GPU Comparison

NVIDIA CMP 70HX and AMD Radeon RX 470 are two very different products from different eras of GPU design, yet they land in a similar performance bracket in the database. The CMP 70HX is a mining-specific Ampere part with no display outputs, while the RX 470 is a 2016-era GCN card designed for gaming. Their average benchmark scores are close, 30,476 for the NVIDIA card versus 28,996 for the AMD card, a gap of about 5.1%. However, the head-to-head results tell a more nuanced story, with the older AMD card winning both shared tests by significant margins. The data reveals a fascinating clash between modern compute density and older, broader driver support.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA CMP 70HX has a higher average benchmark score at 30,476, placing it in the 75th percentile of all GPUs. The AMD Radeon RX 470 scores 28,996 on average, which puts it in the 74th percentile.

Q: How do the two cards compare in the Geekbench OpenCL test?

A: The AMD Radeon RX 470 wins decisively with a score of 33,568 versus the NVIDIA CMP 70HX's 25,135. That is a delta of -25.1% for the NVIDIA card, meaning the RX 470 is roughly 33.5% faster in this specific workload.

Q: Which card has more shading units?

A: The NVIDIA CMP 70HX has 3,840 shading units, while the AMD Radeon RX 470 has 2,048. Despite having nearly double the shading units, the CMP 70HX loses in both shared benchmark tests.

Q: What is the memory configuration difference?

A: The NVIDIA CMP 70HX features 8 GB of GDDR6X memory on a 256-bit bus with 608.3 GB/s bandwidth. The AMD Radeon RX 470 has 4 GB of GDDR5 memory on a 256-bit bus with 211.2 GB/s bandwidth.

Q: Does the AMD Radeon RX 470 support newer graphics APIs?

A: No, the RX 470 supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. The NVIDIA CMP 70HX supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, which are newer API versions.

Q: What is the power requirement for each card?

A: The AMD Radeon RX 470 has a TDP of 120 W and a suggested PSU of 300 W with a single 6-pin connector. The NVIDIA CMP 70HX has no listed TDP but suggests a 200 W PSU and uses a single 12-pin connector.

Architecture Differences

The architectural gap between these two GPUs is substantial. The NVIDIA CMP 70HX is built on the Ampere architecture using the GA104 chip, fabricated on an 8 nm process at Samsung. It packs 17,400 million transistors into a 392 mm² die, yielding a transistor density of 44.4 million per square millimeter. The AMD Radeon RX 470 uses the older GCN 4.0 architecture with the Ellesmere chip, built on a 14 nm process at GlobalFoundries. It contains 5,700 million transistors on a 232 mm² die, with a density of 24.6 million per square millimeter.

The CMP 70HX includes modern feature sets like 30 ray tracing cores and 120 tensor cores, which are completely absent from the RX 470. The NVIDIA card also has more texture mapping units (120 versus 128 for AMD) and double the render output units (64 versus 32). Memory technology differs sharply: the CMP 70HX uses 8 GB of GDDR6X with 608.3 GB/s bandwidth, while the RX 470 uses 4 GB of GDDR5 with 211.2 GB/s bandwidth. The bus widths are identical at 256-bit, but the effective memory clock is much higher on the NVIDIA part, 19 Gbps versus 6.6 Gbps.

The RX 470 retains display outputs (1x HDMI 2.0b and 3x DisplayPort 1.4a), while the CMP 70HX has no display outputs at all, reflecting its mining-only purpose. The bus interface also differs: the CMP 70HX uses PCIe 1.0 x4, a legacy interface that severely limits bandwidth for general compute, while the RX 470 uses PCIe 3.0 x16. This interface difference likely explains some of the benchmark anomalies, as the older card has a much faster connection to the host system.

Head-to-Head Benchmarks

The head-to-head comparison is limited to two tests, and the AMD Radeon RX 470 wins both. In Geekbench OpenCL, the RX 470 scores 33,568 against the CMP 70HX's 25,135. The deltaPct is -25.1% for the NVIDIA card, which means the AMD card outperforms it by about one-third. This is a substantial margin, especially considering the CMP 70HX has 3,840 shading units versus 2,048 on the RX 470, and a significantly higher FP32 throughput of 10.71 TFLOPS versus 4.940 TFLOPS.

