System Details
College of Science and Engineering
Compute Nodes
Compute (8x)
-
CPU: 2x Intel Xeon E5-2620
2GHz, 2.5Ghz Max Turbo
6-Core, HT
15MB Cache
32nm
95W
-
RAM: 64GB
8 x 8GB DDR3-1600 ECC Registered 2R DIMMs
Operating at 1600 MT/s Max
Maximum 256GB (8 x 32GB)
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x IPMI Management
2.0 & KVM over with Dedicated LAN - Integrated
-
-
Storage
-
System (Hot Swap)
1 x 1TB Seagate Constellation ES (6Gb/s, 7.2K RPM, 64MB Cache) 3.5” SATA
-
-
Priority
4 are open / none
4 are Kodner research group
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 24 cpus, ~64GB memory
ParallelSchedulingGroup: “no_ib”
Compute (8x)
-
CPU: 2x Intel Xeon E5-2620
2GHz, 2.5Ghz Max Turbo
6-Core, HT
15MB Cache
32nm
95W
-
RAM: 64GB
8 x 8GB DDR3-1600 ECC Registered 2R DIMMs
Operating at 1600 MT/s Max
Maximum 256GB (8 x 32GB)
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x IPMI Management
2.0 & KVM over with Dedicated LAN - Integrated
-
1x InfiniBand
ConnectX-3 QDR 40Gb/s Controller with QSFP Connector
-
-
Storage
-
System (Hot Swap)
1 x 1TB Seagate Constellation ES (6Gb/s, 7.2K RPM, 64MB Cache) 3.5” SATA
-
-
Notes
Same as above, but with added 40Gb/s InfiniBand
-
Priority
Parallel jobs (but open to all groups)
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 24 cpus, ~64GB memory
ParallelSchedulingGroup: “ib40_1”
Compute (4x)
-
CPU: 2x Intel Xeon E5-2620 v4
2.1GHz Base, 3.0Ghz Max Turbo
8-Core, HT
20MB Cache
14nm
85W
-
RAM: 128GB
8 x 16GB DDR4-2400 ECC Registered 2R DIMMs
Operating at 2400 MT/s Max
Maximum 1TB (8 x 256GB)
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x IPMI Management
2.0 & KVM over with Dedicated LAN - Integrated
-
1x Infiniband
ConnectX-3 QDR 40Gb/s Controller with QSFP Connector
-
-
Storage
-
System (Hot Swap)
1 x 400GB Intel SSD SATA 6Gb/s 2.5”
-
-
Priority
Berger research group
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 32 cpus, ~128GB memory
ParallelSchedulingGroup: “ib40_1”
Graphics Nodes
GPU (2x)
-
CPU: 2x Intel Xeon E5-2620
2GHz, 2.5Ghz Max Turbo
6-Core, HT
15MB Cache
32nm
95W
-
RAM: 128GB
8 x 16GB DDR3-1600 ECC Registered 2R DIMMs
Operating at 1600 MT/s Max
Maximum 128GB (8 x 16GB)
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x InfiniBand
Mellanox ConnectX-2 VPI Single-Port QDR 40Gb/s InfiniBand HCA, PCIe 2.0 x8
-
-
Storage
-
System (Hot Swap)
1 x 1TB Seagate Constellation ES (6Gb/s, 7.2K RPM, 64MB Cache) 3.5” SATA
-
-
GPU: 3x Tesla M2075 GPUs
Architecture: Fermi
CUDA Cores: 448
Memory: 6 GB GDDR5
Clock Speed: 574 Mhz
-
Priority
Open / None
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 24 cpus, ~128GB memory
ParallelSchedulingGroup: “ib40_1”
GPU (1x)
-
CPU: 1x AMD EPYC™ 7742
2.25GHz, 3.4GHz Max Boost Clock
64-Core, 128 Threads
256MB L3 Cache
14nm
225W
-
RAM: 512GB
8 x 64GB SAMSUNG 3200MHZ ECC
Operating at 1600 MT/s Max
Maximum 1TB (8 x 128GB)
-
Networking
-
2x 10GBase-T Ethernet
Dual 10GBase-T LAN with via Broadcom BCM57416
Operating in 1 Gigabit mode
-
1x InfiniBand
Mellanox ConnectX-5 VPI Single-Port EDR 100Gb/s InfiniBand QSFP28, PCIe 3.0 x16
-
-
Storage
-
System (Hot Swap)
1 x 960GB Intel S4610 (6Gb/s, 3 DWPD) 2.5” SATA
-
-
GPU: 2x NVIDIA Tesla T4 GPU
Architecture: Turing
CUDA Cores: 2,560
Tensor Cores: 320
Memory: 16 GB GDDR6
Clock Speed: 585 MHz Base, 1590 MHz Max Boost
Bus: PCIe 3.0 x16
-
Performance
-
Single Precision Performance (FP32)
8.1 TFLOPS
-
Mixed Precision (FP16/FP32)
65 FP16 TFLOPS
-
INT8 Precision
130 INT8 TOPS
-
INT4 Precision
260 INT4 TOPS
-
-
Priority
Berger research group
