Vista only recognizes 1GB for my HD4870X2

CoolGamer48

Distinguished
Aug 20, 2008
59
0
18,630
I recently built my own machine with a Radeon HD 4870 X2. It's supposed to come with 2GB of DDR5 video memory. When I look at the card's info in Vista x64, dedicated video memory is only listed as 1024MB (total available graphics memory: 2815MB, system video memory: 0MB, and shared system memory: 1791MB). I've installed the drivers that came with the card.

Anyone know a way to fix this?

Multas gratias.
 

emp

Distinguished
Dec 15, 2004
2,593
0
20,780
each GPU has access to 1024MB of vRAM, which adds up to a total of 2GB of vRAM onboard, it's a marketing tactic used by both, AMD and nvidia, on their dual GPU solutions. Basically the card does have 2GB of memory on it, but only 1GB is accessible to each GPU.
 

SirCrono

Distinguished
Sep 9, 2006
463
3
18,785
Basically the problem is that both GPU's need to copy the textures in their own memory, so physically there are 2 GB of ram, but each texture is copied twice making 1 GB effective for each GPU.

Its the same emp said, but I think it's easier to understand this way.

The short answer is, it's normal, marketing department screw you over.
 

CoolGamer48

Distinguished
Aug 20, 2008
59
0
18,630
Why does one texture need to be copied into vRAM twice? How exactly do the two GPUs deal with work? They both work on the same thing? (sorry, im a bit of a noob at all of this).
 

dagger

Splendid
Mar 23, 2008
5,624
0
25,780


Basically, they each work on half of the same thing, but stores the full picture. Frame buffer. :p
 

CoolGamer48

Distinguished
Aug 20, 2008
59
0
18,630
Ahh, ok. So you're going to have 1GB of vRAM with a set of textures, and another set of vRAM with the exact same textures, just for a different gpu. So you cards with dual gpus need to have their memory amount cut in half if you were to compare it to card with a single gpu (obviously 2 gpus are better than 1, but as far as strictly comparing memory...)
 

SirCrono

Distinguished
Sep 9, 2006
463
3
18,785


Exactly, this has to be done because if one GPU was to read from the other's vRAM it would be SLOOOOOW.

We don't want slow do we?

And yes, 2 GPUs are usually better than one, but there are some rather funny cases where a single GPU beats two of the same due to drivers issues.

Back to the topic, yes, memory has to be cut in half.