NVIDIA GeForce GTX 660 vs NVIDIA GeForce GTX TITAN BLACK Comparison

# NVIDIA GeForce GTX TITAN BLACK vs NVIDIA GeForce GTX 660

The NVIDIA GeForce GTX TITAN BLACK and the NVIDIA GeForce GTX 660 represent two very different tiers within the Kepler architecture family. The TITAN BLACK, built on the GK110B chip, is a top-tier enthusiast part from the GeForce 700 generation, while the GTX 660, using the GK106 chip, is a mid-range offering from the GeForce 600 series. Benchmark data shows the TITAN BLACK wins all three head-to-head tests, with deltas ranging from 139.9% to 461.5%. However, the GTX 660 holds its own in the broader percentile rankings, sitting at the 45th percentile versus the TITAN BLACK's 46th, indicating that the older card's average score of 9022 is not far from the TITAN BLACK's 9474 average.

Where Each One Wins

The TITAN BLACK dominates every benchmark category where both cards have data. In Geekbench Metal, the TITAN BLACK scores 24172 against the GTX 660's 4305, a massive 461.5% advantage. This gap reflects the TITAN BLACK's far larger compute infrastructure: 2880 shading units versus 960, 240 TMUs versus 80, and 48 ROPs versus 24. The TITAN BLACK also leads in Geekbench OpenCL with 27218 versus 11347 (139.9% ahead) and in Geekbench Vulkan with 29561 versus 11415 (159% ahead). These are not marginal wins; they represent a completely different performance class.

The GTX 660 has no benchmark wins in the head-to-head comparison. However, its PassMark scores, while not directly compared in the head-to-head table, show it is not without merit in legacy workloads. The GTX 660's percentile ranking of 45 versus the TITAN BLACK's 46 suggests that in the broader GPU population, the two cards sit close together. This is notable because the GTX 660 achieves this with only 2 GB of memory, a 192-bit bus, and 1.981 TFLOPS of FP32 performance, compared to the TITAN BLACK's 6 GB, 384-bit bus, and 5.645 TFLOPS.

The TITAN BLACK's wins are most pronounced in compute-oriented tests like Metal and Vulkan, where its 7,080 million transistors and 561 mm² die size provide substantial parallel processing capability. The GTX 660, with 2,540 million transistors on a 221 mm² die, cannot match this throughput. For users running modern API workloads, the TITAN BLACK is the clear choice, while the GTX 660's lower power draw of 140 W versus 250 W makes it more suited to systems with modest power supplies.

Architecture Differences

Both cards share the Kepler architecture and are fabricated on TSMC's 28 nm process, but their chips diverge significantly. The TITAN BLACK uses the GK110B chip, which contains 7,080 million transistors on a 561 mm² die, yielding a transistor density of 12.6M / mm². The GTX 660 uses the GK106 chip, with 2,540 million transistors on a 221 mm² die, giving a density of 11.5M / mm². This density difference, while small, reflects the TITAN BLACK's more complex design with additional execution resources.

Clock speeds tell an interesting story. The GTX 660 actually runs higher base and boost clocks: 980 MHz base and 1032 MHz boost, versus the TITAN BLACK's 889 MHz base and 980 MHz boost. Despite lower clocks, the TITAN BLACK's massive resource count delivers far higher throughput. Memory clocks also differ: the TITAN BLACK runs at 1750 MHz (7 Gbps effective) while the GTX 660 runs at 1502 MHz (6 Gbps effective). The TITAN BLACK's memory bandwidth of 336.0 GB/s is more than double the GTX 660's 144.2 GB/s, enabled by the 384-bit bus versus 192-bit.

Memory capacity and configuration are starkly different. The TITAN BLACK offers 6 GB of GDDR5, while the GTX 660 offers 2 GB. This 4 GB difference matters for high-resolution textures and large compute datasets. The TITAN BLACK's pixel rate of 58.80 GPixel/s and texture rate of 235.2 GTexel/s dwarf the GTX 660's 20.64 GPixel/s and 82.56 GTexel/s. FP32 performance is also lopsided: 5.645 TFLOPS versus 1.981 TFLOPS, a 2.85x difference.

Power and physical specifications show the TITAN BLACK's enthusiast positioning. It has a 250 W TDP with a suggested 600 W PSU and requires both a 6-pin and 8-pin power connector. The GTX 660 has a 140 W TDP, a suggested 300 W PSU, and needs only a single 6-pin connector. The TITAN BLACK is also physically larger at 267 mm length, 111 mm height, and 38 mm width, versus the GTX 660's 241 mm length. Both are dual-slot cards with identical display outputs: 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. Both support PCIe 3.0 x16 and share the same OpenGL 4.6 and Vulkan 1.2.175 APIs, though the TITAN BLACK supports DirectX 12 (11_1) while the GTX 660 only supports DirectX 12 (11_0).

The Verdict

The data is unambiguous: the GTX TITAN BLACK outperforms the GTX 660 in every head-to-head benchmark, with deltas that range from 139.9% in OpenCL to 461.5% in Metal. The TITAN BLACK's 2880 shading units, 240 TMUs, and 48 ROPs, combined with 6 GB of GDDR5 on a 384-bit bus, give it overwhelming advantages in raw throughput and memory bandwidth. Its 5.645 TFLOPS FP32 performance is nearly three times the GTX 660's 1.981 TFLOPS.

