diff --git a/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/accuracy/accuracy.json b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/accuracy/accuracy.json new file mode 100644 index 00000000..66467933 --- /dev/null +++ b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/accuracy/accuracy.json @@ -0,0 +1,8 @@ +{ + "subset_score": 0.4, + "baseline_delta": 0.02, + "valid": true, + "framework": "vLLM", + "precision": "BF16", + "notes": "Integrated accuracy check \u2014 used same vLLM instance as benchmark." +} \ No newline at end of file diff --git a/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/env_info.json b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/env_info.json new file mode 100644 index 00000000..b9ee4c9f --- /dev/null +++ b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/env_info.json @@ -0,0 +1,33 @@ +{ + "collected_at": "2026-08-21T22:45:35.076567+00:00", + "accelerators": [ + { + "index": 0, + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "memory_gb": 48.0, + "driver_version": "610.43.03", + "firmware_version": null, + "compute_capability": "8.9", + "supports_bf16": true + } + ], + "accelerator_platform": "nvidia", + "accelerator_topology": "\tGPU0\tCPU Affinity\tNUMA Affinity\tGPU NUMA ID\nGPU0\t X \t0-13\t0\t\tN/A\n\nLegend:\n\n X = Self\n SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)\n NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node\n PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)\n PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)\n PIX = Connection traversing at most a single PCIe bridge\n NV# = Connection traversing a bonded set of # NVLinks\n", + "intra_node_interconnect": null, + "cpu": { + "model": "AMD EPYC 7542 32-Core Processor", + "physical_cores": 64, + "logical_cores": 14, + "numa_nodes": 2 + }, + "system_memory_gb": 120.0, + "pcie_generation": "PCIe Gen 4", + "cpu_accelerator_bandwidth_gbs": null, + "network_interfaces": null, + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20", + "kernel_version": "6.8.0-90-generic", + "runtime_version": "CUDA 12.4", + "pytorch_version": "2.5.1+cu124" +} \ No newline at end of file diff --git a/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/interactive/result.json b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/interactive/result.json new file mode 100644 index 00000000..3eb229f3 --- /dev/null +++ b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/interactive/result.json @@ -0,0 +1,132 @@ +{ + "schema_version": "1.0", + "suite_id": "suite_F", + "implementation_id": "nvidia_vllm_47f5d58e", + "chip": { + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "count": 1, + "card_count": 1, + "dies_per_card": 1, + "memory_gb": 48.0, + "interconnect_intra_node": null, + "interconnect_inter_node": null + }, + "environment": { + "collected_at": "2026-08-21T22:45:35.076567+00:00", + "accelerators": [ + { + "index": 0, + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "memory_gb": 48.0, + "driver_version": "610.43.03", + "firmware_version": null, + "compute_capability": "8.9", + "supports_bf16": true + } + ], + "accelerator_platform": "nvidia", + "accelerator_topology": "\tGPU0\tCPU Affinity\tNUMA Affinity\tGPU NUMA ID\nGPU0\t X \t0-13\t0\t\tN/A\n\nLegend:\n\n X = Self\n SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)\n NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node\n PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)\n PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)\n PIX = Connection traversing at most a single PCIe bridge\n NV# = Connection traversing a bonded set of # NVLinks\n", + "intra_node_interconnect": null, + "cpu": { + "model": "AMD EPYC 7542 32-Core Processor", + "physical_cores": 64, + "logical_cores": 14, + "numa_nodes": 2 + }, + "system_memory_gb": 120.0, + "pcie_generation": "PCIe Gen 4", + "cpu_accelerator_bandwidth_gbs": null, + "network_interfaces": null, + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20", + "kernel_version": "6.8.0-90-generic", + "runtime_version": "CUDA 