AMD Radeon Pro 555X vs NVIDIA GeForce GTX 680 Comparison

Head-to-Head Benchmarks

The benchmark data splits the two GPUs along API lines. In Geekbench OpenCL, the NVIDIA GeForce GTX 680 scores 16,906 against the AMD Radeon Pro 555X’s 12,628. That is a 33.9% advantage for the NVIDIA card, the largest margin in any test. The Vulkan result follows a similar pattern: NVIDIA scores 17,646, AMD scores 12,855, a 37.3% gap. In both compute-oriented workloads, the older Kepler architecture holds a decisive edge.

The AMD Radeon Pro 555X strikes back in Geekbench Metal. It scores 14,479 while the GTX 680 manages only 7,897. The delta is -45.5% from NVIDIA’s perspective, meaning AMD more than doubles the NVIDIA score in this test. This is not a narrow win; it is a complete reversal of the OpenCL and Vulkan results. The Metal benchmark appears to favor the GCN 4.0 architecture overwhelmingly, likely due to driver optimization and API-level design choices.

Looking at average benchmark scores, the GTX 680 sits at 14,150, while the Radeon Pro 555X averages 13,321. That puts NVIDIA ahead by roughly 6.2% overall. However, the percentile rankings tell a nearly identical story: the GTX 680 lands in the 55th percentile of all GPUs, and the Radeon Pro 555X sits at the 54th percentile. Both cards are clustered in the same performance tier, despite their wildly different results per API.

The nearest rivals confirm this middle-tier positioning. The GTX 680’s closest competitor is the NVIDIA Tesla K10, which averages 14,029, just 0.9% behind. The GeForce GTX 1070 Ti is 0.9% ahead at 14,277, and the Intel Iris Xe MAX Graphics is 1.2% ahead at 14,315. The AMD Radeon Pro 555X is bracketed by the GeForce GTX 480 at 13,300 (0.2% behind) and the AMD FirePro M6100 at 13,354 (0.2% ahead). Both cards are within roughly 1.5% of their nearest neighbors, meaning the head-to-head margin between these two is larger than their separation from adjacent rivals.

Architecture Differences

The NVIDIA GeForce GTX 680 uses the GK104 chip built on TSMC’s 28 nm process. It packs 3,540 million transistors onto a 294 mm² die, yielding a transistor density of 12.0 million per mm². The AMD Radeon Pro 555X uses the Polaris 21 chip fabricated by GlobalFoundries at 14 nm. That die is much smaller at 123 mm², holds 3,000 million transistors, and achieves a significantly higher density of 24.4 million per mm². The process node difference explains the density gap: AMD fits nearly as many transistors into less than half the silicon area.

The GTX 680 is a Kepler-generation part from the GeForce 600 series. It has 1,536 shading units, 128 texture mapping units, and 32 ROPs. The Radeon Pro 555X belongs to GCN 4.0 and the Radeon Pro Mac 500X series. It has 768 shading units, 48 TMUs, and 16 ROPs. NVIDIA’s chip carries double the shading units, more than double the TMUs, and exactly double the ROPs of the AMD part.

Memory configurations differ substantially. The GTX 680 has 2 GB of GDDR5 on a 256-bit bus, delivering 192.3 GB/s of bandwidth. The Radeon Pro 555X has 4 GB of GDDR5 on a 128-bit bus, delivering 94.08 GB/s. AMD has twice the capacity but exactly half the bandwidth. The memory clock is similar — 6 Gbps effective on NVIDIA versus 5.9 Gbps on AMD — so the bus width is the determining factor in the bandwidth gap.

Compute throughput also favors NVIDIA. The GTX 680 outputs 3.250 TFLOPS of FP32 performance, while the Radeon Pro 555X manages 1,393.2 GFLOPS. Pixel rate is 33.86 GPixel/s versus 14.51 GPixel/s, and texture rate is 135.4 GTexel/s versus 43.54 GTexel/s. The AMD card does match its FP32 rate for FP16 at 1:1, whereas the NVIDIA card lists no FP16 capability. However, for general compute workloads, the GTX 680’s raw throughput numbers are substantially higher.

Power and physical design diverge sharply. The GTX 680 has a 195 W TDP, requires a dual-slot cooler, two 6-pin power connectors, and a 450 W suggested power supply. It measures 256 mm in length, 111 mm in height, and 38 mm in width. The Radeon Pro 555X has a 75 W TDP, is an integrated graphics processor with no slot width, no power connectors, and no listed dimensions. It connects via PCIe 3.0 x8, while the GTX 680 uses PCIe 3.0 x16. This is a desktop discrete card versus a laptop-integrated part.

API support shows a generational shift. The GTX 680 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The Radeon Pro 555X supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. AMD has a higher DirectX feature level and a newer Vulkan version, which may explain its Metal performance advantage.

FAQ

Q: Which GPU wins the OpenCL benchmark?

A: The NVIDIA GeForce GTX 680 scores 16,906 in Geekbench OpenCL, which is 33.9% ahead of the AMD Radeon Pro 555X’s 12,628.

Q: How large is the AMD card’s win in Metal?

A: The Radeon Pro 555X scores 14,479 in Geekbench Metal against the GTX 680’s 7,897. The delta is -45.5% from NVIDIA’s perspective, meaning AMD’s score is roughly 83% higher.

Q: Do the two cards have similar overall performance?

A: The GTX 680 averages 14,150 across benchmarks, and the Radeon Pro 555X averages 13,321. The GTX 680 sits at the 55th percentile of all GPUs, while the Radeon Pro 555X sits at the 54th percentile.

Q: Which card has more memory bandwidth?

