MrDeepFakes Forums

Some content may not be available to Guests. Consider registering an account to enjoy unrestricted access to guides, support and tools

  • We are looking for community members who are intested in helping out. See our HELP WANTED post.

[Help]Looking for advice on what hardware to use.

AceTony

DF Vagrant
In the hopes that Ampere launch will send ripples down the market and reduce the price of Turing, which in turn will reduce the price of used 10series, I'm hoping to upgrade from my 1060 3GB to a 1070, 1070ti or 1080ti.

Can you please recommend what card would be the best in terms of value for money and a small budget? Whether a 1070ti or 1080 would have any significant improvements over the 1070?


The rest of my system specs are R5 1600, 16GB RAM and a 7200 RPM HDD. Is getting a SSD worth it? Does running DFL on a SSD cause it to die quicker?

Apologies if this isn't the proper place to post or is a repeated question. Also appreciated if those who reply can provide their specs as well.
 

BeakFakes

DF Pleb
I think you should take the 1080ti. It has 11GB VRAM. Has your PC power supply enough power (W)?

SSD lets you load faster. That means the transfer form HDD/SDD to VRAM is faster. When your data is loaded to VRAM, just at the beginning of the training, there will be no difference in training time (I think so).
 

AceTony

DF Vagrant
BeakFakes said:
I think you should take the 1080ti. It has 11GB VRAM. Has your PC power supply enough power (W)?

SSD lets you load faster. That means the transfer form HDD/SDD to VRAM is faster. When your data is loaded to VRAM, just at the beginning of the training, there will be no difference in training time (I think so).

I have a CX550 so I think 1080ti is doable. Is it better than a 2060S? Doesn't Turing have some fancy AI stuff that's beneficial for machine learning?
 

angeloshredder

DF Admirer
AceTony said:
BeakFakes said:
I think you should take the 1080ti. It has 11GB VRAM. Has your PC power supply enough power (W)?

SSD lets you load faster. That means the transfer form HDD/SDD to VRAM is faster. When your data is loaded to VRAM, just at the beginning of the training, there will be no difference in training time (I think so).

I have a CX550 so I think 1080ti is doable. Is it better than a 2060S? Doesn't Turing have some fancy AI stuff that's beneficial for machine learning?

1080ti needs 2 slots  +  6 pins + 8 pins , not sure if your PSU have both 6 pins + 8 pins , power consumption is 250w

FbYyUqMh.png


j5zR3Svh.png
 

SPT

Moderator
Staff member
Moderator
Verified Video Creator
650W Alim which should be enough (I think, check it) for a 1080ti is only $80/70€ if you take a medium quality one. There's even cheaper models.
 
Top