NVIDIA GeForce GTX 680 vs NVIDIA GeForce GTX TITAN Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX TITAN posts an average benchmark score of 14373, while the NVIDIA GeForce GTX 680 scores 14150. The TITAN leads by roughly 1.6% in aggregate performance.

Q: How do the two cards compare in OpenCL compute workloads?

A: The GTX TITAN delivers a Geekbench OpenCL score of 24873 versus 16906 for the GTX 680. This represents a 47.1% advantage for the TITAN, its largest single-test win.

Q: Does the GTX 680 win any benchmark category?

A: Yes, in Geekbench Vulkan the GTX 680 scores 17646 against the TITAN's 10027. That is a 43.2% margin in favor of the GTX 680, making it the clear winner in that specific API test.

Q: What is the transistor count difference between the two chips?

A: The GTX TITAN uses the GK110 chip with 7,080 million transistors on a 561 mm² die. The GTX 680 uses the GK104 chip with 3,540 million transistors on a 294 mm² die. The TITAN has exactly double the transistor count.

Q: Which card has more memory bandwidth?

A: The GTX TITAN provides 288.4 GB/s of bandwidth through a 384-bit bus with 6 GB of GDDR5 memory. The GTX 680 offers 192.3 GB/s over a 256-bit bus with 2 GB of GDDR5. The TITAN leads by roughly 50% in bandwidth.

Q: How do the cards rank within the overall GPU landscape?

A: The GTX TITAN sits at the 56th percentile of all GPUs, while the GTX 680 sits at the 55th percentile. Both cards are within one percentile point of each other globally, despite their architectural differences.

Architecture Differences

The GTX TITAN and GTX 680 both use NVIDIA's Kepler architecture, but they are built on fundamentally different chips. The TITAN employs the GK110 chip, designed as a high-end compute-oriented processor, while the GTX 680 uses the GK104 chip, originally intended for gaming-class performance. Both are manufactured on TSMC's 28 nm process, but the die sizes diverge sharply: GK110 measures 561 mm² versus 294 mm² for GK104.

Transistor density is nearly identical — 12.6 million transistors per square millimeter for the TITAN and 12.0 million for the GTX 680 — but the absolute transistor counts tell the story. The TITAN packs 7,080 million transistors, twice the 3,540 million found on the GTX 680. This enables a much larger execution resource pool.

The TITAN features 2,688 shading units, 224 texture mapping units, and 48 ROPs. The GTX 680 counters with 1,536 shading units, 128 TMUs, and 32 ROPs. In every execution unit category, the TITAN holds a commanding lead — 75% more shaders, 75% more TMUs, and 50% more ROPs.

Memory subsystems are equally divergent. The TITAN carries 6 GB of GDDR5 across a 384-bit bus, yielding 288.4 GB/s of bandwidth. The GTX 680 has 2 GB of GDDR5 on a 256-bit bus, producing 192.3 GB/s. Clock speeds partially compensate for the smaller GTX 680 architecture: it runs at 1006 MHz base and 1058 MHz boost, while the TITAN runs at 836 MHz base and 876 MHz boost. Both cards use 6 Gbps effective GDDR5 memory.

The TITAN's FP32 throughput is 4.709 TFLOPS, compared to 3.250 TFLOPS for the GTX 680. Pixel and texture rates follow the same pattern: the TITAN achieves 49.06 GPixel/s and 196.2 GTexel/s, versus 33.86 GPixel/s and 135.4 GTexel/s for the GTX 680. Neither card has dedicated ray tracing or tensor cores, as both predate those features.

API support is identical: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Both use PCIe 3.0 x16 interfaces and offer the same display outputs — 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The TITAN draws 250 W TDP with a 600 W suggested PSU and requires 1x 6-pin plus 1x 8-pin power connectors. The GTX 680 draws 195 W TDP with a 450 W suggested PSU and uses 2x 6-pin connectors. Both are dual-slot cards.

Head-to-Head Benchmarks

The head-to-head results reveal a split personality between these two Kepler cards. The TITAN wins two of the three benchmark tests, but the magnitudes of the wins and losses are highly asymmetric.

In Geekbench OpenCL, the TITAN delivers a decisive victory. Its score of 24873 eclipses the GTX 680's 16906 by 47.1%. This is the largest delta between the two cards in any test, and it reflects the TITAN's doubled transistor count and significantly higher FP32 throughput. OpenCL workloads stress raw compute resources, where the TITAN's 4.709 TFLOPS of FP32 performance and 196.2 GTexel/s texture rate provide a substantial advantage over the GTX 680's 3.250 TFLOPS and 135.4 GTexel/s.

The Geekbench Metal test also goes to the TITAN, though by a much narrower margin. The TITAN scores 8218 against the GTX 680's 7897, a 4.1% difference. Metal performance appears to be less sensitive to raw core counts, since the GTX 680's higher clock speeds (1006 MHz base versus 836 MHz) partially offset its smaller execution resource pool.

