Understanding Your Vera Machine#

After you boot your Vera machine, run the sudo ipmitool fru print command and check the information about the NVIDIA Vera CPU or Vera Rubin.

The Field Replaceable Unit (FRU) inventory identifies chassis, board, and product metadata for each hardware module. The exact strings depend on your platform SKU and firmware revision. Large systems expose multiple FRU devices (for example, power distribution, compute trays, and accelerators). To find the records you need, scroll the full ipmitool output or filter by device.

Here is sample output from sudo ipmitool fru print on a dual-socket NVIDIA Vera CPU system. In this example, the Board Product is P5035-A01, and the Product Name is 2U AC MGX C2.

Chassis Part Number   : 670-25036-9200-000
Chassis Serial        : [REDACTED]
Board Mfg Date        : [REDACTED]
Board Mfg             : NVIDIA
Board Product         : P5035-A01
Board Serial          : [REDACTED]
Board Part Number     : 699-25035-0200-TS1
Board Extra           : [REDACTED]
Product Manufacturer  : NVIDIA
Product Name          : 2U AC MGX C2
Product Part Number   : [REDACTED]
Product Version       : X
Product Serial        : [REDACTED]

Here is sample output from sudo ipmitool fru print on a Vera Rubin NVL72 compute tray. In this example, the Board Product is P4107, and the Product Name is VR NVL72 Compute Tray.

FRU Device Description : PDB_0 (ID 141)
 Chassis Type          : Blade
 Chassis Part Number   : 675-24109-0010-TS2
 Chassis Serial        : [REDACTED]
 Chassis Extra         : Midplane PN: 699-24108-0000-TS2
 Chassis Extra         : Midplane SN: [REDACTED]
 Board Mfg Date        : [REDACTED]
 Board Mfg             : NVIDIA
 Board Product         : P4107
 Board Serial          : [REDACTED]
 Board Part Number     : 699-24107-0210-TS2
 Product Name          : VR NVL72 Compute Tray
 Product Part Number   : [REDACTED]
 Product Version       : A.0
 Product Serial        : [REDACTED]

On multi-module servers, expect one FRU record per field-replaceable unit and compare FRU Device Description, Board Product, and Product Name across entries to map trays, CPUs, and GPUs to physical modules.

Checking the CPUs#

The lscpu command-line utility in Linux gets CPU information about the system, fetches the CPU architecture information from the sysfs and /proc/cpuinfo files, and displays the information in a terminal.

After you boot your Vera machine, run lscpu and review sockets, cores, threads, cache sizes, and NUMA layout. The NVIDIA Vera CPU uses NVIDIA Olympus cores and might expose two logical processors per core when Spatial Multithreading (SMT) is enabled.

Here is sample output from a dual-socket NVIDIA Vera CPU system:

Architecture:                aarch64
  CPU op-mode(s):            64-bit
  Byte Order:                Little Endian
CPU(s):                      352
  On-line CPU(s) list:       0-351
Vendor ID:                   NVIDIA
  Model name:                Olympus
    Model:                   0
    Thread(s) per core:      2
    Core(s) per socket:      88
    Socket(s):               2
    Stepping:                0x0
    Frequency boost:         disabled
    CPU(s) scaling MHz:      100%
    CPU max MHz:             3492.0000
    CPU min MHz:             810.0000
    BogoMIPS:                2000.00
    Flags:                   fp asimd evtstrm aes pmull sha1 sha2 crc32
                             atomics fphp asimdhp cpuid asimdrdm jscvt fcma
                             lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve
                             asimdfhm dit uscat ilrcpc flagm ssbs sb paca
                             pacg dcpodp sve2 sveaes svepmull svebitperm
                             svesha3 svesm4 flagm2 frint svei8mm svebf16
                             i8mm bf16 dgh rng bti ecv afp wfxt fpmr lut
                             faminmax f8cvt f8fma f8dp4 f8dp2 f8e4m3 f8e5m2
                             ls64
Caches (sum of all):
  L1d:                       16.5 MiB (176 instances)
  L1i:                       11 MiB (176 instances)
  L2:                        352 MiB (176 instances)
  L3:                        328 MiB (2 instances)
NUMA:
  NUMA node(s):              4
  NUMA node0 CPU(s):         0-87,176-263
  NUMA node1 CPU(s):         88-175,264-351
  NUMA node2 CPU(s):
  NUMA node3 CPU(s):
Vulnerabilities:
  Gather data sampling:      Not affected
  Ghostwrite:                Not affected
  Indirect target selection: Not affected
  Itlb multihit:             Not affected
  L1tf:                      Not affected
  Mds:                       Not affected
  Meltdown:                  Not affected
  Mmio stale data:           Not affected
  Old microcode:             Not affected
  Reg file data sampling:    Not affected
  Retbleed:                  Not affected
  Spec rstack overflow:      Not affected
  Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
  Spectre v1:                Mitigation; __user pointer sanitization
  Spectre v2:                Mitigation; CSV2, BHB
  Srbds:                     Not affected
  Tsa:                       Not affected
  Tsx async abort:           Not affected
  Vmscape:                   Not affected

