NVIDIA GeForce GTX 670M vs NVIDIA RTX A400 Comparison

The NVIDIA GeForce GTX 670M and NVIDIA RTX A400 are separated by twelve years of GPU architecture, yet both occupy similar performance percentiles in the current database. The GTX 670M, a Fermi 2.0-era mobile part from 2012, scores a 6,513 in Geekbench OpenCL, placing it at the 38th percentile of all GPUs. The RTX A400, an Ampere workstation card from 2024, averages 6,078 across all its benchmark scores, sitting at the 35th percentile. Despite the A400’s massive generational advantage in raw specifications, its average score is dragged down by poor DirectX 10/11/12 results, while the GTX 670M has only a single OpenCL benchmark to its name. This creates a fascinating paradox: the newer card is dramatically faster in compute workloads but appears comparable in overall rankings due to benchmark mix.

Where Each One Wins

The RTX A400 dominates compute-oriented tasks. Its Geekbench OpenCL score of 22,844 is 3.5 times higher than the GTX 670M’s 6,513, a 71.5% delta in favor of the Ampere card. This is the A400’s clear territory: general-purpose GPU compute, OpenCL acceleration, and any workload that leverages its 768 shading units against the GTX 670M’s 336. The A400 also wins on memory bandwidth, delivering 96.00 GB/s versus 72.00 GB/s, which matters for data-intensive tasks.

The GTX 670M wins by default in legacy DirectX scenarios — but only because the A400’s PassMark DirectX 10 and 11 scores are astoundingly low at 32 and 37, respectively. The GTX 670M has no DirectX benchmark scores in the FACT PACK, so it cannot be directly compared there. However, the A400’s PassMark DirectX 9 score of 87 and DirectX 12 score of 27 suggest that the GTX 670M, with its older architecture, would likely outperform the A400 in these legacy APIs if tested. The GTX 670M also holds an edge in transistor density efficiency relative to its era, but that is an architectural trait, not a benchmark win.

For professional workstation use, the A400 is the only viable option: it supports DirectX 12 Ultimate (12_2), Vulkan 1.4, and features 6 RT cores and 24 tensor cores. The GTX 670M caps out at DirectX 12 (11_0) and has no Vulkan support listed. The A400’s 4 GB GDDR6 memory doubles the GTX 670M’s 1,536 MB, and its 28.19 GPixel/s pixel rate is 3.4 times higher. In every measurable modern workload, the A400 wins.

Architecture Differences

The GTX 670M uses the GF114 chip on TSMC’s 40 nm process, packing 1,950 million transistors into a 332 mm² die. The RTX A400 uses the GA107 chip on Samsung’s 8 nm process, fitting 8,700 million transistors into just 200 mm². This is a density revolution: 5.9M transistors per mm² for Fermi versus 43.5M per mm² for Ampere — a 7.4x improvement. The A400’s 8 nm node allows for 768 shading units, 24 TMUs, and 16 ROPs, whereas the GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. Interestingly, the GTX 670M has more TMUs and ROPs than the A400, but the A400 compensates with higher clocks (1,417 MHz base, 1,762 MHz boost versus no listed clocks for the GTX 670M) and superior memory technology.

The A400 introduces dedicated hardware the GTX 670M lacks entirely: 6 RT cores for ray tracing and 24 tensor cores for AI acceleration. The GTX 670M’s FP32 throughput is 803.7 GFLOPS, while the A400 delivers 2.706 TFLOPS — a 3.37x advantage. The A400 also offers FP16 at 1:1 ratio (2.706 TFLOPS), while the GTX 670M has no listed FP16 capability. Memory configurations differ drastically: the GTX 670M uses 1,536 MB of GDDR5 on a 192-bit bus, while the A400 uses 4 GB of GDDR6 on a 64-bit bus. The A400’s narrower bus is offset by faster memory (12 Gbps effective versus 3 Gbps), resulting in 96.00 GB/s bandwidth versus 72.00 GB/s.

The interface also tells a generational story: the GTX 670M is an MXM-B (3.0) module with portable-device-dependent display outputs, while the A400 is a single-slot PCIe 4.0 x8 card with four mini-DisplayPort 1.4a outputs. Power draw favors the newer card: 50 W TDP for the A400 versus 75 W for the GTX 670M, despite the A400’s vastly higher compute capability. The A400 also lists a suggested PSU of 250 W, while the GTX 670M has none.

FAQ

Q: Why does the RTX A400 have a lower average benchmark score than the GTX 670M despite being far more powerful?

A: The A400’s average of 6,078 includes several PassMark DirectX scores that are extremely low (27 to 87), which pull its average down. The GTX 670M’s average of 6,513 comes solely from its Geekbench OpenCL score of 6,513. The A400’s OpenCL score alone is 22,844, but the legacy DirectX results disproportionately skew its overall average.

Q: Which card has better memory bandwidth?

A: The RTX A400 wins with 96.00 GB/s from 4 GB of GDDR6 on a 64-bit bus. The GTX 670M offers 72.00 GB/s from 1,536 MB of GDDR5 on a 192-bit bus. The A400 achieves higher bandwidth despite a narrower bus due to its 12 Gbps effective memory clock versus the GTX 670M’s 3 Gbps.

