AMD Radeon 880M vs NVIDIA GeForce GTX 660 Comparison

Head-to-Head Benchmarks

The most striking pattern in the head-to-head data is the sheer scale of the AMD Radeon 880M's victory in both shared tests. In Geekbench OpenCL, the 880M scores 31,285 against the GTX 660's 11,347, a delta of -63.7% from the 880M's perspective — meaning the AMD part is roughly 2.76 times faster. The gap widens further in Geekbench Vulkan, where the 880M hits 40,006 versus 11,415, a -71.5% delta, translating to a 3.5x advantage. These are not marginal wins; they represent generational leaps in compute throughput and API efficiency.

The GTX 660 does not win either of the two head-to-head tests, and the data shows zero wins across all benchmark categories. However, looking at the broader benchmark suite, the GTX 660 still holds its own in legacy APIs. The 880M's Passmark DirectX 9 score of 97 and DirectX 11 score of 73 are strong, while its DirectX 10 (31) and DirectX 12 (32) scores are notably lower — a curious inversion that suggests driver optimizations favor older DX9/11 workloads on the integrated part. The GTX 660's average benchmark score of 9,022 slightly exceeds the 880M's 8,436, despite the 880M's massive wins in Geekbench compute tests. This discrepancy implies the GTX 660 performs better in the specific workloads captured by Passmark's G3D suite, even though the 880M dominates in OpenCL and Vulkan compute.

The percentile rankings tell a nuanced story: the GTX 660 sits at the 45th percentile of all GPUs, while the 880M sits at the 43rd. Despite the 880M's raw compute superiority in modern APIs, its overall standing in the database is marginally lower. The nearest rivals for each card confirm the 880M competes with older mobile parts like the GTX 675MX (delta 0.1%) and MX330 (delta -0.3%), whereas the GTX 660 sits near the TITAN V CEO Edition (delta -0.2%) and GTX 560 (delta -0.4%). This suggests the 880M's strengths are concentrated in specific modern workloads rather than broad-spectrum gaming performance.

The Verdict

From the data alone, the AMD Radeon 880M is the clear winner for anyone running OpenCL or Vulkan compute workloads. Its Geekbench Vulkan score of 40,006 is 3.5x the GTX 660's 11,415, and its OpenCL score of 31,285 is 2.76x higher. For applications that leverage these APIs — think GPU-accelerated rendering, scientific compute, or modern game engines built around Vulkan — the 880M is categorically superior.

However, the GTX 660's higher average benchmark score (9,022 vs 8,436) and higher percentile rank (45 vs 43) indicate it remains competitive in the aggregate database metrics. The Passmark G3D score of 7,615 for the 880M versus the GTX 660's implied performance in that suite (derived from its overall average) suggests the older card may still handle traditional DirectX gaming titles more consistently. The GTX 660 also benefits from being an end-of-life product with mature drivers, whereas the 880M's Active status means ongoing optimization could shift future results.

The 880M is the pick for modern API workloads and efficiency — its 15 W TDP versus the GTX 660's 140 W is a tenfold difference in power draw, making it ideal for portable devices. The GTX 660 is the pick for those relying on legacy DirectX 9/11 titles or needing a dual-slot PCIe card with discrete memory. Neither card dominates the other across all metrics; the choice hinges on workload specificity.

Architecture Differences

The architectural gulf between these two GPUs is vast, even beyond the obvious generational gap. The GTX 660 uses the GK106 chip on TSMC's 28 nm process, packing 2,540 million transistors into a 221 mm² die with a density of 11.5M transistors per mm². In contrast, the 880M uses the Strix Point chip on TSMC's 4 nm process, integrating 34,000 million transistors into a 233 mm² die — a density of 145.9M per mm², roughly 12.7 times denser. The 880M's transistor count is over 13 times higher than the GTX 660's, despite a nearly identical die size, which explains its superior compute capabilities.

The GTX 660 is built on the Kepler architecture, while the 880M uses RDNA 3.5. Kepler was designed for the DirectX 11 era, supporting DirectX 12 (11_0) — a partial implementation that lacks the full feature set of later architectures. The 880M supports DirectX 12 Ultimate (12_2), which includes hardware ray tracing via its 12 RT cores. The GTX 660 has no RT cores; the 880M's ray tracing capability is entirely absent from the older card. The 880M also supports Vulkan 1.4 versus the GTX 660's 1.2.175, and its FP16 throughput matches FP32 at 4.454 TFLOPS (1:1), whereas the GTX 660 has no listed FP16 capability. The 880M's rendering pipeline is also more efficient: 46.40 GPixel/s pixel rate and 139.2 GTexel/s texture rate versus the GTX 660's 20.64 GPixel/s and 82.56 GTexel/s.

