AMD Radeon R7 M380 vs NVIDIA GeForce GTX 660 Comparison

AMD Radeon R7 M380 and NVIDIA GeForce GTX 660 are both end-of-life graphics cards built on a 28 nm TSMC process, yet they target very different segments of the market. The GTX 660 is a desktop GPU from the GeForce 600 generation, while the R7 M380 is a mobile part from AMD's R7 M300 series. Benchmark data shows a clear performance gap, with the GTX 660 winning the only head-to-head test, but the R7 M380's newer feature set and larger memory pool give it distinct advantages in specific scenarios.

Where Each One Wins

The NVIDIA GeForce GTX 660 is the clear winner in raw compute performance. In the Geekbench OpenCL benchmark, it scores 11,347 points compared to the R7 M380's 9,313 points, a decisive 17.9% advantage. This makes the GTX 660 the better choice for any workload that relies heavily on parallel processing, such as GPU-accelerated rendering, video encoding, or compute-heavy scientific applications. Its higher shading unit count, texture mapping units, and memory bandwidth all contribute to this lead.

The AMD Radeon R7 M380, however, wins on memory capacity and API compatibility. It comes equipped with 4 GB of DDR3 memory, double the GTX 660's 2 GB GDDR5 allocation. For modern games or applications that exceed the 2 GB frame buffer limit, the R7 M380 will avoid stuttering and texture thrashing that can plague the GTX 660. Additionally, the R7 M380 supports DirectX 12 (11_1) while the GTX 660 only supports DirectX 12 (11_0). This means the AMD card can take advantage of certain DirectX 11.1 features that NVIDIA's Kepler architecture cannot, potentially offering better compatibility with specific titles or API extensions.

In terms of benchmark percentiles, the two are remarkably close. The R7 M380 sits at the 46th percentile of all GPUs, while the GTX 660 sits at the 45th percentile. This indicates that despite the GTX 660's higher raw OpenCL score, both cards are positioned similarly in the overall performance hierarchy. The GTX 660's average benchmark score of 9,022 is actually lower than the R7 M380's 9,313, suggesting that the OpenCL test is the GTX 660's strongest showing, while other workloads may favor the AMD card.

Architecture Differences

The architectural divide between these two GPUs is fundamental. The AMD Radeon R7 M380 uses the Tropo chip based on the Graphics Core Next (GCN) 1.0 architecture, while the NVIDIA GeForce GTX 660 uses the GK106 chip based on the Kepler architecture. These are completely different design philosophies from the same 28 nm era.

The R7 M380 packs 1,500 million transistors onto a 123 mm² die, achieving a transistor density of 12.2 million per square millimeter. In contrast, the GTX 660 houses 2,540 million transistors on a much larger 221 mm² die, with a slightly lower density of 11.5 million per square millimeter. The GTX 660's larger die and higher transistor count reflect its more complex compute pipeline, which is evident in its superior shading and texture throughput.

Clock speeds also differ significantly. The R7 M380 runs at a base clock of 900 MHz with a boost of 915 MHz, while the GTX 660 operates at a higher base of 980 MHz and boosts to 1032 MHz. The GTX 660's memory runs at 1502 MHz (6 Gbps effective) compared to the R7 M380's 1000 MHz (2 Gbps effective). This clock advantage, combined with a wider 192-bit memory bus versus 128-bit, gives the GTX 660 a massive bandwidth lead.

The GTX 660 features 960 shading units, 80 texture mapping units, and 24 raster operations pipelines. The R7 M380 has only 640 shading units, 40 TMUs, and 16 ROPs. This 50% advantage in shaders and 100% advantage in TMUs for the GTX 660 explains its dominant position in compute benchmarks. However, the R7 M380's GCN architecture is known for better compute efficiency per clock, and its support for newer API versions (Vulkan 1.2.170 vs 1.2.175, both close) keeps it relevant for modern software.

Head-to-Head Benchmarks

The only direct benchmark comparison available is the Geekbench OpenCL test, where the NVIDIA GeForce GTX 660 defeats the AMD Radeon R7 M380 by a margin of 17.9%. The GTX 660 scores 11,347 points versus the R7 M380's 9,313. This is a substantial gap that reflects the GTX 660's superior hardware specifications.

