AMD Radeon RX 9070 GRE vs NVIDIA CMP 90HX Comparison

FAQ

Q: Which GPU has the higher raw compute throughput in FP32?

A: The AMD Radeon RX 9070 GRE delivers 34.28 TFLOPS of FP32 compute, while the NVIDIA CMP 90HX delivers 21.89 TFLOPS. The AMD part is roughly 57% higher in this metric.

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

A: The NVIDIA CMP 90HX features 760.3 GB/s of bandwidth from 10 GB of GDDR6X on a 320-bit bus. The AMD Radeon RX 9070 GRE offers 432.0 GB/s from 12 GB of GDDR6 on a 192-bit bus. NVIDIA holds a substantial 76% bandwidth advantage.

Q: Which card has the higher transistor density?

A: The AMD Radeon RX 9070 GRE, built on TSMC 4 nm, packs 151.0M transistors per mm² across a 357 mm² die. The NVIDIA CMP 90HX, on Samsung 8 nm, achieves 45.1M per mm² on a 628 mm² die.

Q: What is the Geekbench OpenCL score difference?

A: The AMD Radeon RX 9070 GRE scores 109309, while the NVIDIA CMP 90HX scores 69000. The AMD card leads by 36.9% in this benchmark.

Q: Which GPU has the higher boost clock?

A: The AMD Radeon RX 9070 GRE boosts to 2790 MHz, compared to 1710 MHz for the NVIDIA CMP 90HX. The AMD card also has a higher base clock of 1420 MHz versus 1500 MHz for NVIDIA, though NVIDIA’s base is slightly higher.

Q: What is the power draw difference?

A: The NVIDIA CMP 90HX has a 320 W TDP, while the AMD Radeon RX 9070 GRE has a 220 W TDP. The AMD part consumes 100 W less power.

Q: Which card has display outputs?

A: The AMD Radeon RX 9070 GRE includes 1x HDMI 2.1b and 3x DisplayPort 2.1a outputs. The NVIDIA CMP 90HX has no display outputs, as it was designed for mining operations.

Where Each One Wins

The AMD Radeon RX 9070 GRE wins in computational workloads that favor throughput and modern manufacturing. Its FP32 performance of 34.28 TFLOPS is substantially higher than the NVIDIA CMP 90HX’s 21.89 TFLOPS, a 57% advantage. The Geekbench OpenCL result confirms this: the AMD card scores 109309 versus 69000 for NVIDIA, a 36.9% lead. The AMD part also excels in pixel processing with 267.8 GPixel/s and texture work with 535.7 GTexel/s, both roughly double the NVIDIA figures of 136.8 GPixel/s and 342.0 GTexel/s.

The NVIDIA CMP 90HX wins in memory-centric scenarios. Its 760.3 GB/s bandwidth is 76% higher than the AMD card’s 432.0 GB/s. The NVIDIA card also has a wider 320-bit bus compared to 192-bit on AMD. For workloads that are bandwidth-bound, such as large data transfers or certain rendering passes, the NVIDIA part holds a clear edge. Additionally, NVIDIA’s higher ROP count of 80 versus 96 on AMD is actually lower, but the AMD card compensates with higher clock speeds.

In shading units, NVIDIA has 6400 versus 3072 on AMD, but AMD’s higher clocks and newer architecture narrow the gap. The NVIDIA card also has 200 TMUs versus 192 on AMD, a marginal difference. For ray tracing, NVIDIA has 50 RT cores versus 48 on AMD, though AMD’s boost clocks may offset this in real workloads. The AMD card wins in overall average benchmark score percentile, sitting at 87 versus 90 for NVIDIA, though NVIDIA edges slightly higher in percentile ranking.

Architecture Differences

The NVIDIA CMP 90HX uses the GA102 chip on the Ampere architecture, manufactured on Samsung’s 8 nm process. This is a large chip at 628 mm² with 28,300 million transistors. The die size is nearly double that of the AMD part, but the transistor density is only 45.1M per mm², reflecting the older process node.

The AMD Radeon RX 9070 GRE uses the Navi 48 chip on RDNA 4.0 architecture, built on TSMC’s 4 nm process. The die is 357 mm² with 53,900 million transistors, giving a transistor density of 151.0M per mm². This is more than three times the density of the NVIDIA chip, allowing AMD to pack more transistors into a smaller area. The newer process also enables higher clock speeds, with the AMD boost clock reaching 2790 MHz versus 1710 MHz on NVIDIA.

The NVIDIA chip relies on dedicated tensor cores, with 200 present, while the AMD card has no tensor cores listed. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The NVIDIA card is based on the Ampere architecture, which is one generation older than RDNA 4.0. The AMD card has a game clock of 2220 MHz, a feature not present on the NVIDIA part.

The NVIDIA CMP 90HX is part of the Mining GPUs generation and has no display outputs. The AMD card belongs to the Radeon RX 9000 series and is actively sold with full display connectivity. The NVIDIA card is end-of-life, while the AMD card is still in active production.