A: The GTX 680 has 192.3 GB/s of bandwidth on a 256-bit bus. The Radeon Pro 555X has 94.08 GB/s on a 128-bit bus, which is exactly half the bandwidth.

Q: How do the transistor densities compare?

A: The GTX 680 has a density of 12.0M transistors per mm² on a 294 mm² die. The Radeon Pro 555X has 24.4M transistors per mm² on a 123 mm² die, roughly double the density.

Q: Is the Radeon Pro 555X faster in Vulkan?

A: No. The GTX 680 scores 17,646 in Geekbench Vulkan, which is 37.3% ahead of the Radeon Pro 555X’s 12,855.

Specification Differences

• Process Node: NVIDIA uses 28 nm (TSMC); AMD uses 14 nm (GlobalFoundries).
• Transistors: NVIDIA has 3,540 million; AMD has 3,000 million.
• Die Size: NVIDIA is 294 mm²; AMD is 123 mm².
• Transistor Density: NVIDIA is 12.0M / mm²; AMD is 24.4M / mm².
• Memory Size: NVIDIA has 2 GB; AMD has 4 GB.
• Memory Bus Width: NVIDIA is 256-bit; AMD is 128-bit.
• Memory Bandwidth: NVIDIA is 192.3 GB/s; AMD is 94.08 GB/s.
• Shading Units: NVIDIA has 1,536; AMD has 768.
• TMUs: NVIDIA has 128; AMD has 48.
• ROPs: NVIDIA has 32; AMD has 16.
• Pixel Rate: NVIDIA is 33.86 GPixel/s; AMD is 14.51 GPixel/s.
• Texture Rate: NVIDIA is 135.4 GTexel/s; AMD is 43.54 GTexel/s.
• FP32: NVIDIA is 3.250 TFLOPS; AMD is 1,393.2 GFLOPS.
• FP16: NVIDIA lists none; AMD is 1,393.2 GFLOPS (1:1).
• TDP: NVIDIA is 195 W; AMD is 75 W.
• Slot Width: NVIDIA is dual-slot; AMD is IGP.
• Power Connectors: NVIDIA has 2x 6-pin; AMD has none.
• Suggested PSU: NVIDIA is 450 W; AMD lists none.
• Bus Interface: NVIDIA is PCIe 3.0 x16; AMD is PCIe 3.0 x8.
• Display Outputs: NVIDIA has 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2; AMD is portable device dependent.
• DirectX: NVIDIA is 12 (11_0); AMD is 12 (12_0).
• Vulkan: NVIDIA is 1.2.175; AMD is 1.3.
• Dimensions: NVIDIA is 256 mm x 111 mm x 38 mm; AMD has no listed dimensions.
• Release Date: NVIDIA released 2012-03-21; AMD released 2018-07-15.
• Launch MSRP: NVIDIA was 499 USD; AMD has no listed launch MSRP.

The Verdict

The data presents a clear split. The NVIDIA GeForce GTX 680 wins two of the three head-to-head benchmarks, taking OpenCL by 33.9% and Vulkan by 37.3%. Its average score of 14,150 is higher, and its percentile rank of 55 is one point above the Radeon Pro 555X. If the workload uses OpenCL or Vulkan, the GTX 680 is the stronger choice by a wide margin.

The AMD Radeon Pro 555X wins the Metal benchmark decisively, scoring 14,479 versus 7,897. That is the single biggest delta in either direction. For Metal-centric applications, the Radeon Pro 555X is not just competitive; it is dramatically faster. The AMD card also offers twice the memory capacity at 4 GB versus 2 GB, which matters for larger datasets or higher-resolution textures.

Architecturally, the GTX 680 is the more powerful GPU in raw terms. It has double the shading units, more than double the TMUs, double the ROPs, 3.4x the texture rate, 2.3x the pixel rate, and 2.3x the FP32 throughput. The Radeon Pro 555X counters with a smaller, denser 14 nm die, a newer DirectX feature level, a newer Vulkan version, and a 75 W TDP versus 195 W. It is a laptop-integrated part, whereas the GTX 680 is a desktop card requiring external power.

For compute workloads in OpenCL or Vulkan, the GTX 680 is the clear winner. For Metal workloads, the Radeon Pro 555X is the only sensible pick. The overall average favors NVIDIA, but the API-specific results are so lopsided that the choice depends entirely on the target environment.

Where Each One Wins

NVIDIA GeForce GTX 680 wins in OpenCL and Vulkan. The 33.9% OpenCL margin and 37.3% Vulkan margin are consistent and substantial. Raw compute resources back this up: 3.250 TFLOPS FP32, 135.4 GTexel/s texture rate, and 192.3 GB/s bandwidth. Any application using these APIs will see a meaningful performance advantage.

AMD Radeon Pro 555X wins in Metal. The 45.5% delta in NVIDIA’s favor is inverted here, with AMD scoring 14,479 versus 7,897. This is the largest single-test margin between the two cards. The 4 GB memory capacity also gives AMD an edge in memory-heavy scenarios, even though its bandwidth is lower.

NVIDIA wins on compute density. Higher shading unit count, TMU count, and ROP count translate directly to higher pixel and texture rates. The GTX 680’s 33.86 GPixel/s and 135.4 GTexel/s dwarf the Radeon’s 14.51 GPixel/s and 43.54 GTexel/s.

AMD wins on efficiency and integration. The 75 W TDP versus 195 W means the Radeon Pro 555X draws less than 40% of the power. It requires no power connectors, no dual-slot cooler, and fits in an IGP form factor. The GTX 680 needs a 450 W PSU and a 256 mm long, dual-slot card.

NVIDIA wins on overall average. The 14,150 average score tops 13,321, and the 55th percentile rank edges out the 54th. The nearest rival data shows both cards are mid-pack, but the GTX 680 sits slightly higher.