AMD Radeon 660M vs NVIDIA GeForce GTX 480 Comparison

AMD Radeon 660M and NVIDIA GeForce GTX 480 represent two very different eras of GPU design, yet their benchmark scores place them in a surprisingly close contest. The data shows a single head-to-head benchmark result, with the GTX 480 edging out the 660M by a narrow margin. This comparison hinges on interpreting that score alongside each card’s architectural strengths and limitations, rather than expecting a dominant victory from either side.

Head-to-Head Benchmarks

The only direct comparison available in the data is the Geekbench OpenCL test. Here, the NVIDIA GeForce GTX 480 scores 13,300, while the AMD Radeon 660M scores 12,876. This gives the GTX 480 a victory with a delta of -3.2% relative to the 660M, meaning the 660M trails by roughly three percent. While this is a win for the older Fermi-based card, the margin is slim enough to be considered near-parity in raw compute workloads.

Looking at broader context, the 660M’s average benchmark score across all tests is 13,812, which is higher than its OpenCL-only result. This suggests that the 660M performs better in other API workloads, such as Vulkan, where it scores 14,748. The GTX 480 has no Vulkan score listed, so its 13,300 OpenCL result is its sole data point. This asymmetry matters: the 660M’s Vulkan performance is over 14% higher than its own OpenCL score, and it exceeds the GTX 480’s OpenCL score by about 10.9%. However, without a Vulkan result for the GTX 480, a direct head-to-head in that API is impossible.

The percentile rankings reinforce the closeness. The 660M sits at the 55th percentile among all GPUs, while the GTX 480 is at the 53rd percentile. Both cards are essentially mid-pack performers in the overall database, with the 660M holding a slight edge in that metric. The nearest rivals for each card further contextualize their standing: the 660M’s closest competitor is the NVIDIA RTX A2000 Mobile (average score 13,821, delta -0.1%), while the GTX 480’s closest is the AMD Radeon Pro 555X (average score 13,321, delta -0.2%). Neither card is far from its nearest neighbors, indicating that both occupy a crowded performance tier.

Architecture Differences

The architectural gap between these two GPUs is immense, reflecting a 12-year production timeline. The 660M uses the Rembrandt chip built on RDNA 2.0 architecture, fabricated on a 6 nm process at TSMC. In contrast, the GTX 480 uses the GF100 chip with Fermi architecture, on a 40 nm process, also at TSMC. The transistor counts tell the story of density improvements: the 660M packs 13,100 million transistors into a 208 mm² die, yielding a density of 63.0 million transistors per square millimeter. The GTX 480 has only 3,100 million transistors on a much larger 529 mm² die, giving a density of 5.9 million per square millimeter — a 10.7x difference in density in favor of the 660M.

Feature support diverges sharply. The 660M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GTX 480 only supports DirectX 12 (11_0) and OpenGL 4.6, with no Vulkan support listed. The 660M also includes 6 ray tracing cores, a feature entirely absent from the GTX 480, which has no RT cores or tensor cores. Both cards have null tensor cores, but the 660M’s RDNA 2.0 architecture brings modern rendering capabilities that Fermi simply cannot match.

Memory architecture is another fundamental split. The 660M uses system-shared memory, with its size, type, bus width, and bandwidth all listed as "System Shared" or "System Dependent." The GTX 480 has dedicated 1536 MB of GDDR5 memory on a 384-bit bus, delivering 177.4 GB/s of bandwidth. This means the GTX 480 has fixed, high-bandwidth local memory, while the 660M relies on the host system’s RAM, making its bandwidth variable. In terms of raw fill rates, the 660M has a pixel rate of 30.40 GPixel/s and texture rate of 45.60 GTexel/s, while the GTX 480 has 21.03 GPixel/s and 42.06 GTexel/s. The 660M leads in both, despite the GTX 480 having more shading units (480 vs 384), TMUs (60 vs 24), and ROPs (48 vs 16).

Where Each One Wins

The GTX 480 wins the single OpenCL benchmark, indicating a slight advantage in that specific compute workload. Its dedicated 177.4 GB/s memory bandwidth likely helps in memory-intensive tasks, as the data shows a fixed, high-throughput local memory pool. The GTX 480 also has more shading units (480 vs 384), which could benefit certain parallel compute patterns, even if its older architecture is less efficient per clock.

