GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Talk about Hardware used for Deep Learning


Locked
User avatar
fsrenhui
Posts: 2
Joined: Wed Oct 14, 2020 2:54 am
Has thanked: 3 times

GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by fsrenhui »

Hey guys, new to faceswap.

I am currently testing on a older pc with i5 8600, 8GB RAM and a 1060 3GB, which is capable of training at low res only.
I use my laptop with docking station all the time, so the pc is not really getting used much. (I don't game on the pc)

Thinking of upgrade to a "newer" used card since on a budget, between 1070 8GB and 1660 6GB(similar price point)
Couple questions:
1 ) which will perform better for faceswap?
2) How much benefit will I see compared to my 1060 3GB?
3) Will either work using with the more expensive Villian training mode?

Also how much benefit we'll see by upgrading ram from 8gb to 16gb?

Thanks a lot.

User avatar
abigflea
Posts: 182
Joined: Sat Feb 22, 2020 10:59 pm
Answers: 2
Has thanked: 20 times
Been thanked: 62 times

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by abigflea »

System Ram is fine at 8gb, but if running Windows, 16 GB would be plenty of breathing room.

For GPU Ive been looking at this graph to get a general Idea

pic_disp.jpeg
pic_disp.jpeg (71.7 KiB) Viewed 11726 times

:o I dunno what I'm doing :shock:
2X RTX 3090 : RTX 3080 : RTX: 2060 : 2x RTX 2080 Super : Ghetto 1060

User avatar
torzdf
Posts: 2651
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 129 times
Been thanked: 622 times

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by torzdf »

1660 will train faster, 1070 will train bigger models. It really depends where your priorities are. If it were me, and all things being equal, I'd go for the 1070, but it is subjective.

My word is final

User avatar
fsrenhui
Posts: 2
Joined: Wed Oct 14, 2020 2:54 am
Has thanked: 3 times

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by fsrenhui »

Thanks for the replies, guys.
By training bigger models, I suppose you meant higher resolution models such as 128x128?
Hunting for a 1070 now.

User avatar
torzdf
Posts: 2651
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 129 times
Been thanked: 622 times

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by torzdf »

Yes and no. It's more about model complexity. More complex models require more VRAM.

My word is final

User avatar
BrokenClock
Posts: 5
Joined: Mon Aug 31, 2020 5:48 am
Been thanked: 1 time

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by BrokenClock »

I'm running an i5 1660 works fine.

User avatar
Helioses
Posts: 5
Joined: Sat Oct 24, 2020 3:02 pm

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by Helioses »

To expand on this topic.
I’m working on a budget and is thinking about a 2060 (6GB) or a used 1070 or 1080 (8 GB). What’s better, tensorcores or 2 GB extra VRAM.

User avatar
torzdf
Posts: 2651
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 129 times
Been thanked: 622 times

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by torzdf »

That's a tough one. Personally I would take RAM over speed, but someone else might equally say the opposite.

My view is that as long as you can fit the model in VRAM, then it doesn't matter how long it will take.

My word is final

User avatar
bryanlyon
Site Admin
Posts: 793
Joined: Fri Jul 12, 2019 12:49 am
Answers: 44
Location: San Francisco
Has thanked: 4 times
Been thanked: 218 times
Contact:

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by bryanlyon »

I am, as torzdf said, going to disagree with him. I would rather have the speed of the 2060 and using mixed precision more than gets back that lost vram. However, if you can find an extra $50-100 you can get a 2060 super and get both 8gb and tensorcores. (I know that's not always possible, but I list it for those where it is).

User avatar
Helioses
Posts: 5
Joined: Sat Oct 24, 2020 3:02 pm

Re: GPU upgrade on a budge: 1070 8GB vs. 1660 6GB

Post by Helioses »

Thanks for the responses. Not much clarity, but plenty to think about. With divided opinions, it seems I can’t completely wrong with either option so that calmed some worries
Again, thanks for the help.

Locked