Windows 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 reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
A 421-million-parameter Laya decision model sustained 175 typed decisions per second on an NVIDIA H100 NVL in Bhushan Kinge’s 2026 benchmark. That result came from TensorRT FP16 at a selected operating point with about 91 ms p99 latency, within a 130 ms p99 target. On the same target, the RTX PRO 6000 reached 146 decisions per second and the RTX PRO 5000 reached 42. These are workload-specific benchmark results—not guaranteed speeds for every Laya deployment or a measure of generated tokens.
What the benchmark measured
Laya takes a state and typed questions, then returns structured decisions rather than generating prose. Its English checkpoint is based on ModernBERT-large, has 421 million parameters and lists a 512-token context limit. Accordingly, the reported unit is typed decisions per second, not tokens per second. The Laya model card
For the benchmark, each request presented a public SAM.gov contract-opportunity notice and asked three questions: one choice, one score and one yes/no. The author used a frozen sample of 1,000 notices, open-loop Poisson arrivals and a small HTTP server with dynamic batching. A rate counted as sustained capacity only if the system achieved at least 90% of the offered load, remained within the latency target, returned no errors and avoided a growing queue. Benchmark workload and capacity definition
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →How fast each GPU ran at the tested latency targets
The main comparison below reports capacity under two p99 latency budgets. The 130 ms column is the most useful direct comparison across the three GPUs; results depend on the tested model, workload and serving setup.
#1 Best Overall
- PLEASE NOTE: Exporting an NVIDIA RTX Pro 6000 GPU outside the US requires strict adherence to the U.S. Export Administration Regulations (EAR) and issuance of an export license from the Bureau of Industry and Security (BIS). Compliance and Know Your Customer (KYC) screening may be required as a condition of order acceptance. [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations.
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
| GPU and backend | Capacity at p99 ≤ 50 ms | Capacity at p99 ≤ 130 ms |
|---|---|---|
| RTX PRO 5000, TensorRT FP16 | 15 decisions/s | 42 decisions/s |
| RTX PRO 6000, TensorRT FP16 | Not measured; the tested sweep did not establish capacity below 50 requests/s | 146 decisions/s |
| H100 NVL, TensorRT FP16 | 105 decisions/s | 175 decisions/s; measured p99 was about 91 ms at this selected operating point |
The author also converted the sustained rates at the 130 ms target into daily totals: 3.6 million decisions for the RTX PRO 5000, 12.6 million for the RTX PRO 6000 and 15.1 million for the H100 NVL. These figures extrapolate a sustained rate across a day; they were not separate 24-hour tests. Reported GPU capacity results
What the test setup does—and does not—show
Hardware and software coverage
The tested devices included an 8 GB RTX 2000 Ada laptop GPU, a 24 GB RTX PRO 5000 Blackwell laptop GPU, a 96 GB RTX PRO 6000 Blackwell workstation GPU, and a 94 GB H100 NVL. The H100 was tested both as a whole GPU and split into seven 1g.12gb MIG instances.
Backends and configurations included PyTorch eager in FP32, FP16 and BF16; torch.compile with max-autotune; ONNX Runtime CUDA; and TensorRT FP16. The study also provided hosted Jev API and Qwen3.5 vLLM context baselines. This breadth helps explain the reported results, but it does not make the capacity numbers transferable to different prompts, languages, checkpoints or production serving systems. Benchmark configuration and methodology
Latency targets matter
At a stricter p99 target of 50 ms, the H100 NVL sustained 105 decisions per second, compared with 175 at the 130 ms target. A higher latency budget can allow more batching and throughput, but it is only useful if the service can tolerate the resulting tail latency.
Rank #2
- NVIDIA Ampere Architecture-based CUDA Cores - Double-speed processing for single-precision floating point (FP32) operations and improved power efficiency provide significant performance improvements for graphics and simulation workflows, such as complex 3D computer-aided design (CAD) and computer-aided engineering (CAE), on the desktop.
- Second-Generation RT Cores - With up to 2X the throughput over the previous generation and the ability to concurrently run ray tracing with either shading or denoising capabilities, second-generation RT Cores deliver massive speedups for workloads like photorealistic rendering of movie content, architectural design evaluations, and virtual prototyping of product designs. This technology also speeds up the rendering of ray-traced motion blur for faster results with greater visual accuracy.
- Third-Generation Tensor Cores - New Tensor Float 32 (TF32) precision provides up to 5X the training throughput over the previous generation to accelerate AI and data science model training without requiring any code changes. Hardware support for structural sparsity doubles the throughput for inferencing. Tensor Cores also bring AI to graphics with capabilities like DLSS, AI denoising, and enhanced editing for select applications.