The 660M wins in several other categories based on the data. Its Vulkan score of 14,748 is substantially higher than its OpenCL score, and the absence of a Vulkan result for the GTX 480 suggests the 660M supports a modern API that the older card cannot use. The 660M also has higher pixel and texture rates, as noted above, which would favor rasterization-heavy tasks like gaming at lower resolutions. The 660M’s ray tracing cores provide a feature the GTX 480 lacks entirely, making it the only option for any ray-traced workload. Additionally, the 660M’s higher percentile ranking (55th vs 53rd) and higher average benchmark score (13,812 vs 13,300) indicate better overall performance across the full test suite.

Specification Differences

The two cards differ in nearly every specification field where a direct comparison is possible. The process node is 6 nm for the 660M versus 40 nm for the GTX 480. Transistor count is 13,100 million versus 3,100 million, and die size is 208 mm² versus 529 mm². Clock speeds are not directly comparable: the 660M has a base clock of 1500 MHz and boost of 1900 MHz, while the GTX 480 has no base or boost clock listed, only a memory clock of 924 MHz (3.7 Gbps effective).

Memory configuration is completely different: the 660M uses system-shared memory, while the GTX 480 has 1536 MB of GDDR5 on a 384-bit bus with 177.4 GB/s bandwidth. Shading units are 384 for the 660M versus 480 for the GTX 480. TMUs are 24 versus 60, and ROPs are 16 versus 48. The 660M has 6 RT cores; the GTX 480 has none. Power consumption is a major differentiator: the 660M has a TDP of 40 W and no power connectors, while the GTX 480 has a 250 W TDP with 1x 6-pin and 1x 8-pin connectors, plus a suggested PSU of 600 W. The 660M is an IGP with PCIe 4.0 x8 interface, while the GTX 480 is a dual-slot card with PCIe 2.0 x16 and dimensions of 267 mm (10.5 inches). Display outputs also differ: the 660M is portable-device dependent, while the GTX 480 has 2x DVI and 1x mini-HDMI 1.3a.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The AMD Radeon 660M has an average benchmark score of 13,812, while the NVIDIA GeForce GTX 480 has 13,300. The 660M is ahead by 512 points.

Q: Does the GTX 480 support Vulkan?

A: No. The data lists Vulkan as null for the GTX 480. The 660M supports Vulkan 1.4.

Q: What is the memory bandwidth difference?

A: The GTX 480 has a fixed bandwidth of 177.4 GB/s from its GDDR5 memory. The 660M’s bandwidth is listed as "System Dependent," meaning it varies based on the host system’s memory.

Q: How many ray tracing cores does each GPU have?

A: The 660M has 6 ray tracing cores. The GTX 480 has none, as its RT cores field is null.

Q: Which card has a higher pixel fill rate?

A: The 660M has a pixel rate of 30.40 GPixel/s, compared to 21.03 GPixel/s for the GTX 480. The 660M is about 44.6% higher.

Q: What is the TDP difference?

A: The 660M has a TDP of 40 W, while the GTX 480 has a TDP of 250 W. The GTX 480 also requires a 600 W suggested PSU and two power connectors.

The Verdict

Based strictly on the data, the AMD Radeon 660M is the more capable and modern GPU. It wins on average benchmark score, percentile ranking, pixel and texture rates, and feature support, including Vulkan and ray tracing. Its 6 nm process and 13,100 million transistors enable this performance at a fraction of the power draw (40 W vs 250 W). The 660M is the clear choice for anyone needing modern API support, efficiency, or integrated graphics in a portable device.

The NVIDIA GeForce GTX 480, however, wins the single head-to-head OpenCL benchmark and offers dedicated memory with fixed bandwidth. Its 1536 MB of GDDR5 and 384-bit bus provide consistent memory performance that the 660M cannot guarantee with system-shared memory. For users running legacy OpenCL workloads that favor the Fermi architecture’s higher shading unit count (480 vs 384), the GTX 480 remains competitive. Its 250 W TDP and dual-slot design are drawbacks, but its 53rd percentile ranking shows it still holds its own in the database.

In summary, the 660M is the better all-rounder for modern use cases, while the GTX 480 is a niche performer for specific compute tasks. The 3.2% OpenCL delta is too small to overcome the 660M’s advantages in every other measurable category. Choose the 660M for versatility and efficiency; choose the GTX 480 only if a specific workload demands its dedicated memory and older compute profile.