AMD Radeon 660M vs NVIDIA GeForce GTX 1070 Ti Comparison

The NVIDIA GeForce GTX 1070 Ti and the AMD Radeon 660M represent two fundamentally different approaches to graphics processing, separated by nearly half a decade of technological evolution. The head-to-head benchmark data is starkly one-sided: the GTX 1070 Ti wins both available tests, with a 282.3% advantage in Geekbench OpenCL (49,219 vs 12,876) and a 313.7% lead in Geekbench Vulkan (61,006 vs 14,748). Yet the average benchmark scores tell a more nuanced story, with the 1070 Ti averaging 14,277 and the 660M averaging 13,812 — a mere 3.4% gap that defies the individual test results. This contradiction between the head-to-head deltas and the aggregate averages suggests the two cards excel in different workloads, and the data rewards a closer look at what each architecture actually prioritizes.

Head-to-Head Benchmarks

The Geekbench results are unambiguous in their verdict. In OpenCL, the GTX 1070 Ti scores 49,219 against the 660M’s 12,876, a 282.3% margin that dwarfs any other comparison in the data. Vulkan is even more lopsided: 61,006 vs 14,748, a 313.7% advantage. These are not close contests; the Pascal-based discrete card is delivering roughly four times the raw compute throughput in these synthetic workloads. The texture rate differential supports this — the 1070 Ti pushes 255.8 GTexel/s versus 45.60 GTexel/s for the 660M, a 5.6x gap. Pixel rate tells a similar story: 107.7 GPixel/s vs 30.40 GPixel/s, a 3.5x difference. The FP32 throughput is equally decisive: 8.186 TFLOPS against 1,459.2 GFLOPS (approximately 1.46 TFLOPS), a 5.6x advantage.

What makes the aggregate scores so close is revealing. The 1070 Ti’s Passmark G3D score of 14,674 and GPU Compute score of 7,165 pull its average to 14,277, while the 660M’s two Geekbench scores average out to 13,812. The 1070 Ti’s percentile rank of 56 versus the 660M’s 55 places them nearly adjacent in the global GPU hierarchy, yet the head-to-head deltas suggest the 660M’s nearest rivals — the NVIDIA RTX A2000 Mobile at 13,821 (0.1% behind) and the AMD Radeon RX 570X at 13,871 (0.4% behind) — are closer to its actual performance level. The 1070 Ti, sitting at 0.3% behind the Intel Iris Xe MAX Graphics and 0.5% behind the AMD Radeon Vega 11, appears to be grouped with older integrated and entry-level parts in aggregate terms, masking its dominance in specific compute tests.

The Verdict

The data points to a clear split: the GTX 1070 Ti is the compute and rasterization powerhouse, while the 660M is a far more efficient and modern architecture that trades raw speed for capability per watt. The 1070 Ti’s 180 W TDP against the 660M’s 40 W TDP is the defining difference — the NVIDIA card consumes 4.5x the power to deliver roughly 5.6x the texture rate and 3.5x the pixel rate. For users who prioritize peak frame rates and compute-heavy workloads, the 1070 Ti is the obvious choice. For those constrained by power budgets, thermal limits, or form factor, the 660M’s integrated design (labeled IGP, with no power connectors and portable-device-dependent outputs) makes it the only practical option. The 660M’s FP16 performance is a notable exception — its 2.918 TFLOPS at 2:1 ratio vastly exceeds the 1070 Ti’s 127.9 GFLOPS at 1:64, indicating the RDNA 2.0 architecture is designed for workloads that leverage half-precision math, a capability the older Pascal card effectively lacks.

Architecture Differences

The two GPUs are built on opposite ends of the semiconductor spectrum. The GTX 1070 Ti uses the GP104 chip on TSMC’s 16 nm process, packing 7,200 million transistors into a 314 mm² die for a density of 22.9 million transistors per square millimeter. The 660M uses the Rembrandt chip on TSMC’s 6 nm process, fitting 13,100 million transistors into a smaller 208 mm² die — a density of 63.0 million per square millimeter, nearly three times higher. This density advantage is the root of the 660M’s efficiency: it delivers modern features in a fraction of the silicon area. The 1070 Ti’s Pascal architecture is limited to DirectX 12 (12_1), while the 660M’s RDNA 2.0 supports DirectX 12 Ultimate (12_2), including six ray tracing cores (the 1070 Ti has none). Both support OpenGL 4.6 and Vulkan 1.4, but the 660M’s ray tracing capability is a generational leap that the Pascal card cannot match. The 1070 Ti features 2,432 shading units, 152 TMUs, and 64 ROPs; the 660M has 384 shading units, 24 TMUs, and 16 ROPs — a 6.3x, 6.3x, and 4x reduction respectively, mirroring the texture and pixel rate gaps.

FAQ

Q: Which GPU has higher raw compute throughput?

A: The GTX 1070 Ti, with FP32 performance of 8.186 TFLOPS versus the 660M’s 1,459.2 GFLOPS, a 5.6x advantage. In Geekbench OpenCL, the 1070 Ti scores 49,219 against 12,876, a 282.3% lead.