12.4", + "pytorch_version": "2.5.1+cu124" + }, + "software": { + "framework": "vLLM", + "framework_version": "0.7.3", + "driver_version": "610.43.03", + "runtime_version": "CUDA 12.4", + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20" + }, + "model": { + "model_id": "Qwen/Qwen2.5-0.5B-Instruct", + "model_revision": "7ae557604adf67be50417f59c2c2f167def9a775", + "model_name": null, + "model_note": null, + "model_source": "huggingface", + "architecture": "dense", + "parameter_count_b": 0.5, + "precision": "BF16", + "effective_dtype": "bfloat16", + "quantization_method": null, + "model_format": "HuggingFace original" + }, + "task": { + "scenario": "interactive", + "num_runs": 3, + "warmup_runs": 1, + "parallelism": { + "tensor_parallel_size": 1, + "pipeline_parallel_size": 1, + "expert_parallel_size": 1, + "data_parallel_size": 1 + }, + "extra_config": null, + "runtime_metrics": null + }, + "metrics": { + "interactive": { + "ttft_ms_p50": 16.75, + "ttft_ms_p90": 30.95, + "ttft_ms_p99": 34.42, + "tpot_ms_p50": 2.8, + "tpot_ms_p90": 3.63, + "tpot_ms_p99": 3.76, + "peak_memory_gb": null, + "elapsed_seconds_median": 90.2, + "ttft_ms_p99_reliability": { + "n": 3, + "mean": 27.01, + "std": 7.24, + "cv_pct": 26.81, + "stability": "high-variance", + "runs": [ + 24.69, + 21.22, + 35.14 + ] + }, + "power_watts_avg": 203.66, + "power_watts_peak": 219.5 + } + }, + "accuracy": { + "subset_score": null, + "baseline_delta": null, + "valid": false, + "notes": "Run --scenario accuracy to check model accuracy." + }, + "meta": { + "submitted_by": "Keke946", + "submission_type": "individual", + "date": "2026-08-21", + "time": "22:58:23", + "run_id": "d00caeb0", + "run_name": "nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0", + "flagged": null, + "reproduce_script": "runners/nvidia_vllm_47f5d58e/runner.py", + "env_info_file": "../env_info.json", + "log_file": "run.log", + "samples_file": "samples.jsonl", + "notes": null, + "benchmark_start_time": "2026-08-21T22:53:37.326819+00:00", + "benchmark_end_time": "2026-08-21T22:58:23.470727+00:00", + "benchmark_elapsed_minutes": 4.8, + "model_load_seconds": 49.2 + } +} \ No newline at end of file diff --git a/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/offline/result.json b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/offline/result.json new file mode 100644 index 00000000..b248cf57 --- /dev/null +++ b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/offline/result.json @@ -0,0 +1,187 @@ +{ + "schema_version": "1.0", + "suite_id": "suite_F", + "implementation_id": "nvidia_vllm_47f5d58e", + "chip": { + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "count": 1, + "card_count": 1, + "dies_per_card": 1, + "memory_gb": 48.0, + "interconnect_intra_node": null, + "interconnect_inter_node": null + }, + "environment": { + "collected_at": "2026-08-21T22:45:35.076567+00:00", + "accelerators": [ + { + "index": 0, + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "memory_gb": 48.0, + "driver_version": "610.43.03", + "firmware_version": null, + "compute_capability": "8.9", + "supports_bf16": true + } + ], + "accelerator_platform": "nvidia", + "accelerator_topology": "\tGPU0\tCPU Affinity\tNUMA Affinity\tGPU NUMA ID\nGPU0\t X \t0-13\t0\t\tN/A\n\nLegend:\n\n X = Self\n SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)\n NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node\n PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)\n PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)\n PIX = Connection traversing at most a single PCIe bridge\n NV# = Connection traversing a bonded set of # NVLinks\n", + "intra_node_interconnect": null, + "cpu": { + "model": "AMD EPYC 7542 32-Core Processor", + "physical_cores": 64, + "logical_cores": 14, + "numa_nodes": 2 + }, + "system_memory_gb": 120.0, + "pcie_generation": "PCIe Gen 