Merge branch 'master' into wip-2025

data/grid5000/sites/lyon/clusters/hydra/hydra.json

@@ -222,8 +222,8 @@
   ],
   "model": "NVIDIA Grace Hopper",
   "nodes_count": 4,
-  "nodes_description": "1 CPU Grace A02  CPU Unknown, 72 cores/CPU, 480GB RAM, 894GB SSD, 1788GB SSD, 1 x 10Gb Ethernet",
-  "priority": 202402,
+  "nodes_description": "1 CPU Grace A02  CPU Unknown, 72 cores/CPU, 1 GPU GH200, 480GB RAM, 894GB SSD, 1788GB SSD, 1 x 10Gb Ethernet",
+  "priority": 202502,
   "queues": [
     "admin",
     "testing"
@@ -231,5 +231,5 @@
   "redfish": true,
   "type": "cluster",
   "uid": "hydra",
-  "warranty_end": "2027-02-15"
+  "warranty_end": "2027-02-01"
 }
\ No newline at end of file

data/grid5000/sites/lyon/clusters/hydra/nodes/hydra-1.json

@@ -17,9 +17,26 @@
     "manufacturer": "Quanta Cloud Technology Inc.",
     "name": "QuantaGrid S74G-2U 1S7GZ9Z0000",
     "serial": "QTWS7G0240200024",
-    "warranty_end": "2027-02-15"
+    "warranty_end": "2027-02-01"
   },
   "exotic": true,
+  "gpu_devices": {
+    "nvidia0": {
+      "compute_capability": "9.0",
+      "cores": 18176,
+      "cpu_affinity": 1,
+      "device": "/dev/nvidia0",
+      "memory": 102625181696,
+      "microarchitecture": "Hopper",
+      "model": "GH200",
+      "performance": {
+        "fp-16": 19790000000000,
+        "fp-32": 670000000000000,
+        "fp-64": 340000000000000
+      },
+      "vendor": "Nvidia"
+    }
+  },
   "kavlan": {
     "eth0": {
       "kavlan-1": "192.168.201.1",

data/grid5000/sites/lyon/clusters/hydra/nodes/hydra-2.json

@@ -17,9 +17,26 @@
     "manufacturer": "Quanta Cloud Technology Inc.",
     "name": "QuantaGrid S74G-2U 1S7GZ9Z0000",
     "serial": "QTWS7G0240200023",
-    "warranty_end": "2027-02-15"
+    "warranty_end": "2027-02-01"
   },
   "exotic": true,
+  "gpu_devices": {
+    "nvidia0": {
+      "compute_capability": "9.0",
+      "cores": 18176,
+      "cpu_affinity": 1,
+      "device": "/dev/nvidia0",
+      "memory": 102625181696,
+      "microarchitecture": "Hopper",
+      "model": "GH200",
+      "performance": {
+        "fp-16": 19790000000000,
+        "fp-32": 670000000000000,
+        "fp-64": 340000000000000
+      },
+      "vendor": "Nvidia"
+    }
+  },
   "kavlan": {
     "eth0": {
       "kavlan-1": "192.168.201.2",

data/grid5000/sites/lyon/clusters/hydra/nodes/hydra-3.json

@@ -17,9 +17,26 @@
     "manufacturer": "Quanta Cloud Technology Inc.",
     "name": "QuantaGrid S74G-2U 1S7GZ9Z0000",
     "serial": "QTWS7G0240100009",
-    "warranty_end": "2027-02-15"
+    "warranty_end": "2027-02-01"
   },
   "exotic": true,
+  "gpu_devices": {
+    "nvidia0": {
+      "compute_capability": "9.0",
+      "cores": 18176,
+      "cpu_affinity": 1,
+      "device": "/dev/nvidia0",
+      "memory": 102625181696,
+      "microarchitecture": "Hopper",
+      "model": "GH200",
+      "performance": {
+        "fp-16": 19790000000000,
+        "fp-32": 670000000000000,
+        "fp-64": 340000000000000
+      },
+      "vendor": "Nvidia"
+    }
+  },
   "kavlan": {
     "eth0": {
       "kavlan-1": "192.168.201.3",

data/grid5000/sites/lyon/clusters/hydra/nodes/hydra-4.json

@@ -17,9 +17,26 @@
     "manufacturer": "Quanta Cloud Technology Inc.",
     "name": "QuantaGrid S74G-2U 1S7GZ9Z0000",
     "serial": "QTWS7G0240100010",
-    "warranty_end": "2027-02-15"
+    "warranty_end": "2027-02-01"
   },
   "exotic": true,
+  "gpu_devices": {
+    "nvidia0": {
+      "compute_capability": "9.0",
+      "cores": 18176,
+      "cpu_affinity": 1,
+      "device": "/dev/nvidia0",
+      "memory": 102625181696,
+      "microarchitecture": "Hopper",
+      "model": "GH200",
+      "performance": {
+        "fp-16": 19790000000000,
+        "fp-32": 670000000000000,
+        "fp-64": 340000000000000
+      },
+      "vendor": "Nvidia"
+    }
+  },
   "kavlan": {
     "eth0": {
       "kavlan-1": "192.168.201.4",

input/grid5000/sites/lyon/clusters/hydra/hydra.yaml

@@ -11,8 +11,8 @@ nodes:
   hydra-[1-4]:
     nodeset: hydra
     chassis:
-      manufactured_at: 2024-02-15 # TODO: retrieve info from manufacturer
-      warranty_end: 2027-02-15 # TODO: retrieve info from manufacturer
+      manufactured_at: 2024-02-15
+      warranty_end: 2027-02-01
     supported_job_types:
       deploy: true
       besteffort: true
@@ -41,6 +41,13 @@ nodes:
       pci-0006:01:00.0-scsi-0:2:0:0:
         id: disk1
         interface: SCSI
+    gpu_devices:
+      nvidia0:
+        cpu_affinity: 1
+        device: "/dev/nvidia0"
+        memory: 102_625_181_696
+        model: GH200
+        vendor: Nvidia
     software:
       standard-environment: debiangh11-arm64-std
     management_tools:

lib/refrepo/gpu_ref.rb

@@ -378,9 +378,22 @@ class GPURef
          'fp-64' => 141400000000,
        },
      },
+	'GH200' => {
+       'cores'              => 18176,
+       'compute_capability' => '9.0',
+       'short_name'         => 'H200',
+       'alias'              => 'H200',
+       'microarchitecture'  => 'Hopper',
+       'performance'        => {
+         # Performance informations comes from:
+         #   https://www.nvidia.com/en-us/data-center/h200/
+         'fp-16' => 19790000000000, # FP-16 Tensor Core value, no flops available for FP-16.
+         'fp-32' => 670000000000000,
+         'fp-64' => 340000000000000,
+      },
+    }
   }
 
-
   def self.get_cores(model)
     if @@gpus[model]
       return @@gpus[model]['cores']