AlmaLinux Hardware Catalog
  1. Catalog
  2. Run 1703dc7d-75…

NVIDIA L4

GPU only This run is evidence for the GPU above and for nothing else. Measured in Google Google Compute Engine, which it makes no claim about.

Summary

Run type
Validate
Certification
PASS for the GPU only
Claim
GPU only NVIDIA L4
Host
Custom build Google Google Compute Engine
CPU
Intel(R) Xeon(R) CPU @ 2.20GHz · 1 socket · 2 cores / 4 threads
CPU flags
90 feature flags (informational)
All flags, as reported:
3dnowprefetchabmadxaesapicaratarch_capabilitiesavxavx2avx512_vnniavx512bwavx512cdavx512dqavx512favx512vlbmi1bmi2clflushclflushoptclwbcmovconstant_tsccpuidcx16cx8deermsf16cfmafpufsgsbasefxsrhlehthypervisoribpbibrsibrs_enhancedinvpcidlahf_lmlmmcamcemd_clearmmxmovbempxmsrmtrrnonstop_tscnoplnxpaepatpcidpclmulqdqpdpe1gbpgepnipopcntpsepse36rdrandrdseedrdtscprep_goodrtmsepsmapsmepssssbdssesse2sse4_1sse4_2ssse3stibpsyscalltsctsc_adjusttsc_known_freqvmex2apicxgetbv1xsavexsavecxsaveoptxsavesxtopology
GPU
NVIDIA L4 (nvidia 610.57.04) what this run certifies
Memory
16 GB · 1 × RAM
Memory modules
SlotChannelSizeType SpeedRankManufacturerPart number
DIMM 0 Not Specified 16 GB RAM - - - -
1 of 1 slots populated
AlmaLinux
-
Kernel
6.12.0-211.46.1.el10_2.x86_64 tainted (4096) An out-of-tree module was loaded
Suite
0.1.0 (0.1.0-7.el10)
Submitted by
jonathan · 2026-08-19 17:58

Result counts

11 pass

Linked components

Validation tests

TestCategorySeverityStatusDetail
bench.gpu.clpeak gpu pass
bench.gpu.cuda-bandwidth gpu pass
validate.gpu.cuda-bandwidth gpu required pass host to device 10.87 GiB/s, device to host 12.13 GiB/s
validate.gpu.cuda-devicequery gpu required pass CUDA sees 1 device(s): NVIDIA L4 (runtime 13030, driver 13030)
validate.gpu.cuda-vectoradd gpu required pass 16777216 elements computed correctly on the device
validate.gpu.driver gpu informational pass 1 with a driver (nvidia)
validate.gpu.nvidia-smi gpu required pass 1 NVIDIA GPU(s), driver 610.57.04
validate.gpu.opencl-compute gpu required pass 4194304 elements computed correctly on NVIDIA L4
validate.gpu.opencl-devices gpu required pass OpenCL sees 1 hardware device(s): NVIDIA L4
validate.gpu.vulkan-compute gpu required pass 16777216 bytes filled and verified on NVIDIA L4 (discrete)
validate.gpu.vulkan-devices gpu required pass Vulkan sees 1 hardware device(s): NVIDIA L4

Benchmarks

BenchmarkMetricValue
GPU compute bench.gpu.clpeak OpenCL Compute · Single precision opencl_single_precision_compute primary 26,537 GFLOPS ↑ better
GPU compute bench.gpu.clpeak OpenCL Compute · Double precision opencl_double_precision_compute 472 GFLOPS ↑ better
GPU compute bench.gpu.clpeak OpenCL Compute · Integer opencl_integer_compute 11,919 GOPS ↑ better
GPU compute bench.gpu.clpeak OpenCL Bandwidth · Global memory opencl_global_memory_bandwidth 258 GB/s ↑ better
GPU compute bench.gpu.clpeak OpenCL Bandwidth · Local memory opencl_local_memory_bandwidth 12,519 GB/s ↑ better
GPU compute bench.gpu.clpeak OpenCL Bandwidth · Image memory opencl_image_memory_bandwidth 189 GB/s ↑ better
GPU compute bench.gpu.clpeak OpenCL Bandwidth · Host transfer opencl_transfer_bandwidth 12.9 GB/s ↑ better
GPU compute bench.gpu.clpeak OpenCL Latency · Kernel launch opencl_kernel_launch_latency 17.4 us ↓ better
GPU compute bench.gpu.clpeak CUDA Compute · Single precision cuda_single_precision_compute 25,326 GFLOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Compute · Double precision cuda_double_precision_compute 473 GFLOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Compute · Half precision cuda_half_precision_compute 29,941 GFLOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Compute · Mixed precision cuda_mixed_precision_compute 20,985 GFLOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Compute · bfloat16 cuda_bfloat16_compute 23,849 GFLOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Compute · Integer cuda_integer_compute 11,865 GOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Compute · Integer, int8 dot product cuda_integer_compute_int8_dp 31,650 GOPS ↑ better
GPU compute bench.gpu.clpeak CUDA Bandwidth · Global memory cuda_global_memory_bandwidth 247 GB/s ↑ better
GPU compute bench.gpu.clpeak CUDA Bandwidth · Local memory cuda_local_memory_bandwidth 12,011 GB/s ↑ better
GPU compute bench.gpu.clpeak CUDA Bandwidth · Image memory cuda_image_memory_bandwidth 188 GB/s ↑ better
GPU compute bench.gpu.clpeak CUDA Bandwidth · Host transfer cuda_transfer_bandwidth 13 GB/s ↑ better
GPU compute bench.gpu.clpeak CUDA Latency · Kernel launch cuda_kernel_launch_latency 7.14 us ↓ better
GPU compute bench.gpu.clpeak Vulkan Compute · Single precision vulkan_single_precision_compute 24,568 GFLOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Compute · Double precision vulkan_double_precision_compute 469 GFLOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Compute · Half precision vulkan_half_precision_compute 29,354 GFLOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Compute · Mixed precision vulkan_mixed_precision_compute 24,550 GFLOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Compute · bfloat16 vulkan_bfloat16_compute 21,106 GFLOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Compute · Integer vulkan_integer_compute 8,064 GOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Compute · Integer, int8 dot product vulkan_integer_compute_int8_dp 8,405 GOPS ↑ better
GPU compute bench.gpu.clpeak Vulkan Bandwidth · Global memory vulkan_global_memory_bandwidth 247 GB/s ↑ better
GPU compute bench.gpu.clpeak Vulkan Bandwidth · Local memory vulkan_local_memory_bandwidth 11,490 GB/s ↑ better
GPU compute bench.gpu.clpeak Vulkan Bandwidth · Image memory vulkan_image_memory_bandwidth 226 GB/s ↑ better
GPU compute bench.gpu.clpeak Vulkan Bandwidth · Host transfer vulkan_transfer_bandwidth 12.9 GB/s ↑ better
GPU compute bench.gpu.clpeak Vulkan Latency · Kernel launch vulkan_kernel_launch_latency 130 us ↓ better
GPU transfer bench.gpu.cuda-bandwidth device_to_device 107 GiB/s ↑ better
GPU transfer bench.gpu.cuda-bandwidth device_to_host 12 GiB/s ↑ better
GPU transfer bench.gpu.cuda-bandwidth host_to_device primary 11.2 GiB/s ↑ better