AMD Radeon HD 8690M vs NVIDIA Quadro K620M Comparison

The AMD Radeon HD 8690M and NVIDIA Quadro K620M are both end-of-life mobile graphics solutions, but the data shows a clear, if narrow, winner in raw compute performance. Based on the single available benchmark result, the AMD Radeon HD 8690M outperforms the NVIDIA Quadro K620M by 3% in Geekbench OpenCL, scoring 6137 against 5957. This margin is small, yet it places the AMD part in a higher performance percentile (36th vs 34th), indicating a slightly better standing across the entire GPU landscape.

FAQ

Q: Which GPU is faster in the available benchmark data?

A: The AMD Radeon HD 8690M is faster, winning the Geekbench OpenCL test with a score of 6137 compared to the NVIDIA Quadro K620M's 5957, a delta of 3%.

Q: How does each GPU compare to its closest rivals?

A: The AMD Radeon HD 8690M is 0.3% ahead of the Intel Iris Pro Graphics 6200 and 1% ahead of the NVIDIA RTX A400, but it is 1.2% behind the AMD Radeon HD 6950. The NVIDIA Quadro K620M is 0.5% ahead of the Intel UHD Graphics 730, but it trails the AMD Radeon HD 8750M by 0.2% and the NVIDIA Quadro K4000 by 0.4%.

Q: What are the key architectural differences between the two chips?

A: The AMD Radeon HD 8690M uses the GCN 1.0 architecture on a chip codenamed "Sun," while the NVIDIA Quadro K620M uses the Maxwell architecture on the "GM108S" chip. AMD's chip has 690 million transistors on a 56 mm² die, while NVIDIA's has 1,020 million transistors on a 77 mm² die.

Q: Which GPU has a higher memory bandwidth?

A: The AMD Radeon HD 8690M has double the memory bandwidth, offering 32.00 GB/s, while the NVIDIA Quadro K620M provides 16.02 GB/s. This is due to the AMD part using GDDR5 memory versus the NVIDIA part's DDR3.

Q: Do these GPUs support modern graphics APIs?

A: Both support DirectX 12 and OpenGL 4.6. However, they differ in Vulkan support: the AMD Radeon HD 8690M supports Vulkan 1.2.170, while the NVIDIA Quadro K620M supports a newer Vulkan 1.4.

Q: What is the TDP of the NVIDIA Quadro K620M?

A: The NVIDIA Quadro K620M has a specified TDP of 30 W. The AMD Radeon HD 8690M's TDP is not listed in the data.

Architecture Differences

The two GPUs represent fundamentally different design philosophies from their respective manufacturers. The AMD Radeon HD 8690M is built on the Graphics Core Next (GCN) 1.0 architecture, utilizing a chip codenamed "Sun." In contrast, the NVIDIA Quadro K620M employs the Maxwell architecture, built around the "GM108S" chip. Both are fabricated on a 28 nm process at TSMC, but the transistor budgets differ significantly.

The NVIDIA chip contains notably more transistors: 1,020 million compared to AMD's 690 million. This results in a larger die size for NVIDIA at 77 mm², versus AMD's more compact 56 mm². Consequently, the transistor density is slightly higher on the NVIDIA chip at 13.2M transistors per mm², compared to AMD's 12.3M per mm². This suggests NVIDIA packed more logic into a smaller area relative to its die size.

In terms of compute resources, the NVIDIA Quadro K620M has more shading units, with 384 compared to AMD's 320. However, the AMD Radeon HD 8690M has more texture mapping units (TMUs), with 20 versus NVIDIA's 16. Both parts have an identical number of render output units (ROPs) at 8. The memory subsystems also diverge, with AMD opting for 2 GB of GDDR5 on a 64-bit bus, while NVIDIA uses 2 GB of DDR3 on the same 64-bit bus width. Neither GPU features dedicated ray tracing or tensor cores.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, which measures general-purpose compute performance. In this test, the AMD Radeon HD 8690M comes out on top with a score of 6137, beating the NVIDIA Quadro K620M's score of 5957. This represents a 3% performance delta in favor of AMD.

This 3% lead is small enough to be considered a marginal victory, but it is consistent with the percentile ranking. The AMD part sits in the 36th percentile of all GPUs, while the NVIDIA part sits in the 34th percentile. When compared to their nearest rivals, the AMD GPU's score of 6137 is nearly identical to the Intel Iris Pro Graphics 6200's 6117, but it is 1.2% behind the older AMD Radeon HD 6950's 6210. The NVIDIA Quadro K620M's score of 5957 puts it in the same performance class as the AMD Radeon HD 8730M, which scores 5955, and it is slightly behind the AMD Radeon HD 8750M's 5970.

