NVIDIA Quadro RTX 6000 Mobile

Published on 2019-09-03, the NVIDIA Quadro RTX 6000 Mobile is a Turing-architecture mobile GPU manufactured by TSMC on a 12 nm process. The TU102 die contains 18,600 million transistors on a 754 mm² die, yielding a transistor density of 24.7M/mm². The database lists it in the Quadro Turing-M (Tx000) generation with an end-of-life production status. Its predecessor is Quadro Pascal-M and its successor is Ampere-MW. The benchmark record is empty: the benchmarks array contains no entries, avgBenchmarkScore is 0, and percentileVsAllGpus is 50.

Benchmark Performance

Because the benchmarks array contains no entries, there are no measured scores from which to derive performance percentages. The only performance figures available are the specification ceilings. The GPU is equipped with 4608 shading units, 288 texture mapping units, and 96 raster output units. The base clock is 1275 MHz, and the boost clock is 1455 MHz. Those clocks feed a peak FP32 rate of 13.41 TFLOPS, a texture rate of 419.0 GTexel/s, and a pixel rate of 139.7 GPixel/s.

The 13.41 TFLOPS figure is the rated upper limit for conventional shader and FP32 compute work. The 419.0 GTexel/s texture rate is the maximum rate at which the 288 TMUs can process filtered texture data, while the 139.7 GPixel/s pixel rate is the maximum rate at which the 96 ROPs can output pixel data. The 26.82 TFLOPS FP16 (2:1) figure represents a doubled half-precision compute path available for workloads that can use it.

The average benchmark score of 0 is not a measured result; it indicates the absence of logged submissions. The percentileVsAllGpus value of 50 places the product at the midpoint of the database’s global GPU distribution, but with no benchmark entries behind it, that ranking should not be treated as a test-derived performance position. The nearestRivals array is also empty, so no deltaPct values exist to quantify leads or deficits against specific competing parts. In short, the benchmark section for this product is defined by raw resource counts rather than observed application performance.

Memory Subsystem

The memory subsystem is built around a 24 GB GDDR6 frame buffer connected across a 384-bit bus. The memory clock is 1750 MHz, which corresponds to an effective data rate of 14 Gbps, producing 672.0 GB/s of peak bandwidth. These four numbers—24 GB, GDDR6, 384-bit, and 672.0 GB/s—define the memory behavior of the card.

For high-resolution rendering, the 24 GB capacity is the first resource that matters. It determines how much geometry, texture data, and intermediate render target can reside on the GPU simultaneously. If a frame exceeds that capacity, parts of the scene must be streamed or processed in smaller portions, regardless of how fast the memory bus is. The 384-bit bus is the width of the data path between the memory array and the GPU cores. The 672.0 GB/s bandwidth is the rate at which data can cross that path, and it is the ceiling for feeding the 4608 shading units, 288 TMUs, and 96 ROPs.

The 1750 MHz memory clock is the reference frequency, while 14 Gbps effective is the data-rate equivalent on this GDDR6 configuration. When combined with the 139.7 GPixel/s pixel rate and 419.0 GTexel/s texture rate, the memory subsystem must sustain very large transfers for every high-resolution frame. The exact point at which bandwidth becomes limiting is not recorded in the fact pack, but the combination of a wide bus, a 24 GB pool, and 672.0 GB/s of bandwidth is the dataset’s principal evidence for high-resolution capability.

Ray Tracing and Feature Set

The feature set is defined by dedicated hardware blocks and API support. The TU102 die includes 72 RT cores and 576 tensor cores alongside the 4608 shading units. The RT cores are the specified hardware resources for ray-tracing acceleration. The 576 tensor cores are the specified resources for tensor-oriented compute work, providing a path distinct from the standard shader array.

The supported APIs are DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. DirectX 12 Ultimate (12_2) is the highest feature-level descriptor given for this part. OpenGL 4.6 and Vulkan 1.4 describe the cross-platform graphics and compute interfaces available to applications. The host interface is PCIe 3.0 x16, which is the data link between the GPU and the rest of the system.

The power connector field is None, and the slot width is listed as IGP, which fits the mobile chip design. Display outputs are portable-device dependent, meaning external display support is determined by the host system rather than by the GPU itself. The presence of RT cores, tensor cores, and the listed API set makes the product feature-complete for hardware-accelerated ray tracing and tensor workloads, even though no benchmark entries confirm how those features perform in applications.

How It Compares

The database provides no nearestRivals entries for this product. As a result, there are no rival names, scores, or deltaPct values to include in a per-rival comparison. No numerical “ahead of” or “behind” statements can be made based on the fact pack. The only generation markers are the predecessor Quadro Pascal-M, the successor Ampere-MW, and the release date 2019-09-03. Those markers establish position in the product timeline, not performance relative position.

The end-of-life production status indicates that this product is no longer current, but it does not provide a performance comparison. The avgBenchmarkScore of 0 is the only performance aggregate, and it is empty rather than measured. The percentileVsAllGpus of 50 is the only rank-like value, but with no benchmark entries behind it, it cannot be used as a relative delta.

The absolute specification fields provide the only quantitative comparison information: 24 GB of GDDR6, 672.0 GB/s bandwidth, 13.41 TFLOPS FP32, 26.82 TFLOPS FP16 (2:1), 72 RT cores, and 576 tensor cores. These are not deltas. They do not indicate whether one product is 5% or 50% faster than another. The empty nearestRivals field is definitive: this dataset cannot answer the comparison question numerically.

Who Should Consider It

The absence of benchmark scores means any recommendation must be based on the hardware resources documented in the fact pack. The 24 GB GDDR6 frame buffer and 672.0 GB/s bandwidth make this GPU a candidate for high-resolution workloads that need large memory allocations. The 13.41 TFLOPS FP32 throughput and 26.82 TFLOPS FP16 (2:1) throughput provide compute headroom for standard shader work and half-precision compute paths.

The 72 RT cores and 576 tensor cores extend the feature set beyond pure rasterization to ray tracing and tensor-oriented workloads. The API support for DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 covers multiple modern graphics and compute interfaces. The PCIe 3.0 x16 host interface provides the system-level connection, while the IGP slot width and None power connectors reflect the mobile integration constraints.

Users with workloads that fit within 24 GB and can exploit the 384-bit memory bus are the audience indicated by the specification sheet. High-resolution rendering that requires large texture sets and sizable frame-buffer resources is the most defensible use case based on the available data. Workloads that use ray tracing or tensor acceleration have dedicated hardware in the form of 72 RT cores and 576 tensor cores. Workloads that need FP16 performance have an explicitly listed 26.82 TFLOPS (2:1) path.

What the dataset cannot support is any claim about measured frame rates, quality presets, or relative scores against other GPUs. The avgBenchmarkScore is 0, and nearestRivals is empty. The recommendation is therefore resource-based: if the workload requires large memory capacity, high bandwidth, and dedicated RT/tensor resources, the specification list supports consideration. If measured application performance is essential, this database entry currently contains no benchmark data to confirm it.