Q: Does the GTX 670M support ray tracing or tensor cores?

A: No. The GTX 670M has no RT cores or tensor cores listed in its specifications. The RTX A400 includes 6 RT cores and 24 tensor cores, enabling hardware-accelerated ray tracing and AI workloads.

Q: What is the performance gap in OpenCL compute?

A: The RTX A400 scores 22,844 in Geekbench OpenCL, which is 71.5% higher than the GTX 670M’s 6,513. In absolute terms, the A400 is 3.5 times faster in this test.

Q: Are both cards still in production?

A: No. The GTX 670M is marked as end-of-life, having been released on March 21, 2012. The RTX A400 is active, released on April 15, 2024, and remains available.

Q: Which card supports Vulkan?

A: Only the RTX A400 lists Vulkan support (version 1.4). The GTX 670M has no Vulkan version listed in its API specifications.

Specification Differences

The two cards differ across nearly every specification field. The GTX 670M uses a 40 nm process and GF114 chip, while the A400 uses 8 nm and GA107. Transistor counts are 1,950 million versus 8,700 million; die sizes are 332 mm² versus 200 mm². The A400 has a base clock of 1,417 MHz and boost of 1,762 MHz; the GTX 670M has no listed base or boost clocks. Memory speed is 3 Gbps effective for the GTX 670M versus 12 Gbps for the A400. Shading units: 336 versus 768. TMUs: 56 versus 24. ROPs: 24 versus 16. The A400 adds 6 RT cores and 24 tensor cores; the GTX 670M has none. Pixel rate: 8.372 GPixel/s versus 28.19 GPixel/s. Texture rate: 33.49 GTexel/s versus 42.29 GTexel/s. FP32: 803.7 GFLOPS versus 2.706 TFLOPS. FP16: not listed for the GTX 670M versus 2.706 TFLOPS for the A400. TDP: 75 W versus 50 W. Slot width: MXM Module versus Single-slot. Bus interface: MXM-B (3.0) versus PCIe 4.0 x8. Display outputs: Portable Device Dependent versus 4x mini-DisplayPort 1.4a. DirectX support: 12 (11_0) versus 12 Ultimate (12_2). Vulkan: none versus 1.4. Production status: End-of-life versus Active. Release date: March 21, 2012 versus April 15, 2024.

Head-to-Head Benchmarks

The only direct head-to-head benchmark available is Geekbench OpenCL, where the RTX A400 delivers a decisive victory. The A400 scores 22,844 against the GTX 670M’s 6,513, a delta of -71.5% from the A400’s perspective (meaning the GTX 670M is 71.5% slower). This is the largest single-swing result in the comparison, and it underscores the A400’s compute dominance. In FP32 throughput, the A400’s 2.706 TFLOPS versus 803.7 GFLOPS represents a 3.37x advantage. The A400 also wins on pixel rate (28.19 vs 8.372 GPixel/s), texture rate (42.29 vs 33.49 GTexel/s), and memory bandwidth (96.00 vs 72.00 GB/s).

However, the GTX 670M holds structural advantages that do not appear in head-to-head scores. It has 56 TMUs versus the A400’s 24, and 24 ROPs versus 16. Its 192-bit memory bus is three times wider than the A400’s 64-bit bus, though faster GDDR6 compensates. The GTX 670M’s 1,950 million transistors on a larger die suggest a more balanced rasterization pipeline, which likely explains why the A400’s PassMark DirectX scores (32 for DX10, 37 for DX11, 27 for DX12) are so low — the A400 appears to sacrifice legacy raster performance for compute and RT capabilities. The GTX 670M’s sole benchmark score of 6,513 suggests it would outperform the A400 in DirectX 9-12 workloads, but no direct data exists to confirm this.

The Verdict

The data points to a clear split: the RTX A400 is the superior card for any modern compute, OpenCL, ray tracing, or AI-accelerated workload. Its 22,844 OpenCL score, 2.706 TFLOPS FP32, and 96.00 GB/s bandwidth make it the only choice for professionals running CUDA-based or Vulkan applications. The A400’s 50 W TDP and single-slot form factor also make it more deployable in workstations, with a suggested 250 W PSU.

The GTX 670M, despite being end-of-life, holds value only as a legacy part for older DirectX 9/10/11 titles or as a curiosity in MXM-based laptops. Its 336 shading units and 56 TMUs suggest capable rasterization for its era, but its 803.7 GFLOPS FP32 and 72.00 GB/s bandwidth are dwarfed by the A400. The GTX 670M’s 38th percentile ranking versus the A400’s 35th is misleading — the A400’s average is depressed by nonsensical PassMark DirectX scores (32, 37, 27) that do not reflect its compute capability. If you need a modern workstation GPU with RT and tensor cores, the A400 is the obvious pick. If you are restoring a 2012-era laptop, the GTX 670M is the only option that fits. For everything else, the A400’s 3.5x OpenCL lead makes the verdict unambiguous.