The data indicates that while the AMD Radeon HD 8690M wins this specific benchmark, the performance envelope of both GPUs is very tight. The NVIDIA part is not far behind, and its performance is closely clustered with several other mobile and desktop parts from both AMD and NVIDIA.

Specification Differences

The following specifications differ between the AMD Radeon HD 8690M and the NVIDIA Quadro K620M:

• Architecture: GCN 1.0 vs Maxwell
• Chip: Sun vs GM108S
• Transistors: 690 million vs 1,020 million
• Die Size: 56 mm² vs 77 mm²
• Transistor Density: 12.3M / mm² vs 13.2M / mm²
• Base Clock: 925 MHz vs 1029 MHz
• Boost Clock: 975 MHz vs 1124 MHz
• Memory Type: GDDR5 vs DDR3
• Memory Clock: 4 Gbps effective vs 2 Gbps effective
• Memory Bandwidth: 32.00 GB/s vs 16.02 GB/s
• Shading Units: 320 vs 384
• TMUs: 20 vs 16
• Pixel Rate: 7.800 GPixel/s vs 8.992 GPixel/s
• Texture Rate: 19.50 GTexel/s vs 17.98 GTexel/s
• FP32 Performance: 624.0 GFLOPS vs 863.2 GFLOPS
• TDP: Not listed vs 30 W
• Slot Width: Not listed vs MXM Module
• Power Connectors: Not listed vs None
• Bus Interface: PCIe 3.0 x8 vs MXM-A (3.0)
• Display Outputs: Not listed vs Portable Device Dependent
• Vulkan Support: 1.2.170 vs 1.4
• Release Date: 2013-02-28 vs 2015-02-28
• Predecessor: London vs Quadro Fermi-M
• Successor: Gem System vs Quadro Maxwell-M

The Verdict

The benchmark data points to a narrow but definitive win for the AMD Radeon HD 8690M in raw compute performance. Its 3% lead in Geekbench OpenCL, coupled with a higher percentile rank, makes it the stronger choice for tasks that rely on general-purpose compute. The AMD part also offers a substantial advantage in memory bandwidth at 32.00 GB/s, which is exactly double the 16.02 GB/s of the NVIDIA Quadro K620M.

However, the NVIDIA Quadro K620M is not without its strengths. It has a significantly higher FP32 compute rating at 863.2 GFLOPS versus 624.0 GFLOPS, and it also has a higher pixel rate. This suggests that while the AMD GPU wins the specific OpenCL workload, the NVIDIA GPU may have advantages in other types of rendering tasks. The NVIDIA part also has a defined TDP of 30 W, which is a specification not available for the AMD part.

For a user prioritizing the measured OpenCL benchmark score, the AMD Radeon HD 8690M is the superior product. For a user who values the higher theoretical FP32 throughput or the defined power envelope, the NVIDIA Quadro K620M presents a compelling counterpoint.

Where Each One Wins

AMD Radeon HD 8690M:

• Compute Benchmark: Wins the only head-to-head test, Geekbench OpenCL, with a score of 6137 vs 5957.
• Memory Bandwidth: Offers a decisive 2x advantage with 32.00 GB/s compared to NVIDIA's 16.02 GB/s, which can benefit memory-intensive workloads.
• Texture Processing: Has a higher texture rate of 19.50 GTexel/s, suggesting an advantage in fill-rate-bound scenarios.
• Vulkan Support: Supports Vulkan 1.2.170, which is a newer version than the 1.2.170? No, the NVIDIA supports 1.4, so this is a win for NVIDIA. Correction: The AMD part supports Vulkan 1.2.170, which is older than the NVIDIA's Vulkan 1.4. Therefore, the NVIDIA wins on this front. The AMD wins on OpenCL benchmark and memory bandwidth.

NVIDIA Quadro K620M:

• FP32 Performance: Has a higher theoretical FP32 rating of 863.2 GFLOPS compared to AMD's 624.0 GFLOPS, indicating potential strength in compute tasks using single-precision floats.
• Pixel Rate: Has a higher pixel rate of 8.992 GPixel/s vs AMD's 7.800 GPixel/s, suggesting an edge in pixel-heavy rendering.
• Clock Speeds: Operates at higher base and boost clocks (1029 MHz / 1124 MHz) compared to AMD (925 MHz / 975 MHz).
• API Support: Supports a newer version of Vulkan (1.4) and has a DirectX 12 feature level of 11_0, which is one tier below AMD's 11_1.
• Power Specification: Is the only one of the two with a defined TDP of 30 W, making its power requirements explicit.