NVIDIA GRID K2 vs NVIDIA T600 Comparison

Head-to-Head Benchmarks

The recorded data contains one direct head-to-head comparison between the NVIDIA GRID K2 and the NVIDIA T600: the Geekbench OpenCL test. The result is decisive. The T600 scores 27,875 points, while the GRID K2 scores 10,602 points. This translates to a 62% deficit for the GRID K2 relative to the T600, a substantial gap that places the two cards in different performance tiers for compute workloads.

The GRID K2 does not win any of the recorded head-to-head comparisons. The T600 takes the single available win. While the GRID K2 relies on older Kepler architecture with a higher raw transistor count, the T600's Turing architecture delivers far superior OpenCL throughput per unit of hardware. The delta of 62% is not a marginal difference; it indicates a generational leap in compute efficiency.

Beyond the shared test, the broader benchmark database paints a fuller picture. The GRID K2 has an average benchmark score of 8,080 across its recorded tests, placing it at the 42nd percentile of all GPUs. Its nearest rivals include the GeForce GTX 650 Ti Boost (average score 8,067, a 0.2% delta), the GeForce 945M (8,099, a -0.2% delta), and the GeForce GTX 650 Ti (8,053, a 0.3% delta). The GRID K2 sits essentially neck-and-neck with these mid-range parts from its own era.

The T600, by contrast, has an average benchmark score of 7,035, placing it at the 39th percentile. Its nearest rivals include the GeForce GTX 680M (7,023, a 0.2% delta), the AMD Radeon R5 M240 (6,975, a 0.9% delta), and the GeForce GTX 675M (6,946, a 1.3% delta). Notably, the GeForce GTX 970 scores 7,157, which is 1.7% higher than the T600, showing the T600 slots just below that high-end part.

The apparent contradiction is worth examining: the GRID K2 has a higher average benchmark score (8,080 vs 7,035), yet loses the OpenCL head-to-head by 62%. This is because the average includes different test suites. The GRID K2 has only two recorded benchmarks: Geekbench Metal (5,557) and Geekbench OpenCL (10,602). The T600 has nine recorded benchmarks, including multiple Passmark tests (DirectX 9 through 12, G2D, G3D, and GPU compute) which drag its average down. The OpenCL result is the only apples-to-apples comparison, and it shows a clear T600 advantage.

Architecture Differences

The two cards come from different architectural eras. The GRID K2 uses the GK104 chip built on the Kepler architecture, fabricated on a 28 nm process at TSMC. The T600 uses the TU117 chip built on the Turing architecture, also at TSMC but on a 12 nm process. The node shrink is significant: 28 nm to 12 nm allows for a much denser design.

Transistor counts tell an interesting story. The GRID K2 packs 3,540 million transistors on a 294 mm² die, yielding a transistor density of 12.0 million per mm². The T600 packs 4,700 million transistors on a smaller 200 mm² die, yielding a density of 23.5 million per mm². The T600 has roughly one-third more transistors in a smaller physical area, a direct consequence of the newer process node.

The GRID K2 is a dual-slot card with no display outputs, designed for server-side virtualization workloads. It requires a 1x 6-pin plus 1x 8-pin power connector and has a 225 W TDP. The T600 is a single-slot card with 4x mini-DisplayPort 1.4a outputs, requiring no external power connectors and drawing only 40 W. The power efficiency gap is enormous: the T600 delivers its performance at less than one-fifth the power draw.

Memory configurations share the same capacity and bandwidth but differ in type. Both cards have 4 GB of memory and 160.0 GB/s of bandwidth. The GRID K2 uses GDDR5 on a 256-bit bus with 1250 MHz memory clock (5 Gbps effective). The T600 uses GDDR6 on a 128-bit bus with the same 1250 MHz memory clock but 10 Gbps effective, doubling the per-pin data rate to compensate for the narrower bus.

Compute resources differ substantially. The GRID K2 has 1,536 shading units, 128 TMUs, and 32 ROPs. The T600 has 640 shading units, 40 TMUs, and 32 ROPs. Despite having fewer than half the shading units, the T600 achieves higher pixel rate (42.72 GPixel/s vs 23.84 GPixel/s) thanks to its higher clocks. The texture rate favors the GRID K2 at 95.36 GTexel/s versus 53.40 GTexel/s.

FP32 throughput shows the GRID K2 ahead at 2.289 TFLOPS versus 1.709 TFLOPS for the T600. However, the T600 adds FP16 capability at 3.418 TFLOPS (2:1 ratio), which the GRID K2 does not support at all. The T600 also has a boost clock of 1335 MHz and base clock of 735 MHz, while the GRID K2 has no recorded base or boost clocks in the database.

API support differs. The GRID K2 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The T600 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The T600's superior feature levels matter for modern applications.

FAQ

Q: Which card has better OpenCL performance?

A: The NVIDIA T600 wins decisively with a Geekbench OpenCL score of 27,875 versus 10,602 for the GRID K2, a 62% advantage for the T600.

Q: Do both cards have the same memory bandwidth?

