NVIDIA Tesla M6 X2 Mobile

Benchmark Performance

The NVIDIA Tesla M6 X2 Mobile occupies a distinctly mid-pack position in the database, with a percentile ranking of 50 against all GPUs. This means exactly half of all tracked graphics processors deliver higher benchmark scores, while the other half fall below. The average benchmark score for this unit is recorded as 0, which indicates that no standardized benchmark submissions exist for this specific SKU in the current dataset; therefore, all performance interpretation must be derived from its architectural specifications and the percentile placement.

The raw compute figures paint a clear picture of a Maxwell 2.0-era part. The GM204 chip, fabricated on TSMC's 28 nm process, houses 5,200 million transistors across a 398 mm² die, yielding a transistor density of 13.1 million per square millimeter. With 1,536 shading units, 96 texture mapping units, and 64 raster output units, the Tesla M6 X2 achieves a pixel rate of 75.52 GPixel/s and a texture rate of 113.3 GTexel/s. Floating-point performance for FP32 operations reaches 3.625 TFLOPS, a figure that situates this mobile module firmly within the upper-midrange of its generation.

Clock behavior shows a base frequency of 930 MHz with a boost ceiling of 1180 MHz. The boost delta of 250 MHz represents a 26.9% headroom above base clocks, which is substantial for a mobile MXM form-factor part. This boost behavior suggests that thermal and power constraints are the primary governors of sustained performance, as the 100 W TDP must be managed within the portable device chassis. The memory clock runs at 1253 MHz, translating to 5 Gbps effective data rate on the GDDR5 interface.

Who Should Consider It

Given the lack of direct benchmark scores, the suitability analysis must rest on the compute and memory specifications. The 3.625 TFLOPS FP32 throughput places this card in a position where it can handle 1080p gaming at medium-to-high settings for titles released around the 2015-2016 window, which aligns with its August 2015 release date. For 1440p resolutions, the 160.4 GB/s memory bandwidth becomes the limiting factor, as texture-heavy scenes will strain the 256-bit bus. Users targeting 4K should look elsewhere; the pixel rate of 75.52 GPixel/s is insufficient to drive modern titles at that resolution without significant graphical compromises.

The 8 GB GDDR5 frame buffer is generous for the era and allows high-resolution texture packs without exceeding VRAM capacity. This makes the card suitable for workstation tasks such as 3D rendering and video editing where large datasets reside in memory. However, the absence of any ray tracing or tensor core hardware means the card is confined to traditional rasterization workloads. For scientific computing or CUDA-accelerated applications that leverage FP32 performance, the 3.625 TFLOPS figure provides a baseline for expected throughput, though the mobile thermal envelope will prevent sustained peak clocks during extended compute sessions.

Memory Subsystem

The memory configuration is notable for its capacity: 8 GB of GDDR5 across a 256-bit bus. The effective memory clock of 5 Gbps (1253 MHz base) yields a total bandwidth of 160.4 GB/s. This bandwidth figure is modest by later standards but was competitive for a mobile solution at launch. The 256-bit bus width allows for efficient memory access patterns, though the bandwidth per byte of VRAM is relatively low at 20.05 GB/s per GB. This ratio suggests that capacity exceeds what the bandwidth can feed in bandwidth-intensive scenarios, meaning frame buffers will rarely overflow, but texture streaming may bottleneck in open-world titles with large draw distances.

At higher resolutions, the 8 GB capacity becomes an asset, preventing stutter caused by texture swapping. Yet the 160.4 GB/s bandwidth imposes a practical ceiling on fill-rate-heavy effects such as anti-aliasing and high-resolution shadows. For 1440p gaming, the memory subsystem can keep pace with the 3.625 TFLOPS compute throughput in most scenarios, but 4K workloads will see the bandwidth starve the shading units, resulting in frame times that reflect memory latency more than compute capability. The pixel rate of 75.52 GPixel/s, derived from the 64 ROPs at boost clock, reinforces this analysis: 64 ROPs are sufficient for 1080p but become a constraint at higher pixel counts.

