Measure a self-hosted AI assistant with three separate questions: how quickly one person gets a response, how much work the server handles across concurrent requests, and how much electricity and operating cost that useful output consumes. A single tokens-per-second result cannot answer all three. A repeatable benchmark fixes the model and serving configuration, uses a disclosed workload, reports latency and aggregate capacity separately, and ties every energy or cost figure to a clear measurement boundary.
What to measure—and what each result means
Benchmark results are meaningful only when readers can tell whether they describe one user’s experience, the server’s total capacity, or resource efficiency. These are related, but not interchangeable.
As an Amazon Associate I earn from qualifying purchases.
| Measure | What it describes | How to report it |
|---|---|---|
| Time to first token (TTFT) | Wait from request submission until the first streamed output appears. | Report a distribution, such as median and p95; lower usually feels more responsive. |
| Inter-token latency (ITL) | Time gaps between streamed output events, indicating response pacing. | State the tool’s measurement point and formula. With speculative decoding, one output event can contain multiple tokens, affecting comparison with TPOT. |
| Time per output token (TPOT) | Decode time amortized over output tokens after the first token. | Name the engine’s formula. vLLM notes that TPOT and ITL can diverge, including with speculative decoding; see its benchmarking CLI documentation. |
| End-to-end latency | Total time for a request to finish. | Report percentiles, not only a mean. Include p95 or p99 when the number of samples supports them. |
| Aggregate output throughput | Output tokens generated across requests divided by benchmark duration. | Give the request concurrency or rate. This is system capacity, not an individual user’s token speed; NVIDIA’s AIPerf metrics reference distinguishes per-user from aggregate throughput. |
| Energy efficiency | Energy consumed for generated output, or output produced per unit of energy. | Identify whether the reading is GPU-only or whole-system. Possible measures include energy per output token or tokens per joule. |
| Cost per million useful output tokens | An accounting estimate for a defined operating period and workload. | Specify included costs, electricity tariff, utilization, and hardware-amortization assumptions. |
Latency metrics answer how a request feels; aggregate throughput answers how much the server can generate for all users. A system can achieve high aggregate throughput by serving concurrent requests while each request waits longer or streams more slowly. Do not substitute one category for another.
Set up a benchmark that can be repeated
1. Record the configuration
Write down the assistant model and exact revision, tokenizer, serving engine and version, GPU and CPU, RAM, accelerator count, precision or quantization, context limit, and generation settings. Record relevant batching and caching options, too. Keep these fixed when comparing runs unless the setting itself is what you are testing. vLLM’s benchmarking CLI makes choices such as model, endpoint, backend, dataset, and request count explicit.
#1 Best Overall
- EVOLUTION CORE ULTRA 9 285H MINI PC - GMKtec EVO-T1 is the next evolution in AI mini PC Ultra 9 series. The Core Ultra 9 285H offers 16 cores (six P-cores + eight E-cores + two LPE-cores) and 16 threads with a turbo clock of 5.4 GHz. It is currently one of the best value for performance AI mini PC computers.
- AI NPU - The 285H features an Intel AI Boost NPU, capable of up to 13 TOPS (Tera Operations per Second) for INT8 calculations, which is designed to accelerate AI tasks.
- INTEL ARC 140T GAMING PC - The Arc 140T GPU includes 8 Xe cores and supports features like DirectX 12, OpenGL 4.5, and OpenCL 3, making it capable of handling modern games and creative applications. It also supports Quick Sync Video for efficient video encoding and decoding, as well as AV1 encoding and decoding.
- 64GB DDR5 RAM + 1TB SSD - The EVO-T1 is equipped with Dual 32GB (Total 64GB) SO-DIMM DDR5 5600MHz memory sticks. 2TB PCIE 4.0 SSD Drive with 3x M.2 2280 Expansion slots. Each slot capable of reading up to 4TB. (12TB MAX)
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-T1 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and USB Type-C Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
2. Match the workload to real use
Choose prompts and expected answers that resemble the assistant’s actual work. Prompt length, output length, request rate or concurrency, streaming mode, and model configuration all affect the results. Use real prompts only when privacy and licensing allow; otherwise, create a synthetic workload with similar prompt and answer lengths and disclose that choice.
Include a low-load interactive case, then test a range of concurrency or request rates through saturation. Warm up the service, measure a sufficiently long steady interval, and repeat runs. Preserve output-length and sampling settings between comparisons: both influence latency and tokens per second. vLLM supports dataset-driven serving benchmarks and percentile reporting controls in its benchmarking documentation.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall3. Capture request and server-side results
For each load level, record TTFT, ITL, end-to-end latency, aggregate output tokens per second, and request throughput. Include prompt and generated token counts, and report median plus tail percentiles where the sample size makes them useful. Keep per-user speed separate from aggregate throughput. Define ITL or TPOT using the specific engine’s measurement point and formula rather than assuming tools use identical definitions.
