NVIDIA GeForce GTX 660 Ti vs NVIDIA T400 Comparison

Where Each One Wins

The recorded data splits cleanly along API lines. The NVIDIA T400 wins both recorded benchmark disciplines, but the margin tells a more nuanced story. In the Geekbench OpenCL test, the T400 posts a score of 17,039 against the GTX 660 Ti’s 15,113, a 12.7% advantage. That is a substantial gap, one that suggests the T400 is notably stronger in compute workloads that leverage OpenCL’s general-purpose execution model.

The Vulkan result narrows considerably. The T400 scores 15,976, while the GTX 660 Ti manages 15,012, a 6.4% difference. This smaller margin implies that when both cards are pushed through a modern graphics API, the older Kepler architecture holds up better relative to its compute disadvantage. The T400 still wins, but the gap is roughly half of what appears in OpenCL.

For users focused on legacy API performance, the GTX 660 Ti is closer than the average scores might suggest. The database’s average benchmark score for the T400 is 16,508, versus 15,063 for the GTX 660 Ti. That 1,445-point average difference, about 9.6%, reflects the T400’s consistent superiority across both tests. The GTX 660 Ti’s strongest showing is in Vulkan, where its 15,012 score lands within 6.4% of the T400, making it the better choice among the two if Vulkan compatibility is the primary concern.

Architecture Differences

The two cards come from different architectural eras. The T400 uses the TU117 chip on the Turing architecture, built on TSMC’s 12 nm process. The GTX 660 Ti uses the GK104 chip on the Kepler architecture, fabricated on TSMC’s 28 nm node. This process gap is significant: the T400 packs 4,700 million transistors into a 200 mm² die, yielding a transistor density of 23.5 million per square millimeter. The GTX 660 Ti contains 3,540 million transistors across a larger 294 mm² die, for a density of just 12.0 million per square millimeter. The T400’s newer process allows nearly double the density, which explains how a smaller chip can deliver comparable or better performance.

Clock behavior differs sharply. The T400 runs a 420 MHz base clock but boosts to 1,425 MHz, a 1,005 MHz boost range. The GTX 660 Ti starts at 915 MHz base and boosts to 980 MHz, a mere 65 MHz range. The T400’s aggressive boost algorithm is a hallmark of modern graphics cards, while the GTX 660 Ti’s near-static clocks reflect an older design philosophy.

Memory technology also diverges. The T400 uses 2 GB of GDDR6 at 10 Gbps effective, on a 64-bit bus, yielding 80.00 GB/s bandwidth. The GTX 660 Ti uses 2 GB of GDDR5 at 6 Gbps effective, on a 192-bit bus, yielding 144.2 GB/s. The GTX 660 Ti has nearly double the memory bandwidth, a critical advantage for texture-heavy workloads. However, the T400 compensates with newer memory technology and a more efficient architecture.

Feature support differs at the API level. The T400 supports DirectX 12 (12_1) and Vulkan 1.4. The GTX 660 Ti is limited to DirectX 12 (11_0) and Vulkan 1.2.175. The T400’s higher DirectX feature level and newer Vulkan version indicate better support for modern rendering techniques. The T400 also outputs via 3x mini-DisplayPort 1.4a, while the GTX 660 Ti uses 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2.

FAQ

Q: Which card has the higher average benchmark score?

A: The T400 averages 16,508 points, while the GTX 660 Ti averages 15,063 points. The T400 leads by 9.6% based on these two recorded tests.

Q: How much faster is the T400 in OpenCL?

A: The T400 scores 17,039 in Geekbench OpenCL, compared to 15,113 for the GTX 660 Ti. That is a 12.7% advantage.

Q: Is the GTX 660 Ti competitive in any discipline?

A: In Vulkan, the GTX 660 Ti scores 15,012, which is within 6.4% of the T400’s 15,976. This is its closest performance, though it still loses.

Q: What memory bandwidth does each card provide?

A: The T400 offers 80.00 GB/s on a 64-bit bus with GDDR6. The GTX 660 Ti offers 144.2 GB/s on a 192-bit bus with GDDR5.

Q: Which card consumes less power?

A: The T400 is rated at 30 W TDP and requires no power connectors. The GTX 660 Ti is rated at 150 W TDP and requires two 6-pin connectors.

Q: What is the release timeline for these cards?

A: The T400 launched in May 2021, while the GTX 660 Ti launched in August 2012. Both are now end-of-life products.

Specification Differences

The database records several key differences between the two cards. The T400 uses the TU117 chip on a 12 nm process, while the GTX 660 Ti uses the GK104 chip on a 28 nm process. Transistor counts differ: 4,700 million for the T400 versus 3,540 million for the GTX 660 Ti. Die sizes are 200 mm² and 294 mm² respectively.

Clock speeds show the T400’s wider boost range: 420 MHz base to 1,425 MHz boost, versus 915 MHz base to 980 MHz boost for the GTX 660 Ti. Memory clocks also differ: 10 Gbps effective for the T400’s GDDR6, 6 Gbps effective for the GTX 660 Ti’s GDDR5.

