AMD Radeon RX 5700 XT vs NVIDIA GeForce GTX 660 Ti Comparison

The Verdict

The data is unambiguous: the AMD Radeon RX 5700 XT and NVIDIA GeForce GTX 660 Ti are not competing in the same league, despite the GTX 660 Ti's surprising Geekbench OpenCL result. The RX 5700 XT is the decisive choice for anyone running modern APIs like Vulkan, where it leads by 355% in the head-to-head benchmark. The GTX 660 Ti’s only victory arrives in Geekbench OpenCL, where it scores 15113 against the RX 5700 XT’s 9524, a 37% advantage. However, that single win does not offset the broader picture: the RX 5700 XT’s average benchmark score of 16361 versus 15063 for the GTX 660 Ti, and its 59th percentile ranking among all GPUs compared to 57th for the older card.

For a gamer targeting modern titles, the RX 5700 XT is the only rational pick. Its Vulkan score of 68310 dwarfs the GTX 660 Ti’s 15012, and its Passmark G3D score of 16277 demonstrates sustained DirectX performance that the GTX 660 Ti cannot match. The GTX 660 Ti, by contrast, is a legacy part whose OpenCL strength suggests compute-oriented workloads from its era, but its 2 GB memory and 192-bit bus are architectural constraints that modern games will exceed. The verdict is simple: choose the RX 5700 XT for contemporary gaming and API support; choose the GTX 660 Ti only if your workload is specifically OpenCL-bound and you cannot use any other card.

Architecture Differences

The two GPUs come from different technological eras, and the gap shows in every major architectural metric. The RX 5700 XT uses AMD’s RDNA 1.0 architecture on a 7 nm TSMC process, packing 10,300 million transistors into a 251 mm² die. The GTX 660 Ti uses NVIDIA’s Kepler architecture on a 28 nm TSMC process, with 3,540 million transistors on a larger 294 mm² die. The transistor density difference is stark: 41.0M per mm² for the RX 5700 XT versus 12.0M per mm² for the GTX 660 Ti. That density advantage translates directly into compute resources: the RX 5700 XT has 2,560 shading units, 160 TMUs, and 64 ROPs, while the GTX 660 Ti has 1,344 shading units, 112 TMUs, and only 24 ROPs.

Memory architecture also diverges sharply. The RX 5700 XT features 8 GB of GDDR6 on a 256-bit bus, delivering 448.0 GB/s of bandwidth. The GTX 660 Ti has 2 GB of GDDR5 on a 192-bit bus, yielding 144.2 GB/s. The RX 5700 XT’s clock speeds are higher across the board: 1605 MHz base and 1905 MHz boost versus 915 MHz base and 980 MHz boost for the GTX 660 Ti. The RX 5700 XT also supports PCIe 4.0 x16, while the GTX 660 Ti is limited to PCIe 3.0 x16. API support favors the newer card: the RX 5700 XT supports DirectX 12 (12_1) and Vulkan 1.4, while the GTX 660 Ti supports DirectX 12 (11_0) and Vulkan 1.2.175. Both cards lack ray tracing and tensor cores, but the RX 5700 XT’s RDNA 1.0 architecture is fundamentally more modern.

FAQ

Q: Which card has a higher average benchmark score?

A: The AMD Radeon RX 5700 XT has an average benchmark score of 16361, while the NVIDIA GeForce GTX 660 Ti scores 15063. The RX 5700 XT also ranks higher in the overall percentile, at 59th versus 57th for the GTX 660 Ti.

Q: Why does the GTX 660 Ti win in Geekbench OpenCL?

A: The GTX 660 Ti scores 15113 in Geekbench OpenCL, compared to 9524 for the RX 5700 XT, a 37% advantage. This is the only head-to-head benchmark where the GTX 660 Ti wins, and it suggests Kepler’s compute scheduling was particularly efficient for that specific OpenCL workload.

Q: How large is the Vulkan performance gap?

A: The RX 5700 XT scores 68310 in Geekbench Vulkan, while the GTX 660 Ti scores 15012. That is a 355% advantage for the RX 5700 XT, making it the clear choice for any Vulkan-based game or application.

Q: What are the memory specifications of each card?

A: The RX 5700 XT has 8 GB of GDDR6 memory on a 256-bit bus with 448.0 GB/s bandwidth. The GTX 660 Ti has 2 GB of GDDR5 memory on a 192-bit bus with 144.2 GB/s bandwidth.

Q: Which card has a higher transistor density?

A: The RX 5700 XT has a transistor density of 41.0M per mm², while the GTX 660 Ti has 12.0M per mm². This reflects the 7 nm versus 28 nm process node difference.

Q: How do the ROP counts compare?

A: The RX 5700 XT has 64 ROPs, while the GTX 660 Ti has only 24 ROPs. This contributes to the RX 5700 XT’s pixel rate of 121.9 GPixel/s versus 27.44 GPixel/s for the GTX 660 Ti.

