Having 2 Gtx cards and 5820k CPU

[UnrealMan]

I'm planing to buy Gtxtitan X I already have Titan and plan to use then in gpu based render
I knew that 5820k 28 lane CPU cuts second cards performance
so here is my question I will put Titan X in 1 slot and it will compute just fine
but when my pc will use Titan from second slot
will it cut the cuda and memory performance of second card ?
thanks

 

[not you]

I dont think you can use the titan x and a non x titan in the same time for the same application

 

[RobCrezz]

28 lanes is plenty for two or three cards. Many people use dual cards on 16 lane systems.

Yes you can use both of rendering (not SLI), many rendering programs will allow you to use different GPUs.

 

[UnrealMan]

28 lanes is plenty for two or three cards. Many people use dual cards on 16 lane systems.

Yes you can use both of rendering (not SLI), many rendering programs will allow you to use different GPUs.

thanks you for response
How much lanes is used when u use 1 gpu like titan X ?
if you say 28 is planty

 

[RobCrezz]

28 lanes is plenty for two or three cards. Many people use dual cards on 16 lane systems.

Yes you can use both of rendering (not SLI), many rendering programs will allow you to use different GPUs.

thanks you for response
How much lanes is used when u use 1 gpu like titan X ?
if you say 28 is planty

8 lanes @ pci-e 3.0 is sufficent. A single card will run at 16 lanes, but it doesn't give any more performance.

 

[UnrealMan]

28 lanes is plenty for two or three cards. Many people use dual cards on 16 lane systems.

Yes you can use both of rendering (not SLI), many rendering programs will allow you to use different GPUs.

thanks you for response
How much lanes is used when u use 1 gpu like titan X ?
if you say 28 is planty

8 lanes @ pci-e 3.0 is sufficent. A single card will run at 16 lanes, but it doesn't give any more performance.

then I assume lanes should be somehow connected to bandwidth right ? if it is can you please explain how lane works
or maybe 1 lane = something or how to count them
thanks

 

[RobCrezz]

Exactly, on the pci-e 3.0 bus it is 985 MB/s per lane. Graphics cards dont tend to need more than 8x due to having fast vram on the card which is where the GPU has very high bandwidth to.

 

[UnrealMan]

Exactly, on the pci-e 3.0 bus it is 985 MB/s per lane. Graphics cards dont tend to need more than 8x due to having fast vram on the card which is where the GPU has very high bandwidth to.

So then if GTX titan has 288.4Memory Bandwidth (GB/sec)
GB/sec so we calculate 8 lanes 8 * 985 = 7880 7GB almost 8GB/sec
its not enough right ?
or I miss something ?

 

[RobCrezz]

Exactly, on the pci-e 3.0 bus it is 985 MB/s per lane. Graphics cards dont tend to need more than 8x due to having fast vram on the card which is where the GPU has very high bandwidth to.

So then if GTX titan has 288.4Memory Bandwidth (GB/sec)
GB/sec so we calculate 8 lanes 8 * 985 = 7880 7GB almost 8GB/sec
its not enough right ?
or I miss something ?

Yes, you completely misunderstand.

As i said, the reason the gpu has high bandwith is because this is to its own VRAM on the graphics card - This doesnt use the PCI-E bus. The graphics card has its own memory bus and own memory (vram), because even the fastest PCI-E setup isnt fast enough. This is the reason why graphics cards have their own memory.