Specification Differences

The two cards differ across nearly every key specification. The NVIDIA CMP 90HX has 6400 shading units, 200 TMUs, and 80 ROPs. The AMD Radeon RX 9070 GRE has 3072 shading units, 192 TMUs, and 96 ROPs. Despite having fewer shading units, the AMD card achieves higher FP32 throughput due to its architecture and clocks.

Memory differs significantly: NVIDIA uses 10 GB of GDDR6X on a 320-bit bus with 760.3 GB/s bandwidth. AMD uses 12 GB of GDDR6 on a 192-bit bus with 432.0 GB/s. The NVIDIA memory clock is 1188 MHz with 19 Gbps effective, while AMD’s memory clock is 2250 MHz with 18 Gbps effective.

The process node is 8 nm for NVIDIA and 4 nm for AMD. The NVIDIA die is 628 mm², while AMD’s is 357 mm². Transistor counts are 28,300 million for NVIDIA and 53,900 million for AMD. The NVIDIA card has a base clock of 1500 MHz and boost of 1710 MHz, while AMD has a base of 1420 MHz, game of 2220 MHz, and boost of 2790 MHz.

Power consumption is 320 W for NVIDIA and 220 W for AMD. The suggested PSU is 700 W for NVIDIA and 550 W for AMD. Both use dual-slot designs and 2x 8-pin power connectors. The NVIDIA card uses PCIe 1.0 x4 interface, while AMD uses PCIe 5.0 x16. The NVIDIA card has no display outputs, while AMD has 1x HDMI 2.1b and 3x DisplayPort 2.1a. NVIDIA’s dimensions are 285 mm in length and 112 mm in height, while AMD’s dimensions are not listed.

Head-to-Head Benchmarks

The only shared benchmark in the database is Geekbench OpenCL. In this test, the AMD Radeon RX 9070 GRE scores 109309, while the NVIDIA CMP 90HX scores 69000. The AMD card wins by 36.9%, a decisive margin. This result reflects the AMD card’s higher FP32 compute and superior memory efficiency despite lower bandwidth.

The NVIDIA card’s nearest rivals in the database include the Intel Arc A770 at 68809 (0.3% behind), the AMD Radeon Instinct MI25 at 68562 (0.6% behind), and the AMD Radeon Pro WX 8200 at 69870 (1.2% ahead). These comparisons show the NVIDIA card sits in a similar performance band to mid-range workstation GPUs.

The AMD card’s nearest rivals include the Intel Arc A580 at 57756 (0.7% behind), the AMD Radeon RX 5600 OEM at 58085 (1.2% behind), and the AMD Radeon RX 6950 XT at 58392 (1.8% behind). The AMD card’s Geekbench score is nearly double these rivals, indicating it operates in a higher performance tier.

When comparing the two directly, the AMD card leads in the only head-to-head benchmark by 36.9%. The NVIDIA card’s bandwidth advantage does not translate into a win in this synthetic test. The AMD card also has a higher average benchmark score of 57367 versus 69000 for NVIDIA, though this average includes different benchmark sets. The percentile ranking shows NVIDIA at 90 versus AMD at 87, but this is based on all GPUs in the database, not just these two.

The Verdict

The AMD Radeon RX 9070 GRE is the superior choice for general compute and rendering workloads. Its 34.28 TFLOPS FP32 performance, 267.8 GPixel/s pixel rate, and 535.7 GTexel/s texture rate are all significantly higher than the NVIDIA CMP 90HX. The Geekbench OpenCL score of 109309 versus 69000 confirms the AMD card’s dominance in compute tasks. The AMD card also consumes less power at 220 W versus 320 W, making it more efficient per watt.

The NVIDIA CMP 90HX should be considered only for specific memory-bandwidth-bound scenarios. Its 760.3 GB/s bandwidth is 76% higher than AMD’s, and its 320-bit bus is wider. However, the card has no display outputs, is end-of-life, and carries a higher power draw. Its 21.89 TFLOPS FP32 performance is well below the AMD card, and its 136.8 GPixel/s pixel rate is roughly half.

For any user needing a functional graphics card with display support, the choice is clear: the AMD Radeon RX 9070 GRE wins on compute, efficiency, and usability. It is actively produced, has modern display outputs, and delivers nearly 57% higher FP32 throughput. The NVIDIA card was designed for mining and lacks display outputs, making it unsuitable for standard desktop use.

The data shows that the AMD Radeon RX 9070 GRE is the better overall product. It wins the only head-to-head benchmark by 36.9%, offers higher clocks, a more advanced process node, and lower power consumption. The NVIDIA CMP 90HX’s only advantages are memory bandwidth and a slightly higher percentile ranking, but these do not outweigh its compute deficit and lack of display functionality.

Buyers should choose the AMD Radeon RX 9070 GRE for modern gaming, compute, and content creation. The NVIDIA CMP 90HX is a niche product for specialized bandwidth-intensive tasks, and its end-of-life status further limits its appeal. The AMD card’s launch MSRP is 549 USD, and it delivers performance that exceeds the NVIDIA part across most measurable metrics.