Specification Differences

The two cards differ in nearly every specification field. The RX 5700 XT uses a 7 nm process node, while the GTX 660 Ti uses 28 nm. Transistor count is 10,300 million versus 3,540 million, and die size is 251 mm² versus 294 mm². Transistor density is 41.0M per mm² versus 12.0M per mm². Base clock is 1605 MHz versus 915 MHz, and boost clock is 1905 MHz versus 980 MHz. The RX 5700 XT has a game clock of 1755 MHz, which the GTX 660 Ti lacks entirely. Memory speed is 1750 MHz (14 Gbps effective) versus 1502 MHz (6 Gbps effective).

Memory capacity is 8 GB versus 2 GB, type is GDDR6 versus GDDR5, bus width is 256-bit versus 192-bit, and bandwidth is 448.0 GB/s versus 144.2 GB/s. Shading units are 2560 versus 1344, TMUs are 160 versus 112, and ROPs are 64 versus 24. Pixel rate is 121.9 GPixel/s versus 27.44 GPixel/s, texture rate is 304.8 GTexel/s versus 109.8 GTexel/s, and FP32 performance is 9.754 TFLOPS versus 2.634 TFLOPS. The RX 5700 XT has FP16 performance of 19.51 TFLOPS (2:1), while the GTX 660 Ti has no FP16 data.

TDP is 225 W versus 150 W, power connectors are 1x 6-pin + 1x 8-pin versus 2x 6-pin, and suggested PSU is 550 W versus 450 W. Bus interface is PCIe 4.0 x16 versus PCIe 3.0 x16. Display outputs are 1x HDMI 2.0b and 3x DisplayPort 1.4a versus 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. DirectX support is 12 (12_1) versus 12 (11_0), Vulkan is 1.4 versus 1.2.175. The RX 5700 XT is 272 mm long, 111 mm tall, and 36 mm wide, while the GTX 660 Ti is 241 mm long with no height or width data. The RX 5700 XT launched with an MSRP of 399 USD, while the GTX 660 Ti launched with an MSRP of 299 USD.

Head-to-Head Benchmarks

The head-to-head data presents a clear split. In Geekbench OpenCL, the GTX 660 Ti wins decisively with a score of 15113 against the RX 5700 XT’s 9524, a delta of -37% for the AMD card. This is a notable result, as it shows the older Kepler architecture outperforming RDNA 1.0 in a compute-heavy OpenCL workload by more than a third. However, the story reverses completely in Geekbench Vulkan. The RX 5700 XT scores 68310, versus 15012 for the GTX 660 Ti, a 355% advantage. This is not a marginal win; it is a generational obliteration. Vulkan is the modern cross-platform graphics API, and the RX 5700 XT’s lead there is the single most important benchmark result in this comparison.

Looking at broader benchmarks available for each card, the RX 5700 XT shows strength across multiple DirectX tests. Its Passmark DirectX 11 score is 116, DirectX 12 is 64, DirectX 10 is 97, and DirectX 9 is 226. The GTX 660 Ti has no DirectX benchmark entries in the data, so the comparison relies on the head-to-head results and the average scores. The RX 5700 XT’s Passmark G3D score of 16277 and GPU compute score of 7650 indicate robust general-purpose performance. The GTX 660 Ti’s average benchmark score of 15063 is driven entirely by its two Geekbench results, which are both near that average. The RX 5700 XT’s average of 16361 is buoyed by its Vulkan score and its Passmark G3D result, showing a more balanced profile.

Where Each One Wins

The RX 5700 XT wins in every scenario that involves modern graphics APIs or high-resolution gaming. Its Vulkan score is 355% higher, its memory bandwidth is 448.0 GB/s versus 144.2 GB/s, and its FP32 performance is 9.754 TFLOPS versus 2.634 TFLOPS. For any game using Vulkan, DirectX 12, or even DirectX 11, the RX 5700 XT is the superior choice. Its 8 GB of GDDR6 memory also provides headroom for texture-heavy workloads that would exhaust the GTX 660 Ti’s 2 GB frame buffer. The RX 5700 XT’s pixel rate of 121.9 GPixel/s and texture rate of 304.8 GTexel/s make it suitable for high-refresh-rate gaming, while the GTX 660 Ti’s 27.44 GPixel/s and 109.8 GTexel/s are artifacts of its older architecture.

The GTX 660 Ti wins specifically in Geekbench OpenCL, where its 15113 score beats the RX 5700 XT by 37%. This suggests that for OpenCL compute tasks—such as certain physics simulations or legacy scientific workloads—the GTX 660 Ti may still hold a niche advantage. However, this is a narrow victory. The GTX 660 Ti’s lower TDP of 150 W versus 225 W means it draws less power, and its shorter length of 241 mm versus 272 mm could fit in smaller cases. But these are secondary considerations. The data shows the RX 5700 XT wins the overwhelming majority of relevant benchmarks, and its architectural advantages in memory, compute, and API support make it the definitive choice for virtually all modern use cases. The GTX 660 Ti is a historical curiosity whose OpenCL result is an outlier, not a trend.