GPU Comparison

The NVIDIA RTX 2000 Ada Generation and the NVIDIA GeForce GTX 690 represent two very different eras of NVIDIA silicon, yet the recorded database numbers place them closer in overall standing than their age gap might suggest. Both land in the lower-middle of the all-GPU percentile table, with the RTX 2000 Ada at the 63rd percentile and the GTX 690 at the 60th. The head-to-head benchmark data, however, tells a far more lopsided story in the Ada card's favor.

Head-to-Head Benchmarks

Only two tests exist where both cards have recorded scores, and the RTX 2000 Ada Generation wins both decisively.

In Geekbench OpenCL, the RTX 2000 Ada posts 78074 against the GTX 690's 17399, a margin of 348.7 percent. Geekbench Vulkan is even more dramatic: 83360 versus 16675, a 399.9 percent advantage, meaning the Ada board delivers roughly five times the compute throughput of the Kepler flagship on that API. These are not marginal wins; they are generational gaps expressed as multiples.

Context from each card's rival set reinforces how the averages converge despite that gap. The RTX 2000 Ada's average benchmark score of 18954 sits within a fraction of a percent of the NVIDIA Quadro K6000 (19030, delta of -0.4 percent), the AMD Radeon RX 6600 (19036, -0.4 percent), the NVIDIA Tesla K80 (18866, +0.5 percent), and the NVIDIA GeForce RTX 4050 Mobile (19049, -0.5 percent). The GTX 690's average of 17037 is similarly clustered around the AMD Radeon HD 7970M (17019, +0.1 percent), the AMD Radeon RX 7600 XT (17083, -0.3 percent), the NVIDIA Tesla M4 (16932, +0.6 percent), and the NVIDIA GeForce RTX 3070 (17208, -1 percent). The apparent parity in the averages comes from the GTX 690's small recorded test coverage rather than competitive performance: with only two benchmark entries against the Ada card's ten, its average is drawn from a narrow slice of results.

The RTX 2000 Ada's broader record includes strong individual results: 1767 in 3DMark Steel Nomad (DX12), 16927 in PassMark G3D, 1072 in PassMark G2D, and 7834 in PassMark GPU Compute. Its PassMark DirectX suite spans scores of 216 for DirectX 9, 138 for DirectX 11, 82 for DirectX 10, and 71 for DirectX 12. The GTX 690 has no entries in any of these tests, so the only direct comparisons available are the two Geekbench runs.

Where Each One Wins

Strictly from the recorded data, the RTX 2000 Ada wins everything that has been measured. It takes both head-to-head benchmarks with margins near or above 350 percent, and the GTX 690 records zero wins across the shared test set.

For workload selection, the data points clearly. Anyone running GPU compute, whether through OpenCL or Vulkan, gets several multiples of the GTX 690's throughput from the Ada board. The RTX 2000 Ada's additional recorded results in 3DMark Steel Nomad and the full PassMark suite cover gaming and graphics scenarios where the GTX 690 simply has no comparable recorded data, so the database cannot credit it with a win anywhere.

The GTX 690's only quantitative claim to relevance is its percentile placement, 60th against the Ada card's 63rd, a three-point gap across the full GPU population. That is a narrow aggregate gap, but it is built on thin data and does not translate into any recorded head-to-head victory.

Architecture Differences

The architectural distance between these two boards is substantial. The RTX 2000 Ada Generation uses the AD107 chip on the Ada Lovelace architecture, fabricated at TSMC on a 5 nm process. The GTX 690 uses the GK104 chip on Kepler, built at TSMC on a 28 nm node. The transistor counts make the generational leap explicit: 18,900 million transistors on the Ada card versus 3,540 million on the GTX 690. Transistor density differs even more sharply, 118.9M per mm² against 12.0M per mm², roughly a tenfold increase, even though the Ada die is physically smaller at 159 mm² versus 294 mm².

Core resources follow the same pattern. The RTX 2000 Ada carries 2816 shading units, 88 TMUs, 48 ROPs, 22 RT cores, and 88 tensor cores. The GTX 690 has 1536 shading units, 128 TMUs, and 32 ROPs, with no RT cores or tensor cores at all, since those functional blocks did not exist in Kepler. The GTX 690's higher TMU count is a reflection of its dual-chip design philosophy, but its texture rate of 130.4 GTexel/s still trails the Ada card's 187.4 GTexel/s, and pixel fill rate favors the Ada board heavily at 102.2 GPixel/s against 32.61 GPixel/s.

