NVIDIA GeForce GTX 660 vs NVIDIA GeForce GTX 880M Comparison

Head-to-Head Benchmarks

The recorded data splits cleanly between the two NVIDIA Kepler parts. In the Geekbench Metal test, the GTX 880M posts 10,458 points against the GTX 660's 4,305 points. That is a 58.8% deficit for the GTX 660, and it places the mobile GPU in a different performance class for that workload. The gap is substantial enough that no clock speed or driver tuning on the desktop card could close it in that specific API.

The Geekbench OpenCL result flips the script entirely. The GTX 660 scores 11,347 points, while the GTX 880M manages only 5,622 points. The desktop card leads by 101.8%, more than doubling the mobile part's output. This is not a marginal edge; it is a decisive victory for the older, smaller desktop GPU in a compute-oriented benchmark. The two cards split the head-to-head at one win apiece, and the delta percentages reveal how differently each architecture responds to the two test environments.

Looking at the broader database averages, the GTX 660 carries an average benchmark score of 9,022, which places it at the 45th percentile of all GPUs. The GTX 880M averages 8,040, sitting at the 42nd percentile. The desktop card holds a 982-point advantage in average score, roughly 12.2% higher. That average masks the extreme variance between the two Geekbench subtests, but it does confirm that the GTX 660 is the stronger overall performer in the database's aggregate measurements.

The nearest rival data reinforces how close these cards sit to their contemporaries. The GTX 660's closest competitor is the NVIDIA TITAN V CEO Edition at 9,037 average points, a mere 0.2% gap. The GTX 560 trails by 0.4% at 9,058, while the AMD Radeon 550X sits 1.2% behind the GTX 660, and the AMD Radeon Pro WX 5100 is 1.8% behind. These are tight margins, and the GTX 660 effectively trades blows with a much newer flagship in this metric, which says more about the benchmark's sensitivity than about real-world equivalence.

The GTX 880M's nearest rivals cluster even closer. The NVIDIA Quadro P5000 matches it at 8,039, a 0% difference. The GTX 650 Ti is 0.2% behind, the GTX 650 Ti Boost is 0.3% behind, and the NVIDIA GRID K2 is 0.5% behind. The mobile GPU sits in a dense pack of midrange Kepler and Pascal parts, none of which separate themselves by more than half a percent. This suggests the GTX 880M's average score is representative of a well-populated performance tier rather than an outlier.

What the head-to-head data does not show is any overlap between the two tests. The Metal benchmark heavily favors the GTX 880M, while OpenCL heavily favors the GTX 660. Neither card wins both. This pattern indicates that the two GPUs have complementary strengths, and the choice between them depends entirely on which workload matters more to the user.

Architecture Differences

Both GPUs are built on NVIDIA's Kepler architecture and manufactured by TSMC on a 28 nm process node. The similarities end there. The GTX 660 uses the GK106 chip, which contains 2,540 million transistors on a 221 mm² die, yielding a transistor density of 11.5 million per square millimeter. The GTX 880M uses the larger GK104 chip, packing 3,540 million transistors onto a 294 mm² die, for a density of 12.0 million per square millimeter. The GK104 die is 33% larger by area and carries 39.4% more transistors, which explains why the mobile part can offer substantially more execution resources.

The shader configuration differs sharply. The GTX 660 has 960 shading units, 80 texture mapping units, and 24 ROPs. The GTX 880M more than doubles the shader count with 1,536 shading units, alongside 128 TMUs and 32 ROPs. That is a 60% increase in shading units, a 60% increase in TMUs, and a 33% increase in ROPs. These raw resource counts give the GTX 880M higher theoretical throughput limits. The pixel rate for the GTX 880M is 31.78 GPixel/s versus 20.64 GPixel/s for the GTX 660, a 54% advantage. The texture rate is 127.1 GTexel/s versus 82.56 GTexel/s, a 54% advantage. The FP32 compute rating is 3.050 TFLOPS versus 1.981 TFLOPS, a 54% advantage. All of these figures point in the same direction: the GTX 880M has significantly more raw processing capacity.

Clock speeds tell a different story. The GTX 660 runs at a 980 MHz base clock and 1032 MHz boost, while the GTX 880M runs at 954 MHz base and 993 MHz boost. The desktop card is 2.7% higher at base and 3.9% higher at boost. This small clock advantage does not offset the GTX 880M's resource lead, but it does explain why the GTX 660 remains competitive in certain workloads.

Memory configurations diverge substantially. The GTX 660 has 2 GB of GDDR5 on a 192-bit bus, with a memory clock of 1502 MHz and 6 Gbps effective speed, producing 144.2 GB/s of bandwidth. The GTX 880M has 8 GB of GDDR5 on a 256-bit bus, with a memory clock of 1250 MHz and 5 Gbps effective speed, producing 160.0 GB/s of bandwidth. The mobile card offers four times the capacity and 11% more bandwidth, but its memory runs at a lower effective speed. The wider bus compensates for the lower clock, giving the GTX 880M a bandwidth edge despite slower memory chips.

Power and physical specifications reflect their intended environments. The GTX 660 is a dual-slot desktop card rated at 140 W TDP, requiring a single 6-pin power connector and a 300 W suggested power supply. It measures 241 mm or 9.5 inches in length. The GTX 880M is an MXM module rated at 122 W TDP with no power connectors, designed for portable systems. The desktop card draws 14.7% more power, which is notable given that the GTX 880M has far more compute resources. The GTX 660 interfaces via PCIe 3.0 x16, while the GTX 880M uses MXM-B (3.0). Display outputs also differ: the GTX 660 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the GTX 880M's outputs are portable device dependent.

