AMD Radeon HD 8690M vs NVIDIA GeForce GTX 580M Comparison

The NVIDIA GeForce GTX 580M and AMD Radeon HD 8690M represent two distinct approaches to mobile graphics from different eras, and the benchmark data shows a narrow but clear performance gap. In the single available OpenCL benchmark, the GTX 580M scores 6389, which is 4.1% higher than the HD 8690M’s 6137. This places the NVIDIA part at the 37th percentile of all GPUs, while the AMD part sits at the 36th percentile, making both essentially mid-pack performers. The GTX 580M’s nearest rival is the NVIDIA GeForce GTX 460 SE, with an identical average score of 6389, while the HD 8690M sits just above the Intel Iris Pro Graphics 6200, which scores 6117.

FAQ

Q: Which GPU is faster in the Geekbench OpenCL benchmark?

A: The NVIDIA GeForce GTX 580M is faster, scoring 6389 compared to the AMD Radeon HD 8690M’s 6137. This represents a 4.1% advantage for the NVIDIA part.

Q: How do these GPUs compare to their closest competitors?

A: The GTX 580M is exactly matched by the NVIDIA GeForce GTX 460 SE (both score 6389), and it is 0.9% ahead of the AMD Radeon Pro WX 4100. The HD 8690M is 0.3% ahead of the Intel Iris Pro Graphics 6200 and 1.2% behind the AMD Radeon HD 6950.

Q: What are the memory specifications of each card?

A: Both cards feature 2 GB of GDDR5 memory, but the GTX 580M uses a 256-bit bus with 96.00 GB/s bandwidth, while the HD 8690M uses a 64-bit bus with only 32.00 GB/s bandwidth.

Q: What is the manufacturing process for each chip?

A: The GTX 580M uses a 40 nm process at TSMC, while the HD 8690M uses a more advanced 28 nm process, also at TSMC. The die sizes differ significantly: 332 mm² for the GTX 580M versus 56 mm² for the HD 8690M.

Q: Which GPU has higher raw compute throughput?

A: The GTX 580M has significantly higher throughput figures. It delivers 952.3 GFLOPS of FP32 performance, 39.68 GTexel/s of texture fill rate, and 9.920 GPixel/s of pixel rate. The HD 8690M, by contrast, offers 624.0 GFLOPS, 19.50 GTexel/s, and 7.800 GPixel/s.

Q: Do these GPUs support different API levels?

A: Yes. Both support DirectX 12 and OpenGL 4.6, but the HD 8690M has a higher DirectX feature level (12_1 versus 11_0) and also supports Vulkan 1.2.170, while the GTX 580M has no Vulkan support listed.

The Verdict

The benchmark data tells a straightforward story for those choosing between these two mobile GPUs. The NVIDIA GeForce GTX 580M is the performance pick, with a 4.1% lead in the OpenCL benchmark. Its 37th percentile ranking versus the HD 8690M’s 36th percentile reinforces this advantage, though the margin is modest. The GTX 580M’s superiority is more pronounced in raw specifications: it offers nearly double the texture fill rate (39.68 GTexel/s versus 19.50 GTexel/s), triple the memory bandwidth (96.00 GB/s versus 32.00 GB/s), and 52.6% higher FP32 throughput (952.3 GFLOPS versus 624.0 GFLOPS). These are not trivial differences; they suggest the NVIDIA part will handle texture-heavy and bandwidth-intensive workloads noticeably better.

However, the HD 8690M is not without its merits. It is built on a 28 nm process versus 40 nm, making it significantly more transistor-dense (12.3M / mm² versus 5.9M / mm²). It also has a newer architecture (GCN 1.0 versus Fermi 2.0), supports Vulkan 1.2.170, and has a higher DirectX feature level (12_1 versus 11_0). The HD 8690M achieves its score with far fewer resources — 320 shading units, 20 TMUs, and 8 ROPs versus 384, 64, and 32 on the GTX 580M — indicating better architectural efficiency. For users prioritizing modern API support or power efficiency (the HD 8690M has no listed TDP while the GTX 580M draws 100 W), the AMD part is the sensible choice. For pure compute performance, the GTX 580M wins, but the margin is slim enough that real-world differences may be minimal.

Head-to-Head Benchmarks

The only benchmark available for direct comparison is Geekbench OpenCL, where the NVIDIA GeForce GTX 580M emerges victorious with a score of 6389 against the AMD Radeon HD 8690M’s 6137. The 4.1% delta in favor of NVIDIA is consistent with the overall performance tier ranking: the GTX 580M sits at the 37th percentile of all GPUs, while the HD 8690M is at the 36th. This is a narrow but consistent gap.

Looking at the rivals to contextualize this result, the GTX 580M’s score of 6389 is exactly equal to that of the NVIDIA GeForce GTX 460 SE (deltaPct 0%), placing the two NVIDIA parts on par. The GTX 580M is 0.9% ahead of the AMD Radeon Pro WX 4100 (6330) and 1% ahead of the AMD Radeon R7 M350 (6327). Meanwhile, the HD 8690M’s 6137 is 0.3% above the Intel Iris Pro Graphics 6200 (6117) and 1% above both the NVIDIA RTX A400 (6078) and the NVIDIA GeForce MX230 (6077). However, the HD 8690M trails the AMD Radeon HD 6950 (6210) by 1.2%. This means the GTX 580M not only beats the HD 8690M directly but also holds a higher position relative to a broader field of comparable parts.

