AMD Radeon RX Vega 11 vs NVIDIA GeForce GTX 660 Ti Comparison
AMD Radeon RX Vega 11
GeForce GTX 660 Ti
PERFORMANCE BENCHMARKS
Analysis: AMD Radeon RX Vega 11 vs NVIDIA GeForce GTX 660 Ti
The NVIDIA GeForce GTX 660 Ti and the AMD Radeon RX Vega 11 represent two very different approaches to graphics processing: a dedicated, add-in board from the Kepler era versus an integrated GPU (IGP) built into a modern APU. The benchmark data shows a clear overall winner in the GTX 660 Ti, which takes both head-to-head tests, but the RX Vega 11 holds its own in specific API workloads, making the comparison more nuanced than a simple performance ranking. The GTX 660 Ti posts an average benchmark score of 15063, placing it in the 57th percentile of all GPUs, while the RX Vega 11 scores 14385 on average, landing in the 56th percentile. These percentile rankings are close, but the raw scores and the architectural context reveal distinct strengths and weaknesses.
Where Each One Wins
The GTX 660 Ti is the definitive winner in raw compute performance as measured by OpenCL and Vulkan. In the Geekbench OpenCL test, the GTX 660 Ti scores 15113 against the RX Vega 11’s 12525, a commanding 20.7% lead. This is the largest single performance gap between the two cards, indicating that the dedicated GPU’s higher shading unit count and dedicated memory bandwidth provide a substantial advantage in general-purpose compute tasks. For workloads that rely heavily on OpenCL—such as video encoding, physics simulations, and certain rendering tasks—the GTX 660 Ti is clearly the superior choice.
In the Vulkan API test, the GTX 660 Ti again wins, scoring 15012 versus 13543 for the RX Vega 11, a 10.8% advantage. Vulkan is a low-overhead API that benefits from raw processing power and memory bandwidth, both of which favor the GTX 660 Ti. While the lead is smaller than in OpenCL, it still demonstrates that the dedicated GPU maintains its edge in modern graphics APIs, making it the better option for gaming and other real-time graphics workloads that leverage Vulkan.
However, the RX Vega 11 has one notable victory: the Geekbench Metal test. Here, it scores 17086, a figure that is not part of the head-to-head comparison because the GTX 660 Ti does not support Metal. This score is significantly higher than the GTX 660 Ti’s OpenCL and Vulkan scores, suggesting that the RX Vega 11’s GCN 5.0 architecture is highly efficient in Apple’s Metal API. For users in macOS environments or applications that specifically use Metal for compute, the RX Vega 11 is the superior choice, despite its overall lower average score. The data shows that while the GTX 660 Ti wins on two of three API benchmarks, the RX Vega 11’s Metal performance is a clear outlier that cannot be ignored, as it outperforms the GTX 660 Ti’s best result by 13.8 points.
Architecture Differences
The two GPUs are built on fundamentally different architectures and process nodes. The GTX 660 Ti uses NVIDIA’s Kepler architecture, fabricated by TSMC on a 28 nm process, with a die size of 294 mm² and a transistor count of 3,540 million. In contrast, the RX Vega 11 uses AMD’s GCN 5.0 architecture, built on a 12 nm process by GlobalFoundries, with a smaller die of 210 mm² but a higher transistor count of 4,940 million. This difference in process node and transistor density—12.0M / mm² for the GTX 660 Ti versus 23.5M / mm² for the RX Vega 11—shows that the AMD chip is far denser, packing more transistors into less space, which is typical for a newer, more integrated design.
The GTX 660 Ti is a discrete GPU with its own dedicated memory subsystem: 2 GB of GDDR5 on a 192-bit bus, providing 144.2 GB/s of bandwidth. The RX Vega 11, being an IGP, relies on system memory, with its size, type, bus width, and bandwidth all listed as “System Dependent.” This is a critical architectural difference; the GTX 660 Ti’s dedicated memory is far faster than any shared system memory configuration, which explains its significant lead in memory-intensive benchmarks. The GTX 660 Ti also has a much larger execution resource pool: 1344 shading units, 112 texture mapping units (TMUs), and 24 raster output units (ROPs), compared to the RX Vega 11’s 704 shading units, 44 TMUs, and 8 ROPs.
Clock speeds tell another story. The GTX 660 Ti runs at a base clock of 915 MHz with a boost of 980 MHz, while the RX Vega 11 has a much lower base clock of 300 MHz but a significantly higher boost clock of 1400 MHz. This aggressive boost behavior allows the RX Vega 11 to reach high peak performance when thermals and power allow, but its low base clock and shared memory architecture limit sustained performance. The RX Vega 11 also supports FP16 compute at 3.942 TFLOPS (2:1), whereas the GTX 660 Ti has no listed FP16 capability, only FP32 at 2.634 TFLOPS. The GTX 660 Ti’s FP32 output is higher than the RX Vega 11’s 1.971 TFLOPS, reinforcing its lead in single-precision compute tasks.
FAQ
Q: Which GPU has a higher average benchmark score?
A: The NVIDIA GeForce GTX 660 Ti has a higher average benchmark score of 15063, compared to the AMD Radeon RX Vega 11’s 14385, a difference of 678 points.
Q: Does the AMD Radeon RX Vega 11 win any benchmark tests?
