Menu
Back previous/next report: k j

Intel 1-node-4x-BMG-B50

Submitted by Intelon 2026-02-27 Published on 2026-03-16

Submission Info
Submitter Intel
Submission ID intel-Intel-submission-2
Submission Status
Submission Date 2026-02-27
Publication Date 2026-03-16
Division Closed
Model
Model Name Llama-3.1-8B
Division Closed
Model Precision INT4
Calibration Link GitHub
Model Notes INT4 precision weights
Datasets
Dataset Name CNN_DailyMail
Dataset Type Accuracy + Performance
Average Input Tokens 870
Dataset Link Hugging Face

System TPS vs. Tokens/s per User

0102030System TPS010203040Tokens/s per User

System TPS vs. Concurrency

0102030System TPS05101520# of queries

Time to First Token P95 vs. Concurrency

05001.0k1.5k2.0k2.5kTTFT P95 (s)05101520# of queries

Tokens/s per User vs. Concurrency

010203040Tokens/s per User05101520# of queries
SUT Summary
Intel
System Name 1-node-4x-BMG-B50
System Availability Available
System Category Datacenter · —
System Size 4x Arc B50
Node Types 1
Total Node Count 1
Max Supported Concurrency
Node Type ID 1
Node Count 1
Software
Ubuntu
Framework VLLM-0.11, vllm-xpu-kernels-0.0.1
Serving Framework VLLM-0.11, vllm-xpu-kernels-0.0.1
Inference Backend
Driver
Link to Container
Operating System Ubuntu 25.04
Filesystem EXT4
Other Software OneAPI® 2025.2
Software Notes
Hardware

Processor

Processor Model Name Intel(R) Xeon(R) Platinum 8480CTDX
Processors per Node 2
Cores Per Processor 56
VCPUs Per Processor 112

Accelerator

Accelerator Model Name Intel(R) Arc Pro(R) B50
Accelerators per Node 4
Memory Type GDDR6
Memory Capacity 16GB
Accelerator Interconnect
Host-Accelerator Interconnect PCIe Gen5 x8

Host / Storage

Host Memory Capacity 1TB
Memory Configuration 16x64GB DDR5 4800MTS
Storage Capacity 1.1TB
Storage Type SSD, EXT4 mounted disk storage

Networking

Network Interfaces 1, 1GB/s
Network Details Ethernet (on-board)
Network Topology

Other

Other Hardware
Cooling Air-cooled
Hardware Notes
Run Data
Field NameRun 1Run 2Run 3Run 4Run 5Run 6Run 7Run 8Run 9Run 10Run 11
Run Date02/25/202602/25/202602/25/202602/24/202602/24/202602/24/202602/24/202602/24/202602/24/202602/24/202602/24/2026
Concurrency0000011351021
System Tokens/Second0.81.53.05.69.815.822.628.933.034.334.1
Tokens/Second per User37.537.236.834.430.524.517.511.26.43.31.7
TTFT P95 (ms)22283.722669.023326.825250.927298.629959.834743.143103.561597.1863516.72319031.9
Utilization2.2%4.4%8.7%16.2%28.7%46.2%65.9%84.5%96.4%100.0%99.5%
Configuration SummaryDP=1 TP=1DP=2 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1DP=4 TP=1
Config Type
Config Region
Target QPS
Result State
Measured Accuracy Score
Benchmark Version
Result Version
Archive URI
QPS0.00590.01170.02320.04340.07690.12360.17640.22610.25800.26770.2665
Total Output Tokens1,711,1021,711,1021,711,1021,711,1021,711,1021,711,1021,711,1021,711,1021,711,1021,711,1021,711,102
Average Output Tokens
Run Duration (s)45,583.4623,023.5811,622.336,211.963,505.222,182.111,528.921,192.601,045.121,007.161,011.90
Total Queries13,36813,36813,36813,36813,36813,36813,36813,36813,36813,36813,368
Total Queries Issued
Total Queries Failed
Time to First Token (TTFT) P95 (ms)
Minimum2853.02789.62948.33196.03766.44002.54724.46043.57391.18280.5139440.2
Average11170.811588.512018.113256.114154.015585.618785.524378.038803.5216163.11791993.7
P509548.710727.410837.111690.212511.813337.216099.519949.728654.071737.51778153.4
P9018411.819277.920267.021765.223382.825727.029788.236443.750932.1728659.32210943.7
P9522283.722669.023326.825250.927298.629959.834743.143103.561597.1863516.72319031.9
P9926908.727402.529185.231673.736086.239429.647732.668134.4451034.71194731.53113226.7
P99930889.431295.534471.343249.755020.2176489.8277677.9430841.1773465.81561933.03510064.9
Maximum97148.2138855.5135940.0131553.2150596.1227363.1333904.2516537.2855429.31646303.73696981.3
Time Per Output Token (TPOT) (ms)
Minimum1206.41210.31207.71209.11225.31270.71484.31610.41812.02560.52736.4
Average1243.01253.21262.71346.21525.31913.82704.54257.57482.813160.015477.1
P501219.51227.61229.71324.61509.41895.52681.74238.97568.013714.615686.2
P901308.51324.51353.01481.11713.62179.53155.85076.59076.216005.415964.9
P951415.31377.51399.11530.11773.22268.23300.65326.99663.216136.616310.0
P991569.11514.21489.71626.81889.92440.23602.55869.111001.316527.017347.7
P9991619.21673.81600.21725.62014.62689.93963.46498.412398.017795.918150.7
Maximum1648.71746.01730.91846.32208.83111.74213.56725.112907.318637.719061.6
Query Latency (ms)
Minimum156356.0156856.4157695.9159173.1163560.2174309.1212693.1228987.9252268.4363538.81944601.4
Average169025.7170747.1172380.9184225.2207867.9258640.5362257.6565084.4989118.21887483.03757582.9
P50165713.2168435.6169881.0181998.5205649.8256044.2358827.1560736.8996886.71882921.63765010.8
P90181493.3183122.0186334.3202945.5233677.7294345.7422170.0671468.71198526.42709295.24202160.8
P95191650.0188899.7192346.3209639.4242036.7306292.2441772.7705595.01283463.22858369.64300685.2
P99211857.3204773.7204855.8224111.4258775.9331074.9492153.7801334.91521684.73038176.85084278.0
P999224347.5226696.7219617.3242265.8289769.4443582.8642870.11039406.81799376.33344532.95485008.0
Maximum251524.2314254.8313454.2329896.9362202.9523346.8680619.31113839.41897840.03441031.15658673.0