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 128 cpus, ~512GB memory
ParallelSchedulingGroup: “ib40_1”
-
Required extra settings:
+WantPreempt=True if not in Berger research group
GPU (1x)
-
CPU: 2x AMD EPYC™ 7543 32-Core 2.80GHz
2.80GHz, 3.7GHz Max Boost Clock
32-Core, 64 Threads
256MB L3 Cache
7nm+
225W
-
RAM: 512GB
16 x 32GB SAMSUNG 3200MHZ ECC
Maximum 2TB (16 x 128GB)
-
Networking
-
2x 10GBase-T Ethernet
Dual 10GBase-T LAN with via Intel® X550-AT2 controller
Operating in 1 Gigabit mode
-
1x InfiniBand
Mellanox ConnectX-5 VPI Single-Port EDR 100Gb/s InfiniBand QSFP28, PCIe 3.0 x16
-
-
Storage
-
System (Hot Swap)
1 x 960GB Intel S4610 (6Gb/s, 3 DWPD) 2.5” SATA
-
Scratch space
1 x 960GB Intel S4610 (6Gb/s, 3 DWPD) 2.5” SATA
-
-
GPU: 1x NVIDIA A100 GPU
Architecture: Ampere
CUDA Cores: 6,912
Tensor Cores: 432
Memory: 40 GB HBM2
Clock Speed: 765 MHz Base, 1410 MHz Max Boost
Bus: PCIe 4.0 x16
Multi-Instance GPU: Up to 7 MIGs @ 5 GB
Performance
-
Double Precision Performance (FP64)
9.7 TFLOPS
-
Double Precision Tensor Core Performance (FP64)
19.5 TFLOPS
-
Single Precision Performance (FP32)
19.5 TFLOPS
-
Tensor Float 32 (TF32)
156 TFLOPS
-
BFLOAT16 Tensor Core
312 TFLOPS
-
FP16 Tensor Core
312 TFLOPS
-
INT8 Tensor Core
624 TOPS
-
Priority
Ramasubramanian research group
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 64 cpus, ~512GB memory
ParallelSchedulingGroup: “no_ib”
-
Required extra settings:
+WantPreempt=True if not in Ramasubramanian research group
Support Systems & Miscellaneous
Head Node (x1)
-
CPU: 2 x Intel Xeon E5-2620
2GHz
6-Core, HT
15MB Cache
32nm
95W
-
RAM: 128GB
16 x 8GB DDR3-1600 ECC Registered 2R DIMMs
Operating at 1600 MT/s Max
Maximum 512GB (16 x 32GB)
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x IPMI Management
2.0 & KVM with Dedicated LAN - Integrated
-
1x Infiniband
Mellanox ConnectX-2 VPI Single-Port QDR 40Gb/s InfiniBand HCA, PCIe 2.0 x8
-
-
Storage
-
System - RAID 1
2 x 1TB OS drives in fixed bays
-
-
Usage
This system is used to submit and monitor jobs. No jobs should be run directly on this system.
This is the only system that is reachable directly. You will connect to this system via Secure Shell (SSH) and leverage HTCondor to submit and monitor your jobs. Please see the Getting Started Guide for a quick introduction.
Network
InfiniBand
-
18 port QDR 40Gb/s InfiniBand
Currently 14 nodes are connected
The connected nodes are part of the ParallelSchedulingGroup “ib40_1”
Ethernet
1x Cisco C9200L-48T-4X
Computer Science
Compute Nodes
Compute (8x)
-
CPU: 2 x Intel Xeon Gold 6130
2.1GHz Base, 3.7GHz Max Turbo
16-Core, 32 Thread
22MB L3 Cache
14nm
125W
-
RAM: 192GB
12 x 16GB DDR4 2666MHz ECC Registered DIMMs
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x InfiniBand
ConnectX-5 VPI EDR 100Gb/s Controller with QSFP28 Connector
-
-
Storage
1x 960GB Micron 5200 PRO Series (6Gb/s SATA, 540 MB/s Read, 520 MB/s Write) 2.5” SATA
-
Priority
Open / None
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 64 cpus, ~192GB memory
ParallelSchedulingGroup: N/A
Graphics Nodes
GPU (3x)
-
CPU: 2x Intel Xeon Gold 6130
2.1GHz Base, 3.7GHz Max Turbo
16-Core, 32 Thread
22MB L3 Cache
14nm
125W
-
RAM: 192GB
12x 16GB DDR4 2666MHz ECC Registered DIMMs
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x InfiniBand
ConnectX-5 VPI EDR 100Gb/s Controller with QSFP28 Connector
-
-
Storage
1x 960GB Micron 5200 PRO Series (6Gb/s SATA, 540 MB/s Read, 520 MB/s Write) 2.5” SATA
-
GPU: 4x Nvidia GeForce RTX 2080 TI
Architecture: Turing
CUDA Cores: 4,352
