AMD Radeon RX 6600 LE vs NVIDIA CMP 90HX Comparison

AMD Radeon RX 6600 LE and NVIDIA CMP 90HX represent two radically different approaches to GPU hardware, one designed for conventional graphics workloads and the other for dedicated compute. The benchmark data shows a near-statistical tie in raw compute performance, but the specifications and feature sets diverge sharply, making the choice between them dependent entirely on the intended use case.

Head-to-Head Benchmarks

The only direct benchmark comparison available between the two cards is the Geekbench OpenCL test, which measures general-purpose compute performance. In this test, the AMD Radeon RX 6600 LE scores 69,229 points while the NVIDIA CMP 90HX scores 69,000 points. The delta is a razor-thin 0.3% in favor of the AMD card, a margin that falls well within typical run-to-run variance. For all practical purposes, these two GPUs deliver identical raw compute throughput in this specific workload.

Looking at the broader benchmark context, the AMD Radeon RX 6600 LE posts an average benchmark score of 70,829, placing it in the 91st percentile of all GPUs. Its nearest rivals include the NVIDIA RTX A3000 Mobile (70,140, 1% slower), the NVIDIA Quadro P6000 (69,986, 1.2% slower), and the AMD Radeon RX 6650M (71,768, 1.3% faster). The RX 6600 LE sits in a tightly packed cluster where a few percent separates it from both its mobile workstation and desktop competitors.

The NVIDIA CMP 90HX, meanwhile, achieves an average benchmark score of 69,000, ranking in the 90th percentile of all GPUs. Its nearest rivals include the Intel Arc A770 (68,809, 0.3% slower), the AMD Radeon Instinct MI25 (68,562, 0.6% slower), and the AMD Radeon Pro WX 8200 (69,870, 1.2% faster). The CMP 90HX is similarly positioned in a competitive band, with its closest competitor being effectively tied with it.

The single head-to-head result of 0.3% in favor of the RX 6600 LE is the only numerical comparison that separates these two cards directly. However, the performance context diverges when considering their different positions relative to their own rival sets. The RX 6600 LE edges out the Quadro P6000 by 1.2%, while the CMP 90HX trails the same Quadro P6000 by 1.4%. That 2.6% swing in relative positioning suggests that the AMD card has a slight edge in compute efficiency, though the absolute numbers remain nearly identical.

The Verdict

The data presents a clear but nuanced picture. For any workload that requires a display output, the NVIDIA CMP 90HX is immediately disqualified — it has no display outputs whatsoever. The AMD Radeon RX 6600 LE, by contrast, offers 1x HDMI 2.1 and 3x DisplayPort 1.4a outputs, making it the only viable option for gaming, content creation, or any task that requires visual feedback.

For compute-heavy workloads where display output is irrelevant, the choice hinges on other factors. The benchmark scores are effectively tied, with the RX 6600 LE holding a 0.3% advantage in the sole head-to-head test. However, the specifications tell a different story. The CMP 90HX offers substantially higher raw compute resources: 6,400 shading units versus 1,792, 200 tensor cores versus none, and 21.89 TFLOPS FP32 versus 8.942 TFLOPS. Yet the benchmark results show these theoretical advantages do not translate into real-world compute gains in the measured workload.

The CMP 90HX does offer significantly more memory bandwidth — 760.3 GB/s versus 224.0 GB/s — and twice the memory capacity at 10 GB versus 8 GB. This could matter for memory-bound workloads. However, the CMP 90HX is end-of-life and carries a 320 W TDP with a suggested 700 W power supply, compared to the RX 6600 LE's 132 W TDP and 300 W suggested PSU. The RX 6600 LE is also actively in production.

The verdict from the data: the AMD Radeon RX 6600 LE is the more versatile and practical choice for nearly all users. It matches the CMP 90HX in compute performance, provides display outputs, consumes less power, and remains a current product. The NVIDIA CMP 90HX is only defensible for specialized compute workloads that can take advantage of its larger memory pool and bandwidth, and where its lack of display outputs is not a hindrance.

FAQ

Q: Which GPU has a higher benchmark score in the head-to-head test?

A: The AMD Radeon RX 6600 LE wins the Geekbench OpenCL test with a score of 69,229 versus 69,000 for the NVIDIA CMP 90HX, a difference of 0.3%.

Q: Can the NVIDIA CMP 90HX be used for gaming with a monitor?

A: No. The NVIDIA CMP 90HX has no display outputs, so it cannot connect to a monitor. The AMD Radeon RX 6600 LE, with its HDMI 2.1 and DisplayPort 1.4a outputs, is required for any display-based workload.

Q: How do the two cards compare in terms of memory bandwidth?

A: The NVIDIA CMP 90HX has significantly higher memory bandwidth at 760.3 GB/s, compared to 224.0 GB/s for the AMD Radeon RX 6600 LE. The CMP 90HX also has more memory (10 GB vs 8 GB) and uses faster GDDR6X memory.

Q: What are the power requirements for each card?

A: The AMD Radeon RX 6600 LE has a TDP of 132 W and requires a 300 W power supply, while the NVIDIA CMP 90HX has a TDP of 320 W and requires a 700 W power supply. The CMP 90HX also needs two 8-pin power connectors versus one for the RX 6600 LE.

Q: Which GPU has better production status?

A: The AMD Radeon RX 6600 LE is listed as "Active" in production, while the NVIDIA CMP 90HX is listed as "End-of-life". The RX 6600 LE was also released later, with a release date of December 2023 compared to July 2021 for the CMP 90HX.