API support is identical. Both cards support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Neither card has ray tracing or tensor cores, as those features did not exist in the Kepler generation. The GTX 660 launched on September 5, 2012, while the GTX 880M launched on March 11, 2014, roughly 18 months later. The GTX 660's predecessor is the GeForce 500 series and its successor is the GeForce 700 series. The GTX 880M's predecessor is the GeForce 700M series and its successor is the GeForce 900M series. Both cards are end-of-life production status.

The Verdict

The data supports a clear split decision. For compute workloads measured by OpenCL, the GTX 660 is the dominant choice. Its 11,347 score doubles the GTX 880M's 5,622, and its average benchmark score of 9,022 sits 12.2% above the mobile part's 8,040. The desktop card's 45th percentile placement versus the GTX 880M's 42nd percentile reinforces that the GTX 660 is the stronger all-around performer in the database's aggregate metrics.

For Metal workloads, the GTX 880M is the clear winner. Its 10,458 score is more than double the GTX 660's 4,305, and the 58.8% delta is the largest gap recorded in either direction between these two cards. Users who prioritize Metal performance should choose the GTX 880M without hesitation.

The architecture comparison favors the GTX 880M on paper. More shading units, more TMUs, more ROPs, higher pixel rate, higher texture rate, higher FP32 throughput, and more memory capacity all point to the mobile part having superior raw hardware. Yet the benchmark data shows the GTX 660 winning the OpenCL test and the overall average. This suggests that the GTX 880M's extra resources are not fully utilized in every workload, or that the desktop card's higher clocks and different memory timing provide an advantage in certain compute patterns.

The power envelope is also relevant. The GTX 880M delivers its larger resource pool at 122 W, which is 18 W less than the GTX 660's 140 W rating. That is an unusual outcome: the more powerful GPU on paper draws less power. For portable systems, this makes the GTX 880M the more efficient option. For desktop builders, the GTX 660's power requirement is modest and easily handled by a 300 W PSU.

There is no universal winner. The GTX 660 wins the compute workload and the overall average. The GTX 880M wins the Metal workload and offers more memory, more bandwidth, and more raw shader throughput. The choice depends on the target application and platform.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 660 has an average benchmark score of 9,022, compared to the GTX 880M's 8,040. That is a 982-point difference in favor of the desktop card.

Q: How do the two cards compare in the Metal benchmark?

A: The GTX 880M scores 10,458 in Geekbench Metal, while the GTX 660 scores 4,305. The GTX 880M leads by 58.8% in that specific test.

Q: How do they compare in OpenCL?

A: The GTX 660 scores 11,347 in Geekbench OpenCL, while the GTX 880M scores 5,622. The GTX 660 leads by 101.8%, more than doubling the mobile part's score.

Q: Which GPU has more shading units?

A: The GTX 880M has 1,536 shading units, while the GTX 660 has 960. The GTX 880M also has 128 TMUs versus 80, and 32 ROPs versus 24.

Q: What memory configurations do these cards use?

A: The GTX 660 has 2 GB of GDDR5 on a 192-bit bus with 144.2 GB/s bandwidth. The GTX 880M has 8 GB of GDDR5 on a 256-bit bus with 160.0 GB/s bandwidth.

Q: Are both cards based on the same architecture?

A: Yes, both use NVIDIA's Kepler architecture on a 28 nm TSMC process. The GTX 660 uses the GK106 chip, and the GTX 880M uses the GK104 chip.

Where Each One Wins

The GTX 660 wins in compute-heavy scenarios. Its OpenCL score of 11,347 versus 5,622 demonstrates a 101.8% advantage that makes it the clear choice for OpenCL-based applications, GPU compute tasks, and any workload that scales with the desktop card's clock speed and memory efficiency. Its 45th percentile placement also gives it a higher standing in the database's overall rankings. The GTX 660's 1.981 TFLOPS FP32 rating, while lower than the GTX 880M's 3.050 TFLOPS, does not prevent it from winning the compute benchmark that matters in the recorded data. The card's 241 mm length and dual-slot design fit standard desktop cases, and its 140 W TDP with a single 6-pin connector is straightforward for most PSUs rated at 300 W or higher.

The GTX 880M wins in Metal-based workloads. Its 10,458 Metal score dwarfs the GTX 660's 4,305, making it the superior option for Metal-accelerated applications on platforms that support that API. The mobile card also wins on memory capacity with 8 GB versus 2 GB, which matters for large textures and datasets. Its 160.0 GB/s bandwidth edges out the GTX 660's 144.2 GB/s, and its 256-bit bus provides a wider path for memory-heavy operations. The GTX 880M's 31.78 GPixel/s pixel rate and 127.1 GTexel/s texture rate indicate stronger fill-rate performance, which can benefit resolution-heavy rendering despite the card's lower OpenCL showing. Its 122 W TDP makes it a more power-efficient option for portable systems, and its MXM form factor suits laptops and compact mobile workstations.

For users who need a desktop card for general compute and OpenCL tasks, the GTX 660 is the better choice based on the recorded scores. For users who need a mobile GPU with more memory and better Metal performance, the GTX 880M is the better fit. The two cards are complementary rather than directly substitutable, and the benchmark data reflects that division. Users should match the card to the workload, not the other way around.