AMD Radeon Pro 555 vs NVIDIA GeForce GTX 660 Ti Comparison

Head-to-Head Benchmarks

The recorded data includes two common workloads for these cards: Geekbench OpenCL and Geekbench Vulkan. In both, the NVIDIA GeForce GTX 660 Ti comes out ahead, and the margins are substantial. In OpenCL, the GTX 660 Ti scores 15113 against 11682 for the AMD Radeon Pro 555, a delta of 29.4%. In Vulkan, the gap narrows slightly but remains decisive: 15012 versus 12303, a 22% advantage. That means the NVIDIA card is roughly a quarter to nearly a third faster in these compute-oriented tests, depending on the API.

The AMD card does have a Metal score of 16236, which is its strongest recorded result, but there is no equivalent Metal score for the GTX 660 Ti in the database, so a direct comparison on that API is not possible. The NVIDIA card also has a higher average benchmark score of 15063 across its recorded tests, while the AMD card’s average sits at 13407. That is a difference of over 1600 points, or roughly 12% on the average. The percentile rankings tell a similar story: the GTX 660 Ti sits at the 57th percentile of all GPUs, while the Radeon Pro 555 sits at the 54th. Neither card is near the top of the database, but the NVIDIA part is consistently ahead.

The nearest rival data reinforces the NVIDIA card’s positioning. The GTX 660 Ti is within 0.1% of the AMD Radeon 560X, within 0.7% of the RX 7600, and within 0.9% of the RTX 3050 OEM. The AMD Radeon 680M trails by 1.4%. So the GTX 660 Ti is effectively trading blows with much newer cards, which is remarkable for a 2012 part. The Radeon Pro 555, by contrast, is within 0.2% of the AMD Radeon 8950M, within 0.4% of the RX 5500M, and within 0.4% of the AMD FirePro M6100, while the NVIDIA P106-090 is 0.5% ahead of it. The Radeon Pro 555 is clustered with older mobile and workstation parts, not with the modern desktop GPUs the GTX 660 Ti is matching.

Where Each One Wins

The GTX 660 Ti wins both comparative benchmarks in the database. OpenCL and Vulkan are the only two tests where both cards have scores, and the NVIDIA card takes both. That is a clean sweep in the head-to-head data. The wins are not marginal either; 29.4% in OpenCL and 22% in Vulkan are large enough to matter in real workloads that rely on these APIs.

The Radeon Pro 555 has no head-to-head wins, but it does have one unique strength: Metal. Its Metal score of 16236 is higher than any score the GTX 660 Ti has in the database. Since the GTX 660 Ti has no recorded Metal score, the database cannot tell us how they compare on that API. But for any workload that is Metal-centric, the Radeon Pro 555 is the only one of the two with recorded data showing its capability. The card also has a lower TDP of 75 W versus 150 W for the NVIDIA card, which suggests it is intended for more power-constrained environments. The NVIDIA card requires a 450 W suggested PSU and two 6-pin connectors, while the AMD card uses no power connectors and is listed as an IGP (integrated graphics processor) with a slot width of "IGP," meaning it is designed to be embedded in a portable device. So the Radeon Pro 555 wins on integration and power envelope, not on raw compute.

For compute-heavy tasks using OpenCL or Vulkan, the GTX 660 Ti is the clear choice. For Metal-specific workloads or systems where power draw and physical integration matter more than raw throughput, the Radeon Pro 555 has the recorded data to back its role.

Architecture Differences

The two cards come from completely different design philosophies and eras. The GTX 660 Ti uses the GK104 chip, built on NVIDIA’s Kepler architecture, and belongs to the GeForce 600 generation. It is fabricated on a 28 nm process at TSMC, with 3,540 million transistors on a 294 mm² die. That yields a transistor density of 12.0 million per mm². The Radeon Pro 555 uses the Polaris 21 chip, built on AMD’s GCN 4.0 architecture, and belongs to the Radeon Pro Mac (500 Series) generation. It is fabricated on a 14 nm process at GlobalFoundries, with 3,000 million transistors on a much smaller 123 mm² die. That yields a transistor density of 24.4 million per mm², roughly double the density of the NVIDIA chip.

The Kepler chip is older, larger, and less dense, but it packs more compute hardware. The GTX 660 Ti has 1344 shading units, 112 texture mapping units, and 24 ROPs. The Radeon Pro 555 has 768 shading units, 48 TMUs, and 16 ROPs. The NVIDIA card also has higher clock rates: a base clock of 915 MHz and a boost clock of 980 MHz, while the AMD card’s base and boost clocks are not recorded in the database. The memory clocks differ as well, with the NVIDIA card running at 1502 MHz (6 Gbps effective) versus 1275 MHz (5.1 Gbps effective) for the AMD card. The memory bus is wider on the NVIDIA card at 192 bit versus 128 bit, and the bandwidth reflects that: 144.2 GB/s for the GTX 660 Ti versus 81.60 GB/s for the Radeon Pro 555.

