NVIDIA GeForce GTX 670MX vs NVIDIA GeForce GTX 980M Comparison

The NVIDIA GeForce GTX 980M and the NVIDIA GeForce GTX 670MX are both end-of-life mobile graphics solutions from NVIDIA, but they represent two very different eras of the company’s laptop GPU design. The GTX 980M is built on the Maxwell 2.0 architecture and hails from the GeForce 900M generation, while the GTX 670MX uses the older Kepler architecture from the GeForce 600M generation. The data available for direct comparison is limited to two compute-oriented tests, but the results are decisive: the GTX 980M leads the GTX 670MX by a massive margin in both Geekbench OpenCL and Vulkan workloads. This analysis breaks down those head-to-head results, explores where each GPU holds an advantage, and details the architectural and specification gaps that explain the performance chasm.

Head-to-Head Benchmarks

The only two benchmark tests shared between the GTX 980M and the GTX 670MX are Geekbench OpenCL and Geekbench Vulkan. In both instances, the GTX 980M is the clear winner, and the margins are substantial. In the Geekbench OpenCL test, the GTX 980M scores 23,832 points, while the GTX 670MX manages just 6,125 points. This translates to a deltaPct of 289.1%, meaning the GTX 980M is roughly three times faster in this compute workload. The gap is not a small incremental improvement; it is a generational leap that dwarfs any architectural refinements seen between the two parts.

The Vulkan result tells a similar story. The GTX 980M posts a score of 17,703, against 5,316 for the GTX 670MX. The deltaPct here is 233%, which again places the GTX 980M at more than three times the performance level of the older GPU. It is importantly the GTX 980M’s Vulkan score is lower than its OpenCL score, while the GTX 670MX shows a similar pattern but with a smaller absolute difference. The overall picture is unambiguous: in every measured head-to-head test, the GTX 980M wins, and it wins by a wide margin.

Looking at the broader benchmark context, the average benchmark score for the GTX 980M is 5,308, while the GTX 670MX averages 5,721. This is a curious inversion, as the GTX 670MX actually holds a higher average score despite losing the two direct comparisons. The explanation lies in the limited benchmark sets. The GTX 980M has a far more extensive suite of tests, including PassMark DirectX 9/10/11/12, PassMark G2D, PassMark G3D, and PassMark GPU Compute, none of which are present for the GTX 670MX. The GTX 670MX’s average is based solely on its two Geekbench scores, which are relatively high compared to its overall capability. In contrast, the GTX 980M’s average is dragged down by low PassMark DirectX scores (35 for DX10, 57 for DX11, 31 for DX12, 125 for DX9) despite a strong PassMark G3D score of 7,338. The percentile rankings also reflect this: the GTX 670MX sits at the 33rd percentile against all GPUs, while the GTX 980M is at the 31st percentile, a marginal difference that does not align with the head-to-head dominance.

Where Each One Wins

The data is one-sided, but a use-case split still emerges from the available benchmarks. The GTX 980M wins every direct comparison, so it is the obvious choice for any workload that leverages OpenCL or Vulkan compute. The Geekbench OpenCL score of 23,832 suggests strong general-purpose GPU compute performance, which is relevant for tasks like video encoding, physics simulations, or any application that offloads parallel math to the GPU. The Vulkan score of 17,703 indicates that the GTX 980M is also well-suited for modern graphics APIs that rely on low-level access to the hardware, which is common in recent game engines and compute-heavy visual effects.

The GTX 670MX does not win any of the head-to-head benchmarks, so there is no measured scenario where it outperforms the GTX 980M. However, its higher average benchmark score (5,721 vs. 5,308) and its 33rd percentile ranking suggest that in the limited set of tests it does run, it delivers respectable compute performance for its era. The GTX 670MX’s Geekbench OpenCL score of 6,125 and Vulkan score of 5,316 are not trivial numbers; they place it in the same ballpark as the GTX 980M’s nearest rivals in the overall database. For instance, the GTX 980M’s nearest rival, the NVIDIA GeForce 930A, has an average score of 5,317, which is nearly identical to the GTX 670MX’s Vulkan score. This implies that the GTX 670MX is competitive with entry-level GPUs from later generations, but it is simply outclassed by the high-end GTX 980M.

For a user prioritizing legacy software that is optimized for Kepler-era compute, the GTX 670MX might still function adequately, but the data does not provide any test where it wins. The only sensible conclusion is that the GTX 980M is the superior part for all measured workloads, and the GTX 670MX is best suited for older applications where its lower performance is acceptable.

Architecture Differences

The architectural gap between these two GPUs is vast. The GTX 980M uses the GM204 chip, built on the Maxwell 2.0 architecture, while the GTX 670MX uses the GK104 chip, based on the older Kepler architecture. Both are manufactured on a 28 nm process at TSMC, but the similarities end there. The GM204 die is significantly larger at 398 mm², packing 5,200 million transistors, which yields a transistor density of 13.1M per mm². The GK104 is smaller at 294 mm² with 3,540 million transistors, resulting in a lower density of 12.0M per mm². This difference in transistor count and density is a primary driver of the performance gap, as the GTX 980M simply has more hardware resources to work with.

