To raise LLM inference throughput, tune the amount of token work and the number of active requests scheduled per iteration—but measure the resulting latency at realistic load. Larger limits can improve GPU utilization and aggregate tokens per second, yet may worsen time to first token (TTFT) or the gaps between generated tokens. The right settings depend on your model, hardware, prompt and output lengths, traffic pattern, cache behavior, and latency targets.
What continuous batching controls
Continuous batching is an online scheduling approach: requests arrive and finish at different times, and the server decides which prefill and decode work to run in each iteration. Unlike a fixed batch that waits for every sequence to finish, it can add new work as capacity becomes available. TensorRT-LLM calls this in-flight batching and also describes it as continuous or iteration-level batching. Its documentation says the implementation uses packed inputs with padding removed: NVIDIA TensorRT-LLM in-flight batching.
Two limits matter, but they are not interchangeable:
- Token-work limit: In vLLM,
max_num_batched_tokenscaps the tokens processed in one iteration. TensorRT-LLM’smax_num_tokenscaps packed input tokens in a batch after padding is removed. - Active-request or sequence limit: vLLM’s
max_num_seqscaps sequences processed in an iteration. TensorRT-LLM’smax_batch_sizecontrols the number of runtime requests the engine can schedule.
These similarly named controls have engine-specific semantics; do not assume the values map one-to-one. In vLLM, queued-request and queued-prompt-token limits are separate admission controls, not per-iteration batch limits. See the vLLM CLI reference.
#1 Best Overall
- Read Before You Buy — No Video Output: These adapters support charging and USB 2.0 data transfer, but cannot transmit video signals. Except for standard USB webcams (which use USB data only), they are not compatible with HDMI/DisplayPort cables, video-capable USB-C hubs, or docking stations with video output.
- Convert USB-A Ports to USB-C: Designed to connect USB-C earphones, cables, flash drives, card readers, and other USB-C accessories to standard USB-A ports. Plug-and-play with no drivers or software required.
- Aluminum Alloy Housing: Built with a sturdy aluminum alloy shell that aids in heat dissipation and protects against daily wear and scratches. Designed to maintain a stable and secure connection.
- Compact & Travel-Friendly: The ultra-compact design allows the adapter to stay plugged into your device without blocking adjacent ports or adding bulk, reducing wear and tear on your original USB ports.
- 12-Month Warranty: Backed by a 12-month manufacturer warranty for peace of mind. Designed to meet strict quality control standards for reliable everyday performance.
Establish a baseline before changing limits
Record the deployment configuration and workload so that a change can be compared fairly. At minimum, capture:
- Serving framework and exact version, model and precision, GPU type and count, and tensor or pipeline parallelism.
- Prompt and output token-length distributions, request arrival pattern, target concurrency, and whether prefix or other cache reuse is expected.
- Current scheduler limits and admission controls.
- Output-token throughput and request throughput, alongside TTFT, inter-token latency (ITL) or time per output token (TPOT), and relevant tail percentiles.
- The service-level objectives (SLOs) that the configuration must meet.
Use the same model, hardware, workload, arrival pattern, concurrency, and cache conditions when comparing settings. Otherwise, a throughput difference may reflect a changed test rather than a scheduler improvement. The vLLM benchmarking guide emphasizes that benchmark metric terminology is not standardized, so check what each metric measures and where timing begins and ends.
Rank #2
- 5-in-1 USB-C Hub: Experience comprehensive connectivity featuring a Power Delivery input, two USB-A 2.0 ports, a USB-A 3.0 port, and an HDMI port. (Note: The USB-C power delivery input port is only for connecting an external wall charger to power your laptop and cannot power peripheral devices.)
- 90W Pass-Through Charging: Achieve optimal charging with 90W pass-through power to your laptop, supported by a total input of 100W, with the hub reserving 10W for operational efficiency. (Note: Wall charger not included.)
- Quick Data Transfers: Accelerate your productivity with rapid data transfers using a high-speed 5Gbps USB 3.0 port and two 480Mbps USB 2.0 ports.
- 4K HDMI Display: Enhance your visual experience with a hub capable of delivering 4K resolution at 30Hz in both mirror and extend modes. Please note that this hub is compatible with MacBook (macOS 12 and newer), Windows 10 and 11, ChromeOS, and laptops equipped with DP Alt Mode and Power Delivery. Note: This device is not compatible with Linux.