Rank #2
- LOW ENERGY HIGH PERFORMANCE MINI PC - The Intel Core Ultra 5 125U is part of the Ultra 5 lineup, using the Meteor Lake architecture with BGA 2049. Intel Hyper-Threading technology is available and effectly doubles the core-count of the P-Cores, to a total of 14 threads. Core Ultra 5 125U has 12 MB of L3 cache and operates at 1300 MHz by default, but can boost up to 4.3 GHz, depending on the workload. With a TDP of 15 W, the Core Ultra 5 125U consumes very little energy but outputs high performance efficiency
- 32GB DDR5 RAM + 512GB SSD - The K15 mini computer is equipped with Dual 16GB (Total 32GB) SO-DIMM DDR5 4800MHz memory sticks. 512GB PCIE 4.0 SSD Drive with 3x M.2 2280 Expansion slots. Each slot capable of reading up to 8TB. (24TB MAX)
- QUAD SCREEN 4K DISPLAY SUPPORT - K15 Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and USB Type-C Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support
- OCULINK PORT - The Oculink port on the rear interface enables higher bandwidth capabilities, better frame rates and lower lag. The standard also operates at PCIe x4 speeds, compared to Thunderbolt's x3. Gamers and content creators can benefit from Oculink's higher bandwidth, resulting in better performance and lower lag for eGPU setups
- DUAL NIC FAST 2.5GBE + WIFI 6E + BT 5.2 - Dual Ethernet 2.5GbE LAN port design provides more applications, such as firewall, multichannel aggregation, soft routing, file storage server. Built-in WIFI 6E / Bluetooth 5.2 is more stable and efficient to connect multiple wireless devices such as projector, printer, monitor, speakers and etc
Use server telemetry to explain results, not just rank systems. Queue time, prefill time, decode time, token counts, and successful, failed, or aborted requests can reveal where time and capacity go. NVIDIA documents examples of such measures in its AIPerf server metrics reference. Telemetry names and interfaces are implementation-specific, so record the engine and version alongside them.
Measure electricity and calculate cost with a clear boundary
Separate GPU energy from whole-system electricity
GPU power or energy describes the accelerator, not the electricity used by the full host. For GPU-level efficiency, collect average GPU power or total GPU energy and calculate a compatible measure such as output tokens per joule, energy per output token, or tokens per second per watt. NVIDIA’s AIPerf metrics reference documents GPU-level power and energy metrics, including total GPU energy, energy per output token, and output tokens per second per watt.
To estimate the host’s electricity use, measure whole-system wall energy across a representative interval with a suitable wall meter or metered power source. State what the interval includes: idle periods, warm-up, steady serving, and service overhead can materially change the result. Apply the electricity tariff that actually applies to the reader’s location and plan; do not treat a GPU-only reading as the host’s bill.
Recommended Free Tools
Rank #3
- Entry-level NAS Personal Storage：UGREEN NAS DH2300 is your first and best NAS made easy. It is designed for beginners who want a simple, private way to store videos, photos and personal files, which is intuitive for users moving from cloud storage or external drives and move away from scattered date across devices. This entry-level NAS 2-bay perfect for personal entertainment, photo storage, and easy data backup (doesn't support Docker or virtual machines).
- Set Your Devices Free, Expand Your Digital World: This unified storage hub supports massive capacity up to 64TB.*Storage drives not included. Stop Deleting, Start Storing. You can store 22 million 3MB images, or 2 million 30MB songs, or 43K 1.5GB movies or 67 million 1MB documents! UGREEN NAS is a better way to free up storage across all your devices such as phones, computers, tablets and also does automatic backups across devices regardless of the operating system—Window, iOS, Android or macOS.
- The Smarter Long-term Way to Store: Unlike cloud storage with recurring monthly fees, a UGREEN NAS enclosure requires only a one-time purchase for long-term use. For example, you only need to pay $459.98 for a NAS, while for cloud storage, you need to pay $719.88 per year, $2,159.64 for 3 years, $3,599.40 for 5 years. You will save $6,738.82 over 10 years with UGREEN NAS! *NAS cost based on DH2300 + 12TB HDD; cloud cost based on 12TB plan (e.g. $59.99/month).
- Blazing Speed, Minimal Power: Equipped with a high-performance processor, 1GbE port, and 4GB RAM on Board, this NAS handles multiple tasks with ease. File transfers reach up to 125MB/s—a 1GB file takes only 8 seconds. Don't let slow clouds hold you back; they often need over 100 seconds for the same task. The difference is clear.