The GTX 680 strikes back decisively in Geekbench Vulkan. Its score of 17646 dwarfs the TITAN's 10027, representing a 43.2% margin in favor of the older card. This result is surprising given the TITAN's hardware advantages, but the data is unambiguous. The Vulkan test likely favors the GTX 680's different driver optimization path or scheduling behavior, though the FACT PACK does not provide a specific explanation.

Averaging the three tests, the TITAN's aggregate score of 14373 versus 14150 for the GTX 680 confirms a narrow overall lead — about 1.6%. The TITAN's two wins (47.1% and 4.1%) weigh against one large loss (43.2%), yielding a modest net positive. The winsA counter shows 2 for the TITAN and 1 for the GTX 680.

Specification Differences

The two cards differ across nearly every hardware specification. The TITAN uses the GK110 chip, while the GTX 680 uses GK104. The TITAN belongs to the GeForce 700 generation with a predecessor of GeForce 600 and successor of GeForce 900; the GTX 680 belongs to the GeForce 600 generation with a predecessor of GeForce 500 and successor of GeForce 700. Release dates differ by roughly 11 months: the TITAN launched on 2013-02-18, the GTX 680 on 2012-03-21.

Transistor count is 7,080 million for the TITAN versus 3,540 million for the GTX 680. Die size is 561 mm² versus 294 mm². Transistor density is 12.6M per mm² versus 12.0M per mm². Base clock is 836 MHz for the TITAN and 1006 MHz for the GTX 680. Boost clock is 876 MHz versus 1058 MHz. Memory speed is identical at 1502 MHz (6 Gbps effective).

Memory capacity is 6 GB for the TITAN and 2 GB for the GTX 680. Bus width is 384-bit versus 256-bit. Bandwidth is 288.4 GB/s versus 192.3 GB/s. Shading units number 2688 versus 1536. TMUs number 224 versus 128. ROPs number 48 versus 32. Pixel rate is 49.06 GPixel/s versus 33.86 GPixel/s. Texture rate is 196.2 GTexel/s versus 135.4 GTexel/s. FP32 performance is 4.709 TFLOPS versus 3.250 TFLOPS.

TDP is 250 W for the TITAN and 195 W for the GTX 680. Power connectors are 1x 6-pin plus 1x 8-pin for the TITAN versus 2x 6-pin for the GTX 680. Suggested PSU is 600 W versus 450 W. Card length is 267 mm (10.5 inches) for the TITAN versus 256 mm (10.1 inches) for the GTX 680. Height and width are identical at 111 mm (4.4 inches) and 38 mm (1.5 inches). The launch MSRP for the TITAN is 999 USD; the launch MSRP for the GTX 680 is 499 USD.

The Verdict

The data presents a clear but nuanced picture. The GTX TITAN is the stronger card overall, winning two of three benchmarks and posting a higher average score (14373 versus 14150). Its 47.1% OpenCL advantage is the single largest performance gap in either direction, and its 4.1% Metal lead confirms broad compute superiority. For users prioritizing compute workloads — OpenCL-heavy applications, rendering, or scientific tasks — the TITAN is the unambiguous choice from this data.

However, the GTX 680's 43.2% Vulkan victory cannot be ignored. For users whose primary workload runs through the Vulkan API, the GTX 680 delivers dramatically better performance despite having half the transistors, fewer shaders, less memory, and lower bandwidth. The GTX 680 also runs at higher clocks (1006 MHz base versus 836 MHz), draws 55 W less power, and requires a less demanding PSU (450 W versus 600 W).

The percentile rankings are nearly identical — 56th for the TITAN, 55th for the GTX 680 — suggesting that in a broad mix of workloads, neither card dominates the other by a meaningful margin. The TITAN's average score advantage of 223 points is small relative to the variance between individual tests.

Where Each One Wins

The GTX TITAN wins in OpenCL and Metal workloads. Its doubled transistor count, 75% more shading units, 75% more texture units, and 50% more ROPs provide the raw resources needed for compute-heavy tasks. The 6 GB memory capacity and 288.4 GB/s bandwidth also make it better suited for large datasets or high-resolution textures. Users running OpenCL-based compute, GPU-accelerated rendering, or Metal-compatible applications should expect the TITAN to outperform the GTX 680 by margins ranging from 4.1% to 47.1%.

The GTX 680 wins decisively in Vulkan workloads. Its 17646 Vulkan score versus the TITAN's 10027 represents a 43.2% advantage — a massive reversal of the hardware hierarchy. This makes the GTX 680 the better choice for Vulkan-focused gaming or applications that leverage this API heavily. The GTX 680 also consumes less power (195 W versus 250 W), requires a smaller PSU (450 W versus 600 W), and uses simpler power connectors (2x 6-pin versus 1x 6-pin + 1x 8-pin), which may matter for system compatibility.

For a balanced workload mix, the TITAN edges ahead based on its average benchmark score and two-of-three test wins. But the margin is thin — roughly 1.6% in aggregate — and the specific application determines the winner. The GTX 680 is a targeted pick for Vulkan users; the TITAN is the broader compute performer.