CPU Outcomes (Common Time in ms)<\/strong><\/td>\n| HP EliteDesk 8 Mini g1a (AMD Ryzen AI 7 Professional 350) (Radeon 860M)<\/b><\/td>\n | Lenovo ThinkStation P3 Extremely SFF Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 4000 Ada)<\/strong><\/td>\nLenovo ThinkStation P3 Tower Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 5000 Ada Technology)<\/strong><\/td>\nLenovo ThinkStation P3 Tiny Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX A1000)<\/strong><\/td>\n<\/tr>\n\n| AI Laptop Imaginative and prescient General Rating<\/td>\n | 103<\/td>\n | 157<\/td>\n | 196<\/td>\n | 184<\/td>\n<\/tr>\n | \n| MobileNet V3<\/td>\n | 1.11 ms<\/td>\n | 0.99 ms<\/td>\n | 0.94 ms<\/td>\n | 0.98 ms<\/td>\n<\/tr>\n | \n| ResNet 50<\/td>\n | 11.97 ms<\/td>\n | 8.52 ms<\/td>\n | 6.83 ms<\/td>\n | 7.47 ms<\/td>\n<\/tr>\n | \n| Inception V4<\/td>\n | 37.27 ms<\/td>\n | 25.57 ms<\/td>\n | 19.66 ms<\/td>\n | 18.94 ms<\/td>\n<\/tr>\n | \n| DeepLab V3<\/td>\n | 58.75 ms<\/td>\n | 31.42 ms<\/td>\n | 24.89 ms<\/td>\n | 24.97 ms<\/td>\n<\/tr>\n | \n| YOLO V3<\/td>\n | 93.94 ms<\/td>\n | 58.88 ms<\/td>\n | 43.43 ms<\/td>\n | 44.46 ms<\/td>\n<\/tr>\n | \n| REAL-ESRGAN<\/td>\n | 4,542.25 ms<\/td>\n | 2,565.34 ms<\/td>\n | 2,019.39 ms<\/td>\n | 2,586.87 ms<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/span>UL Procyon: AI Textual content Technology<\/span><\/h3>\nThe\u00a0Procyon AI Textual content Technology Benchmark<\/a> streamlines AI LLM efficiency testing by offering a concise and constant analysis methodology. It permits repeated testing throughout a number of LLM fashions whereas minimizing the complexity of enormous fashions and the variety of variables. Developed in collaboration with main AI {hardware} suppliers, it optimizes the usage of native AI accelerators to ship extra dependable, environment friendly efficiency assessments. The outcomes measured beneath had been examined utilizing TensorRT.<\/p>\nTextual content-generation efficiency outcomes are fairly related, with decrease scores throughout Phi, Mistral, Llama2, and Llama3 as a result of built-in GPU and lower-power NPU. The EliteDesk Mini g1a produces outputs extra slowly, significantly in time-to-first-token and tokens-per-second. Regardless of that, it passes each check cleanly and delivers stable general efficiency for a compact enterprise desktop.<\/p>\n \n\n\nUL Procyon: AI Textual content Technology<\/strong><\/td>\n| HP EliteDesk 8 Mini g1a (AMD Ryzen AI 7 Professional 350) (Radeon 860M)<\/b><\/td>\n | Lenovo ThinkStation P3 Extremely SFF Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 4000 Ada)<\/strong><\/td>\nLenovo ThinkStation P3 Tower Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 5000 Ada Technology)<\/strong><\/td>\nLenovo ThinkStation P3 Tiny Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX A1000)<\/strong><\/td>\n<\/tr>\n\n| Phi General Rating<\/td>\n | 301<\/td>\n | 2,303<\/td>\n | 4,344<\/td>\n | 1,117<\/td>\n<\/tr>\n | \n| Phi Output Time To First Token<\/td>\n | 7.692 s<\/td>\n | 0.395 s<\/td>\n | 0.218 s<\/td>\n | 0.808 s<\/td>\n<\/tr>\n | \n| Phi Output Tokens Per Second<\/td>\n | 27.219 tokens\/s<\/td>\n | 81.784 tokens\/s<\/td>\n | 160.757 tokens\/s<\/td>\n | 39.379 tokens\/s<\/td>\n<\/tr>\n | \n| Phi General Length<\/td>\n | 152.699 s<\/td>\n | 35.869 s<\/td>\n | 18.356 s<\/td>\n | 27.734 s<\/td>\n<\/tr>\n | \n| Mistral General Rating<\/td>\n | 242<\/td>\n | 1,977<\/td>\n | 4,136<\/td>\n | 3,045<\/td>\n<\/tr>\n | \n| Mistral Output Time To First Token<\/td>\n | 13.678 s<\/td>\n | 0.666 s<\/td>\n | 