A Performance/Cost Evaluation for a GPU-Based Drug Discovery Application on Volunteer Computing

Table 1

Hardware features for our local test-bed infrastructure.

Intel system
Processor	Intel Xeon E5620 @ 2.4 GHz
GPU 0	Nvidia 7300GT
Memory	16 GB DDR3 @ 1333 MHz
Maximum power draw	80 W
Experimental idle power	38 W
GPU 1: Nvidia GTX 465
GPU family	GF100
Manufacturing process	40 nm
Core clock	607 MHz
Memory size	1024 MB
Memory clock	2 × 1603 MHz
Memory bus width	256 bits
Memory bandwidth	102.6 GB/sec
Stream processors	352
Maximum power draw	200 W
Experimental idle power	24 W
GPU 2: Nvidia GTX 480
GPU family	GF100
Manufacturing process	40 nm.
Core clock	700 MHz
Memory size	1536 MB
Memory clock	2 × 1848 MHz
Memory bus width	384 bits
Memory bandwidth	177.4 GB/sec
Stream processors	480
Maximum power draw	250 W
Experimental idle power	37 W
GPU 3: Nvidia Tesla C2070
GPU family	GF100
Process	40 nm.
Core clock	573.5 MHz
Memory size	6143 MB
Memory clock	2 × 1494 MHz
Memory bus width	384 bits
Memory bandwidth	143.4 GB/sec
Stream processors	448
Maximum power draw	247 W
Experimental idle power	107 W
GPU 4: Nvidia GTX 590
GPU family	GF100
Manufacturing process	40 nm.
Core clock	1215 MHz
Memory size	2 × 1536 MB
Memory clock	2 × 1707 MHz
Memory bus width	2 × 384 bits
Memory bandwidth	2 × 327.7 GB/sec
Stream processors	1024
Maximum power draw	365 W
Experimental idle power	140 W
GPU 5: Nvidia Tesla K20c
GPU family	GK110
Manufacturing process	28 nm.
Core clock	705 MHz
Memory size	5120 MB
Memory clock	2 × 2600 MHz
Memory bus width	320 bits
Memory bandwidth:	208 GB/sec
Stream processors	2496
Maximum power draw	225 W
Experimental idle power	27 W