Aviation FAN :
RobCrezz :
Aviation FAN :
RobCrezz :
Aviation FAN :
RobCrezz :
Because the bandwidth they are talking about on the graphics card is between the VRAM and the GPU - this is all on the card so the speed is based on the vram bus width and the Vram speed.
The graphics card doesnt need to communicate as quickly with the rest of the system through the PCI-E Bus, hence it not being a bottleneck in performance.
So it's the VRAM bandwidth? If so, I've never seen the mention of the VRAM bandwidth... So what would be the VRAM bandwidth of the GTX 750?
vram or memory bandwidth (same thing) is 80GB/s
http://www.geforce.co.uk/hardware/desktop-gpus/geforce-gtx-750/specifications
.......If it is 80GB/s and my motherboard is PCI-e 2.0 @ x16 which is total 16GB/s why isn't it bottlenecking? Like say that PCI-E 3.0 is double pcie 2.0 which has max bandwidth of 32gb/s but high end cards have like 300+gb/s. If PCI-E have such limited bandwidth, why isn't there any bottlenecking?
As I said in my first post, the high bandwidth is required between the GPU and the memory for the GPU (the vram). This is all on the graphics card, so its possible for it to have high bandwidth between the two (due to graphics cards having much faster ram than system ram and much wider bus). This is where the bandwidth is needed as the graphics card stores all the information it needs to access very quickly here.
The max bandwidth of PCI-E isnt a bottleneck as the graphics card doesn't need as much bandwidth to communicate with the other pc components.
So how do you calculate how much bandwidth IS required to communicate with the other pc components?
For gaming, its not a big deal as everything needed is usually loaded into the graphics cards fast vram during the loading process, so its available at high speed.
Its generally accepted that PCI-E 2.0 8x is the minimum required before any noticable performance decrease. PCI-E 2.0 is 500mb per lane, so thats 4Gb/s