thanks in advance for your help!

 

-Mahmoud

Link to comment
https://linustechtips.com/topic/1267011-deep-learning-build-in-2020/
Share on other sites
Link to post
Share on other sites
Posted
6 minutes ago, mahmoud.tabikh said:

Budget (including currency): 2000 Euros 

Country: Germany

Games, programs or workloads that it will be used for:  

 

Hi Guys,

I want to buy a PC to run my machine learning models on and i am really confused by how my choices would play out with running models and parsing data.
My budget is around 2000 euros max and i wonder how could my options affect my efficiency working with Deep Learning.
I cannot say for sure what models i will be working on, as i am still a novice but i will be running deep learning models like CNNs.

Here's what i am thinking about:

  • CPU: Ryzen 5800x vs i7-10700K
  • RAM: 16 gb vs 16 gb Ram 3200 MHz
  • MoBo: X570 vs B550 (+BT)
  • GPU: NVIDIA RTX 3070 vs RTX 3080
  • Storage: 1 TB M.2 SSD vs M.2 NVME
  • PSU: 750 W vs 850 W (Modular)

Is this system enough ?
What is you advice for choosing, changing or adding ?

Thanks in advance for your help!

 

-Mahmoud

In my experiences with neural networks, Intel and nVidia have the tools you need (software and hardware) that are much more mature than AMD's offerings.

NOTE: I no longer frequent this site. If you really need help, PM/DM me and my e.mail will alert me. 

Link to post
Share on other sites
Posted
6 minutes ago, mahmoud.tabikh said:

Budget (including currency): 2000 Euros 

Country: Germany

Games, programs or workloads that it will be used for:  

 

Hi Guys,

I want to buy a PC to run my machine learning models on and i am really confused by how my choices would play out with running models and parsing data.
My budget is around 2000 euros max and i wonder how could my options affect my efficiency working with Deep Learning.
I cannot say for sure what models i will be working on, as i am still a novice but i will be running deep learning models like CNNs.

Here's what i am thinking about:

  • CPU: Ryzen 5800x vs i7-10700K
  • RAM: 16 gb vs 16 gb Ram 3200 MHz
  • MoBo: X570 vs B550 (+BT)
  • GPU: NVIDIA RTX 3070 vs RTX 3080
  • Storage: 1 TB M.2 SSD vs M.2 NVME
  • PSU: 750 W vs 850 W (Modular)

Is this system enough ?
What is you advice for choosing, changing or adding ?

Thanks in advance for your help!

 

-Mahmoud

Here's a list I made. 60 dollars over budget, but this should keep your whole system very cool :) 

PCPartPicker Part List
Type Item Price
CPU AMD Ryzen 9 5900X 3.7 GHz 12-Core Processor €550.00
CPU Cooler NZXT Kraken X53 73.11 CFM Liquid CPU Cooler €119.99 @ Computeruniverse
Motherboard Asus TUF GAMING X570-PLUS ATX AM4 Motherboard €204.89 @ Alza
Memory G.Skill Aegis 32 GB (4 x 8 GB) DDR4-3200 CL16 Memory €125.10 @ Computeruniverse
Storage Crucial P1 500 GB M.2-2280 NVME Solid State Drive €59.79 @ Computeruniverse
Storage Seagate Barracuda Compute 2 TB 3.5" 7200RPM Internal Hard Drive €57.99 @ Mindfactory
Video Card NVIDIA GeForce RTX 3080 10 GB Founders Edition Video Card €700.00
Case Lian Li Lancool II Mesh ATX Mid Tower Case €95.12 @ Amazon Deutschland
Power Supply Corsair RMx (2018) 850 W 80+ Gold Certified Fully Modular ATX Power Supply €148.69 @ Alternate
  Prices include shipping, taxes, rebates, and discounts  
  Total €2061.57
  Generated by PCPartPicker 2020-11-09 14:47 CET+0100  

 

geometry is hard
b550 > x570

Link to post
Share on other sites
Posted
  • Author
1 hour ago, Downkey said:

Here's a list I made. 60 dollars over budget, but this should keep your whole system very cool :) 