A: Yes, the RX Vega 11 scores 17086 in the Geekbench Metal test, which is not available for the GTX 660 Ti. This score is higher than the GTX 660 Ti’s best result of 15113 in OpenCL.
Q: What is the largest performance gap in the head-to-head benchmarks?
A: The largest gap is in the Geekbench OpenCL test, where the GTX 660 Ti leads by 20.7%, scoring 15113 against the RX Vega 11’s 12525.
Q: How does the memory configuration differ between the two?
A: The GTX 660 Ti has 2 GB of dedicated GDDR5 memory on a 192-bit bus with 144.2 GB/s bandwidth. The RX Vega 11 uses system shared memory, with bandwidth described as “System Dependent.”
Q: What is the transistor density difference?
A: The RX Vega 11 has a much higher transistor density at 23.5M / mm², while the GTX 660 Ti has 12.0M / mm², despite the GTX 660 Ti having a larger die area of 294 mm² versus 210 mm².
Q: Which GPU has a higher boost clock?
A: The AMD Radeon RX Vega 11 has a significantly higher boost clock of 1400 MHz, compared to the GTX 660 Ti’s 980 MHz. However, the GTX 660 Ti’s base clock is higher at 915 MHz versus 300 MHz.
Specification Differences
The following table highlights the key specification differences between the two GPUs, based solely on the data provided.
| Specification | NVIDIA GeForce GTX 660 Ti | AMD Radeon RX Vega 11 |
|---|---|---|
| Architecture | Kepler | GCN 5.0 |
| Process Node | 28 nm | 12 nm |
| Foundry | TSMC | GlobalFoundries |
| Transistors | 3,540 million | 4,940 million |
| Die Size | 294 mm² | 210 mm² |
| Transistor Density | 12.0M / mm² | 23.5M / mm² |
| Base Clock | 915 MHz | 300 MHz |
| Boost Clock | 980 MHz | 1400 MHz |
| Memory Size | 2 GB | System Shared |
| Memory Type | GDDR5 | System Shared |
| Memory Bus Width | 192 bit | System Shared |
| Memory Bandwidth | 144.2 GB/s | System Dependent |
| Shading Units | 1344 | 704 |
| TMUs | 112 | 44 |
| ROPs | 24 | 8 |
| Pixel Rate | 27.44 GPixel/s | 11.20 GPixel/s |
| Texture Rate | 109.8 GTexel/s | 61.60 GTexel/s |
| FP32 | 2.634 TFLOPS | 1.971 TFLOPS |
| FP16 | N/A | 3.942 TFLOPS (2:1) |
| TDP | 150 W | 15 W |
| Slot Width | Dual-slot | IGP |
| Power Connectors | 2x 6-pin | None |
| Suggested PSU | 450 W | N/A |
| Bus Interface | PCIe 3.0 x16 | IGP |
| Display Outputs | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 | Motherboard Dependent |
| DirectX Version | 12 (11_0) | 12 (12_1) |
| Vulkan Version | 1.2.175 | 1.3 |
| Release Date | 2012-08-15 | 2019-07-06 |
| Launch MSRP | 299 USD | N/A |
Head-to-Head Benchmarks
The head-to-head comparison consists of two Geekbench tests, both of which are won by the NVIDIA GeForce GTX 660 Ti. The first test, Geekbench OpenCL, shows the GTX 660 Ti scoring 15113 against the RX Vega 11’s 12525. This 20.7% delta is the most decisive result in the comparison. The GTX 660 Ti’s dedicated 144.2 GB/s of memory bandwidth and 1344 shading units allow it to process OpenCL workloads far more efficiently than the RX Vega 11, which is constrained by system shared memory and only 704 shading units. The RX Vega 11’s lower FP32 throughput of 1.971 TFLOPS, compared to 2.634 TFLOPS for the GTX 660 Ti, further explains this gap in compute-heavy tasks.
In the second test, Geekbench Vulkan, the GTX 660 Ti again wins, with a score of 15012 versus 13543 for the RX Vega 11. The delta here is 10.8%, a smaller margin than in OpenCL but still a clear victory. Vulkan’s low-overhead nature tends to reward raw throughput, and the GTX 660 Ti’s higher texture rate of 109.8 GTexel/s and pixel rate of 27.44 GPixel/s, versus 61.60 GTexel/s and 11.20 GPixel/s for the RX Vega 11, contribute to this advantage. However, the RX Vega 11’s higher boost clock of 1400 MHz and support for DirectX 12 (12_1) and Vulkan 1.3 suggest that in properly optimized scenarios, it can narrow the gap, as evidenced by the smaller delta in Vulkan compared to OpenCL.
While the RX Vega 11 loses both head-to-head tests, its performance in the Geekbench Metal test—scoring 17086—is not captured in the head-to-head data. This score is 13.8% higher than the GTX 660 Ti’s best OpenCL score, indicating that in Metal-specific workloads, the RX Vega 11 is the superior part. The data shows that the GTX 660 Ti wins decisively in traditional compute APIs, but the RX Vega 11’s architectural efficiency in Metal makes it a viable alternative for users in that ecosystem. The overall wins tally stands at 2 for the GTX 660 Ti and 0 for the RX Vega 11 in the head-to-head tests, but the Metal result adds a critical caveat to the narrative.