Raw compute diverges massively. The RTX 2000 Ada delivers 12.00 TFLOPS of FP32 and 12.00 TFLOPS of FP16 at a 1:1 rate; the GTX 690 manages 3.130 TFLOPS of FP32 and records no FP16 figure. Clocks are correspondingly higher on the newer card, with a 1620 MHz base and 2130 MHz boost against 915 MHz base and 1019 MHz boost.

Memory is another sharp dividing line. The RTX 2000 Ada has 16 GB of GDDR6 on a 128-bit bus running at 2000 MHz (16 Gbps effective), yielding 256.0 GB/s of bandwidth. The GTX 690 has 2 GB of GDDR5 on a 256-bit bus at 1502 MHz (6 Gbps effective), for 192.3 GB/s. The wider bus of the GTX 690 cannot compensate for its slower memory technology, and the 2 GB capacity is severely limiting by modern standards.

Power and platform characteristics also differ. The RTX 2000 Ada has a 70 W TDP, no auxiliary power connectors, and a suggested PSU of 250 W; the GTX 690 draws 300 W, requires 2x 8-pin connectors, and calls for a 700 W PSU. The Ada card is also more compact, at 168 mm long and 69 mm tall versus 279 mm and 111 mm. Both are dual-slot. Connectivity favors the newer card: PCIe 4.0 x8 versus PCIe 3.0 x16, and 4x mini-DisplayPort 1.4a versus 3x DVI plus 1x mini-DisplayPort 1.2.

Software support separates them too. The RTX 2000 Ada reports DirectX 12 Ultimate (12_2) and Vulkan 1.4; the GTX 690 reports DirectX 12 (11_0) feature level and Vulkan 1.2.175. Both list OpenGL 4.6. Production status reflects their lifecycle positions: the RTX 2000 Ada is an active product released on 2024-02-11, sitting between its predecessor (Workstation Ampere) and successor (Blackwell PRO W), while the GTX 690 is end-of-life, released 2012-05-02 between the GeForce 500 and GeForce 700 series.

FAQ

Q: How much faster is the RTX 2000 Ada than the GTX 690 in the shared benchmarks?

A: It wins Geekbench OpenCL by 348.7 percent (78074 vs 17399) and Geekbench Vulkan by 399.9 percent (83360 vs 16675). That is roughly four to five times the performance in both tests.

Q: Do their overall database averages suggest they are close in performance?

A: Superficially yes, with average scores of 18954 and 17037 and percentiles of 63 and 60. But the GTX 690's average is built from only two recorded results, while the Ada card has ten, so the averages are not directly comparable indicators of real-world performance.

Q: Which card is more power efficient based on the recorded data?

A: The RTX 2000 Ada, by a wide margin. Its TDP is 70 W with no power connectors and a 250 W suggested PSU, versus the GTX 690's 300 W TDP, 2x 8-pin connectors, and 700 W suggested PSU, all while delivering several times the benchmark performance.

Q: Does the GTX 690 support ray tracing or tensor operations?

A: No. The database lists no RT cores or tensor cores for the GK104 chip. The RTX 2000 Ada has 22 RT cores and 88 tensor cores.

Q: How do their memory configurations compare?

A: The RTX 2000 Ada has 16 GB of GDDR6 with 256.0 GB/s of bandwidth on a 128-bit bus; the GTX 690 has 2 GB of GDDR5 with 192.3 GB/s on a 256-bit bus.

Q: Which GPUs does the database list as nearest rivals for each card?

A: The RTX 2000 Ada's closest rivals are the Quadro K6000, Radeon RX 6600, Tesla K80, and RTX 4050 Mobile, all within roughly half a percent of its average. The GTX 690's are the Radeon HD 7970M, Radeon RX 7600 XT, Tesla M4, and RTX 3070, all within about one percent.

The Verdict

The recorded data leaves no ambiguity. The RTX 2000 Ada Generation wins every shared benchmark by margins approaching 400 percent, offers modern API support at DirectX 12 Ultimate and Vulkan 1.4, includes dedicated ray tracing and tensor hardware the GTX 690 lacks entirely, and does so at a 70 W TDP that needs no auxiliary power. Its 16 GB of memory, active production status, and dense 5 nm implementation make it the only defensible choice of the two for any current workload.

The GTX 690's case rests on history rather than measurements. Its dual-chip Kepler design, 28 nm node, and 2 GB framebuffer place it firmly in the end-of-life category, and it records zero head-to-head wins in the database. Unless the requirement involves specifically legacy hardware, the data points exclusively to the RTX 2000 Ada Generation.