GPU Memory Interface vs. Memory Bandwidth

[jshhndsd]

Hi Guys,

 

Can you explain the difference between the GPU memory interface and it's memory bandwidth as well as stream processors and TFLOPS? Can someone also explain how to setup or access  DirectGMA  for AMD graphics cards?

 

 

Thanks in advance.

[Electronics Wizardy]

3 minutes ago, jshhndsd said:

Can someone also explain how to setup or access  DirectGMA  for AMD graphics cards?

Do you have a firepro?

 

3 minutes ago, jshhndsd said:

GPU memory interface and it's memory bandwidth

bandwidth is how much data it can carry, interface is how its connected.

 

4 minutes ago, jshhndsd said:

well as stream processors and TFLOPS?

TFLOPS is how many floating point in can do in a second. Stream units are like cores

 

[typographie]

Memory bandwidth is the overall measure of how quickly data is read from or written to memory, usually expressed in bytes per second. Both the memory frequency (in Hz) and the memory bus width contribute to memory bandwidth.

 

"Stream processors" are the GPU's cores. Nvidia refers to them as "CUDA cores" and AMD uses the stream processor nomenclature. The architectures differ but basically that's what they are.

 

FLOPS is a measure of floating point operations (that is, mathematical calculations using decimal numbers) per second. The tera- prefix means trillion. So TFLOPS is how many trillion floating point calculations the GPU can do per second.

[jshhndsd]

14 minutes ago, Electronics Wizardy said:

Do you have a firepro?

 

bandwidth is how much data it can carry, interface is how its connected.

 

TFLOPS is how many floating point in can do in a second. Stream units are like cores

 

Hi, thanks for your reply. Yes to FirePro