Q: Does the 660M have any performance advantage?

A: Yes, in FP16 compute. The 660M delivers 2.918 TFLOPS at a 2:1 ratio, while the 1070 Ti manages only 127.9 GFLOPS at a 1:64 ratio — a 22.8x difference in half-precision throughput.

Q: What is the power consumption difference?

A: The GTX 1070 Ti has a TDP of 180 W and requires a single 8-pin power connector with a 450 W suggested PSU. The 660M has a 40 W TDP, no power connectors, and integrates as an IGP with portable-device-dependent display outputs.

Q: How do their memory architectures differ?

A: The 1070 Ti has 8 GB of GDDR5 on a 256-bit bus with 256.3 GB/s bandwidth. The 660M uses system-shared memory with system-dependent bandwidth, meaning its memory performance varies with the host system.

Q: Which has better ray tracing support?

A: The 660M includes 6 ray tracing cores and supports DirectX 12 Ultimate (12_2). The 1070 Ti has no ray tracing cores and is limited to DirectX 12 (12_1).

Q: Are their overall benchmark scores similar?

A: Yes, the average benchmark scores are close — 14,277 for the 1070 Ti and 13,812 for the 660M — placing them at the 56th and 55th percentiles of all GPUs respectively, despite the 1070 Ti’s dominance in specific head-to-head tests.

Where Each One Wins

The GTX 1070 Ti wins in every scenario that demands raw rasterization throughput. Its 255.8 GTexel/s texture rate and 107.7 GPixel/s pixel rate make it suitable for high-resolution gaming at high detail settings, where texture fill and pixel output are the bottlenecks. The 1070 Ti’s 8 GB GDDR5 memory with 256.3 GB/s bandwidth provides consistent frame pacing in memory-intensive titles, and its 64 ROPs ensure strong anti-aliasing performance. The Geekbench Vulkan score of 61,006 indicates strong low-level API performance, benefiting modern game engines that leverage Vulkan’s direct hardware access. The 1070 Ti’s Passmark scores reinforce this: a DirectX 11 score of 107 and a DirectX 12 score of 51, alongside a G3D score of 14,674, show broad API compatibility.

The 660M wins in scenarios where power efficiency and modern features matter more than absolute performance. Its 40 W TDP makes it viable for thin-and-light laptops where the 1070 Ti’s 180 W draw is physically impossible. The 6 nm process and 63.0M transistors per square millimeter density mean the 660M can deliver ray tracing and DirectX 12 Ultimate support in an integrated package, enabling hardware-accelerated ray tracing in a form factor that a discrete card cannot match. The 660M’s FP16 throughput of 2.918 TFLOPS suits machine learning inference and compute workloads that use half-precision arithmetic, where the 1070 Ti’s 1:64 FP16 ratio is a severe handicap. The 660M’s PCIe 4.0 x8 interface also provides faster host communication than the 1070 Ti’s PCIe 3.0 x16, which can benefit workloads that stream data from system memory.

Specification Differences

The two cards differ in nearly every measurable specification. The 1070 Ti uses the Pascal architecture on a 16 nm process with 7,200 million transistors on a 314 mm² die; the 660M uses RDNA 2.0 on 6 nm with 13,100 million transistors on a 208 mm² die. Clock speeds: the 1070 Ti runs at 1607 MHz base and 1683 MHz boost, while the 660M runs at 1500 MHz base and 1900 MHz boost. Memory is the starkest split — the 1070 Ti has 8 GB GDDR5 on a 256-bit bus with 256.3 GB/s bandwidth, while the 660M uses system-shared memory with system-dependent bandwidth. The 1070 Ti has 2,432 shading units, 152 TMUs, and 64 ROPs; the 660M has 384 shading units, 24 TMUs, and 16 ROPs — a 6.3x reduction in both shading units and TMUs, and a 4x reduction in ROPs. The 1070 Ti outputs 107.7 GPixel/s and 255.8 GTexel/s; the 660M outputs 30.40 GPixel/s and 45.60 GTexel/s. FP32: 8.186 TFLOPS vs 1,459.2 GFLOPS. FP16: 127.9 GFLOPS (1:64) vs 2.918 TFLOPS (2:1). TDP: 180 W vs 40 W. The 1070 Ti is dual-slot with a 1x 8-pin connector and 450 W suggested PSU; the 660M is IGP with no connectors. The 1070 Ti has fixed display outputs (1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.4a) and measures 267 mm x 112 mm x 40 mm; the 660M’s outputs are portable-device-dependent with no listed dimensions. The 1070 Ti supports DirectX 12 (12_1), while the 660M supports DirectX 12 Ultimate (12_2) with 6 ray tracing cores. The 1070 Ti’s launch MSRP is 399 USD; the 660M has no launch MSRP. Both support OpenGL 4.6 and Vulkan 1.4, and both are end-of-life production status.