GPU memory
Hello,
Is there a link between the "Memory global" of the GPU and the maximum "Allocated mermory" ?
I am surprised that, with 2048MB of memory, a simulation with 294MB allocated memory in GPU doesn't work.
As you can see in the following examples, the computation works with 127.09MB (Case 1), but not with 294.78MB (Case 2).
Case 1 :
RunMode="Single-Gpu"
Allocated memory in CPU: 34324968 (32.73 MB)
Allocated memory in GPU: 133263588 (127.09 MB)
Part_0000 1007672 particles successfully stored
[Initialising simulation (rd0fhfr8) 06-07-2015 14:26:53]
PART PartTime TotalSteps Steps Time/Seg Finish time
========= ============ ============ ======= ========= ===================
Part_0001 0.000053 1 1 21898.57 06-07-2015 14:26:54
Case 2 :
RunMode="Single-Gpu"
Allocated memory in CPU: 79557378 (75.87 MB)
Allocated memory in GPU: 309099468 (294.78 MB)
Part_0000 2338037 particles successfully stored
[Initialising simulation (rm9dmdmt) 06-07-2015 14:28:17]
PART PartTime TotalSteps Steps Time/Seg Finish time
========= ============ ============ ======= ========= ===================
*** bad allocation
Here are the caracteristics of the GPU card :
[Select CUDA Device]
Device 0: "Quadro 1000M"
Compute capbility: 2.1
Multiprocessors: 2 (96 cores)
Memory global: 2048 MB
Clock rate: 1.40 GHz
Run time limit on kernels: Yes
ECC support enabled: No
Have anyone an explanation ?
Thank you in advance.
Regards
Is there a link between the "Memory global" of the GPU and the maximum "Allocated mermory" ?
I am surprised that, with 2048MB of memory, a simulation with 294MB allocated memory in GPU doesn't work.
As you can see in the following examples, the computation works with 127.09MB (Case 1), but not with 294.78MB (Case 2).
Case 1 :
RunMode="Single-Gpu"
Allocated memory in CPU: 34324968 (32.73 MB)
Allocated memory in GPU: 133263588 (127.09 MB)
Part_0000 1007672 particles successfully stored
[Initialising simulation (rd0fhfr8) 06-07-2015 14:26:53]
PART PartTime TotalSteps Steps Time/Seg Finish time
========= ============ ============ ======= ========= ===================
Part_0001 0.000053 1 1 21898.57 06-07-2015 14:26:54
Case 2 :
RunMode="Single-Gpu"
Allocated memory in CPU: 79557378 (75.87 MB)
Allocated memory in GPU: 309099468 (294.78 MB)
Part_0000 2338037 particles successfully stored
[Initialising simulation (rm9dmdmt) 06-07-2015 14:28:17]
PART PartTime TotalSteps Steps Time/Seg Finish time
========= ============ ============ ======= ========= ===================
*** bad allocation
Here are the caracteristics of the GPU card :
[Select CUDA Device]
Device 0: "Quadro 1000M"
Compute capbility: 2.1
Multiprocessors: 2 (96 cores)
Memory global: 2048 MB
Clock rate: 1.40 GHz
Run time limit on kernels: Yes
ECC support enabled: No
Have anyone an explanation ?
Thank you in advance.
Regards
Comments
I am still wondering why I can't allocate more than 200MB with my 2048MB graphic card.
Does anyone have an idea ?
Regards.
Maybe the number of cells in which the domain is getting divided is too high and causes an allocation error?
EDIT: Sorry, my mistake... The memory for cells is already included in the value you posted...
Is it you display card (i.e. the one displaying to your monitor?)
I bet it is since it is a ***M card
and the rest of your memory is used by your system...
Cheers,
George