4", + "cpu_accelerator_bandwidth_gbs": null, + "network_interfaces": null, + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20", + "kernel_version": "6.8.0-90-generic", + "runtime_version": "CUDA 12.4", + "pytorch_version": "2.5.1+cu124" + }, + "software": { + "framework": "vLLM", + "framework_version": "0.7.3", + "driver_version": "610.43.03", + "runtime_version": "CUDA 12.4", + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20" + }, + "model": { + "model_id": "Qwen/Qwen2.5-0.5B-Instruct", + "model_revision": "7ae557604adf67be50417f59c2c2f167def9a775", + "model_name": null, + "model_note": null, + "model_source": "huggingface", + "architecture": "dense", + "parameter_count_b": 0.5, + "precision": "BF16", + "effective_dtype": "bfloat16", + "quantization_method": null, + "model_format": "HuggingFace original" + }, + "task": { + "scenario": "offline", + "num_runs": 3, + "warmup_runs": 1, + "parallelism": { + "tensor_parallel_size": 1, + "pipeline_parallel_size": 1, + "expert_parallel_size": 1, + "data_parallel_size": 1 + }, + "extra_config": null, + "runtime_metrics": null + }, + "metrics": { + "offline": { + "results_by_concurrency": [ + { + "client_concurrency": 4, + "throughput_tokens_per_sec": 10860.17, + "throughput_tokens_per_sec_per_chip": 10860.17, + "throughput_tokens_per_sec_total": 16145.21, + "elapsed_seconds_median": 3.9, + "peak_memory_gb": 40.36, + "power_watts_avg": 161.83, + "power_watts_peak": 190.2, + "oom": false, + "throughput_tokens_per_sec_reliability": { + "n": 3, + "mean": 10485.91, + "std": 743.29, + "cv_pct": 7.09, + "stability": "noisy", + "runs": [ + 10860.17, + 9629.88, + 10967.67 + ] + }, + "_throughput_note": "output_only", + "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs." + }, + { + "client_concurrency": 16, + "throughput_tokens_per_sec": 11026.74, + "throughput_tokens_per_sec_per_chip": 11026.74, + "throughput_tokens_per_sec_total": 16392.85, + "elapsed_seconds_median": 3.8, + "peak_memory_gb": 40.36, + "power_watts_avg": 165.89, + "power_watts_peak": 196.9, + "oom": false, + "throughput_tokens_per_sec_reliability": { + "n": 3, + "mean": 10548.13, + "std": 882.8, + "cv_pct": 8.37, + "stability": "high-variance", + "runs": [ + 11088.27, + 11026.74, + 9529.38 + ] + }, + "_throughput_note": "output_only", + "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs." + }, + { + "client_concurrency": 64, + "throughput_tokens_per_sec": 10820.87, + "throughput_tokens_per_sec_per_chip": 10820.87, + "throughput_tokens_per_sec_total": 16086.8, + "elapsed_seconds_median": 3.9, + "peak_memory_gb": 40.36, + "power_watts_avg": 169.31, + "power_watts_peak": 198.3, + "oom": false, + "throughput_tokens_per_sec_reliability": { + "n": 3, + "mean": 10761.02, + "std": 161.99, + "cv_pct": 1.51, + "stability": "stable", + "runs": [ + 10884.57, + 10577.62, + 10820.87 + ] + }, + "_throughput_note": "output_only", + "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs." + } + ] + } + }, + "accuracy": { + "subset_score": null, + "baseline_delta": null, + "valid": false, + "notes": "Run --scenario accuracy to check model accuracy." + }, + "meta": { + "submitted_by": "Keke946", + "submission_type": "individual", + "date": "2026-08-21", + "time": "22:48:32", + "run_id": "d00caeb0", + "run_name": "nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0", + "flagged": null, + "reproduce_script": "runners/nvidia_vllm_47f5d58e/runner.py", + "env_info_file": "../env_info.json", + "log_file": "run.log", + "samples_file": "samples.jsonl", + "notes": null, + "benchmark_start_time": "2026-08-21T22:47:42.908336+00:00", + "benchmark_end_time": "2026-08-21T22:48:32.501802+00:00", + "benchmark_elapsed_minutes": 0.8, + "model_load_seconds": 52.6 + } +} \ No newline at end of file diff --git a/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/online/result.json b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/online/result.json new file mode 100644 index 00000000..8abeb6be --- /dev/null +++ b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/online/result.json @@ -0,0 +1,166 @@ +{ + "schema_version": "1.0", + "suite_id": "suite_F", + "implementation_id": "nvidia_vllm_47f5d58e", + "chip": { + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "count": 1, + "card_count": 1, + "dies_per_card": 1, + "memory_gb": 48.0, + "interconnect_intra_node": null, + "interconnect_inter_node": null + }, + "environment": { + "collected_at": "2026-08-21T22:45:35.076567+00:00", + "accelerators": [ + { + "index": 0, + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "memory_gb": 48.0, + "driver_version": "610.43.03", + "firmware_version": null, + "compute_capability": "8.9", + "supports_bf16": true + } + ], + "accelerator_platform": "nvidia", + "accelerator_topology": "\tGPU0\tCPU Affinity\tNUMA Affinity\tGPU NUMA ID\nGPU0\t X \t0-13\t0\t\tN/A\n\nLegend:\n\n X = Self\n SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)\n NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node\n PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)\n PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)\n PIX = Connection traversing at most a single PCIe bridge\n NV# = Connection traversing a bonded set of # NVLinks\n", + "intra_node_interconnect": null, + "cpu": { + "model": "AMD EPYC 7542 32-Core Processor", + "physical_cores": 64, + "logical_cores": 14, + "numa_nodes": 2 + }, + "system_memory_gb": 120.0, + "pcie_generation": "PCIe Gen 4", + "cpu_accelerator_bandwidth_gbs": null, + "network_interfaces": null, + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20", + "kernel_version": "6.8.0-90-generic", + "runtime_version": "CUDA 12.4", + "pytorch_version": "2.5.1+cu124" + }, + "software": { + "framework": "vLLM", + "framework_version": "0.7.3", + "driver_version": "610.43.03", + "runtime_version": "CUDA 12.4", + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20" + }, + "model": { + "model_id": "Qwen/Qwen2.5-0.5B-Instruct", + "model_revision": "7ae557604adf67be50417f59c2c2f167def9a775", + "model_name": null, + "model_note": null, + "model_source": "huggingface", + "architecture": "dense", + "parameter_count_b": 0.5, + "precision": "BF16", + "effective_dtype": "bfloat16", + "quantization_method": null, + "model_format": "HuggingFace original" + }, + "task": { + "scenario": "online", + "num_runs": 3, + "warmup_runs": 1, + "parallelism": { + "tensor_parallel_size": 1, + "pipeline_parallel_size": 1, + "expert_parallel_size": 1, + "data_parallel_size": 1 + }, + "extra_config": null, + "runtime_metrics": null + }, + "metrics": { + "online": { + "sla_ttft_ms": 500, + "max_valid_qps": 10, + "results_by_qps": [ + { + "target_qps": 10, + "achieved_qps": 10.0, + "ttft_ms_p50": 26.63, + "ttft_ms_p90": 54.54, + "ttft_ms_p99": 74.03, + "tpot_ms_p50": 5.59, + "tpot_ms_p90": 11.86, + "tpot_ms_p99": 14.52, + "elapsed_seconds_median": 32.3, + "sla_met": true, + "power_watts_avg": 150.62, + "power_watts_peak": 195.7, + "ttft_ms_p99_reliability": { + "n": 3, + "mean": 232.36, + "std": 302.02, + "cv_pct": 129.98, + "stability": "high-variance", + "runs": [ + 38.8, + 580.37, + 77.91 + ] + } + }, + { + "target_qps": 40, + "achieved_qps": 40.0, + "ttft_ms_p50": 75.21, + "ttft_ms_p90": 121.15, + "ttft_ms_p99": 3224.45, + "tpot_ms_p50": 69.14, + "tpot_ms_p90": 89.08, + "tpot_ms_p99": 136.22, + "elapsed_seconds_median": 22.2, + "sla_met": false, + "power_watts_avg": 99.1, + "power_watts_peak": 123.0, + "ttft_ms_p99_reliability": { + "n": 3, + "mean": 3088.26, + "std": 475.95, + "cv_pct": 15.41, + "stability": "high-variance", + "runs": [ + 3481.14, + 3224.63, + 2559.02 + ] + } + } + ] + } + }, + "accuracy": { + "subset_score": null, + "baseline_delta": null, + "valid": false, + "notes": "Run --scenario accuracy to check model