However, the GTX 660's 45th percentile ranking versus the TITAN BLACK's 46th shows that in the broader GPU ecosystem, these cards are closer than their direct comparison suggests. The GTX 660's average benchmark score of 9022 is only 4.8% behind the TITAN BLACK's 9474. This is likely because the GTX 660's lower power draw of 140 W and smaller physical footprint make it viable in more systems, while the TITAN BLACK's 250 W TDP and 600 W PSU requirement limit its deployment.

For users who need maximum compute performance, especially in Metal or Vulkan workloads, the TITAN BLACK is the only choice. Its 6 GB memory capacity and 336.0 GB/s bandwidth handle large datasets that would exhaust the GTX 660's 2 GB allocation. For users with power or space constraints, the GTX 660's lower TDP and shorter length make it a more practical option, though its performance ceiling is significantly lower. The TITAN BLACK sits at the 46th percentile of all GPUs, while the GTX 660 sits at the 45th, indicating both are mid-pack performers in the modern landscape.

FAQ

Q: How much faster is the GTX TITAN BLACK than the GTX 660 in Geekbench Metal?

A: The TITAN BLACK scores 24172 versus the GTX 660's 4305, a 461.5% advantage. This is the largest delta in any head-to-head test.

Q: Does the GTX 660 have higher clock speeds than the TITAN BLACK?

A: Yes. The GTX 660 runs at 980 MHz base and 1032 MHz boost, while the TITAN BLACK runs at 889 MHz base and 980 MHz boost. Despite lower clocks, the TITAN BLACK's larger chip delivers far higher performance.

Q: What is the memory bandwidth difference between these two cards?

A: The TITAN BLACK has 336.0 GB/s of bandwidth from its 384-bit bus, while the GTX 660 has 144.2 GB/s from its 192-bit bus. This is a 2.33x difference.

Q: Which card requires more power connectors?

A: The TITAN BLACK requires one 6-pin and one 8-pin connector, while the GTX 660 requires only a single 6-pin connector. The TITAN BLACK's TDP is 250 W versus 140 W for the GTX 660.

Q: Are both cards from the same architecture generation?

A: Both use the Kepler architecture and are built on TSMC's 28 nm process, but the TITAN BLACK is from the GeForce 700 generation while the GTX 660 is from the GeForce 600 generation. The TITAN BLACK uses the GK110B chip; the GTX 660 uses GK106.

Q: How do these cards compare in DirectX support?

A: The TITAN BLACK supports DirectX 12 (11_1), while the GTX 660 supports DirectX 12 (11_0). Both support OpenGL 4.6 and Vulkan 1.2.175.

Head-to-Head Benchmarks

The Geekbench Metal test shows the most extreme difference between the two cards. The TITAN BLACK scores 24172, while the GTX 660 scores 4305. This 461.5% delta is a direct result of the TITAN BLACK's 2880 shading units versus 960, and its 48 ROPs versus 24. Metal workloads that leverage compute shaders scale nearly linearly with shading unit count, so this result aligns with the hardware specifications.

In Geekbench OpenCL, the TITAN BLACK scores 27218 against the GTX 660's 11347, a 139.9% advantage. This smaller delta compared to Metal suggests that OpenCL workloads are less sensitive to raw shading unit count and more dependent on memory bandwidth and clock speed. The TITAN BLACK's 336.0 GB/s bandwidth versus 144.2 GB/s helps, but the GTX 660's higher boost clock of 1032 MHz versus 980 MHz narrows the gap somewhat.

Geekbench Vulkan shows the TITAN BLACK at 29561 and the GTX 660 at 11415, a 159% difference. Vulkan's explicit API benefits from the TITAN BLACK's larger texture rate of 235.2 GTexel/s versus 82.56 GTexel/s, as texture-bound operations are common in Vulkan workloads. The TITAN BLACK's 240 TMUs versus 80 provide a 3x texture throughput advantage, though the final score delta is smaller due to other bottlenecks.

The TITAN BLACK's average benchmark score of 9474 places it just 0.2% ahead of the GTX 760 (9458) and 1.7% ahead of the AMD Radeon R7 M380 (9313). It trails the Quadro K5000 (9637) by 1.7% and the GTX 960M (9645) by 1.8%. The GTX 660's average score of 9022 is 0.2% behind the NVIDIA TITAN V CEO Edition (9037) and 0.4% behind the GTX 560 (9058), while leading the AMD Radeon 550X (8918) by 1.2% and the AMD Radeon Pro WX 5100 (8863) by 1.8%.

These nearest-rival comparisons show that while the TITAN BLACK is far ahead of the GTX 660, both cards sit in a crowded mid-field of GPUs. The TITAN BLACK's 46th percentile and the GTX 660's 45th percentile mean neither card is exceptional by modern standards, but the TITAN BLACK's head-to-head dominance is indisputable. The data supports a clear hierarchy: the TITAN BLACK is a high-end part that has aged into mid-pack status, while the GTX 660 is a mid-range part that remains competitive in its niche.