From this output you can read the number of sockets, cores per socket, hardware threads per core (including SMT), and CPU frequency limits. Cache capacities reflect the per-core L2 and shared system-level cache (L3) on Vera.

The following sample output is from a dual-socket NVIDIA Vera Rubin system. When SMT is enabled, the logical CPU indices span all hardware threads on the Olympus cores.

Architecture:                aarch64
  CPU op-mode(s):            64-bit
  Byte Order:                Little Endian
CPU(s):                      352
  On-line CPU(s) list:       0-351
Vendor ID:                   NVIDIA
  Model name:                Olympus
    Model:                   0
    Thread(s) per core:      2
    Core(s) per socket:      88
    Socket(s):               2
    Stepping:                0x0
    Frequency boost:         disabled
    CPU(s) scaling MHz:      100%
    CPU max MHz:             3492.0000
    CPU min MHz:             810.0000
    BogoMIPS:                2000.00
    Flags:                   fp asimd evtstrm aes pmull sha1 sha2 crc32
                             atomics fphp asimdhp cpuid asimdrdm jscvt fcma
                             lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve
                             asimdfhm dit uscat ilrcpc flagm ssbs sb paca
                             pacg dcpodp sve2 sveaes svepmull svebitperm
                             svesha3 svesm4 flagm2 frint svei8mm svebf16
                             i8mm bf16 dgh rng bti ecv afp wfxt fpmr lut
                             faminmax f8cvt f8fma f8dp4 f8dp2 f8e4m3 f8e5m2
                             ls64
Caches (sum of all):
  L1d:                       16.5 MiB (176 instances)
  L1i:                       11 MiB (176 instances)
  L2:                        352 MiB (176 instances)
  L3:                        328 MiB (2 instances)
NUMA:
  NUMA node(s):              36
  NUMA node0 CPU(s):         0-87,176-263
  NUMA node1 CPU(s):         88-175,264-351
  NUMA node2 CPU(s):
  NUMA node3 CPU(s):
  NUMA node4 CPU(s):
  NUMA node5 CPU(s):
  NUMA node6 CPU(s):
  NUMA node7 CPU(s):
  NUMA node8 CPU(s):
  NUMA node9 CPU(s):
  NUMA node10 CPU(s):
  NUMA node11 CPU(s):
  NUMA node12 CPU(s):
  NUMA node13 CPU(s):
  NUMA node14 CPU(s):
  NUMA node15 CPU(s):
  NUMA node16 CPU(s):
  NUMA node17 CPU(s):
  NUMA node18 CPU(s):
  NUMA node19 CPU(s):
  NUMA node20 CPU(s):
  NUMA node21 CPU(s):
  NUMA node22 CPU(s):
  NUMA node23 CPU(s):
  NUMA node24 CPU(s):
  NUMA node25 CPU(s):
  NUMA node26 CPU(s):
  NUMA node27 CPU(s):
  NUMA node28 CPU(s):
  NUMA node29 CPU(s):
  NUMA node30 CPU(s):
  NUMA node31 CPU(s):
  NUMA node32 CPU(s):
  NUMA node33 CPU(s):
  NUMA node34 CPU(s):
  NUMA node35 CPU(s):
Vulnerabilities:
  Gather data sampling:      Not affected
  Ghostwrite:                Not affected
  Indirect target selection: Not affected
  Itlb multihit:             Not affected
  L1tf:                      Not affected
  Mds:                       Not affected
  Meltdown:                  Not affected
  Mmio stale data:           Not affected
  Old microcode:             Not affected
  Reg file data sampling:    Not affected
  Retbleed:                  Not affected
  Spec rstack overflow:      Not affected
  Spec store bypass:         Not affected
  Spectre v1:                Mitigation; __user pointer sanitization
  Spectre v2:                Mitigation; CSV2, but not BHB
  Srbds:                     Not affected
  Tsa:                       Not affected
  Tsx async abort:           Not affected
  Vmscape:                   Not affected

Note

On a Vera Rubin system, lscpu shows multiple NUMA nodes. Typically, one node enumerates the Vera CPU cores, and the additional nodes correspond to each Rubin GPU and to the optional NVIDIA Multi-Instance GPU (MIG) instances. The exact count and numbering depend on the driver, the firmware, and the MIG configuration.

