NVIDIA GeForce GTX 670M vs NVIDIA GeForce GTX 880M Comparison

Where Each One Wins

The benchmark data splits these two mobile GPUs in a way that might surprise anyone expecting the newer, larger product to dominate outright. The NVIDIA GeForce GTX 880M carries the newer architecture and far higher specifications on paper, yet the only direct head-to-head test in the database, Geekbench OpenCL, records a win for the older NVIDIA GeForce GTX 670M. The GTX 670M scores 6,513 against the GTX 880M's 5,622, a delta of 13.7% in favor of the 2012 part. That result defines the use-case split: if your workload is OpenCL compute, the GTX 670M is the stronger choice based on recorded measurements.

The GTX 880M, however, claims the higher overall standing in the database. Its average benchmark score sits at 8,040, placing it in the 42nd percentile of all GPUs. The GTX 670M averages 6,513, which lands in the 38th percentile. The newer card also has a second recorded benchmark, Geekbench Metal, where it posts 10,458, a test the GTX 670M does not appear in at all. For anyone running Metal-accelerated tasks, the GTX 880M is the only one of the two with measured results. So the split is clear: the GTX 880M wins on overall database standing and Metal performance, while the GTX 670M wins the one shared compute test.

The practical takeaway is that neither card is a modern powerhouse, both are end-of-life products. But the data suggests different strengths. The GTX 880M belongs in a system where general GPU compute in Metal matters, or where the higher average score across all tests matters more than a single OpenCL result. The GTX 670M, despite being older and less capable on paper, delivers better results in the specific OpenCL workload that both were measured on. That is a narrow but real victory.

Architecture Differences

The two chips come from different generations of NVIDIA design. The GTX 880M uses the GK104 chip, built on the Kepler architecture, fabricated on a 28 nm process at TSMC. The GTX 670M uses the GF114 chip, built on the Fermi 2.0 architecture, also from TSMC but on a larger 40 nm node. The manufacturing difference is significant: the newer Kepler chip packs 3,540 million transistors into a 294 mm² die, giving a transistor density of 12.0 million per square millimeter. The older Fermi chip contains 1,950 million transistors across a larger 332 mm² die, with a density of just 5.9 million per square millimeter. In simple terms, the GTX 880M crams almost twice the transistors into a smaller physical area.

That transistor advantage translates directly into compute resources. The GTX 880M carries 1,536 shading units, 128 texture mapping units, and 32 raster operation units. The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. The newer card has more than four times the shading units and more than double the texture units. Clock speeds also differ. The GTX 880M runs a base clock of 954 MHz with a boost up to 993 MHz. The GTX 670M has no base or boost clock listed in the database, so the comparison rests on the memory clock and the effective throughput rates.

Memory architecture diverges sharply. The GTX 880M uses 8 GB of GDDR5 on a 256 bit bus, yielding 160.0 GB/s of bandwidth. The GTX 670M has 1536 MB of GDDR5 on a 192 bit bus, producing 72.00 GB/s. That is a 2.2x bandwidth advantage for the newer card, alongside more than five times the memory capacity. The GTX 880M also records a pixel rate of 31.78 GPixel/s and a texture rate of 127.1 GTexel/s, against 8.372 GPixel/s and 33.49 GTexel/s for the GTX 670M. The FP32 throughput difference is stark: 3.050 TFLOPS versus 803.7 GFLOPS, nearly four times in favor of the GTX 880M.

API support shows a similar generational gap. Both cards support DirectX 12 (11_0) and OpenGL 4.6, but the GTX 880M adds Vulkan 1.2.175 support, while the GTX 670M has no Vulkan support recorded. Power draw differs as well, with the GTX 880M rated at 122 W and the GTX 670M at 75 W. Both use MXM modules with no power connectors and portable-device-dependent display outputs.

Head-to-Head Benchmarks

The database records exactly one shared benchmark between these two GPUs: Geekbench OpenCL. The GTX 670M wins it decisively. Its score of 6,513 beats the GTX 880M's 5,622, a margin of 13.7%. That result is counterintuitive given the massive compute advantage listed in the specifications. The GTX 880M has roughly 3.8 times the FP32 throughput, 2.3 times the texture rate, and 2.2 times the memory bandwidth, yet it loses the OpenCL test by a wide margin. The data does not explain why, but the numbers are what they are. In the only test where both cards ran the same workload, the older Fermi part outperformed the newer Kepler part.

The GTX 880M does have a second benchmark result, Geekbench Metal, where it scores 10,458. The GTX 670M has no Metal score recorded, so no direct comparison exists for that API. The GTX 880M's average benchmark score of 8,040 comes from averaging its two results, while the GTX 670M's average of 6,513 is simply its single OpenCL result. The percentile rankings reflect this: the GTX 880M sits at 42nd percentile against all GPUs, the GTX 670M at 38th. The delta between their average scores is roughly 23% in favor of the GTX 880M, but that comparison mixes different test sets.