PCPartPicker Part List
Type Item Price
CPU AMD Ryzen 9 5900X 3.7 GHz 12-Core Processor €550.00
CPU Cooler NZXT Kraken X53 73.11 CFM Liquid CPU Cooler €119.99 @ Computeruniverse
Motherboard Asus TUF GAMING X570-PLUS ATX AM4 Motherboard €204.89 @ Alza
Memory G.Skill Aegis 32 GB (4 x 8 GB) DDR4-3200 CL16 Memory €125.10 @ Computeruniverse
Storage Crucial P1 500 GB M.2-2280 NVME Solid State Drive €59.79 @ Computeruniverse
Storage Seagate Barracuda Compute 2 TB 3.5" 7200RPM Internal Hard Drive €57.99 @ Mindfactory
Video Card NVIDIA GeForce RTX 3080 10 GB Founders Edition Video Card €700.00
Case Lian Li Lancool II Mesh ATX Mid Tower Case €95.12 @ Amazon Deutschland
Power Supply Corsair RMx (2018) 850 W 80+ Gold Certified Fully Modular ATX Power Supply €148.69 @ Alternate
  Prices include shipping, taxes, rebates, and discounts  
  Total €2061.57
  Generated by PCPartPicker 2020-11-09 14:47 CET+0100  

 

Thanks! i appreciate the effort.

Seeing how expensive the rtx 3080 options are now in germany, i think this build would cost me well over 2400 €. The FE is almost non existent.

Why use 4x8 instead of 2x16 though?

Link to post
Share on other sites
Posted
  • Author
1 hour ago, Radium_Angel said:

In my experiences with neural networks, Intel and nVidia have the tools you need (software and hardware) that are much more mature than AMD's offerings.

That is true when it comes to NVIDIA; but i do not think intel have any software that would impact my use.

Link to post
Share on other sites
Posted
2 minutes ago, mahmoud.tabikh said:

Thanks! i appreciate the effort.

Seeing how expensive the rtx 3080 options are now in germany, i think this build would cost me well over 2400 €. The FE is almost non existent.

Why use 4x8 instead of 2x16 though?

Zen 3 benefits from quad channel. Here's a great video on it 

 

 

geometry is hard
b550 > x570

Link to post
Share on other sites
Posted
  • Solution
9 hours ago, mahmoud.tabikh said:

Budget (including currency): 2000 Euros 

Country: Germany

Games, programs or workloads that it will be used for:  

 

Hi Guys,

I want to buy a PC to run my machine learning models on and i am really confused by how my choices would play out with running models and parsing data.
My budget is around 2000 euros max and i wonder how could my options affect my efficiency working with Deep Learning.
I cannot say for sure what models i will be working on, as i am still a novice but i will be running deep learning models like CNNs.

Here's what i am thinking about:

  • CPU: Ryzen 5800x vs i7-10700K
  • RAM: 16 gb vs 16 gb Ram 3200 MHz
  • MoBo: X570 vs B550 (+BT)
  • GPU: NVIDIA RTX 3070 vs RTX 3080
  • Storage: 1 TB M.2 SSD vs M.2 NVME
  • PSU: 750 W vs 850 W (Modular)

Is this system enough ?
What is you advice for choosing, changing or adding ?

Thanks in advance for your help!

 

-Mahmoud

Are you really sure about only 16gb of ram? I currently have 28gb and often end up using over 50gb for some reasonable models (like Resnet100+). I'd go with 2x32gb sticks, which would allow for a later upgrade to 128gb if you ever need.

 

I'd say to go with a cheaper CPU, but it's not like you'd be able to use the extra money to do anything else. Instead, I'd recommend a ryzen 3900x due to the extra cores.

 

Are you in a hurry about buying this system? It's rumored that nvidia will release a new revision of their cards with more VRAM, and more vram is always better for ML, a 3070 16gb rould do better than a 3080 10gb due to the increased batch sizes you'd be able to use.

 

I recommend you to read some posts on that topic, specially the ones from Tim Dettmers. (here's another one).

 

If you are just beginning, you could go with a 2060 Super or 2070 Super, they'd be able to handle most stuff a beginner would throw at them, and would be neat placeholders until you get the hang of things and can decide for yourself what you actually need (more power vs more vram, or even multiple GPUs for increased data parallelism).

 

I'd say to go with a 850W PSU just in case, since the newer cards do like to sip tons of power, or in case you want to run 2x older cards (like the 2080Ti).

 

Go for an nvme as your main boot drive and get a regular M.2 sata one for data.

 

9 hours ago, Radium_Angel said:

Intel and nVidia have the tools you need (software and hardware) that are much more mature than AMD's offerings.

True for GPU, irrelevant for CPU. He could do just fine with an i3 or a ryzen 3.

 

8 hours ago, Downkey said:

Zen 3 benefits from quad channel. Here's a great video on it 

Consumer platforms don't have quad channel support, you actually meant dual rank on each channel. That's achievable with high density sticks too, and it's also irrelevant for machine learning tasks.

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share
Go to topic listing New Builds and Planning

×