A: Yes, both cards offer 160.0 GB/s of bandwidth and 4 GB of memory. However, the GRID K2 uses GDDR5 on a 256-bit bus, while the T600 uses GDDR6 on a 128-bit bus with a higher effective memory clock of 10 Gbps versus 5 Gbps.

Q: Which card consumes less power?

A: The T600 has a 40 W TDP and requires no external power connectors. The GRID K2 has a 225 W TDP and requires a 1x 6-pin plus 1x 8-pin power connector. The T600 is far more power-efficient.

Q: Can either card output video to displays?

A: The T600 has 4x mini-DisplayPort 1.4a outputs. The GRID K2 has no display outputs, as it is designed for server-side virtualization without direct display connectivity.

Q: What is the transistor density difference?

A: The T600 has a transistor density of 23.5 million per mm² on a 12 nm process, while the GRID K2 has 12.0 million per mm² on a 28 nm process. The T600 packs 4,700 million transistors on a 200 mm² die, versus 3,540 million on a 294 mm² die for the GRID K2.

Q: Which card supports newer API features?

A: The T600 supports DirectX 12 (12_1) and Vulkan 1.4, while the GRID K2 supports DirectX 12 (11_0) and Vulkan 1.2.175. The T600 has a higher feature level in both DirectX and Vulkan.

Specification Differences

| Specification | NVIDIA GRID K2 | NVIDIA T600 |

|---|---|---|

| Architecture | Kepler | Turing |

| Chip | GK104 | TU117 |

| Process Node | 28 nm | 12 nm |

| Transistors | 3,540 million | 4,700 million |

| Die Size | 294 mm² | 200 mm² |

| Transistor Density | 12.0M / mm² | 23.5M / mm² |

| Base Clock | Not recorded | 735 MHz |

| Boost Clock | Not recorded | 1335 MHz |

| Memory Type | GDDR5 | GDDR6 |

| Memory Bus Width | 256 bit | 128 bit |

| Effective Memory Clock | 5 Gbps | 10 Gbps |

| Shading Units | 1536 | 640 |

| TMUs | 128 | 40 |

| ROPs | 32 | 32 |

| Pixel Rate | 23.84 GPixel/s | 42.72 GPixel/s |

| Texture Rate | 95.36 GTexel/s | 53.40 GTexel/s |

| FP32 | 2.289 TFLOPS | 1.709 TFLOPS |

| FP16 | Not supported | 3.418 TFLOPS (2:1) |

| TDP | 225 W | 40 W |

| Slot Width | Dual-slot | Single-slot |

| Power Connectors | 1x 6-pin + 1x 8-pin | None |

| Suggested PSU | 550 W | 200 W |

| Display Outputs | No outputs | 4x mini-DisplayPort 1.4a |

| DirectX Support | 12 (11_0) | 12 (12_1) |

| Vulkan Support | 1.2.175 | 1.4 |

| Release Date | 2013-05-10 | 2021-04-11 |

Where Each One Wins

The GRID K2 wins in raw compute resource counts. It has 1,536 shading units versus 640, 128 TMUs versus 40, and a higher FP32 throughput at 2.289 TFLOPS versus 1.709 TFLOPS. Its texture rate of 95.36 GTexel/s is nearly double the T600's 53.40 GTexel/s. For workloads that scale with shading unit count and texture fill, the older card retains an advantage.

The T600 wins in almost every other measurable category. Its pixel rate of 42.72 GPixel/s is nearly double the GRID K2's 23.84 GPixel/s, driven by much higher clock speeds. Its FP16 capability at 3.418 TFLOPS opens up workloads the GRID K2 cannot handle. The T600's OpenCL score of 27,875 crushes the GRID K2's 10,602. Its power draw of 40 W versus 225 W makes it suitable for systems with modest power supplies, and its single-slot design with four display outputs enables workstation deployment scenarios that the GRID K2 cannot serve.

The GRID K2's higher average benchmark score of 8,080 versus 7,035 reflects its two strong Geekbench results, but this average does not include the diverse Passmark suite that the T600 is measured against. The T600's percentile ranking of 39 versus the GRID K2's 42 is close, but the head-to-head OpenCL result is the more reliable indicator of relative compute capability.

The Verdict

The data indicates a clear split based on workload type. The NVIDIA T600 is the better choice for modern compute, graphics, and workstation tasks. Its 62% OpenCL lead over the GRID K2, combined with FP16 support, superior pixel rate, newer API support, and dramatically lower power requirements, makes it the more capable and versatile card for most applications.

The NVIDIA GRID K2 retains relevance only in specific legacy scenarios. Its higher FP32 throughput and texture rate, along with its 256-bit memory bus, may benefit certain older workloads that were optimized for Kepler-era hardware. However, its lack of display outputs limits it to server-side virtualization roles, and its 225 W TDP with dual power connectors makes deployment more demanding.

For a user choosing between these two cards today, the T600 is the rational pick for almost any purpose. It offers better measured compute performance, modern feature support, lower power consumption, and actual display connectivity. The GRID K2's sole advantages are in raw FP32 and texture throughput, which do not translate to better results in the recorded benchmarks. The benchmark database shows the T600 as the superior product overall.