0.314 s<\/td>\n | 1.398 s<\/td>\n<\/tr>\n | \n| Mistral Output Tokens Per Second<\/td>\n | 17.534 tokens\/s<\/td>\n | 57.102 tokens\/s<\/td>\n | 117.784 tokens\/s<\/td>\n | 27.734 tokens\/s<\/td>\n<\/tr>\n | \n| Mistral General Length<\/td>\n | 249.621 s<\/td>\n | 51.942 s<\/td>\n | 25.142 s<\/td>\n | 107.361 s<\/td>\n<\/tr>\n | \n| Llama3 General Rating<\/td>\n | 206<\/td>\n | 1,759<\/td>\n | 3,611<\/td>\n | 837<\/td>\n<\/tr>\n | \n| Llama3 Output Time To First Token<\/td>\n | 12.731 s<\/td>\n | 0.628 s<\/td>\n | 0.301 s<\/td>\n | 1.412 s<\/td>\n<\/tr>\n | \n| Llama3 Output Tokens Per Second<\/td>\n | 13.308 tokens\/s<\/td>\n | 47.884 tokens\/s<\/td>\n | 96.565 tokens\/s<\/td>\n | 24.323 tokens\/s<\/td>\n<\/tr>\n | \n| Llama3 General Length<\/td>\n | 293.514 s<\/td>\n | 61.189 s<\/td>\n | 30.285 s<\/td>\n | 121.980 s<\/td>\n<\/tr>\n | \n| Llama2 General Rating<\/td>\n | 234<\/td>\n | 1,770<\/td>\n | 3,942<\/td>\n | \u2013<\/td>\n<\/tr>\n | \n| Llama2 Output Time To First Token<\/td>\n | 21.500 s<\/td>\n | 1.106 s<\/td>\n | 0.502 s<\/td>\n | \u2013<\/td>\n<\/tr>\n | \n| Llama2 Output Tokens Per Second<\/td>\n | 8.732 tokens\/s<\/td>\n | 25.617 tokens\/s<\/td>\n | 57.623 tokens\/s<\/td>\n | \u2013<\/td>\n<\/tr>\n | \n| Llama2 General Length<\/td>\n | 461.126 s<\/td>\n | 112.746 s<\/td>\n | 50.269 s<\/td>\n | \u2013<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/span>UL Procyon: AI Picture Technology<\/span><\/h3>\nThe Procyon AI Picture Technology Benchmark<\/a> offers a constant and correct methodology for measuring AI inference efficiency throughout a spread of {hardware}, from low-power NPUs to high-end GPUs. It consists of three checks: Steady Diffusion XL (FP16) for high-end GPUs, Steady Diffusion 1.5 (FP16) for reasonably highly effective GPUs, and Steady Diffusion 1.5 (INT8) for low-power units. The benchmark makes use of the optimum inference engine for every system, guaranteeing truthful and comparable outcomes.<\/p>\nSteady Diffusion 1.5 FP16 completes with out challenge, although the general rating of 169 and the lengthy era time imply Mini G1a is principally supposed as a light-duty image-generation machine. FP16 rendering takes a number of minutes per check go, and the system lacks INT8 and SDXL outcomes resulting from {hardware} class. Nonetheless, the EliteDesk reveals steady outcomes all through the whole image-generation workflow.<\/p>\n \n\n\nUL Procyon:\u00a0AI Picture Technology<\/strong><\/td>\n| HP EliteDesk 8 Mini g1a (AMD Ryzen AI 7 Professional 350) (Radeon 860M)<\/b><\/td>\n | Lenovo ThinkStation P3 Extremely SFF Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 4000 Ada)<\/strong><\/td>\nLenovo ThinkStation P3 Tower Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 5000 Ada Technology)<\/strong><\/td>\nLenovo ThinkStation P3 Tiny Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX A1000)<\/strong><\/td>\n<\/tr>\n\n| Steady Diffusion 1.5 (FP16) \u2013 General Rating<\/td>\n | 169<\/td>\n | 1,316<\/td>\n | 3,734<\/td>\n | 558<\/td>\n<\/tr>\n | \n| Steady Diffusion 1.5 (FP16) \u2013 General Time<\/td>\n | 590.630 s<\/td>\n | 75.986 s<\/td>\n | 26.779 s<\/td>\n | 179.144 s<\/td>\n<\/tr>\n | \n| Steady Diffusion 1.5 (FP16) \u2013 Picture Technology Pace<\/td>\n | 36.914<\/td>\n | 4.749 s\/picture<\/td>\n | 1.674 s\/picture<\/td>\n | 11.196 s\/picture<\/td>\n<\/tr>\n | \n| Steady Diffusion 1.5 (INT8) \u2013 General Rating<\/td>\n | \u2013<\/td>\n | 17,263<\/td>\n | 45,881<\/td>\n | 6,793<\/td>\n<\/tr>\n | \n| Steady Diffusion 1.5 (INT8) \u2013 General Time<\/td>\n | \u2013<\/td>\n | 14.482 s<\/td>\n | 5.449 s<\/td>\n | 36.798 s<\/td>\n<\/tr>\n | \n| Steady Diffusion 1.5 (INT8) \u2013 Picture Technology Pace<\/td>\n | \u2013<\/td>\n | 1.810 s\/picture<\/td>\n | 0.681 s\/picture<\/td>\n | 4.600 s\/picture<\/td>\n<\/tr>\n | \n| Steady Diffusion XL (FP16) \u2013 General Rating<\/td>\n | \u2013<\/td>\n | \u00a01,061<\/td>\n | 3,293<\/td>\n | \u2013<\/td>\n<\/tr>\n | \n| Steady Diffusion XL (FP16) \u2013 General Time<\/td>\n | \u2013<\/td>\n | 565.015<\/td>\n | 182.179 s<\/td>\n | \u2013<\/td>\n<\/tr>\n | \n| Steady Diffusion XL (FP16) \u2013 Picture Technology Pace<\/td>\n | \u2013<\/td>\n | 35.313 s\/picture<\/td>\n | 11.386 s\/picture<\/td>\n | \u2013<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/span>SPECworkstation 4<\/strong><\/span><\/h3>\nThe SPECworkstation 4.0 benchmark is a complete device for evaluating all key elements of workstation efficiency. It offers a real-world measure of CPU, graphics, accelerator, and disk efficiency, enabling professionals to make knowledgeable choices about their {hardware} investments. The benchmark features a devoted set of checks targeted on AI\/ML workloads, together with information science duties and ONNX runtime-based inference checks, reflecting the rising significance of AI\/ML in workstation environments. It encompasses seven trade verticals and 4 {hardware} subsystems, offering an in depth, related measure of as we speak\u2019s workstation efficiency.<\/p>\n SPECworkstation numbers stay towards the decrease finish of every class, as anticipated for an iGPU small-form-factor system. Scores round 1.0 in most verticals present that the EliteDesk is constructed for normal productiveness and light-weight workstation duties.<\/p>\n \n\n\nSPECworkstation 4.0.0 (Greater is best)<\/strong><\/td>\n| HP EliteDesk 8 Mini g1a (AMD Ryzen AI 7 Professional 350) (Radeon 860M)<\/b><\/td>\n | Lenovo ThinkStation P3 Extremely SFF Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 4000 Ada)<\/strong><\/td>\nLenovo ThinkStation P3 Tower Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 5000 Ada Technology)<\/strong><\/td>\nLenovo ThinkStation P3 Tiny Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX A1000)<\/strong><\/td>\n<\/tr>\n<\/thead>\n\n\nPower<\/strong><\/td>\n| 1.01<\/td>\n | 2.11<\/td>\n | 2.49<\/td>\n | 1.33<\/td>\n<\/tr>\n | \nMonetary Companies<\/strong><\/td>\n| 1.02<\/td>\n | 1.52<\/td>\n | 2.44<\/td>\n | 1.62<\/td>\n<\/tr>\n | \nLife Sciences<\/strong><\/td>\n| 1.02<\/td>\n | 2.51<\/td>\n | 2.84<\/td>\n | 1.80<\/td>\n<\/tr>\n | \nMedia & Leisure<\/strong><\/td>\n| 1.06<\/td>\n | 2.28<\/td>\n | 2.69<\/td>\n | 1.89<\/td>\n<\/tr>\n | \nProduct Design<\/strong><\/td>\n| 0.98<\/td>\n | 2.28<\/td>\n | 2.46<\/td>\n | 1.77<\/td>\n<\/tr>\n | \nProductiveness & Improvement<\/strong><\/td>\n| 0.64<\/td>\n | 1.29<\/td>\n | 1.50<\/td>\n | 1.29<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/span>LuxMark<\/span><\/h3>\nLuxMark is a GPU benchmark that makes use of LuxRender, an open-source ray-tracing renderer, to guage a system\u2019s efficiency with extremely detailed 3D scenes. This benchmark is related for assessing the graphical rendering capabilities of servers and workstations, particularly for visible results and architectural visualization functions, the place correct mild simulation is essential.<\/p>\n LuxMark closely favors stronger GPUs, so the EliteDesk\u2019s Hallbench rating of 1,461 and Meals rating of 706 land close to the underside of the chart. The scenes render efficiently, however the instances spotlight the Radeon 860M\u2019s limits in ray-traced workloads. Stable scores for its class, although.