NVIDIA GeForce GTX 670 vs NVIDIA GeForce GTX TITAN Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX TITAN leads with an average benchmark score of 14,373, compared to the GTX 670's 12,773. That puts the TITAN at the 56th percentile of all GPUs in the database, while the GTX 670 sits at the 52nd percentile.

Q: How large is the performance gap in compute workloads?

A: In the Geekbench OpenCL test, the TITAN scores 24,873 versus 15,341 for the GTX 670, a 62.1% advantage. This is the largest single-test difference between the two cards.

Q: Does the GTX 670 win any benchmark?

A: Yes. In the Geekbench Vulkan test, the GTX 670 scores 15,553 against the TITAN's 10,027, a 35.5% lead for the GTX 670. The TITAN wins the other two recorded tests.

Q: What are the memory specifications of each card?

A: The TITAN has 6 GB of GDDR5 on a 384-bit bus with 288.4 GB/s of bandwidth. The GTX 670 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s of bandwidth. Both run their memory at 1502 MHz, or 6 Gbps effective.

Q: How do their thermal and power requirements compare?

A: The TITAN has a 250 W TDP and requires a 600 W suggested PSU, with 1x 6-pin plus 1x 8-pin power connectors. The GTX 670 has a 170 W TDP, a 450 W suggested PSU, and uses 2x 6-pin connectors. Both are dual-slot cards.

Q: Which card is physically larger?

A: The TITAN measures 267 mm in length, 111 mm in height, and 38 mm in width. The GTX 670 is 241 mm long with the same 111 mm height and 38 mm width. Both use the same display outputs: 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2.

Architecture Differences

Both cards are built on NVIDIA's Kepler architecture and manufactured by TSMC on a 28 nm process. The similarities end there. The TITAN uses the GK110 chip, a massive 561 mm² die housing 7,080 million transistors, giving a transistor density of 12.6M per mm². The GTX 670 uses the smaller GK104 chip, a 294 mm² die with 3,540 million transistors and a density of 12.0M per mm². In other words, the TITAN packs roughly double the silicon area and double the transistor count.

The compute resources diverge sharply. The TITAN carries 2,688 shading units, 224 texture mapping units, and 48 ROPs. The GTX 670 has 1,344 shading units, 112 TMUs, and 32 ROPs. That is exactly half the shaders and TMUs, and two-thirds the ROPs. The TITAN's pixel rate is 49.06 GPixel/s and its texture rate is 196.2 GTexel/s, while the GTX 670 manages 27.44 GPixel/s and 109.8 GTexel/s. The FP32 compute throughput tells a similar story: 4.709 TFLOPS for the TITAN versus 2.634 TFLOPS for the GTX 670.

Clock behavior is the one area where the GTX 670 counters the TITAN's hardware advantage. The GTX 670 runs a base clock of 915 MHz and a boost of 980 MHz, while the TITAN sits at 836 MHz base and 876 MHz boost. The GTX 670 also belongs to an earlier generation, the GeForce 600 series, whereas the TITAN is part of the GeForce 700 series. Both cards are end-of-life products, and both expose the same API set: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Neither card has ray tracing cores or tensor cores.

Head-to-Head Benchmarks

The database records three head-to-head tests between these two cards, and the results are not uniform. The TITAN dominates compute-style workloads, while the GTX 670 takes an unexpected win in the Vulkan test.

Starting with Geekbench OpenCL, the TITAN posts 24,873 against the GTX 670's 15,341. That is a 62.1% lead, the largest margin in any test. The reason is straightforward: the TITAN has twice the shading units, twice the TMUs, and nearly double the FP32 throughput. The OpenCL test is highly parallel and scales with raw compute resources, which is exactly where the TITAN's GK110 chip excels. This result is consistent with the TITAN's nearest rival positioning, where its average score of 14,373 sits within 0.4% of cards like the AMD Radeon RX Vega 11, the NVIDIA GeForce GTX 965M, and the Intel Iris Xe MAX Graphics.

In Geekbench Metal, the TITAN wins again, but by a much smaller margin. The TITAN scores 8,218 versus 7,424 for the GTX 670, a 10.7% advantage. Metal is a lower-level graphics API, and here the TITAN's extra ROPs and memory bandwidth (288.4 GB/s versus 192.3 GB/s) help, but the gap is far less dramatic than in OpenCL. The GTX 670's higher clocks (980 MHz boost versus 876 MHz boost) likely narrow the difference.

The Geekbench Vulkan test flips the script. The GTX 670 scores 15,553, while the TITAN manages only 10,027. That gives the GTX 670 a 35.5% lead. This is a surprising outcome given the TITAN's hardware superiority on paper. The recorded data shows the TITAN's nearest rivals in this score range are the AMD Radeon RX Vega 11 and the Intel Iris Xe MAX Graphics, both around 14,300 to 14,400 average. The GTX 670, by contrast, sits near the AMD Radeon RX 7600M XT and the NVIDIA GeForce GTX 590 in its own average score range. The Vulkan result suggests that driver maturity or workload characteristics in the Vulkan path did not favor the GK110 chip at the time of testing.