Memory: 11 GB GDDR6
Clock Speed: 1350-1545 Mhz
-
Priority
Open / None
-
HTCondor Provisioning
Partitionable (Dynamic)
Max: 64 cpus, 4 gpus, ~192GB memory
Must request at least 1 GPU
ParallelSchedulingGroup: N/A
GPU (2x)
-
CPU: 2x Intel Xeon Gold 5218T
2.1GHz Base, 3.8GHz Max Turbo
16-Core, 32 Thread
22MB L3 Cache
14nm
105W
-
RAM: 384GB
12x 32GB DDR4 2666MHz ECC Registered DIMMs
-
Networking
-
2x 10-Gigabit Ethernet
Intel Dual-Port X722 10-Gigabit Ethernet Controller - Integrated (10Gbase-T)
-
2x 10-Gigabit Ethernet
Intel Dual-Port X710 10-Gigabit Ethernet Controller (SFP+)
-
1x InfiniBand
ConnectX-5 VPI EDR 100Gb/s Controller with QSFP28 Connector
-
-
Storage
1x 984GB Intel SSD SATA 6Gb/s 2.5”
-
GPU: 8x Nvidia GeForce RTX 2080 TI
Architecture: Turing
CUDA Cores: 4,352
Memory: 11 GB GDDR6
Clock Speed: 1350-1545 Mhz
-
Priority
Open / None
-
HTCondor Provisioning
Static
Max: 8 cpus, 1 gpu, ~46GB memory
Must request at least 1 GPU
ParallelSchedulingGroup: N/A
GPU (1x)
-
CPU: 2x Intel Xeon Gold 6434
3.7GHz Base, 4.10GHz Max Turbo
8-Core, 16 Thread
22.5MB L3 Cache
10nm
195W
-
RAM: 1TB
16x 64GB PC5-38400 4800MHz DDR5 ECC Registered DIMMs
-
Networking
-
2x Gigabit Ethernet
Intel® X710-AT2 Gigabit Ethernet Controller - Integrated
-
-
Storage
1x 960GB Micron 5400 PRO Series (6Gb/s SATA, 540 MB/s Read, 520 MB/s Write) 2.5” SATA
1x 3.2TB Micron 7450 MAX Series (6800 MB/s Read, 5300 MB/s Write ) U.3 PCIe 4.0 x4 NVMe Solid State Drive
-
GPU: 3x Nvidia H100 GPU
Architecture: Hopper
CUDA Cores: 14,592
Tensor Cores: 456
Memory: 80GB HBM2e
Clock Speed: 1065 MHz Base, 1755 MHz Max Boost
Bus: PCIe 5.0 x16
Multi-Instance GPU: Up to 7 MIGs
Performance
-
Double Precision Performance (FP64)
25.6 TFLOPS
-
Double Precision Tensor Core Performance (FP64)
51 TFLOPS
-
Single Precision Performance (FP32)
51 TFLOPS
-
Tensor Float 32 (TF32)
756 TFLOPS
-
BFLOAT16 Tensor Core
1513 TFLOPS
-
FP16 Tensor Core
1513 TFLOPS
-
FP8 Tensor Core
3026 TFLOPS
-
INT8 Tensor Core
3026 TOPS
-
Priority
-
Slot 1
Hutchinson research group
Idriss research group
Wehrwein research group
-
Slot 2
Open / None
-
-
HTCondor Provisioning
-
Slot 1
Partitionable (Dynamic)
Max: 16 cpus, 1 gpus, ~502GB memory
Max run time before preemption: 3 days
-
Required extra settings:
+WantH100=True
ParallelSchedulingGroup: N/A
-
Slot 2
Partitionable (Dynamic)
Max: 2 gpus
Max per requested GPU: 8 cpus, 257423MB (~251GB) memory
Max run time before preemption: 3 days
-
Required extra settings:
+WantH100=True
ParallelSchedulingGroup: N/A
-
Support Systems & Miscellaneous
Head Node (1x)
-
CPU: 2 x Intel Xeon E5-2620
2GHz
6-Core, HT
15MB Cache
32nm
95W
-
RAM: 128GB
16 x 8GB DDR3-1600 ECC Registered 2R DIMMs
Operating at 1600 MT/s Max
Maximum 512GB (16 x 32GB)
-
Networking
-
2x Gigabit Ethernet
Intel Dual-Port I350 Gigabit Ethernet Controller - Integrated
-
1x InfiniBand
Mellanox ConnectX-2 VPI Single-Port QDR 40Gb/s InfiniBand HCA, PCIe 2.0 x8
-
-
Storage
-
System
-
RAID 1
2 x 1TB OS drives in fixed bays
-
-
-
Usage
This system is used to submit and monitor jobs. No jobs should be run directly on this system.
This is the only system that is reachable directly. You will connect to this system via Secure Shell (SSH) and leverage HTCondor to submit and monitor your jobs. Please see the Getting Started Guide for a quick introduction.
Network
InfiniBand
-
36 port EDR 100Gb/s InfiniBand
All CSCI nodes except g-3-0 are connected via InfiniBand
There is no ParallelSchedulingGroup because they are all connected to the same switch
Ethernet
4x Cisco C9200L-48T-4X