Memory bus widths are 64-bit for the T400 and 192-bit for the GTX 660 Ti. Bandwidth figures are 80.00 GB/s versus 144.2 GB/s. Both have 2 GB of memory, but the type and bus configuration differ.

Compute resources vary widely: the T400 has 384 shading units, 24 TMUs, and 16 ROPs. The GTX 660 Ti has 1,344 shading units, 112 TMUs, and 24 ROPs. Despite having fewer units, the T400 achieves higher clock speeds. Pixel rates are 22.80 GPixel/s for the T400 and 27.44 GPixel/s for the GTX 660 Ti. Texture rates are 34.20 GTexel/s versus 109.8 GTexel/s. FP32 compute is 1,094.4 GFLOPS for the T400 and 2.634 TFLOPS for the GTX 660 Ti. The T400 supports FP16 at 2.189 TFLOPS (2:1), while the GTX 660 Ti has no recorded FP16 support.

Power requirements differ dramatically: 30 W TDP for the T400 with no power connectors, 150 W TDP for the GTX 660 Ti with two 6-pin connectors. Suggested PSU ratings are 200 W for the T400 and 450 W for the GTX 660 Ti. Slot widths are single-slot for the T400 and dual-slot for the GTX 660 Ti. The GTX 660 Ti has a recorded length of 241 mm, while the T400 has no recorded dimensions.

API support differs: DirectX 12 (12_1) for the T400 versus DirectX 12 (11_0) for the GTX 660 Ti. Vulkan support is 1.4 versus 1.2.175. Both support OpenGL 4.6.

Head-to-Head Benchmarks

The two recorded head-to-head tests both favor the T400. In Geekbench OpenCL, the T400’s 17,039 beats the GTX 660 Ti’s 15,113 by 1,926 points, a 12.7% delta. This is the larger of the two gaps and indicates the T400’s compute architecture handles OpenCL workloads more efficiently. The GTX 660 Ti’s raw FP32 throughput is higher at 2.634 TFLOPS versus 1,094.4 GFLOPS, yet it loses this test, suggesting that raw throughput alone does not determine OpenCL performance. The T400’s newer memory type and higher boost clock likely compensate for its lower shader count.

In Geekbench Vulkan, the T400 scores 15,976 against the GTX 660 Ti’s 15,012, a 964-point margin, or 6.4%. This smaller delta reflects the GTX 660 Ti’s stronger relative showing under a modern graphics API. The Kepler architecture’s higher TMU count (112 versus 24) and texture rate (109.8 GTexel/s versus 34.20 GTexel/s) may help narrow the gap in graphics-centric workloads. Still, the T400’s Vulkan 1.4 support versus the GTX 660 Ti’s Vulkan 1.2.175 support gives it a software stack advantage that translates into a win.

Looking at the rivals around each card adds context. The T400’s nearest rival is the NVIDIA GeForce RTX 5090 D V2, with an average score of 16,504, a 0% delta. The AMD Radeon PRO W7500 scores 16,415 (0.6% behind), and the NVIDIA RTX PRO 6000 Blackwell scores 16,408 (0.6% behind). This places the T400 in a performance tier with much newer and more expensive workstation cards. The GTX 660 Ti’s nearest rivals include the AMD Radeon Pro 560X at 15,082 (0.1% behind), the AMD Radeon RX 7600 at 15,171 (0.7% ahead), and the NVIDIA GeForce RTX 3050 OEM at 15,199 (0.9% ahead). The GTX 660 Ti sits at the lower end of this cluster, with the Radeon 680M at 15,270 leading by 1.4%.

The Verdict

The data supports a clear split in use cases. For modern compute workloads, especially those using OpenCL, the T400 is the stronger card by a 12.7% margin. Its 30 W power draw, single-slot design, and no power connectors make it suitable for low-power or compact systems. Its newer Turing architecture supports DirectX 12 (12_1) and Vulkan 1.4, which positions it better for current software. The T400’s 60th percentile ranking among all GPUs places it above the GTX 660 Ti’s 57th percentile.

For legacy compatibility or memory-bandwidth-sensitive tasks, the GTX 660 Ti has arguments in its favor. Its 144.2 GB/s bandwidth more than doubles the T400’s 80.00 GB/s, and its 2.634 TFLOPS FP32 throughput is more than double the T400’s 1,094.4 GFLOPS. In Vulkan, the gap narrows to 6.4%, suggesting the GTX 660 Ti can still handle modern graphics workloads reasonably well. Its dual-slot cooler and 150 W TDP require a more substantial power supply, rated at 450 W versus 200 W for the T400.

Users who prioritize compute efficiency, modern API support, and low power draw should choose the T400. Users who need to run older software or who have workloads that benefit from high memory bandwidth might still consider the GTX 660 Ti, but the benchmark data shows the T400 winning both recorded tests. The GTX 660 Ti’s only recorded advantage is in memory bandwidth and raw shader count, neither of which translates into a benchmark victory. The T400 wins 2 out of 2 head-to-head tests, making it the superior choice based on the available measurements.