New blog post on maximizing GROMACS throughput on GPUs

User discussions
,
1

Dear GROMACS users,

To address a frequently asked question, Szilárd Páll and I have just published a blog article that shows how to maximize GROMACS throughput on GPUs when running multiple simulations. It shows that, when running many small simulations, quite dramatic throughput improvements can be achieved by running multiple simulations per GPU, provided you use the NVIDIA MPS and/or MIG facilities. It gives detailed instructions on how to do this. I hope it is useful for some of you; the link is here:

Best regards,

Alan Gray

1 Like
2

I tried running this with a few V100 GPUs on LSF Spectrum but GROMACs is running slow on all simulations. I tried 2 GPUs with 4 simulations each.

$ lscpu
Architecture:        x86_64
CPU op-mode(s):      32-bit, 64-bit
Byte Order:          Little Endian
Address sizes:       46 bits physical, 48 bits virtual
CPU(s):              80
On-line CPU(s) list: 0-79
Thread(s) per core:  2
Core(s) per socket:  20
Socket(s):           2
NUMA node(s):        2
Vendor ID:           GenuineIntel
CPU family:          6
Model:               85
Model name:          Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping:            4
CPU MHz:             2835.937
CPU max MHz:         3700.0000
CPU min MHz:         1000.0000
BogoMIPS:            4800.00
Virtualization:      VT-x
L1d cache:           1.3 MiB
L1i cache:           1.3 MiB
L2 cache:            40 MiB
L3 cache:            55 MiB
NUMA node0 CPU(s):   0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s):   1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp l
                     m constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm
                     2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch e
                     pb cat_l3 cdp_l3 invpcid_single intel_ppin intel_pt ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx
                     2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 cqm_
                     llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke md_clear spec_ctrl intel_stibp flush_l1d arch_capabilities

$ nvidia-smi -L
GPU 0: Tesla V100-SXM2-32GB (UUID: GPU-25d6b76e-b656-2e2c-26e4-6e818175cd9d)
GPU 1: Tesla V100-SXM2-32GB (UUID: GPU-d4c1bbd2-c164-8759-b433-6845a5a5f6d6)
3

Hi,

Sometimes there can be conflicts when using a queuing system that allows node sharing, please see the discussion at Maximizing GROMACS Throughput with Multiple Simulations per GPU Using MPS and MIG - #7 by akshaychenna - Technical Blog - NVIDIA Developer Forums

I suggest to try the recommendations mentioned there and/or request exclusive use of the node.

If you still have problems, please provide the md.log file for simulation running in isolation, and the equivalent when running in conjunction with other simulations, and I can take a look.

Best regards,

Alan