- What You Get: Anker USB-C Hub (5-in-1, 4K HDMI), welcome guide, 18-month warranty, and our friendly customer service.
Tune the token budget for your workload
The token budget controls how much work can be scheduled in an iteration. A smaller budget can limit prefill work that competes with ongoing decode, while a larger one can advance more prompt tokens and increase aggregate throughput. The best balance depends on the relative amount of prefill and decode in your traffic.
In its v0.22.1 optimization guide, vLLM gives 2,048 as an example of a smaller max_num_batched_tokens value that favors ITL by limiting competing prefill work. The same guide says higher values allow more prefill tokens per batch and can improve TTFT; it recommends values above 8,192 for optimal throughput, especially for smaller models on large GPUs. Those are version-specific documentation recommendations, not universal thresholds or guarantees for another model or serving stack. Consult the vLLM optimization guide for the behavior of that version.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Rank #3
- Sleek 7-in-1 USB-C Hub: Features an HDMI port, two USB-A 3.0 ports, and a USB-C data port, each providing 5Gbps transfer speeds. It also includes a USB-C PD input port for charging up to 100W and dual SD and TF card slots, all in a compact design.
- Flawless 4K@60Hz Video with HDMI: Delivers exceptional clarity and smoothness with its 4K@60Hz HDMI port, making it ideal for high-definition presentations and entertainment. (Note: Only the HDMI port supports video projection; the USB-C port is for data transfer only.)
- Double Up on Efficiency: The two USB-A 3.0 ports and a USB-C port support a fast 5Gbps data rate, significantly boosting your transfer speeds and improving productivity.
- Fast and Reliable 85W Charging: Offers high-capacity, speedy charging for laptops up to 85W, so you spend less time tethered to an outlet and more time being productive.
- What You Get: Anker USB-C Hub (7-in-1), welcome guide, 18-month warranty, and our friendly customer service.
For TensorRT-LLM, its batching guidance says a larger max_num_tokens can let more requests run together and raise GPU utilization. But utilization eventually plateaus, and excessively high values may hurt TTFT and end-to-end latency. Choose a token ceiling high enough to improve token throughput and hardware utilization without violating the latency SLO.
Use chunked prefill when prompt work competes with decode
A long prompt can consume substantial prefill work. Without chunking, that work can take up an iteration while requests already generating wait for GPU time. Chunked prefill divides prompt processing into smaller pieces so prompt work can share iterations with decode work.
Rank #4
- Dual Converters, Infinite Potential：Includes 2× USB C male to USB A female adapters and 2× USB A male to USB C female adapters. Perfect for a wide range of uses—tablets with Bluetooth keyboards, expand USB ports on macbook, and more. Two different converters for all your daily needs
- Next-Level 10Gbps & 3A Charging: No more slow 480Mbps, this usb to usb c adapter has a transfer speed of up to 10Gbps, allowing you to do more transferring in less time. This usb adapter fits both USB A and USB C charger, supporting up to 3A fast charging
- Upgraded Exquisite Craftsmanship: With an aluminum alloy housing and metal connector, the usbc to usb adapter is extremely durable and sturdy. Rigorously tested to withstand more than 10,000 times of plugging and unplugging, ensuring long-lasting performance
- Broad Compatible: The usb c to usb adapter widely supports all USB C/ USB A devices like laptops, tablets, cellphones, car chargers, and phone chargers. Such as compatible with MacBook Pro/Air 2023/2022, Thunderbolt 4/3 Devices,Apple MagSafe Watch 9/8/7/SE/Ultra, iPad Pro 2022/2021, Samsung Galaxy S23/S20/S10, and iPhone 17/16/15 Pro. Plug and play
- Please Note: To reach 10Gbps speed, keep the cable under 3.3 ft. For USB A Male to USB C adapters, try flipping the USB C connector. USB C Male to USB A adapters support bidirectional 10Gbps transfer within 3.3 ft
The vLLM v0.22.1 guide describes chunked prefill as a way to balance compute-bound prefill with memory-bound decode. For the V1 policy described on that page, pending decode requests are prioritized and prefill uses the remaining token budget. This can help mixed workloads, but the details are version-specific; verify the policy and configuration supported by the vLLM release you deploy.