The pixel rate and texture rate also favor the NVIDIA card. The GTX 660 Ti produces 27.44 GPixel/s and 109.8 GTexel/s, while the Radeon Pro 555 produces 13.60 GPixel/s and 40.80 GTexel/s. FP32 throughput is 2.634 TFLOPS for the NVIDIA card versus 1,305.6 GFLOPS for the AMD card. The AMD card does have FP16 at 1,305.6 GFLOPS with a 1:1 ratio, but the NVIDIA card has no recorded FP16 performance. The GTX 660 Ti does not support DirectX 12 (11_0) at the same level as the AMD card, which supports DirectX 12 (12_0). Both support OpenGL 4.6, but the Vulkan versions differ: the NVIDIA card supports Vulkan 1.2.175, while the AMD card supports Vulkan 1.3. The AMD card also has a different bus interface (PCIe 3.0 x8 versus PCIe 3.0 x16) and its display outputs are listed as "Portable Device Dependent," reflecting its laptop-oriented design.

FAQ

Q: Which card is faster in OpenCL?

A: The NVIDIA GeForce GTX 660 Ti scores 15113 in Geekbench OpenCL, while the AMD Radeon Pro 555 scores 11682. That is a 29.4% advantage for the NVIDIA card.

Q: Which card is faster in Vulkan?

A: The NVIDIA GeForce GTX 660 Ti scores 15012 in Geekbench Vulkan, while the AMD Radeon Pro 555 scores 12303. The NVIDIA card leads by 22%.

Q: Does the AMD card have any benchmark where it wins?

A: The AMD Radeon Pro 555 has a Geekbench Metal score of 16236, which is its highest recorded score, but there is no Metal score recorded for the GTX 660 Ti, so no direct comparison exists on that API.

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

A: The GTX 660 Ti is within 0.1% of the AMD Radeon 560X and within 0.9% of the RTX 3050 OEM. The Radeon Pro 555 is within 0.2% of the AMD Radeon HD 8950M and within 0.5% of the NVIDIA P106-090.

Q: What are the power requirements?

A: The GTX 660 Ti has a TDP of 150 W, requires a 450 W suggested PSU, and uses two 6-pin power connectors. The Radeon Pro 555 has a TDP of 75 W and uses no power connectors.

Q: Which card has better memory bandwidth?

A: The GTX 660 Ti has 144.2 GB/s of bandwidth on a 192-bit bus, while the Radeon Pro 555 has 81.60 GB/s on a 128-bit bus.

Specification Differences

The two cards differ across nearly every measured specification. The most obvious are the process node: 28 nm for the GTX 660 Ti versus 14 nm for the Radeon Pro 555. Transistor counts are close (3,540 million versus 3,000 million), but die size is not: 294 mm² versus 123 mm². Transistor density is 12.0M per mm² versus 24.4M per mm². Clock speeds differ in memory clock (1502 MHz versus 1275 MHz effective 5.1 Gbps), and the NVIDIA card has recorded base and boost clocks (915 MHz and 980 MHz) while the AMD card has none recorded. Memory size is the same at 2 GB and both use GDDR5, but the bus width differs (192 bit versus 128 bit, bandwidth 144.2 GB/s versus 81.60 GB/s). Shading units are 1344 versus 768, TMUs are 112 versus 48, and ROPs are 24 versus 16. Pixel rate is 27.44 GPixel/s versus 13.60 GPixel/s, texture rate is 109.8 GTexel/s versus 40.80 GTexel/s, and FP32 is 2.634 TFLOPS versus 1,305.6 GFLOPS. The AMD card has FP16 at 1,305.6 GFLOPS, while the NVIDIA card has none recorded. TDP is 150 W versus 75 W. The GTX 660 Ti is dual-slot with two 6-pin connectors and a 450 W suggested PSU; the Radeon Pro 555 is an IGP with no connectors and no suggested PSU. Bus interface is PCIe 3.0 x16 versus PCIe 3.0 x8. Display outputs are 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 for the NVIDIA card, versus "Portable Device Dependent" for the AMD card. The API support differs in DirectX version (12 (11_0) versus 12 (12_0, Vulkan 1.2.175 versus 1.3, with OpenGL 4.6 on both). Release dates differ as well, with the GTX 660 Ti released on 2012-08-15T17:00:00.000Z and the AMD card on 2017-06-04T17:00:00.000Z. The GTX 660 Ti has a launch MSRP of 299 USD; the AMD card has no launch MSRP recorded.

The Verdict

The data points in one direction for raw performance: the NVIDIA GeForce GTX 660 Ti wins both recorded head-to-head benchmarks by 29.4% in OpenCL and 22% in Vulkan. Its average benchmark score is 15063 versus 13407 for the AMD Radeon Pro 555, and it sits at the 57th percentile of all GPUs versus the 54th for the AMD card. If the task is compute-heavy and uses OpenCL or Vulkan, the GTX 660 Ti is the stronger choice.

The Radeon Pro 555 is not without its own case. It has a Metal score of 16236, the highest single score recorded for either card, and it does so at a 75 W TDP with no external power connectors. It is an IGP designed for portable devices, with a 14 nm process and over twice the transistor density of the NVIDIA chip. For a system where power draw, physical footprint, and Metal compatibility are priorities, the Radeon Pro 555 is the only one of the two with recorded data supporting that role.

The GTX 660 Ti is the faster card in every direct comparison available. The Radeon Pro 555 is the more efficient and more integrated option, but its only unique advantage in the database is Metal support, where no competing score exists for the NVIDIA card. Pick the GTX 660 Ti for compute performance; pick the Radeon Pro 555 for Metal-centric or power-constrained scenarios.