The victory for the GTX 580M is notable because it achieves this score with older, less efficient technology. The Fermi 2.0 architecture uses a 40 nm process and a 332 mm² die, while the HD 8690M’s GCN 1.0 uses a 28 nm process and a 56 mm² die. Despite the modern advantages of the AMD part, it still cannot overcome the GTX 580M’s larger complement of execution resources — 384 shading units versus 320, and 64 TMUs versus 20. The GTX 580M’s 256-bit memory bus and 96.00 GB/s bandwidth are also decisive advantages over the HD 8690M’s 64-bit bus and 32.00 GB/s, likely explaining the OpenCL score difference in memory-bound workloads.

Specification Differences

The two GPUs diverge sharply across nearly every specification. The NVIDIA GeForce GTX 580M is built on the GF114 chip using the Fermi 2.0 architecture, while the AMD Radeon HD 8690M uses the Sun chip with GCN 1.0 architecture. The process nodes differ: the GTX 580M uses 40 nm, the HD 8690M uses 28 nm, both at TSMC. Transistor counts are dramatically different — the GTX 580M packs 1,950 million transistors on a 332 mm² die (5.9M / mm² density), while the HD 8690M has 690 million on 56 mm² (12.3M / mm²). This makes the HD 8690M over twice as dense.

Clock speeds are only listed for the HD 8690M, with a base of 925 MHz and boost of 975 MHz; the GTX 580M’s base and boost clocks are not provided. Memory clocks differ as well: the GTX 580M runs at 750 MHz (3 Gbps effective), while the HD 8690M runs at 1000 MHz (4 Gbps effective). However, the GTX 580M’s 256-bit bus versus the HD 8690M’s 64-bit bus means the former achieves 96.00 GB/s bandwidth versus 32.00 GB/s, despite the lower memory clock.

Compute resources are heavily in favor of the GTX 580M: 384 shading units, 64 TMUs, and 32 ROPs versus the HD 8690M’s 320, 20, and 8. Consequently, pixel rate is 9.920 GPixel/s versus 7.800 GPixel/s, texture rate is 39.68 GTexel/s versus 19.50 GTexel/s, and FP32 throughput is 952.3 GFLOPS versus 624.0 GFLOPS. The GTX 580M has a TDP of 100 W, while the HD 8690M has no listed TDP. The GTX 580M uses an MXM-B (3.0) interface and MXM Module slot width, while the HD 8690M uses PCIe 3.0 x8. Display outputs are listed as portable device dependent for the GTX 580M, with no listing for the HD 8690M.

Architecture Differences

The architectural gap between these two GPUs reflects a generational shift. The GTX 580M is based on Fermi 2.0, NVIDIA’s second iteration of the Fermi architecture, built on a 40 nm process. It features 1,950 million transistors on a 332 mm² die, which is a massive chip for a mobile part. The HD 8690M, in contrast, uses AMD’s GCN 1.0 architecture, built on a 28 nm process with 690 million transistors on a 56 mm² die. This makes the AMD chip far more efficient in terms of transistor density — 12.3M / mm² versus 5.9M / mm² — and likely more power-efficient overall, though the GTX 580M’s 100 W TDP is the only explicit power figure available.

The core configurations tell a story of brute force versus efficiency. The GTX 580M has 384 shading units, 64 TMUs, and 32 ROPs, giving it a theoretical peak of 952.3 GFLOPS FP32. The HD 8690M has 320 shading units, 20 TMUs, and 8 ROPs, peaking at 624.0 GFLOPS FP32. The GTX 580M’s advantage in TMUs and ROPs is particularly stark — 3.2x more TMUs and 4x more ROPs — which explains its much higher texture and pixel rates despite the architecture being older.

Memory architecture also differs fundamentally. The GTX 580M uses a 256-bit memory bus with 2 GB of GDDR5 at 750 MHz (3 Gbps effective), achieving 96.00 GB/s. The HD 8690M uses a 64-bit bus with 2 GB of GDDR5 at 1000 MHz (4 Gbps effective), achieving only 32.00 GB/s. This threefold bandwidth advantage for the NVIDIA part is a critical architectural difference that likely contributes to its higher benchmark score.

API support shows the HD 8690M’s newer lineage. Both support DirectX 12 and OpenGL 4.6, but the HD 8690M supports DirectX feature level 12_1 versus 11_0 on the GTX 580M. The HD 8690M also supports Vulkan 1.2.170, while the GTX 580M has no Vulkan support listed. This means the AMD part is better positioned for modern software that can leverage newer API features, even though its raw compute is lower. The release dates corroborate this: the GTX 580M launched on 2011-06-27, while the HD 8690M launched on 2013-02-28, making the AMD part nearly two years newer. Both are now end-of-life products, but the HD 8690M’s architectural advantages in process node and API support provide a more future-proof feature set, even if the GTX 580M retains the performance lead in the available benchmark.