If MIG is disabled or unused, MIG-related NUMA nodes can be ignored for many CPU-centric workflows. NVIDIA recommends that you verify the affinity recommendations for GPU workloads.

On large nodes with two Vera Rubins (four Rubin GPUs in total), the NUMA node count is higher than on a single one. Ensure that you always reconcile lscpu and numactl -H with your hardware guide.

Checking the Non-Uniform Memory Access Settings#

The lscpu output includes basic information about the Non-Uniform Memory Access (NUMA) settings on your Vera machine.

To understand more about the NUMA settings, run the numactl -H command, and here is the sample output from a dual-socket NVIDIA Vera CPU machine:

available: 4 nodes (0-3)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263
node 0 size: 768035 MB
node 0 free: 623820 MB
node 1 cpus: 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
node 1 size: 785180 MB
node 1 free: 769935 MB
node 2 cpus:
node 2 size: 0 MB
node 2 free: 0 MB
node 3 cpus:
node 3 size: 0 MB
node 3 free: 0 MB
node distances:
node   0   1   2   3
  0:  10  40  11  40
  1:  40  10  40  11
  2: 255 255  10 255
  3: 255 255 255  10

On a dual-socket Vera CPU system, nodes 0 and 1 contain the allocatable CPUs and LPDDR5X memory. Nodes 2 and 3 are intentional, memoryless proximity domains used by Memory System Resource Partitioning and Monitoring (MPAM) Memory System Component (MSC) monitors to distinguish remote traffic to CPU-attached memory (CMEM). One domain is associated with each socket. These nodes contain no CPUs or allocatable memory, are not GPU memory nodes, and do not indicate a hardware failure. The distance matrix associates node 2 with node 0 and node 3 with node 1.

Use nodes 0 and 1 for application CPU and memory placement. Do not pass nodes 2 or 3 to numactl --cpunodebind or numactl --membind. For example, bind a workload to the CPUs and local memory on socket 0 as follows:

$ numactl --cpunodebind=0 --membind=0 <command>

Topology-aware scripts must not assume that every node that are reported as available by numactl -H contain CPUs or memory. The kernel exposes the nodes that contain these resources in sysfs. On the system in this example, both files report nodes 0 and 1:

$ cat /sys/devices/system/node/has_cpu
0-1
$ cat /sys/devices/system/node/has_memory
0-1

Select nodes that appear in both masks before assigning CPU and memory affinity. Depending on the system firmware, nodes 2 and 3 might remain visible when MPAM bandwidth controls are disabled. Their presence alone does not indicate whether an MPAM policy is active.

The output also shows the node distances, which help the kernel scheduler execute application threads on CPU cores that are closest to memory-resident data.

On dual-socket Vera systems, CPU numbering interleaves the two SMT threads per Olympus core across the NUMA node that owns the physical core (for example, logical CPUs 0-87 and 176-263 on one node, and 88-175 and 264-351 on the other in the sample above).

The following sample output is from a dual-socket NVIDIA Vera Rubin system with SMT disabled:

available: 36 nodes (0-35)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87
node 0 size: 658402 MB
node 0 free: 648852 MB
node 1 cpus: 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
node 1 size: 736094 MB
node 1 free: 724517 MB
node 2 cpus:
node 2 size: 0 MB
node 2 free: 0 MB
node 3 cpus:
node 3 size: 0 MB
node 3 free: 0 MB
node 4 cpus:
node 4 size: 0 MB
node 4 free: 0 MB
node 5 cpus:
node 5 size: 0 MB
node 5 free: 0 MB
node 6 cpus:
node 6 size: 0 MB
node 6 free: 0 MB
node 7 cpus:
node 7 size: 0 MB
node 7 free: 0 MB
node 8 cpus:
node 8 size: 0 MB
node 8 free: 0 MB
node 9 cpus:
node 9 size: 0 MB
node 9 free: 0 MB
node 10 cpus:
node 10 size: 0 MB
node 10 free: 0 MB
node 11 cpus:
node 11 size: 0 MB
node 11 free: 0 MB
node 12 cpus:
node 12 size: 0 MB
node 12 free: 0 MB
node 13 cpus:
node 13 size: 0 MB
node 13 free: 0 MB
node 14 cpus:
node 14 size: 0 MB
node 14 free: 0 MB
node 15 cpus:
node 15 size: 0 MB
node 15 free: 0 MB
node 16 cpus:
node 16 size: 0 MB
node 16 free: 0 MB
node 17 cpus:
node 17 size: 0 MB
node 17 free: 0 MB
node 18 cpus:
node 18 size: 0 MB
node 18 free: 0 MB
node 19 cpus:
node 19 size: 0 MB
node 19 free: 0 MB
node 20 cpus:
node 20 size: 0 MB
node 20 free: 0 MB
node 21 cpus:
node 21 size: 0 MB
node 21 free: 0 MB
node 22 cpus:
node 22 size: 0 MB
node 22 free: 0 MB
node 23 cpus:
node 23 size: 0 MB
node 23 free: 0 MB
node 24 cpus:
node 24 size: 0 MB
node 24 free: 0 MB
node 25 cpus:
node 25 size: 0 MB
node 25 free: 0 MB
node 26 cpus:
node 26 size: 0 MB
node 26 free: 0 MB
node 27 cpus:
node 27 size: 0 MB
node 27 free: 0 MB
node 28 cpus:
node 28 size: 0 MB
node 28 free: 0 MB
node 29 cpus:
node 29 size: 0 MB
node 29 free: 0 MB
node 30 cpus:
node 30 size: 0 MB
node 30 free: 0 MB
node 31 cpus:
node 31 size: 0 MB
node 31 free: 0 MB
node 32 cpus:
node 32 size: 0 MB
node 32 free: 0 MB
node 33 cpus:
node 33 size: 0 MB
node 33 free: 0 MB
node 34 cpus:
node 34 size: 0 MB
node 34 free: 0 MB
node 35 cpus:
node 35 size: 0 MB
node 35 free: 0 MB
node distances:
node   0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35
  0:  10  116  91  91  91  91  91  91  91  91  91  91  91  91  91  91  91  91  206  206  206  206  206  206  206  206  206  206  206  206  206  206  206  206  11  116
  1:  116  10  206  206  206  206  206  206  206  206  206  206  206  206  206  206  206  206  91  91  91  91  91  91  91  91  91  91  91  91  91  91  91  91  116  11
  2:  77  254  10  42  42  42  42  42  42  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  3:  77  254  42  10  42  42  42  42  42  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  4:  77  254  42  42  10  42  42  42  42  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  5:  77  254  42  42  42  10  42  42  42  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  6:  77  254  42  42  42  42  10  42  42  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  7:  77  254  42  42  42  42  42  10  42  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  8:  77  254  42  42  42  42  42  42  10  42  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
  9:  77  254  42  42  42  42  42  42  42  10  255  255  255  255  255  255  255  255  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 10:  77  254  255  255  255  255  255  255  255  255  10  42  42  42  42  42  42  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 11:  77  254  255  255  255  255  255  255  255  255  42  10  42  42  42  42  42  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 12:  77  254  255  255  255  255  255  255  255  255  42  42  10  42  42  42  42  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 13:  77  254  255  255  255  255  255  255  255  255  42  42  42  10  42  42  42  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 14:  77  254  255  255  255  255  255  255  255  255  42  42  42  42  10  42  42  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 15:  77  254  255  255  255  255  255  255  255  255  42  42  42  42  42  10  42  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 16:  77  254  255  255  255  255  255  255  255  255  42  42  42  42  42  42  10  42  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 17:  77  254  255  255  255  255  255  255  255  255  42  42  42  42  42  42  42  10  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  77  254
 18:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  10  42  42  42  42  42  42  42  255  255  255  255  255  255  255  255  254  77
 19:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  10  42  42  42  42  42  42  255  255  255  255  255  255  255  255  254  77
 20:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  42  10  42  42  42  42  42  255  255  255  255  255  255  255  255  254  77
 21:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  42  42  10  42  42  42  42  255  255  255  255  255  255  255  255  254  77
 22:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  42  42  42  10  42  42  42  255  255  255  255  255  255  255  255  254  77
 23:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  42  42  42  42  10  42  42  255  255  255  255  255  255  255  255  254  77
 24:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  42  42  42  42  42  10  42  255  255  255  255  255  255  255  255  254  77
 25:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  42  42  42  42  42  42  42  10  255  255  255  255  255  255  255  255  254  77
 26:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  10  42  42  42  42  42  42  42  254  77
 27:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  10  42  42  42  42  42  42  254  77
 28:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  42  10  42  42  42  42  42  254  77
 29:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  42  42  10  42  42  42  42  254  77
 30:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  42  42  42  10  42  42  42  254  77
 31:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  42  42  42  42  10  42  42  254  77
 32:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  42  42  42  42  42  10  42  254  77
 33:  254  77  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  187  255  255  255  255  255  255  255  255  42  42  42  42  42  42  42  10  254  77
 34:  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  10  255
 35:  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  255  10