How It Compares

The nearestRivals array is empty in the FACT PACK, which means no direct comparative scores or delta percentages are available from the database. This absence is itself informative: the Tesla M6 X2 Mobile sits in a segment where few direct contemporaries share the same MXM form factor and thermal envelope. The production status of end-of-life, with a release date of August 29, 2015, places it between the Tesla Kepler predecessor and the Tesla Pascal successor. The architectural jump from Kepler to Maxwell 2.0 brought improved power efficiency per transistor, while the subsequent Pascal generation would introduce simultaneous multi-projection and improved memory compression. Without rival scores, the percentile ranking of 50 must be interpreted as a median position in the full GPU landscape, not against a specific competitive set.

The 28 nm process node, shared with the Kepler generation, means the M6 X2 does not benefit from a die shrink advantage over its immediate predecessor. However, the Maxwell 2.0 architecture's reorganized scheduler and improved occupancy handling deliver higher instructions per clock for the same transistor budget. The lack of benchmark entries in the database suggests this part was aimed at OEM portable workstations rather than the enthusiast retail market, which explains the absence of standardized gaming benchmarks in the dataset.

FAQ

Q: What is the FP32 compute performance of the Tesla M6 X2 Mobile?

A: The card delivers 3.625 TFLOPS of FP32 performance, driven by 1,536 shading units at a boost clock of 1180 MHz.

Q: How much VRAM does the card have and what is its memory bandwidth?

A: It features 8 GB of GDDR5 memory on a 256-bit bus, with a 5 Gbps effective data rate yielding 160.4 GB/s of bandwidth.

Q: What is the pixel and texture fillrate?

A: The pixel rate is 75.52 GPixel/s from 64 ROPs, and the texture rate is 113.3 GTexel/s from 96 TMUs.

Q: Does the card support modern graphics APIs?

A: It supports DirectX 12 (feature level 12_1), OpenGL 4.6, and Vulkan 1.4, covering the API requirements for most titles through the late 2010s.

Q: What is the power draw of this module?

A: The TDP is rated at 100 W, which is the thermal design power for the MXM module in a portable chassis.

Q: When was this product released and what is its current status?

A: It launched on August 29, 2015, and is now listed as end-of-life, with the Tesla Kepler as its predecessor and Tesla Pascal as its successor.

Power and Cooling

The Tesla M6 X2 Mobile carries a TDP of 100 W, a figure that dictates the thermal solution required in any portable device. As an MXM module, the card does not use a standard PCIe slot power connector; the power connectors field is listed as "None," meaning all power is delivered through the MXM connector itself. This simplifies integration but ties the card to proprietary laptop or workstation motherboards. The slot width is designated as "MXM Module," confirming its form factor is not a standard desktop expansion card.

The bus interface is PCIe 3.0 x16, which provides adequate bandwidth for the 8 GB frame buffer and the 160.4 GB/s memory bandwidth. No suggested PSU rating is provided in the database, which is consistent with a mobile part that draws power from a laptop's power brick rather than a desktop PSU. The absence of external power connectors means that peak power delivery is limited by the MXM slot's specifications, which historically cap around 100 W for this form factor. This aligns with the TDP figure, suggesting the card is designed to run at or near its thermal limit under sustained load.

Cooling requirements are therefore substantial for a 100 W module in a compact chassis. The boost clock behavior from 930 MHz to 1180 MHz indicates that thermal headroom directly translates to performance, meaning devices with superior cooling solutions will sustain higher average clocks. The 5,200 million transistors on a 398 mm² die produce a power density that requires efficient heat pipes or vapor chambers in the host system. Given the end-of-life status, replacement thermal paste or fan servicing may be necessary for continued operation in aging portable workstations. The display outputs are listed as "Portable Device Dependent," meaning the video output configuration varies by the host laptop or mobile workstation, further emphasizing the card's role as an OEM-integrated component rather than a user-swappable desktop GPU.