- Let AI Better Organize Your Memories: UGREEN NAS uses AI to tag faces, locations, texts, and objects—so you can effortlessly find any photo by searching for who or what's in it in seconds. It also automatically finds and deletes similar or duplicate photo, backs up live photos and allows you to share them with your friends or family with just one tap. Everything stays effortlessly organized, powered by intelligent tagging and recognition.
Make the cost-per-token calculation auditable
Define the numerator before calculating cost per million useful output tokens. An electricity-only estimate uses measured whole-system energy multiplied by the applicable tariff, divided by useful output over the same interval. A broader operational estimate can also include hardware amortization over an explicitly assumed service life and recurring hosting or maintenance costs. Keep electricity-only and fully loaded estimates separate, and disclose utilization and every included cost.
There is no universal self-hosted cost-per-token figure: tariff, utilization, workload, hardware, and accounting boundary differ. A benchmark cost from another deployment should not be transferred to a different machine or operating pattern without recalculating it.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Include quality before choosing a winning configuration
Faster generation is not an improvement if the assistant no longer performs the intended task adequately. Run a fixed task-quality set alongside load tests and state the evaluation method and acceptance threshold. Then choose an operating point on the throughput, latency, energy, and quality trade-off that meets the assistant’s real response-time and task requirements.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Rank #4
- [Powerful PC] Gaming PC equipped with Core i9-14900F, 24 Cores 32 Threads, 36M Cache, Max Turbo Frequency: 5.8GHz, Windows 11 pro (64 Bit). With GeForce RTX 50 Series GPUs. Adopting DLSS 4 technology, it dramatically improves frame rate performance, supports FP4 low-precision computing, and doubles the efficiency of AI inference. SD graph generation speed is 3 times faster than RTX 4070 Super, significantly increasing creative productivity. Graphics work productivity has increased significantly.
- [High Speed DDR5 RAM & PCIE4.0 SSD] The desktop computer is equipped with Dual-DDR5 RAM (dual channel DDR5 high-speed memory, which can support up to 128GB RAM), 1 x M.2 2280 PCIE4.0 high-speed SSD, and support add 2 x 2.5-inch SATA HDD/SSD(not include) is enough to accommodate system files and massive games, Excellent reading and writing speed greatly shortening your boot time.
- [8K@60Hz Quad-Display] Desktop PC with GeForce RTX 5070 12G GDDR7, supporting DLSS 4, ray tracing, and AI cores. Easily connect 4 monitors via 1×HDMI 2.1 + 3×DP 1.4a — all ports support 8K@60Hz. Delivers stunning visuals and ultra-smooth performance for home entertainment, live streaming, video editing, AI workloads, 3D rendering, and AAA gaming.
- [Functional Interfaces] Mini computer is equipped with 4 x USB 3.2, 4 x USB2.0, 1 x HDMI2.1 port, 3 x DP ports, 2xRJ-45 Gigabit Network Ethernet, 1 x Fiber Optic PORT, 1 x Audio in/out. Built-in Bluetooth 5.4 and IEEE 802.11be wifi 7, Higher transfer rates and lower latency. Mini PC supports multiple device connection and can be used with servers, monitoring equipment, office equipment, projectors, televisions, etc, Mini desktop computer support automatic power on and Wake On Lan.
- [Warranty & Liquid Cooling] Warrant: 2 year/24 months. The compact computer size: 11.6*9.3*3.9in, 9.25lb, Chassis built-in 2 large copper fans, built-in liquid cooling device, to further enhance the computer heat dissipation, and at the same time can reduce noise, give full play to the overall performance of the computer.
Published benchmark results are comparable only with their workload and quality context attached. NVIDIA’s MLPerf AI Benchmarks and inference performance results illustrate the importance of retaining system, dataset, and target-accuracy context alongside throughput.
Compare systems on the same terms
When evaluating two or more configurations, use the same workload, measurement interval, quality test, and reporting boundary. Compare the results in context rather than collapsing them into one tokens-per-second ranking.
- Interactive response: TTFT and tail latency at low or typical user load.
- Serving capacity: aggregate output throughput and request throughput at stated concurrency or request rate.
- Efficiency and cost: energy per useful output and cost with electricity, utilization, and included operating costs specified.
- Task performance: quality method and target accuracy or acceptance threshold.
- Operational fit: memory capacity, stability, and the configuration needed to meet the workload.
No first-hand result or universal cost figure is established here; the useful output of the method is a reproducible measurement with its workload, system, quality, and energy boundaries made explicit.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →
