NVIDIA GeForce GTX 970M

The NVIDIA GeForce GTX 970M is a Maxwell 2.0 portable-device GPU built on TSMC's 28 nm process. Its GM204 chip houses 5,200 million transistors on a 398 mm² die, for a transistor density of 13.1M per mm², and it sits in the GeForce 900M generation between the GeForce 800M and GeForce 10 Mobile. Released 2014-10-06 and now end-of-life, this MXM-B (3.0) module carries an average benchmark score of 4,655 and sits at the 25th percentile of all GPUs. Its nearest rivals are all within 0.6% in either direction, making the surrounding competitive field unusually tight.

Who Should Consider It

The benchmark pattern points strongly to older API workloads. PassMark DirectX 9 returns 99, while DirectX 10, 11, and 12 return 28, 42, and 24. The DirectX 9 result is the highest of those four by a wide margin, which suggests this GPU is not built around modern API demands. Geekbench OpenCL at 18,991 and Geekbench Vulkan at 18,516 are close, so the compute side is more balanced than the API scores alone suggest. Still, 3DMark Steel Nomad DX12 is only 472, and PassMark G3D is 5,704 against an average of 4,655. Users who want to push very high resolutions should not look here; the 120.3 GB/s memory bandwidth and the low modern API scores place a clear limit on that use case. The more plausible audience is someone running lighter or older titles at moderate settings, where the high DirectX 9 result can be used.

Power and Cooling

The fact pack lists no TDP and no suggested PSU. The only power connector entry is "None." The physical interface is an MXM Module using MXM-B (3.0), and display outputs are portable device dependent. Because no external power connectors appear, the MXM socket is the only electrical interface mentioned. Cooling, likewise, is not specified in the data; it depends on whatever portable device hosts the module. This is not a standalone add-in card with discrete power requirements.

Benchmark Performance

The available benchmark scores are:

| Benchmark | Score |

|---|---|

| 3DMark Steel Nomad DX12 | 472 |

| Geekbench OpenCL | 18,991 |

| Geekbench Vulkan | 18,516 |

| PassMark DirectX 10 | 28 |

| PassMark DirectX 11 | 42 |

| PassMark DirectX 12 | 24 |

| PassMark DirectX 9 | 99 |

| PassMark G2D | 381 |

| PassMark G3D | 5,704 |

| PassMark GPU Compute | 2,289 |

The average benchmark score is 4,655. The nearest rivals are as close as the average suggests: AMD Radeon R8 M445DX averages 4,670 with a delta of -0.3%; NVIDIA Quadro M3000M averages 4,635 with a delta of +0.4%; AMD Radeon R5 M255 averages 4,680 with a delta of -0.5%; and AMD Radeon R7 M260 averages 4,630 with a delta of +0.6%. No rival in this group is more than 0.6% away. The GPU is statistically tied with all four. What separates this GPU from its rivals is not the average, but the shape of the workload scores. Geekbench OpenCL and Vulkan are nearly identical, while PassMark GPU Compute is 2,289 and PassMark G3D is 5,704; the compute result is much lower. The data implies that compute performance lags behind pure 3D rasterization.

FAQ

Q: What APIs does the GTX 970M support?

A: It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

Q: Does it support ray tracing or tensor acceleration?

A: No RT cores and no tensor cores are listed in the data, so hardware ray tracing and tensor acceleration are not part of its feature set.

Q: How much memory and bandwidth does it have?

A: It has 6 GB of GDDR5 memory on a 192-bit bus, with 120.3 GB/s of bandwidth and a memory clock of 1253 MHz / 5 Gbps effective.

Q: What kind of slot does it use?

A: It is an MXM Module with an MXM-B (3.0) bus interface. The power connectors field is "None."

Q: How far is it from its nearest rivals?

A: The nearest rival average scores are 4,670, 4,635, 4,680, and 4,630, with deltas of -0.3%, +0.4%, -0.5%, and +0.6%, respectively.

Q: Is it still in production?

A: No. The production status is end-of-life. It was released 2014-10-06, and the GeForce 10 Mobile is listed as its successor.

Ray Tracing and Feature Set

This is a Maxwell 2.0 architecture part with a GM204 chip. The process is TSMC 28 nm, using 5,200 million transistors on a 398 mm² die, for 13.1M transistors per mm². There are 1,280 shading units, 80 TMUs, and 48 ROPs. The pixel rate is 49.82 GPixel/s, the texture rate is 83.04 GTexel/s, and FP32 throughput is 2.657 TFLOPS. No RT cores and no tensor cores are listed. API support includes DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. Display outputs are portable device dependent, reinforcing that this is a module tied to the host system. The absence of RT and tensor cores is the defining feature-set limitation for any modern ray-traced workload.

Memory Subsystem

The GTX 970M is equipped with 6 GB of GDDR5 memory on a 192-bit bus. The memory clock is 1253 MHz, listed as 5 Gbps effective, and the listed bandwidth is 120.3 GB/s. For high resolutions, the relationship between the 6 GB capacity and the 192-bit bus is what matters. Six gigabytes can hold a large working set, but 120.3 GB/s of bandwidth is the figure that determines how quickly that working set can be delivered. The low modern API scores in PassMark and Steel Nomad suggest that memory bandwidth becomes more important in recent workloads. In short, capacity is not the bottleneck; bandwidth is.

How It Compares

AMD Radeon R8 M445DX: The R8 M445DX has an average score of 4,670, and the GTX 970M trails by 0.3%. With such a small delta, the two are effectively indistinguishable in average benchmark terms.

NVIDIA Quadro M3000M: The Quadro M3000M averages 4,635, with the 970M leading by 0.4%. This is the closest NVIDIA rival in the set, and the average difference is negligible.

AMD Radeon R5 M255: The R5 M255 averages 4,680, putting the 970M 0.5% behind. It is the highest-scoring rival in the list, but the delta is still far too small to be decisive.

AMD Radeon R7 M260: The R7 M260 averages 4,630, and the 970M is 0.6% ahead. This is the largest delta in the group, but it still represents a near-tie in real-world performance terms.