accuracy." + }, + "meta": { + "submitted_by": "Keke946", + "submission_type": "individual", + "date": "2026-08-21", + "time": "22:52:25", + "run_id": "d00caeb0", + "run_name": "nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0", + "flagged": null, + "reproduce_script": "runners/nvidia_vllm_47f5d58e/runner.py", + "env_info_file": "../env_info.json", + "log_file": "run.log", + "samples_file": "samples.jsonl", + "notes": null, + "benchmark_start_time": "2026-08-21T22:49:35.004169+00:00", + "benchmark_end_time": "2026-08-21T22:52:25.528759+00:00", + "benchmark_elapsed_minutes": 2.8, + "model_load_seconds": 39.9 + } +} \ No newline at end of file diff --git a/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/result.json b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/result.json new file mode 100644 index 00000000..a9b26d2f --- /dev/null +++ b/results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/result.json @@ -0,0 +1,284 @@ +{ + "schema_version": "1.0", + "suite_id": "suite_F", + "implementation_id": "nvidia_vllm_47f5d58e", + "chip": { + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "count": 1, + "card_count": 1, + "dies_per_card": 1, + "memory_gb": 48.0, + "interconnect_intra_node": null, + "interconnect_inter_node": null + }, + "environment": { + "collected_at": "2026-08-21T22:45:35.076567+00:00", + "accelerators": [ + { + "index": 0, + "name": "NVIDIA GeForce RTX 4090 D", + "vendor": "NVIDIA", + "memory_gb": 48.0, + "driver_version": "610.43.03", + "firmware_version": null, + "compute_capability": "8.9", + "supports_bf16": true + } + ], + "accelerator_platform": "nvidia", + "accelerator_topology": "\tGPU0\tCPU Affinity\tNUMA Affinity\tGPU NUMA ID\nGPU0\t X \t0-13\t0\t\tN/A\n\nLegend:\n\n X = Self\n SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)\n NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node\n PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)\n PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)\n PIX = Connection traversing at most a single PCIe bridge\n NV# = Connection traversing a bonded set of # NVLinks\n", + "intra_node_interconnect": null, + "cpu": { + "model": "AMD EPYC 7542 32-Core Processor", + "physical_cores": 64, + "logical_cores": 14, + "numa_nodes": 2 + }, + "system_memory_gb": 120.0, + "pcie_generation": "PCIe Gen 4", + "cpu_accelerator_bandwidth_gbs": null, + "network_interfaces": null, + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20", + "kernel_version": "6.8.0-90-generic", + "runtime_version": "CUDA 12.4", + "pytorch_version": "2.5.1+cu124" + }, + "software": { + "framework": "vLLM", + "framework_version": "0.7.3", + "driver_version": "610.43.03", + "runtime_version": "CUDA 12.4", + "os": "Ubuntu 22.04.5 LTS", + "python_version": "3.10.20" + }, + "model": { + "model_id": "Qwen/Qwen2.5-0.5B-Instruct", + "model_revision": "7ae557604adf67be50417f59c2c2f167def9a775", + "model_name": null, + "model_note": null, + "model_source": "huggingface", + "architecture": "dense", + "parameter_count_b": 0.5, + "precision": "BF16", + "effective_dtype": "bfloat16", + "quantization_method": null, + "model_format": "HuggingFace original" + }, + "task": { + "scenarios_run": [ + "offline", + "online", + "interactive" + ], + "parallelism": { + "tensor_parallel_size": 1, + "pipeline_parallel_size": 1, + "expert_parallel_size": 1, + "data_parallel_size": 1 + }, + "num_runs": 3, + "extra_config": null + }, + "metrics": { + "derived": { + "tokens_per_sec_per_watt": 66.4702, + "tokens_per_joule": 66.4702, + "energy_joules": 630.38 + }, + "offline": { + "results_by_concurrency": [ + { + "client_concurrency": 4, + "throughput_tokens_per_sec": 10860.17, + "throughput_tokens_per_sec_per_chip": 10860.17, + "throughput_tokens_per_sec_total": 16145.21, + "elapsed_seconds_median": 3.9, + "peak_memory_gb": 40.36, + "power_watts_avg": 161.83, + "power_watts_peak": 190.2, + "oom": false, + "throughput_tokens_per_sec_reliability": { + "n": 3, + "mean": 10485.91, + "std": 743.29, + "cv_pct": 7.09, + "stability": "noisy", + "runs": [ + 10860.17, + 9629.88, + 10967.67 + ] + }, + "_throughput_note": "output_only", + "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs." + }, + { + "client_concurrency": 