<\/p>\n \n\n\nLuxMark (larger is best)<\/strong><\/td>\nHP EliteDesk 8 Mini g1a (AMD Ryzen AI 7 Professional 350) (Radeon 860M)<\/b><\/strong><\/td>\nLenovo ThinkStation P3 Extremely SFF Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 4000 Ada)<\/strong><\/td>\nLenovo ThinkStation P3 Tower Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 5000 Ada Technology)<\/strong><\/td>\nLenovo ThinkStation P3 Tiny Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX A1000)<\/strong><\/td>\n<\/tr>\n<\/thead>\n\n\nHallbench<\/strong><\/td>\n| 1,461<\/td>\n | 15,054<\/td>\n | 26,792<\/td>\n | 7,423<\/td>\n<\/tr>\n | \nMeals<\/strong><\/td>\n| 706<\/td>\n | 5,612<\/td>\n | 12,562<\/td>\n | 2,860<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/span>7-Zip Compression<\/span><\/h3>\nThe 7-Zip Compression Benchmark evaluates CPU efficiency throughout compression and decompression, measuring GIPS (Giga Directions Per Second) and CPU utilization. Greater GIPS and environment friendly CPU utilization point out superior efficiency.<\/p>\n Right here, the EliteDesk posts stable CPU numbers, with complete scores round 90 GIPS throughout compression and decompression. CPU utilization stays excessive and constant throughout each workloads, displaying good sustained efficiency for its measurement. The system trails the bigger machines, however its outcomes are sturdy for a compact workplace desktop with an iGPU.<\/p>\n \n\n\n7-Zip Compression Benchmark (Greater is Higher)<\/strong><\/td>\n| HP EliteDesk 8 Mini g1a (AMD Ryzen AI 7 Professional 350) (Radeon 860M)<\/b><\/td>\n | Lenovo ThinkStation P3 Extremely SFF Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 4000 Ada)<\/strong><\/td>\nLenovo ThinkStation P3 Tower Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX 5000 Ada Technology)<\/strong><\/td>\nLenovo ThinkStation P3 Tiny Gen 2 (Intel Core Extremely 9 285) (NVIDIA RTX A1000)<\/strong><\/td>\n<\/tr>\n\nCompression<\/em><\/strong><\/td>\n<\/tr>\n\nPresent CPU Utilization<\/strong><\/td>\n| 1,501%<\/td>\n | 2,129%<\/td>\n | 1,899%<\/td>\n | 2,030%<\/td>\n<\/tr>\n | \nPresent Ranking\/Utilization<\/strong><\/td>\n| 5.697 GIPS<\/td>\n | 7.590 GIPS<\/td>\n | 6.970 GIPS<\/td>\n | 5.369 GIPS<\/td>\n<\/tr>\n | \nPresent Ranking<\/strong><\/td>\n| 85.524 GIPS<\/td>\n | 161.558 GIPS<\/td>\n | 132.461 GIPS<\/td>\n | 109.008 GIPS<\/td>\n<\/tr>\n | \nEnsuing CPU Utilization<\/strong><\/td>\n| 1,518%<\/td>\n | 2,116%<\/td>\n | 1,894%<\/td>\n | 2,007%<\/td>\n<\/tr>\n | \nEnsuing Ranking\/Utilization<\/strong><\/td>\n| 5.691 GIPS<\/td>\n | 7.689 GIPS<\/td>\n | 7.115 GIPS<\/td>\n | 5,492 GIPS<\/td>\n<\/tr>\n | \nEnsuing Ranking<\/strong><\/td>\n| 86.365 GIPS<\/td>\n | 162.405 GIPS<\/td>\n | 134.748 GIPS<\/td>\n | 110.189 GIPS<\/td>\n<\/tr>\n | \nDecompression<\/em><\/strong><\/td>\n<\/tr>\n\nPresent CPU Utilization<\/strong><\/td>\n| 1,401%<\/td>\n | 2,263%<\/td>\n | 2,191%<\/td>\n | 2,243%<\/td>\n<\/tr>\n | \nPresent Ranking\/Utilization<\/strong><\/td>\n| 6.674 GIPS<\/td>\n | 7.184 GIPS<\/td>\n | 6.937 GIPS<\/td>\n | 6.049 GIPS<\/td>\n<\/tr>\n | \nPresent Ranking<\/strong><\/td>\n| 93.511 GIPS<\/td>\n | 162.532 GIPS<\/td>\n | 152.009 GIPS<\/td>\n | 135.698 GIPS<\/td>\n<\/tr>\n | \n| Ensuing CPU Utilization<\/strong><\/td>\n | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | ![\"HP](\"https:\/\/www.storagereview.com\/wp-content\/uploads\/2025\/12\/StorageReview-HP-Elitedesk-8-Mini-G1a-1-1024x599.jpg\")
<\/p>\n