Counting wins, the TITAN takes 2 out of 3 benchmarks. The GTX 670 wins 1. But the magnitude of the GTX 670's Vulkan win is substantial, and anyone targeting Vulkan-heavy workloads should pay attention to that result.

Specification Differences

The following fields differ between the two cards:

• Chip: GK110 (TITAN) versus GK104 (GTX 670)
• Generation: GeForce 700 (TITAN) versus GeForce 600 (GTX 670)
• Transistors: 7,080 million versus 3,540 million
• Die Size: 561 mm² versus 294 mm²
• Transistor Density: 12.6M / mm² versus 12.0M / mm²
• Base Clock: 836 MHz versus 915 MHz
• Boost Clock: 876 MHz versus 980 MHz
• Memory Size: 6 GB versus 2 GB
• Memory Bus Width: 384 bit versus 256 bit
• Memory Bandwidth: 288.4 GB/s versus 192.3 GB/s
• Shading Units: 2,688 versus 1,344
• TMUs: 224 versus 112
• ROPs: 48 versus 32
• Pixel Rate: 49.06 GPixel/s versus 27.44 GPixel/s
• Texture Rate: 196.2 GTexel/s versus 109.8 GTexel/s
• FP32: 4.709 TFLOPS versus 2.634 TFLOPS
• TDP: 250 W versus 170 W
• Power Connectors: 1x 6-pin + 1x 8-pin versus 2x 6-pin
• Suggested PSU: 600 W versus 450 W
• Length: 267 mm versus 241 mm
• Release Date: 2013-02-18 versus 2012-05-09
• Predecessor: GeForce 600 versus GeForce 500
• Successor: GeForce 900 versus GeForce 700

Fields that are identical include the 28 nm process node, TSMC as foundry, GDDR5 memory type, 1502 MHz memory clock (6 Gbps effective), PCIe 3.0 x16 interface, dual-slot width, 111 mm height, 38 mm width, display outputs, and the full API list (DirectX 12 (11_0), OpenGL 4.6, Vulkan 1.2.175). Neither card has ray tracing or tensor cores.

Where Each One Wins

The TITAN is the clear pick for compute-heavy workloads. Its 62.1% OpenCL advantage is decisive, and its 10.7% Metal lead shows it also holds up in graphics API tests. The 6 GB memory buffer and 384-bit bus give it far more headroom for large textures, high-resolution frame buffers, or data sets that exceed 2 GB. The 250 W TDP and 600 W suggested PSU are the cost of that capability, but the database shows the TITAN at the 56th percentile of all GPUs, above the GTX 670's 52nd percentile.

The GTX 670 wins in Vulkan, and it wins big. A 35.5% margin is not a rounding error. The GTX 670 also runs at higher clocks (915 MHz base, 980 MHz boost), which likely contributes to its Vulkan showing. Its 170 W TDP and 450 W suggested PSU make it easier to slot into an existing system without upgrading the power supply. It is also 26 mm shorter, which helps in smaller cases. For users whose primary API is Vulkan, the GTX 670's score of 15,553 versus the TITAN's 10,027 is the single most relevant data point in this comparison.

The TITAN's nearest rivals in the database, the AMD Radeon RX Vega 11 and the NVIDIA GeForce GTX 965M, sit within 0.4% of its average score. The GTX 670's nearest rivals, the NVIDIA GeForce GTX 590 and the AMD Radeon Pro 455, sit within 0.6% of its average. Both cards are essentially bracketed by integrated or mobile parts in the overall rankings, which reflects their age and end-of-life status.

The Verdict

The data paints a split picture. The NVIDIA GeForce GTX TITAN wins the overall average score (14,373 versus 12,773) and takes 2 of 3 head-to-head benchmarks. Its OpenCL result is overwhelming, and its Metal result is a solid but narrower win. If the workload is compute-oriented, or if the application can use more than 2 GB of memory, the TITAN is the obvious choice.

The NVIDIA GeForce GTX 670 wins the Vulkan test by a 35.5% margin. That is not a small gap, and it is the kind of result that matters if the user's software stack is built around Vulkan. The GTX 670 also draws less power (170 W versus 250 W), needs a smaller PSU (450 W versus 600 W), and takes up less space in a case.

For the general-purpose user, the TITAN's higher average score and its dominance in OpenCL make it the stronger card overall. The GTX 670's Vulkan win is important, but it is one test against two losses. The TITAN is also the newer part, released in February 2013 versus May 2012, and it belongs to the next generation (GeForce 700 versus GeForce 600). The launch MSRP of the TITAN was 999 USD, while the GTX 670 launched at 399 USD, but the benchmark data does not show any performance-per-dollar metric, so any cost comparison would be speculative.

Choose the TITAN for compute performance, memory capacity, and overall benchmark strength. Choose the GTX 670 for Vulkan-specific workloads, lower power draw, and easier physical fit. Do not choose either card expecting modern ray tracing or tensor core features, because neither has them.