16, + "throughput_tokens_per_sec": 11026.74, + "throughput_tokens_per_sec_per_chip": 11026.74, + "throughput_tokens_per_sec_total": 16392.85, + "elapsed_seconds_median": 3.8, + "peak_memory_gb": 40.36, + "power_watts_avg": 165.89, + "power_watts_peak": 196.9, + "oom": false, + "throughput_tokens_per_sec_reliability": { + "n": 3, + "mean": 10548.13, + "std": 882.8, + "cv_pct": 8.37, + "stability": "high-variance", + "runs": [ + 11088.27, + 11026.74, + 9529.38 + ] + }, + "_throughput_note": "output_only", + "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs." + }, + { + "client_concurrency": 64, + "throughput_tokens_per_sec": 10820.87, + "throughput_tokens_per_sec_per_chip": 10820.87, + "throughput_tokens_per_sec_total": 16086.8, + "elapsed_seconds_median": 3.9, + "peak_memory_gb": 40.36, + "power_watts_avg": 169.31, + "power_watts_peak": 198.3, + "oom": false, + "throughput_tokens_per_sec_reliability": { + "n": 3, + "mean": 10761.02, + "std": 161.99, + "cv_pct": 1.51, + "stability": "stable", + "runs": [ + 10884.57, + 10577.62, + 10820.87 + ] + }, + "_throughput_note": "output_only", + "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs." + } + ] + }, + "online": { + "sla_ttft_ms": 500, + "max_valid_qps": 10, + "results_by_qps": [ + { + "target_qps": 10, + "achieved_qps": 10.0, + "ttft_ms_p50": 26.63, + "ttft_ms_p90": 54.54, + "ttft_ms_p99": 74.03, + "tpot_ms_p50": 5.59, + "tpot_ms_p90": 11.86, + "tpot_ms_p99": 14.52, + "elapsed_seconds_median": 32.3, + "sla_met": true, + "power_watts_avg": 150.62, + "power_watts_peak": 195.7, + "ttft_ms_p99_reliability": { + "n": 3, + "mean": 232.36, + "std": 302.02, + "cv_pct": 129.98, + "stability": "high-variance", + "runs": [ + 38.8, + 580.37, + 77.91 + ] + } + }, + { + "target_qps": 40, + "achieved_qps": 40.0, + "ttft_ms_p50": 75.21, + "ttft_ms_p90": 121.15, + "ttft_ms_p99": 3224.45, + "tpot_ms_p50": 69.14, + "tpot_ms_p90": 89.08, + "tpot_ms_p99": 136.22, + "elapsed_seconds_median": 22.2, + "sla_met": false, + "power_watts_avg": 99.1, + "power_watts_peak": 123.0, + "ttft_ms_p99_reliability": { + "n": 3, + "mean": 3088.26, + "std": 475.95, + "cv_pct": 15.41, + "stability": "high-variance", + "runs": [ + 3481.14, + 3224.63, + 2559.02 + ] + } + } + ] + }, + "interactive": { + "ttft_ms_p50": 16.75, + "ttft_ms_p90": 30.95, + "ttft_ms_p99": 34.42, + "tpot_ms_p50": 2.8, + "tpot_ms_p90": 3.63, + "tpot_ms_p99": 3.76, + "peak_memory_gb": null, + "elapsed_seconds_median": 90.2, + "ttft_ms_p99_reliability": { + "n": 3, + "mean": 27.01, + "std": 7.24, + "cv_pct": 26.81, + "stability": "high-variance", + "runs": [ + 24.69, + 21.22, + 35.14 + ] + }, + "power_watts_avg": 203.66, + "power_watts_peak": 219.5 + } + }, + "accuracy": { + "subset_score": 0.4, + "baseline_delta": 0.02, + "valid": true, + "framework": "vLLM", + "precision": "BF16", + "notes": "Integrated accuracy check \u2014 used same vLLM instance as benchmark." + }, + "meta": { + "submitted_by": "Keke946", + "submission_type": "individual", + "date": "2026-08-21", + "time": "22:48:32", + "run_id": "d00caeb0", + "run_name": "nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0", + "flagged": null, + "reproduce_script": "runners/nvidia_vllm_47f5d58e/runner.py", + "env_info_file": "../env_info.json", + "log_file": "run.log", + "samples_file": "samples.jsonl", + "notes": null, + "benchmark_start_time": "2026-08-21T22:47:42.908336+00:00", + "benchmark_end_time": "2026-08-21T22:48:32.501802+00:00", + "benchmark_elapsed_minutes": 8.4, + "model_load_seconds": 52.6, + "benchmark_elapsed_minutes_note": "Total across ['offline', 'online', 'interactive'] scenarios.", + "scenario_dirs": { + "offline": "results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/offline", + "online": "results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/online", + "interactive": "results/community/nvidia_geforce_rtx_4090_dx1_suite_F_nvidia_vllm_47f5d58e_d00caeb0/interactive" + } + } +} \ No newline at end of file