- Third-Generation NVIDIA NVLink - Increased GPU-to-GPU interconnect bandwidth provides a single scalable memory to accelerate graphics and compute workloads and tackle larger datasets.
- 48 Gigabytes (GB) of GPU Memory - Ultra-fast GDDR6 memory, scalable up to 96 GB with NVLink, gives data scientists, engineers, and creative professionals the large memory necessary to work with massive datasets and workloads like data science and simulation.
The RTX PRO 6000’s capacity at the 50 ms target is unknown: the sweep did not test below 50 requests per second. It would be incorrect to infer a 50 ms result from its 146 decisions per second at the more permissive 130 ms target.
A replay illustrates why operating headroom matters
In one RTX PRO 6000 replay, 138,863 requests—10 million individual decisions—completed with zero errors and an overall p99 of 111 ms. Two peak-hour segments had p99 latency of 132 and 143 ms. The benchmark author estimated that keeping a continuously enforced 130 ms limit at that volume would require roughly 25% headroom. This is a single replay, not a guarantee about ongoing production performance. Replay and serving details
Why backend rankings change with serving conditions
A backend that wins a fixed-shape throughput test may not win in a dynamic server, where request timing, batching and latency limits affect the result. In fixed-shape tests at larger batches, torch.compile max-autotune FP16 was about 1.3–1.7 times faster than eager FP16 on each card, according to the benchmark author. But torch.compile was not tested as a serving backend.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesDynamic-serving results varied by GPU and workload. On the H100, TensorRT raised capacity at p99 ≤ 130 ms from 93 to 175 decisions per second. On the RTX PRO 6000, eager FP16 and TensorRT both reached 146. For the Blackwell laptop’s multilingual checkpoint, eager FP16 outperformed TensorRT under the same service-level objective. That last result also shows why the English procurement workload cannot stand in for multilingual performance generally. Backend comparisons and serving qualifications
Rank #3
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
What the H100 MIG result means
Splitting the H100 NVL into seven 1g.12gb MIG instances provided isolation, but the configuration did not reliably meet the benchmark’s capacity gate on documents longer than 400 tokens. At the lowest tested aggregate load, the slices reached about 49 decisions per second at p99 127 ms but failed the achieved-rate requirement. The whole H100 reached 175 decisions per second within the 130 ms target.
This result is specific to the tested long-document workload and eager FP16 MIG configuration. It does not establish that MIG is generally slower or unsuitable; shorter prompts could behave differently. MIG configuration and results
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Output fidelity is not task accuracy
Before timing, the author checked whether optimized backends reproduced upstream FP32 answers on a parity set of 16 cases and 63 typed questions. Across four GPUs, three checkpoints, multiple precisions and five backends, all 74 backend-and-device rows passed that suite. Checks also covered public JSON equality, finite outputs and steady-state allocator stability. Parity and benchmark limitations
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Passing parity means the tested configurations matched the upstream model outputs on that limited suite. It does not show that Laya’s answers are correct for real procurement decisions or other real-world tasks.
Rank #4
- Discrete graphics card memory 40 GB
- Memory bandwidth (max) 1555 GB/s
- Graphics processor family NVIDIA
- Graphics processor A100
How to read the cost estimates
The benchmark estimated self-hosted costs per million decisions at $0.67 for the RTX PRO 5000, $0.66 for the RTX PRO 6000 and $1.86 for the H100 NVL. Those scenario estimates assume three-year card amortization, 100% utilization and electricity at $0.12 per kWh. The estimated Jev API cost was $6.80–$8.20 per million decisions at list token pricing for this workload.
The comparison is not like-for-like on network path: the Jev estimate includes the public internet route from Arizona, while self-hosted Laya does not include a comparable network path. Hosted use also avoids hardware ownership and operational work. Under the benchmark’s cost model, a hosted service may still make economic sense below roughly one million decisions per day. Treat these numbers as assumption-dependent estimates, not current quotes or universal break-even points. Cost model and assumptions
Benchmark limits to keep in mind
- The study used one English federal-procurement workload, not a representative mix of tasks, languages or prompt lengths.
- Each server sweep and replay used one run per configuration, so the figures are not multi-run averages.
- The serving setup was a compact asyncio dynamic batcher over loopback, not Triton; production networking and deployment behavior may differ.
- The test measured output fidelity against upstream FP32 on a small parity set, not real-world task accuracy.
- The H100 figure is a measured operating point, not a theoretical GPU ceiling or a speed guarantee for purchased hardware.
The benchmark is independent, not an official Laya or Convai result. Benchmark author’s disclosure
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.
