Difference between revisions of "Nexus/Tron"

From UMIACS
Jump to navigation Jump to search
 
(9 intermediate revisions by the same user not shown)
Line 2: Line 2:
  
 
= Hardware =
 
= Hardware =
Not all hardware is available yet.  The full configuration, when finally received and powered on, will include 62 nodes with a total of 48 A6000, 128 A5000 and 160 A4000 Nvidia GPUs.  This computationally provides 2.5 million CUDA cores, 8TB of GPU memory, and more than 70,000 TensorFLOPS of aggregate performance.  In addition to the GPUs, this resource provides 1,344 CPU cores and 10.7 TeraBytes of main memory across the nodes.
+
The full configuration includes 62 nodes with a total of 48 A6000, 128 A5000, and 160 A4000 [https://www.nvidia.com/en-us/design-visualization/desktop-graphics/ NVIDIA RTX A-Series GPUs].  This computationally provides 2.5 million CUDA cores, 8TB of GPU memory, and more than 70,000 TensorFLOPS of aggregate performance.  In addition to the GPUs, this computationally provides 1,344 CPU cores and 10.7TB of main memory across the nodes.
  
{| class="wikitable"
+
{| class="wikitable sortable"
! style="text-align:center;"| Quantity
+
! Nodenames
 
! Type
 
! Type
! CPUs
+
! Quantity
! Memory
+
! CPUs per node
! GPUs
+
! Memory per node
! Nodes
+
! GPUs per node
 
|-
 
|-
 +
|tron[00-05]
 +
|A6000 GPU Node
 
|6
 
|6
|A6000 GPU Node
 
 
|32
 
|32
 
|256GB
 
|256GB
 
|8
 
|8
|tron[00-05]
 
 
|-
 
|-
 +
|tron[06-45]
 +
|A4000 GPU Node
 +
|40
 
|16
 
|16
 +
|128GB
 +
|4
 +
|-
 +
|tron[46-61]
 
|A5000 GPU Node
 
|A5000 GPU Node
 +
|16
 
|32
 
|32
 
|256GB
 
|256GB
 
|8
 
|8
|
+
|- class="sortbottom"
|-
+
|tron[00-61]
|40
 
|A4000 GPU Node
 
|16
 
|128GB
 
|4
 
|tron[06-21]
 
|-
 
 
!Total
 
!Total
 
|62
 
|62
Line 38: Line 39:
 
|10.7 TB
 
|10.7 TB
 
|336
 
|336
|tron[00-21]
 
 
|}
 
|}

Latest revision as of 16:19, 7 July 2022

The Tron partition is a subset of resources available in the Nexus.

Hardware

The full configuration includes 62 nodes with a total of 48 A6000, 128 A5000, and 160 A4000 NVIDIA RTX A-Series GPUs. This computationally provides 2.5 million CUDA cores, 8TB of GPU memory, and more than 70,000 TensorFLOPS of aggregate performance. In addition to the GPUs, this computationally provides 1,344 CPU cores and 10.7TB of main memory across the nodes.

Nodenames Type Quantity CPUs per node Memory per node GPUs per node
tron[00-05] A6000 GPU Node 6 32 256GB 8
tron[06-45] A4000 GPU Node 40 16 128GB 4
tron[46-61] A5000 GPU Node 16 32 256GB 8
tron[00-61] Total 62 1,344 10.7 TB 336