PDA

View Full Version : VRam on GPU, is there a big difference between 1.5gb and 3gb ?



yoshiii
02-17-2014, 07:08 PM
Hello

I am going to buy a GTX 580 for use in Lightwave and Octane. I am wondering if there is a big difference between a 1.5gb model and the 3gb model.
Which is better?

Will the 1.5gb model still work well?

spherical
02-17-2014, 07:33 PM
As Waves Of Light mentioned in the other thread, the 580 will be better for viewport manipulation, the 6xx and 7xx are better for GPU renderers like Octane and Thea. He also said that the 3GB would be better. More RAM is always better.

yoshiii
02-17-2014, 07:43 PM
As Waves Of Light mentioned in the other thread, the 580 will be better for viewport manipulation, the 6xx and 7xx are better for GPU renderers like Octane and Thea. He also said that the 3GB would be better. More RAM is always better.

So the 580 for use in Lightwave? and the 600's and 700's for Octane and Thea? I read a bunch of posts that basically says the 580, Titan and the 780 are best for Octane. Is this true?

spherical
02-17-2014, 10:39 PM
OK, I'll try to explain this another way, just to make sure we're on the same page.


580/590 cards are not crippled, so perform generally the same as a Quadro in viewport manipulation—only.
GPU renderers use CUDA cores. The more, the better.
A 580 has 512 CUDA cores running on a 722MHz clock.
A 670 has 1344 CUDA cores running on a 915MHz clock.
A 690 has 3072 CUDA cores running on a 915MHz clock. (2 GPUs = 1536 CUDA cores each)
A 780 has 2304 CUDA cores running on a 863MHz clock.
A TITAN has 2688 CUDA cores running on a 837MHz clock.
A K5000 has 1536 CUDA cores running on a [who knows?]MHz clock.

Which do you think will run faster when rendering using Octane or Thea?

Note that there are other factors, such as total RAM on each card and memory interface width, that play into the card's performance for a given task. The TITAN's 6GB, as opposed to the K5000's 4GB, will make a difference.

Waves of light
02-18-2014, 04:25 AM
OK, I'll try to explain this another way, just to make sure we're on the same page.


580/590 cards are not crippled, so perform generally the same as a Quadro in viewport manipulation—only.
GPU renderers use CUDA cores. The more, the better.
A 580 has 512 CUDA cores running on a 722MHz clock.
A 670 has 1344 CUDA cores running on a 915MHz clock.
A 690 has 3072 CUDA cores running on a 915MHz clock. (2 GPUs = 1536 CUDA cores each)
A 780 has 2304 CUDA cores running on a 863MHz clock.
A TITAN has 2688 CUDA cores running on a 837MHz clock.
A K5000 has 1536 CUDA cores running on a [who knows?]MHz clock.

Which do you think will run faster when rendering using Octane or Thea?

Note that there are other factors, such as total RAM on each card and memory interface width, that play into the card's performance for a given task. The TITAN's 6GB, as opposed to the K5000's 4GB, will make a difference.

That, right there, should be a sticky.

yoshiii
02-18-2014, 05:54 AM
OK, I'll try to explain this another way, just to make sure we're on the same page.


580/590 cards are not crippled, so perform generally the same as a Quadro in viewport manipulation—only.
GPU renderers use CUDA cores. The more, the better.
A 580 has 512 CUDA cores running on a 722MHz clock.
A 670 has 1344 CUDA cores running on a 915MHz clock.
A 690 has 3072 CUDA cores running on a 915MHz clock. (2 GPUs = 1536 CUDA cores each)
A 780 has 2304 CUDA cores running on a 863MHz clock.
A TITAN has 2688 CUDA cores running on a 837MHz clock.
A K5000 has 1536 CUDA cores running on a [who knows?]MHz clock.

Which do you think will run faster when rendering using Octane or Thea?

Note that there are other factors, such as total RAM on each card and memory interface width, that play into the card's performance for a given task. The TITAN's 6GB, as opposed to the K5000's 4GB, will make a difference.

Thank you

I understand now.

Can two cards be used in the same machine?

Waves of light
02-18-2014, 06:00 AM
Thank you

I understand now.

Can two cards be used in the same machine?

Yes, as long as you have a decent PSU. Be careful, some of the GTX 580 (Asus brand) require 2x8 pin PCIe cables from your PSU. So if you had two installed you would need 4x8 pin PCIe cables.

yoshiii
02-18-2014, 06:08 AM
Yes, as long as you have a decent PSU. Be careful, some of the GTX 580 (Asus brand) require 2x8 pin PCIe cables from your PSU. So if you had two installed you would need 4x8 pin PCIe cables.

I mean can I have two different cards in one machine?

Waves of light
02-18-2014, 06:10 AM
As far as I'm aware, yes. But again, make sure your PSU and motherboard can handle it.

Danner
02-18-2014, 06:17 AM
The PCgamer.com staff built one with 4 Titans http://www.pcgamer.com/2013/11/18/introducing-the-large-pixel-collider-a-computer-of-uncomfortable-power/
That'd be a sweet Octane machine.

yoshiii
02-18-2014, 06:42 AM
As far as I'm aware, yes. But again, make sure your PSU and motherboard can handle it.

good

Do you know how I would set the drivers for that? or set which card to use for what?

spherical
02-18-2014, 03:10 PM
Depends upon the cards and what you are using them for. If you are going to use a GPU renderer, you can dedicate one (or more) cards to it and deselect another card that will be used only for screen display. So, you get a 560/580/590 or low end Quadro and run the displays with it and use the 6xx/7xx/TITAN(s) for rendering. The renderers do not need SLI in order to use multiple cards.

yoshiii
02-18-2014, 06:15 PM
Depends upon the cards and what you are using them for. If you are going to use a GPU renderer, you can dedicate one (or more) cards to it and deselect another card that will be used only for screen display. So, you get a 560/580/590 or low end Quadro and run the displays with it and use the 6xx/7xx/TITAN(s) for rendering. The renderers do not need SLI in order to use multiple cards.

Thats good.

I am using Windows 8, do you know how I would select which cards to use after installation of the cards?