Specification Differences

The most obvious difference is memory architecture. The GTX 660 has 2 GB of dedicated GDDR5 on a 192-bit bus with 144.2 GB/s bandwidth. The 880M uses System Shared memory — no dedicated VRAM, no fixed bus width, and bandwidth described as "System Dependent." This means the 880M's memory performance varies with the host system's RAM speed and configuration, whereas the GTX 660 offers predictable, fixed bandwidth.

Clock speeds also differ dramatically. The GTX 660 runs at a 980 MHz base and 1032 MHz boost, with memory at 1502 MHz (6 Gbps effective). The 880M has a 400 MHz base but boosts to 2900 MHz — a 2.8x higher boost clock. Its shading units number 768 versus the GTX 660's 960, but the 880M's higher clocks and RDNA 3.5 efficiency more than compensate. The 880M has 48 TMUs and 16 ROPs versus 80 TMUs and 24 ROPs on the GTX 660 — the older card actually has more of both, yet the 880M still achieves higher fill rates due to its clock advantage.

Power and physical specs are polar opposites. The GTX 660 draws 140 W TDP, requires a 300 W PSU, uses a 1x 6-pin power connector, and occupies a dual-slot 241 mm (9.5 inches) card. The 880M is an IGP with 15 W TDP, no power connectors, and no dedicated slot — its display outputs are "Portable Device Dependent." The GTX 660 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2; the 880M's outputs depend entirely on the host laptop or mini-PC. Bus interfaces also differ: PCIe 3.0 x16 for the GTX 660 versus PCIe 4.0 x8 for the 880M.

The GTX 660 has a launch MSRP of 229 USD (stated once here per database convention). The 880M has no launch MSRP listed. The GTX 660 was released on 2012-09-05 and is end-of-life; the 880M released on 2024-07-14 and is active. The GTX 660's predecessor is GeForce 500 and successor is GeForce 700; the 880M's predecessor is Navi II IGP with no successor listed.

FAQ

Q: Which card is faster in Vulkan compute?

A: The AMD Radeon 880M scores 40,006 in Geekbench Vulkan versus the GTX 660's 11,415, a 71.5% delta in favor of the 880M — roughly 3.5x faster.

Q: Does the GTX 660 have ray tracing support?

A: No. The GTX 660 has no RT cores listed, while the 880M has 12 RT cores. The 880M also supports DirectX 12 Ultimate (12_2), while the GTX 660 is limited to DirectX 12 (11_0).

Q: How do their overall benchmark averages compare?

A: The GTX 660 has an average benchmark score of 9,022, slightly above the 880M's 8,436. The GTX 660 ranks at the 45th percentile of all GPUs; the 880M ranks at the 43rd.

Q: What is the power draw difference?

A: The GTX 660 has a 140 W TDP and requires a 300 W PSU with a 1x 6-pin connector. The 880M has a 15 W TDP and uses no power connectors, being an integrated GPU.

Q: Which card has more shading units?

A: The GTX 660 has 960 shading units versus the 880M's 768. However, the 880M's 2900 MHz boost clock versus the GTX 660's 1032 MHz boost clock leads to higher FP32 performance: 4.454 TFLOPS versus 1.981 TFLOPS.

Q: Can the 880M be used in a desktop PCIe slot?

A: The 880M is an IGP with PCIe 4.0 x8 interface, but its display outputs are "Portable Device Dependent," meaning it is designed for integrated use in laptops or mini-PCs rather than as a standalone card with fixed display connectors.

Where Each One Wins

The AMD Radeon 880M wins decisively in modern compute APIs. Its Geekbench OpenCL score of 31,285 and Vulkan score of 40,006 dwarf the GTX 660's 11,347 and 11,415 respectively. It also wins on efficiency by every measure: 15 W TDP versus 140 W, no power connector required, and integrated form factor. For ray tracing workloads, the 880M's 12 RT cores and DirectX 12 Ultimate support give it capabilities the GTX 660 simply lacks. Its FP16 throughput of 4.454 TFLOPS (1:1) makes it suitable for AI and machine learning tasks that the GTX 660 cannot accelerate.

The GTX 660 wins in legacy DirectX scenarios and predictable memory performance. Its dedicated 2 GB GDDR5 with 144.2 GB/s bandwidth ensures consistent frame pacing in older games, whereas the 880M's System Shared memory is dependent on host RAM. The GTX 660's higher average benchmark score (9,022 vs 8,436) and better percentile ranking (45 vs 43) suggest it excels in the specific workloads captured by the database's aggregate metrics. Its 960 shading units and 80 TMUs — more than the 880M's 768 and 48 — indicate a wider parallel pipeline for shader-heavy DX9/11 titles. The dual-slot, 241 mm card form factor also makes it a drop-in replacement for older desktops, while the 880M is locked to portable devices. For anyone running a 2012-era game library or needing fixed VRAM bandwidth, the GTX 660 remains the data-supported choice.