Keep admission limits separate from scheduler limits
A long queue does not mean the per-iteration token or sequence limits are too low. In vLLM, queued-request and queued-prompt-token controls affect which incoming requests the API server admits or keeps waiting; they do not replace max_num_batched_tokens or max_num_seqs. Tune admission limits for overload handling and capacity or quality-of-service policy, and tune iteration limits for scheduling work already admitted.
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 →Best Value
- 5-in-1 Connectivity: Equipped with a 4K HDMI port, a 5 Gbps USB-C data port, two 5 Gbps USB-A ports, and a USB C 100W PD-IN port. Note: The USB C 100W PD-IN port supports only charging and does not support data transfer devices such as headphones or speakers.
- Powerful Pass-Through Charging: Supports up to 85W pass-through charging so you can power up your laptop while you use the hub. Note: Pass-through charging requires a charger (not included). Note: To achieve full power for iPad, we recommend using a 45W wall charger.
- Transfer Files in Seconds: Move files to and from your laptop at speeds of up to 5 Gbps via the USB-C and USB-A data ports. Note: The USB C 5Gbps Data port does not support video output.
- HD Display: Connect to the HDMI port to stream or mirror content to an external monitor in resolutions of up to 4K@30Hz. Note: The USB-C ports do not support video output.
- What You Get: Anker 332 USB-C Hub (5-in-1), welcome guide, our worry-free 18-month warranty, and friendly customer service.
Benchmark settings under matched conditions
- Fix the workload. Use a representative request set with realistic prompt and output lengths. Decide whether cache reuse is part of the intended workload and keep that condition consistent.
- Control cache state between runs. The vLLM benchmarking guide describes changing the seed, resetting or restarting the server, or using its serving sweep tool to reset caches between runs.
- Match offered load and concurrency. Use the same arrival rate and concurrency for each candidate. vLLM’s serving benchmark supports an infinite request rate for maximum-throughput stress, as well as finite rates with burstiness controls for more controlled or production-like arrivals. Its
max-concurrencyoption can model a gateway or load-balancer limit. - Sweep a small set of limits. Change the token budget and active-sequence/request capacity in measured steps. Keep other settings fixed so you can attribute differences to the limits being tested.
- Compare the tradeoff, not one headline number. Record aggregate output tokens per second and requests per second alongside TTFT, ITL or TPOT, and tail latency. Keep only configurations that meet the relevant latency targets; among those, compare throughput.
For vLLM’s benchmark guide, TTFT is measured from sending a request to receiving its first streamed output. ITL is the gap between consecutive streamed outputs. TPOT is calculated per request as (end-to-end latency minus TTFT) divided by (output tokens minus one). A one-token request can produce different TPOT treatment in benchmark summaries and Prometheus histograms: the guide says benchmark statistics exclude those requests, while the histogram records TPOT as zero. Interpret results using the metric definition and measurement point, not just the metric name.
Interpret maximum-throughput results carefully
TensorRT-LLM’s benchmark workflow prepares a dataset, builds an engine where required, then runs either a maximum-throughput or a low-latency test. Its maximum-throughput tool submits requests as fast as possible in offline mode and describes the result as an upper-bound throughput figure. That test can help identify capacity, but it is not a substitute for measuring finite arrival rates and user-facing latency SLOs.
One NVIDIA documentation example reports 28,390.4265 tokens/sec and 221.8002 requests/sec for Llama 3.1 8B using TensorRT-LLM 0.17.0. The example used 3,000 requests averaging 128 input tokens and 128 output tokens, displayed a maximum runtime batch size of 4,096 and maximum runtime token count of 8,192, and has a log date of 2025-01-18. It illustrates why a throughput figure needs its workload and configuration attached; it is not a general performance expectation or a current-release claim. See the TensorRT-LLM benchmarking guide.
Choose a production setting from the latency-throughput tradeoff
Test candidate settings at a load and request mix that resemble the deployment, then choose a point that meets TTFT and token-latency targets while improving aggregate throughput. If the service misses latency objectives as the token budget rises, lower it or revisit chunking and admission policy. If utilization and throughput have already plateaued, a higher ceiling may add latency without delivering useful capacity. Re-run the comparison after changing the model, hardware, serving release, or workload mix, because the result is specific to that stack.
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.
