2 Gpu=More vRam?

[Werdna]

So if i put 2 gtx 980 4gb, will it mean i will have 8gb vram?

[Fate]

No

[Juganog]

nope

[StarFire]

No, you're limited to the ram of 1, as they each handle different frames, rather than the same at once.

[Interfectorem]

Nope, VRAM does not stack or add up in SLI or Crossfire configurations.

[Guest Philosophy]

No sir!

[soaringchicken]

Nope

[Enderman]

Yes. You have 8GB.

BUT they contain the exact same data.

That means that you only have 4GB of usable space, because everything is duplicated for each GPU.

[thekeemo]

can anybody explain why they dont add up why cant the data be stored on just one of them

[MightyUnit]

Yes. You have 8GB.

BUT they contain the exact same data.

That means that you only have 4GB of usable space, because everything is duplicated for each GPU.

 SLI or CFX uses AFR, Alternate Frame Rendering. GPU0 renders even frames and GPU1 renders odd. Each GPU temporarily stores the frame that is renders in its own memory before it heads off to your display. So no, you are not correct, the OP only has 4GB of ram available to store each frame. 

[Enderman]

 SLI or CFX uses AFR, Alternate Frame Rendering. GPU0 renders even frames and GPU1 renders odd. Each GPU temporarily stores the frame that is renders in its own memory before it heads off to your display. So no, you are not correct, the OP only has 4GB of ram available to store each frame. 

If that was true then you would have 8GB of usable memory buy putting two 4GB cards in SLI/CFX.

What each GPU does is it loads ALL the data into BOTH the GPUs vram, then each GPU renders an alternate frame.

Both GPUs have the same data stored in their vram at any given time. They just take turns rendering each frame.

 

It would be nice if the world worked the way you think it does, with each GPU loading only the data it needs into its vram. Then we would be able to have 24GB of usable vram by using 4 titans. Sadly this is not how the world works, so no, you are wrong.

[MightyUnit]

If that was true then you would have 8GB of usable memory buy putting two 4GB cards in SLI/CFX.

What each GPU does is it loads ALL the data into BOTH the GPUs vram, then each GPU renders an alternate frame.

Both GPUs have the same data stored in their vram at any given time. They just take turns rendering each frame.

 

It would be nice if the world worked the way you think it does, with each GPU loading only the data it needs into its vram. Then we would be able to have 24GB of usable vram by using 4 titans. Sadly this is not how the world works, so no, you are wrong.

 

Please explain further and if possible link me to source of your info. I ask  because when I read the SLI FAQ found here: http://www.geforce.com/hardware/technology/sli/faq#c26 it says "No. In SLI mode, each graphics card uses its own frame buffer memory to render a 3D application. The operating system will report a graphics card frame buffer memory size that is found on a single graphics board".

 

Also I do not agree that my understanding of AFR would increase vram as the number of GPUs increase. It does not make sense if you read what I wrote. Each card renders a frame and drops it in its own vram. so 4GB for frame 0. Then GPU1 renders frame 1 and puts it in its 4GB frame buffer.... only 4GB for each frame.

 

EDIT: Now reading the SLI White paper... I have just started but I thought say that so far I am wrong about this. Sorry for the mis info!!!!!

[MightyUnit]

Ok. Here it is from the horses mouth:

 

"In all SLI-rendering modes, local memory is duplicated across all GPUs. This means that on 

an SLI system with two 256MB video cards, there is still only 256MB of video memory 

available to applications. Additionally, any change to local memory on one GPU (for 

example, dynamic texture updates) will often require a data broadcast to other GPUs."

 

http://http.download.nvidia.com/developer/SDK/Individual_Samples/DEMOS/Direct3D9/src/slibestpract/docs/slibestpract.pdf

[Enderman]

Please explain further and if possible link me to source of your info. I ask  because when I read the SLI FAQ found here: http://www.geforce.com/hardware/technology/sli/faq#c26 it says "No. In SLI mode, each graphics card uses its own frame buffer memory to render a 3D application. The operating system will report a graphics card frame buffer memory size that is found on a single graphics board".

 

Also I do not agree that my understanding of AFR would increase vram as the number of GPUs increase. It does not make sense if you read what I wrote. Each card renders a frame and drops it in its own vram. so 4GB for frame 0. Then GPU1 renders frame 1 and puts it in its 4GB frame buffer.... only 4GB for each frame.

 

http://www.tomshardware.com/forum/339028-33-does-crossfire-double-dedicated-memory

 

The data is mirrored by both cards so two 1GB cards will still result in 1GB of VRAM being available.

 

http://www.techpowerup.com/forums/threads/why-cant-crossfire-or-sli-use-the-sum-of-vram-from-all-gpus.167301/

 

SLI and CrossfireX usually use AFR (alternate frame rendering) which makes your gpus draw frames in turns, let's say one draws all the odd frames and the other all the even frames.

 

To be able to do that you must have a copy of everything readily available to the GPUs in questions, for a GPU to access VRAM that does not directly belong to itself it would be very hard to implement software wise and ESPECIALLY bandwidth wise, you would need a very very fast link to provide textures that are only available on another GPU VRAM on that system.

 

http://neogaf.com/forum/showthread.php?t=592231

 

in SLI/Crossfire the ram is mirrored, not doubled. So with 2 3gb cards in crossfire you still have 3gb for games.

[Enderman]

 

Ok. Here it is from the horses mouth:

 

"In all SLI-rendering modes, local memory is duplicated across all GPUs. This means that on 

an SLI system with two 256MB video cards, there is still only 256MB of video memory 

available to applications. Additionally, any change to local memory on one GPU (for 

example, dynamic texture updates) will often require a data broadcast to other GPUs."

 

http://http.download.nvidia.com/developer/SDK/Individual_Samples/DEMOS/Direct3D9/src/slibestpract/docs/slibestpract.pdf

 

I see you have enlightened yourself before I finished writing my post, lol

[MightyUnit]

Ya I HATE getting it wrong and desperately wanted to know what was correct.