The nearest rival data puts both cards in perspective. The GTX 880M sits within 0.5% of the NVIDIA Quadro P5000 (8,039 average), the GTX 650 Ti (8,053), the GTX 650 Ti Boost (8,067), and the GRID K2 (8,080). It is essentially tied with those parts. The GTX 670M sits within 0.6% of the GT 555M (6,493), the Quadro M5000M (6,481), the Radeon Vega 10 Mobile (6,476), and the Intel UHD Graphics P750 (6,554). The older card is also effectively tied with its nearest competitors. Neither card stands out dramatically against its own peer group, but the GTX 670M's peer group is a full tier lower in average score.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The GTX 880M averages 8,040 across its recorded tests, compared to 6,513 for the GTX 670M. The newer card sits in the 42nd percentile of all GPUs, while the older card sits in the 38th.

Q: Why does the GTX 670M win the only shared benchmark?

A: In the Geekbench OpenCL test, the GTX 670M scores 6,513 against 5,622 for the GTX 880M, a 13.7% advantage. The database records this as the sole head-to-head result, and the older card wins it despite having far lower specifications on paper.

Q: Does the GTX 880M have any benchmark that the GTX 670M lacks?

A: Yes. The GTX 880M has a Geekbench Metal score of 10,458. The GTX 670M has no Metal result recorded in the database, so there is no comparison available for that API.

Q: What are the memory capacity and bandwidth differences?

A: The GTX 880M has 8 GB of GDDR5 on a 256 bit bus with 160.0 GB/s bandwidth. The GTX 670M has 1536 MB of GDDR5 on a 192 bit bus with 72.00 GB/s bandwidth. The newer card offers over five times the capacity and more than double the bandwidth.

Q: Which card supports Vulkan?

A: Only the GTX 880M lists Vulkan support, with version 1.2.175. The GTX 670M has no Vulkan support recorded. Both cards support DirectX 12 (11_0) and OpenGL 4.6.

Q: How do these cards compare to their nearest rivals?

A: The GTX 880M is within 0.5% of the Quadro P5000, GTX 650 Ti, GTX 650 Ti Boost, and GRID K2. The GTX 670M is within 0.6% of the GT 555M, Quadro M5000M, Radeon Vega 10 Mobile, and Intel UHD Graphics P750.

Specification Differences

The two cards differ across nearly every major specification. The GTX 880M uses the GK104 chip on Kepler architecture, while the GTX 670M uses GF114 on Fermi 2.0. Process nodes differ: 28 nm versus 40 nm, both from TSMC. Transistor counts are 3,540 million versus 1,950 million, with die sizes of 294 mm² versus 332 mm². Transistor density is 12.0M per mm² versus 5.9M per mm².

Compute resources are dramatically different. The GTX 880M has 1,536 shading units, 128 TMUs, and 32 ROPs. The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. Clock speeds: the GTX 880M runs at 954 MHz base and 993 MHz boost, while the GTX 670M has no base or boost clock recorded. Memory clocks: 1250 MHz with 5 Gbps effective for the GTX 880M, 750 MHz with 3 Gbps effective for the GTX 670M.

Memory configuration: 8 GB GDDR5 on a 256 bit bus for the newer card, 1536 MB GDDR5 on a 192 bit bus for the older. Bandwidth: 160.0 GB/s versus 72.00 GB/s. Pixel rates: 31.78 GPixel/s versus 8.372 GPixel/s. Texture rates: 127.1 GTexel/s versus 33.49 GTexel/s. FP32: 3.050 TFLOPS versus 803.7 GFLOPS.

Power and API: the GTX 880M draws 122 W, the GTX 670M draws 75 W. Both use MXM-B (3.0) interfaces and MXM modules with no power connectors. The GTX 880M supports Vulkan 1.2.175, the GTX 670M does not. Both support DirectX 12 (11_0) and OpenGL 4.6. Release dates differ: the GTX 880M launched in March 2014, the GTX 670M in March 2012. The GTX 880M's predecessor is the GeForce 700M series, its successor the 900M series. The GTX 670M's predecessor is the 500M series, its successor the 700M series.

The Verdict

The data points in two directions, and the honest answer depends on which workload matters more. If OpenCL compute is the priority, the GTX 670M is the pick. It wins the only direct comparison by 13.7%, scoring 6,513 against 5,622. That is a decisive margin in a real, recorded test. No amount of specification superiority on the GTX 880M changes that measured outcome. The older Fermi card simply delivers better results in that specific API.

If Metal performance or overall database standing matters more, the GTX 880M is the choice. It posts a 10,458 Metal score, which the GTX 670M cannot match because it has no Metal result at all. Its average benchmark score of 8,040 versus 6,513 places it higher in the overall rankings, at the 42nd percentile versus the 38th. The newer card also brings Vulkan support, far more memory, and roughly four times the FP32 throughput, which matters for workloads that are not captured by the single OpenCL test.

For a builder choosing between these two end-of-life mobile GPUs, the decision is about which software environment you live in. The GTX 670M is the safer bet for OpenCL-centric applications, based on direct evidence. The GTX 880M is the better all-rounder, with a higher average score, Metal support, and a newer feature set. The specifications strongly favor the GTX 880M, but the one benchmark where both were measured favors the GTX 670M. Trust the test result for OpenCL, trust the broader record for everything else.