I have a Gigabyte MZ31-AR0 motherboard with 2 Tesla K80 cards connected at PCIE-Express 3.0 x16 (slots 1 & 6) and when i run a .simpleP2P test within one Tesla K80 card, i get a bandwidth (7 GB/s) of nearly half of what i expected (16 GB/s). ACS control is disabled, as well as IOMMU. If i run across different K80 cards, the bandwidth gets even less at 3 GB/s. The APM EPYC 7251 has 4 numa nodes and the GPU topology given are: PIX and SYS. Is that normal to have 50% bandwidth reduction within the same card and 25% reduction across NUMA nodes?? Any help would be highly appreciated.
Related topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
K80 peer-to-peer transfers: Slow bandwidth and high latency. | 7 | 3691 | August 31, 2016 | |
Low host<->device bandwidth for one of two cards | 1 | 6481 | March 22, 2011 | |
P2P peer communication is slower than the bandwidth between GPU and CPU | 0 | 3575 | June 5, 2011 | |
TESLA bandwidthTest results | 5 | 2891 | January 19, 2010 | |
Tesla Bandwidth Question | 3 | 4999 | January 22, 2008 | |
memory bandwidth of M2090 less than GTX 580? | 2 | 2517 | May 8, 2012 | |
Installing driver for Tesla K80 | 3 | 3096 | July 20, 2019 | |
Bandwidthtest on Tesla S1070 | 0 | 699 | February 16, 2011 | |
Basic question about 2-in-1 GPUs (ie GTX Titan Z or K80) | 2 | 1774 | November 28, 2014 | |
very low bandwidth bandwidth problem with Tesla C870 | 5 | 3746 | January 29, 2008 |