NVIDIA GeForce RTX 3070 Max-Q

The NVIDIA GeForce RTX 3070 Max-Q is a GeForce 30-series mobile GPU from NVIDIA, built on the Ampere architecture with the GA104 chip. Samsung's 8 nm process houses 17,400 million transistors on a 392 mm² die, for a transistor density of 44.4M/mm². The database marks it as End-of-life, with a release date of 2021-01-11, an immediate predecessor of GeForce 20 Mobile, and no listed successor.

Memory Subsystem — VRAM size/type, bus width, bandwidth and what it means for high resolutions

The RTX 3070 Max-Q pairs 8 GB of GDDR6 memory with a 256-bit memory bus. Its memory clock is 1500 MHz, translating to a 12 Gbps effective data rate, and the aggregate bandwidth is 384.0 GB/s. This memory configuration is defined entirely by those quantities in the FACT PACK: the VRAM size, the memory type, the bus width, and the bandwidth are all explicit.

For high-resolution workloads, the 8 GB capacity is the hard working-set limit. Textures, geometry, and render targets that exceed 8 GB cannot remain fully resident in the frame buffer. The 256-bit bus is what allows the GPU to reach 384.0 GB/s; a narrower bus would lower that bandwidth figure. The 80 ROPs combine with the memory system to produce a pixel rate of 103.2 GPixel/s, meaning the GPU can fill a high-resolution image while being fed through a 384.0 GB/s pipe. The 12 Gbps effective memory speed is the data-rate figure attached to the GDDR6 chips themselves.

No benchmark scores are stored for this GPU, so the memory subsystem cannot be connected to measured frame rates at specific resolution presets. The data does, however, establish the resources: 8 GB of VRAM, a 256-bit memory path, and 384.0 GB/s of bandwidth. Those are the concrete numbers that any high-resolution analysis would need.

Ray Tracing and Feature Set

Ray tracing is handled by 40 dedicated RT cores, while matrix and AI-type work is handled by 160 tensor cores. The underlying architecture is Ampere, using the GA104 chip manufactured at 8 nm on Samsung's process. The API support in the FACT PACK is DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

The feature set also includes a full rasterization pipeline: 5,120 shading units, 160 TMUs, and 80 ROPs. Texture rate is 206.4 GTexel/s, and pixel rate is 103.2 GPixel/s. Compute throughput is 13.21 TFLOPS for both FP32 and FP16, at a 1:1 ratio. The PCIe 4.0 x16 bus interface is the host connection. Display outputs are listed as Portable Device Dependent, so the physical monitor connections are determined by the portable device rather than by the GPU definition.

The 40 RT cores and 160 tensor cores are the hardware blocks that support the ray-traced and tensor-oriented features of the DirectX 12 Ultimate API set. The data does not list any vendor-specific feature names, so any deeper feature comparison would have to rely on those core counts and API versions alone.

Benchmark Performance

The benchmark data for this SKU is minimal. The average benchmark score is 0, and the nearestRivals array is empty. Because no rival names, scores, or deltaPct values are present, exact percentage comparisons cannot be calculated from the FACT PACK. There is no basis for a statement such as "product X is 20% faster than product Y" because the database entry provides no rival data.

The only positional field is percentileVsAllGpus, which is 50. This places the RTX 3070 Max-Q at the median of the database's GPU distribution. However, the accompanying average benchmark score is 0, so the percentile is not validated by a stored sample average. The raw throughput fields are the only numeric performance indicators: 13.21 TFLOPS FP32, 13.21 TFLOPS FP16 at 1:1, 206.4 GTexel/s texture rate, and 103.2 GPixel/s pixel rate.

The base clock is 780 MHz, and the boost clock is 1290 MHz. The data does not list a game clock. Combined with 5,120 shading units, these clock values produce the listed 13.21 TFLOPS figure. Without nearestRivals entries, no percentage lead or deficit can be stated, and no specific rival score can be cited.

Power and Cooling

The RTX 3070 Max-Q has a TDP of 80 W. Its power connectors field is None, meaning no auxiliary PCIe power connectors are part of the GPU definition. The database records no suggested PSU, which is consistent with a mobile module rather than a standalone card installation.

The bus interface is PCIe 4.0 x16. No slot width is listed, and no dimensions for length, height, or width are recorded. Display outputs are Portable Device Dependent. Because no cooler model or dimension set is recorded, cooling is defined only by the 80 W TDP and the absence of discrete power connectors. The power delivery path is therefore a board-level design decision inside the portable device, not a user-configured PSU and cable arrangement.

The absence of a suggested PSU value in the FACT PACK means no wattage recommendation can be provided from this database entry. The only concrete power-related numbers are 80 W TDP and the connector count of None.

Who Should Consider It

This GPU is aimed at portable systems with an 80 W power envelope that still need 8 GB of GDDR6 on a 256-bit bus with 384.0 GB/s bandwidth. The 40 RT cores and 160 tensor cores make it applicable to ray-traced and tensor-focused workloads within that power class, and the 5,120 shading units provide a 13.21 TFLOPS FP32/FP16 compute ceiling.

Because the database contains no benchmark scores and no nearest rivals, this page cannot recommend a specific resolution-and-settings combination based on measured frame rates. The data instead supports a resource-based assessment: 8 GB VRAM, 256-bit memory bus, 384.0 GB/s bandwidth, and 103.2 GPixel/s pixel rate are the available image-generation assets. The 50th percentile rank is a midpoint position, but it carries no nonzero score, so it should not be read as a performance guarantee.

The End-of-life production status and 2021-01-11 release date also indicate that this is an older 30-series mobile offering. The predecessor is GeForce 20 Mobile, and no successor is listed in the database.

FAQ

Q: What memory configuration does the RTX 3070 Max-Q have?

A: It uses 8 GB of GDDR6 on a 256-bit bus, with a 1500 MHz memory clock, a 12 Gbps effective data rate, and 384.0 GB/s bandwidth.

Q: How many RT cores and tensor cores are present?

A: The GPU has 40 RT cores and 160 tensor cores.

Q: Which graphics APIs are supported?

A: The listed APIs are DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What is the TDP and what power connectors are needed?

A: TDP is 80 W, power connectors are None, and no suggested PSU is recorded.

Q: What are the raw compute performance figures?

A: FP32 and FP16 are both 13.21 TFLOPS at a 1:1 ratio; texture rate is 206.4 GTexel/s and pixel rate is 103.2 GPixel/s.

Q: Is benchmark data available for this GPU?

A: The average benchmark score is 0, the nearestRivals list is empty, and the percentile versus all GPUs is 50.

How It Compares

The nearestRivals array in the FACT PACK is empty. This means there are no rival names, no rival scores, and no deltaPct values available for comparison. The database entry therefore cannot place the RTX 3070 Max-Q ahead of or behind any specific competing product.

The only comparison-adjacent data is the predecessor field, which lists GeForce 20 Mobile, and the successor field, which records no successor. The 50th percentile rank is a broader positional signal against all GPUs in the database, but it is not tied to any benchmar score or to a nearest-rival delta. Without nearestRivals data, any specific comparison paragraph would require information outside the FACT PACK, so the only supported conclusion is that no direct rival comparison is recorded for this SKU.