The Maxwell 2.0 architecture introduces several efficiency improvements over Kepler, particularly in terms of instruction scheduling and memory compression. The GTX 980M features 1,536 shading units, 96 texture mapping units (TMUs), and 64 render output units (ROPs), compared to the GTX 670MX’s 960 shading units, 80 TMUs, and 24 ROPs. The ROP count is especially telling: the GTX 980M has nearly three times the pixel processing capability, which directly contributes to its higher pixel rate of 72.13 GPixel/s versus the GTX 670MX’s 12.02 GPixel/s. Similarly, the texture rate is more than double, with the GTX 980M achieving 108.2 GTexel/s against 48.08 GTexel/s for the GTX 670MX.

The floating-point performance also reflects the generational leap. The GTX 980M delivers 3.462 TFLOPS of FP32 compute, while the GTX 670MX manages just 1,153.9 GFLOPS (approximately 1.15 TFLOPS). That is a threefold difference in raw compute throughput, aligning closely with the Geekbench OpenCL deltaPct of 289.1%. API support also differs: the GTX 980M supports DirectX 12_1 and Vulkan 1.4, whereas the GTX 670MX is limited to DirectX 12 (11_0) and Vulkan 1.2.175. This means the GTX 980M is better equipped for modern titles that use DirectX 12 features like conservative rasterization, while the GTX 670MX may struggle with such workloads.

Specification Differences

Several key specifications differ between the two GPUs, and these numbers directly explain the benchmark results. The most obvious difference is in memory configuration. The GTX 980M comes with 8 GB of GDDR5 memory on a 256-bit bus, delivering a bandwidth of 160.4 GB/s. The GTX 670MX has only 3 GB of GDDR5 on a 192-bit bus, yielding 67.20 GB/s of bandwidth. This is a 2.4x difference in memory bandwidth, which is critical for high-resolution textures and compute workloads that frequently access large datasets. The GTX 980M’s memory clock is also higher, with an effective data rate of 5 Gbps versus 2.8 Gbps for the GTX 670MX.

Clock speeds further widen the gap. The GTX 980M runs at a base clock of 1038 MHz with a boost up to 1127 MHz, while the GTX 670MX is locked at 601 MHz for both base and boost. This lower clock speed, combined with fewer shading units, means the GTX 670MX is at a severe disadvantage in any clock-for-clock calculation. The power draw also differs, though the GTX 980M’s TDP is not listed in the data. The GTX 670MX has a TDP of 75 W, which is relatively modest for a GPU of its era. The GTX 980M’s power consumption is not specified, but given its higher performance, it likely draws more power, though this cannot be quantified from the pack.

The bus interface is another notable difference: the GTX 980M uses an MXM-B (3.0) connector, while the GTX 670MX uses PCIe 3.0 x16. This reflects the mobile design philosophy of each generation, where the GTX 980M is intended for modular laptop upgrades. The GTX 980M’s slot width is listed as MXM Module, whereas the GTX 670MX has no slot width data. Both GPUs have no power connectors and use portable device dependent display outputs. The GTX 980M was released on 2014-10-06, with the GTX 670MX following a later release on 2012-09-30, which explains the architectural advantages of the newer part. The GTX 980M’s predecessor is GeForce 800M, while the GTX 670MX’s predecessor is GeForce 500M.

FAQ

Q: Which GPU is faster in Geekbench OpenCL?

A: The NVIDIA GeForce GTX 980M scores 23,832, while the GTX 670MX scores 6,125. The deltaPct is 289.1%, making the GTX 980M nearly three times faster.

Q: What is the memory bandwidth difference between the two?

A: The GTX 980M has 160.4 GB/s bandwidth, while the GTX 670MX has 67.20 GB/s. The GTX 980M’s bandwidth is more than double that of the GTX 670MX.

Q: How do their shading unit counts compare?

A: The GTX 980M has 1,536 shading units, while the GTX 670MX has 960. This means the GTX 980M has 60% more shading units.

Q: What is the average benchmark score for each GPU?

A: The GTX 980M has an average benchmark score of 5,308, while the GTX 670MX has an average of 5,721. The GTX 670MX’s higher average is due to its limited benchmark set.

Q: Which GPU supports a newer version of Vulkan?

A: The GTX 980M supports Vulkan 1.4, whereas the GTX 670MX supports Vulkan 1.2.175. The GTX 980M also supports DirectX 12_1, compared to DirectX 12 (11_0) for the GTX 670MX.

Q: What is the pixel rate for each GPU?

A: The GTX 980M achieves 72.13 GPixel/s, while the GTX 670MX achieves 12.02 GPixel/s. This is a sixfold difference in pixel throughput.

The Verdict

The data is unequivocal: the NVIDIA GeForce GTX 980M is the superior GPU in every measured benchmark. It wins both head-to-head tests by margins of 289.1% and 233%, and it holds commanding advantages in shading units, ROPs, memory bandwidth, and clock speeds. For any user running OpenCL or Vulkan workloads, the GTX 980M is the only rational choice. Its 8 GB of memory and 256-bit bus make it more suitable for modern games and compute applications that demand high bandwidth. The GTX 670MX, while it has a higher average benchmark score due to its limited test set, cannot compete in any direct comparison. It might still serve in legacy applications that do not stress the GPU, but the GTX 980M’s architecture is simply in a different league. The GTX 980M should be selected by anyone needing strong mobile compute performance, while the GTX 670MX is only acceptable for users with no performance requirements beyond basic compute. The percentile rankings (31st for GTX 980M, 33rd for GTX 670MX) are misleading given the head-to-head results, but the direct numbers tell the true story: the GTX 980M dominates.