Breaking down the numbers, the GTX 660's texture rate of 82.56 GTexel/s is more than double the R7 M380's 36.60 GTexel/s. Its pixel rate of 20.64 GPixel/s also outstrips the R7 M380's 14.64 GPixel/s. The FP32 compute throughput tells a similar story: 1.981 TFLOPS for the GTX 660 versus 1,171.2 GFLOPS (1.171 TFLOPS) for the R7 M380. This represents a 69% advantage in raw floating-point performance for the NVIDIA card.

Memory bandwidth is where the gap is most pronounced. The GTX 660's 144.2 GB/s bandwidth is over four times higher than the R7 M380's 32.00 GB/s. This is a critical differentiator because many compute workloads and modern games are bandwidth-limited. Even though the R7 M380 has double the memory capacity, its DDR3 interface and narrow 128-bit bus severely constrain data throughput.

Interestingly, the GTX 660's OpenCL score of 11,347 is its highest benchmark result, while its Metal score (4,305) and Vulkan score (11,415) show different strengths. The R7 M380 has no other benchmark results beyond OpenCL, so a full comparison across APIs is not possible from the data. However, the Vulkan score for the GTX 660 (11,415) is nearly identical to its OpenCL score, suggesting consistent compute performance across APIs.

Specification Differences

Beyond the architectural and performance differences, several key specifications set these two cards apart. The most obvious is form factor: the GTX 660 is a dual-slot desktop card measuring 241 mm (9.5 inches) in length, requiring a 300 W power supply and a single 6-pin power connector. Its TDP is rated at 140 W. The R7 M380, being a mobile GPU, has no listed TDP, slot width, or power connector requirements, making it suitable for laptop integration.

Memory configuration is another major divergence. The R7 M380 uses 4 GB of DDR3 memory on a 128-bit bus, while the GTX 660 uses 2 GB of GDDR5 on a 192-bit bus. This affects not only bandwidth but also power efficiency and latency characteristics. GDDR5 is significantly faster than DDR3, which contributes to the GTX 660's superior bandwidth.

Display outputs differ as well. The GTX 660 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the R7 M380 has no listed display outputs. This reflects their different target markets: the GTX 660 is designed for desktop systems with multiple monitors, while the R7 M380 is meant to be integrated into laptops where the display connection is handled by the motherboard.

Release dates are far apart. The GTX 660 was released on September 5, 2012, while the R7 M380 came out on May 4, 2015. Despite being nearly three years newer, the R7 M380 cannot match the GTX 660's performance. The GTX 660 had a launch MSRP of 229 USD, while the R7 M380 has no listed launch price, consistent with its mobile OEM positioning.

API support shows the R7 M380's newer design. It supports DirectX 12 (11_1) and Vulkan 1.2.170, while the GTX 660 supports DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6. The R7 M380's newer API feature level could provide better compatibility with certain modern titles, though the GTX 660's Vulkan support is slightly newer.

FAQ

Q: Which GPU is faster in OpenCL compute?

A: The NVIDIA GeForce GTX 660 is significantly faster, scoring 11,347 points in Geekbench OpenCL versus the AMD Radeon R7 M380's 9,313 points, a 17.9% advantage for the GTX 660.

Q: Does the R7 M380 have more memory than the GTX 660?

A: Yes, the R7 M380 has 4 GB of DDR3 memory, double the GTX 660's 2 GB of GDDR5. However, the GTX 660's GDDR5 memory on a 192-bit bus provides 144.2 GB/s bandwidth versus only 32.00 GB/s for the R7 M380.

Q: Which card has better API support?

A: The R7 M380 supports DirectX 12 (11_1), while the GTX 660 only supports DirectX 12 (11_0). Both cards support OpenGL 4.6 and Vulkan, with the GTX 660 having a slightly newer Vulkan version (1.2.175 vs 1.2.170).

Q: What are the physical dimensions of the GTX 660?

A: The GTX 660 is a dual-slot card measuring 241 mm (9.5 inches) in length. It requires a 300 W power supply and uses a single 6-pin power connector, with a TDP of 140 W.

Q: How do these cards compare in overall benchmark percentile?

A: They are very close, with the R7 M380 at the 46th percentile of all GPUs and the GTX 660 at the 45th percentile. The R7 M380's average benchmark score is 9,313, while the GTX 660's average is 9,022.

Q: Which card is newer?

A: The AMD Radeon R7 M380 was released on May 4, 2015, while the NVIDIA GeForce GTX 660 was released on September 5, 2012. Despite being newer, the R7 M380 cannot match the GTX 660's raw performance.