As in Checking the CPUs, if MIG is not used, the trailing NUMA nodes that exist only for MIG can often be ignored for host-side CPU tuning. GPU workloads still require correct GPU affinity.

On systems with two Vera Rubins, expect separate host-memory NUMA nodes for each Vera socket and additional nodes for each Rubin GPU and MIG slice. To map nodes to hardware, use your platform documentation with numactl -H.

Checking the GPU#

Running nvidia-smi displays GPU status when the NVIDIA driver is loaded.

The following sample output is from an NVIDIA Vera Rubin system with four Rubin GPUs that are visible to the OS:

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 615.23                 KMD Version: 615.23        CUDA UMD Version: 13.4     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA Graphics Device         On  |   00000002:81:00.0 Off |                    0 |
| N/A   29C    P5            375W / 2300W |       0MiB / 286524MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   1  NVIDIA Graphics Device         On  |   00000002:C1:00.0 Off |                    0 |
| N/A   29C    P5            381W / 2300W |       0MiB / 286524MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   2  NVIDIA Graphics Device         On  |   0000000A:81:00.0 Off |                    0 |
| N/A   29C    P5            381W / 2300W |       0MiB / 286524MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+
|   3  NVIDIA Graphics Device         On  |   0000000A:E1:00.0 Off |                    0 |
| N/A   28C    P5            384W / 2300W |       0MiB / 286524MiB |      0%      Default |
|                                         |                        |             Disabled |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

The exact GPU names, memory sizes, power limits, and bus IDs depend on your SKU and driver.

Checking the Memory#

One common way to inspect attached memory is to run sudo dmidecode -t memory.

The following sample output is from a dual-socket NVIDIA Vera CPU system using LPDDR5X memory (capacity and speed vary by SKU):

The physical memory array contains multiple DIMMs (or die-level memory devices, depending on platform reporting). The following sample lists only the first Memory Device entry for brevity. Run sudo dmidecode -t memory on your machine to see the full list of devices for that system.

# dmidecode 3.5
Getting SMBIOS data from sysfs.
SMBIOS 3.6.0 present.
# SMBIOS implementations newer than version 3.5.0 are not
# fully supported by this version of dmidecode.

Handle 0x0026, DMI type 16, 23 bytes
Physical Memory Array
  Location: System Board Or Motherboard
  Use: System Memory
  Error Correction Type: Multi-bit ECC
  Maximum Capacity: 3 TB
  Error Information Handle: No Error
  Number Of Devices: 16

Handle 0x0027, DMI type 17, 100 bytes
Memory Device
  Array Handle: 0x0026
  Error Information Handle: Not Provided
  Total Width: 1080 bits
  Data Width: 960 bits
  Size: 192 GB
  Form Factor: Die
  Set: None
  Locator: LP5x_0
  Bank Locator: LP5x_0
  Type: LPDDR5
  Type Detail: Synchronous Unbuffered (Unregistered)
  Speed: 7500 MT/s
  Manufacturer: Samsung
  Serial Number: 18446744072896524802
  Asset Tag: Not Specified
  Part Number: Not Specified
  Rank: 3
  Configured Memory Speed: 7500 MT/s
  Minimum Voltage: 1.1 V
  Maximum Voltage: 1.1 V
  Configured Voltage: 1.1 V
  Memory Technology: DRAM
  Memory Operating Mode Capability: Volatile memory
  Firmware Version: Not Specified
  Module Manufacturer ID: Bank 1, Hex 0xCE
  Module Product ID: Unknown
  Memory Subsystem Controller Manufacturer ID: Unknown
  Memory Subsystem Controller Product ID: Unknown
  Non-Volatile Size: None
  Volatile Size: 192 GB
  Cache Size: None
  Logical Size: None