Q: How do the two cards compare in terms of compute resources?

A: The NVIDIA CMP 90HX has far more shading units (6,400 vs 1,792), more texture mapping units (200 vs 112), more ray tracing cores (50 vs 28), and 200 tensor cores, while the AMD card has none. The CMP 90HX also has much higher FP32 performance at 21.89 TFLOPS versus 8.942 TFLOPS for the RX 6600 LE.

Specification Differences

The two cards differ across nearly every major specification category. The AMD Radeon RX 6600 LE is built on a 7 nm TSMC process with 11,060 million transistors on a 237 mm² die, while the NVIDIA CMP 90HX uses Samsung's 8 nm process with 28,300 million transistors on a much larger 628 mm² die. Transistor density is nearly identical at 46.7M per mm² versus 45.1M per mm².

Clock speeds show the AMD card running substantially higher: a base of 1626 MHz and boost of 2495 MHz compared to 1500 MHz and 1710 MHz for the CMP 90HX. The AMD card also has a game clock of 2045 MHz, while the NVIDIA card has no game clock specified. Memory clocks differ as well, with the RX 6600 LE at 1750 MHz (14 Gbps effective) and the CMP 90HX at 1188 MHz (19 Gbps effective).

The memory subsystem is a major point of divergence. The RX 6600 LE has 8 GB of GDDR6 on a 128-bit bus with 224.0 GB/s bandwidth. The CMP 90HX has 10 GB of GDDR6X on a 320-bit bus with 760.3 GB/s bandwidth. Pixel rate favors the AMD card at 159.7 GPixel/s versus 136.8 GPixel/s, while texture rate favors the NVIDIA card at 342.0 GTexel/s versus 279.4 GTexel/s.

Power and physical specifications also differ markedly. The RX 6600 LE draws 132 W with a 300 W suggested PSU, while the CMP 90HX draws 320 W with a 700 W suggested PSU. The AMD card is 190 mm long and 40 mm wide, while the NVIDIA card is 285 mm long and 112 mm high. The RX 6600 LE connects via PCIe 4.0 x8, while the CMP 90HX uses PCIe 1.0 x4. The AMD card has display outputs; the NVIDIA card has none.

Architecture Differences

The architectural divide is fundamental. The AMD Radeon RX 6600 LE is based on the RDNA 2.0 architecture with the Navi 23 chip, part of the Navi II (RX 6000) generation. The NVIDIA CMP 90HX uses the Ampere architecture with the GA102 chip, from the Mining GPUs generation. These are entirely different design philosophies.

The RX 6600 LE is a graphics-first architecture, optimized for gaming with its RDNA 2.0 design. It has 28 ray tracing cores and supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The CMP 90HX, despite being a mining-focused card, also supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, but its 50 ray tracing cores and 200 tensor cores suggest a compute-heavy design. The AMD card has no tensor cores, while the NVIDIA card dedicates significant silicon to them.

Transistor counts tell a story of scale: the CMP 90HX packs 28,300 million transistors versus 11,060 million for the RX 6600 LE. The NVIDIA card is more than twice the die size at 628 mm² versus 237 mm². FP16 performance also differs: the RX 6600 LE achieves 17.88 TFLOPS with a 2:1 ratio, while the CMP 90HX achieves 21.89 TFLOPS with a 1:1 ratio, indicating different compute paths for half-precision workloads.

The foundry choice also differs, with TSMC producing the 7 nm AMD chip and Samsung producing the 8 nm NVIDIA chip. These architectural differences explain the divergent clock speeds and efficiency profiles, with the AMD card achieving higher clocks on a smaller, more efficient process.

Where Each One Wins

The AMD Radeon RX 6600 LE wins in every scenario that requires a display. It has HDMI 2.1 and DisplayPort 1.4a outputs, making it the clear choice for gaming, desktop productivity, or any interactive workload. Its higher clock speeds and lower power draw make it more suitable for systems with modest power supplies. It is also the only card of the two that is still in active production, which matters for availability and support.

The NVIDIA CMP 90HX wins in memory-bandwidth-sensitive compute workloads. Its 760.3 GB/s bandwidth and 10 GB memory capacity provide a substantial advantage over the RX 6600 LE's 224.0 GB/s and 8 GB. For compute tasks that saturate memory bandwidth — such as large data processing or certain rendering tasks — the CMP 90HX has the theoretical edge, even if the benchmark data shows a tie in the measured workload.

For raw compute throughput, the CMP 90HX's 21.89 TFLOPS FP32 and 21.89 TFLOPS FP16 (1:1) suggest it should outperform the RX 6600 LE's 8.942 TFLOPS FP32 and 17.88 TFLOPS FP16 (2:1). The benchmark data contradicts this expectation, however, showing near-identical OpenCL performance. This suggests that the CMP 90HX's architectural advantages are not being fully utilized in the measured test, or that other bottlenecks limit its real-world performance.

The RX 6600 LE wins on efficiency, drawing 132 W versus 320 W and requiring a 300 W versus 700 W PSU. It also wins on physical size at 190 mm versus 285 mm in length. For users building compact systems or with power constraints, the AMD card is the clear choice. The CMP 90HX, as a mining-specific product with no display outputs and end-of-life status, only makes sense for dedicated compute rigs where its memory bandwidth and capacity can be exploited and where its size, power draw, and lack of outputs are acceptable trade-offs.