In Geekbench Vulkan, the RX 470 again takes the win with a score of 39,884 versus 35,817 for the CMP 70HX. The deltaPct here is -10.2% for the NVIDIA card. While not as lopsided as the OpenCL result, this still shows the older AMD architecture outperforming the newer Ampere chip in this specific workload. The CMP 70HX's lack of display outputs and its PCIe 1.0 x4 interface may hinder its ability to fully utilize its compute resources in these synthetic tests.

The wins tally is 0 for the NVIDIA CMP 70HX and 2 for the AMD Radeon RX 470. However, it is notably the RX 470 has additional benchmark entries in the database, a 3DMark Steel Nomad DX12 score of 842 and a Geekbench Metal score of 41,690, which the CMP 70HX does not have. The average benchmark score still favors the NVIDIA part, suggesting that the CMP 70HX performs better in other untested workloads or that the RX 470's additional Metal score helps balance its average.

Specification Differences

The specification gap between the two cards is stark. The NVIDIA CMP 70HX uses an 8 nm process, while the AMD Radeon RX 470 uses 14 nm. Transistor count is 17,400 million versus 5,700 million, and die size is 392 mm² versus 232 mm². Transistor density is 44.4M per mm² for NVIDIA versus 24.6M per mm² for AMD.

Clock speeds differ significantly. The CMP 70HX has a base clock of 1365 MHz and a boost clock of 1395 MHz. The RX 470 is much lower, with 926 MHz base and 1206 MHz boost. Memory clocks also vary widely: 1188 MHz (19 Gbps effective) for the NVIDIA card versus 1650 MHz (6.6 Gbps effective) for the AMD card.

The CMP 70HX has more of nearly everything: 3,840 shading units versus 2,048, 120 TMUs versus 128 (AMD actually has more here), 64 ROPs versus 32, and it adds 30 RT cores and 120 tensor cores that the RX 470 lacks. Pixel rate is 89.28 GPixel/s versus 38.59 GPixel/s, and texture rate is 167.4 GTexel/s versus 154.4 GTexel/s. FP32 compute is 10.71 TFLOPS versus 4.940 TFLOPS.

Memory capacity is 8 GB versus 4 GB, with GDDR6X versus GDDR5. The TDP for the RX 470 is 120 W, while the CMP 70HX lists no TDP but suggests a 200 W PSU versus 300 W for the AMD card. Power connectors are 1x 12-pin versus 1x 6-pin. The CMP 70HX has no display outputs, while the RX 470 has 1x HDMI 2.0b and 3x DisplayPort 1.4a. The CMP 70HX is longer at 267 mm versus 240 mm. The RX 470 has a launch MSRP of 179 USD.

The Verdict

The benchmark data presents a confusing picture. The NVIDIA CMP 70HX has a higher average score (30,476 versus 28,996) and sits in a higher percentile (75th versus 74th). Its nearest rivals include the AMD Radeon RX 6700 with a deltaPct of 0.1%, the AMD Radeon RX 6800 at 1.3%, and the NVIDIA GeForce RTX 3070 Ti at 1.8%. The AMD Radeon RX 470's nearest rivals include the AMD Radeon RX 6800M at 0.4%, Intel Arc A370M at -0.6%, AMD Radeon RX Vega M GH at -0.7%, and AMD FirePro W8000 at -0.7%.

However, in the direct head-to-head tests, the RX 470 wins both. This contradiction suggests the CMP 70HX may excel in workloads not represented by the two shared tests, but the data we have shows the RX 470 as the better performer in OpenCL and Vulkan. The CMP 70HX's PCIe 1.0 x4 interface is a severe limitation, potentially bottlenecking data transfer despite its superior compute hardware. The RX 470's PCIe 3.0 x16 interface gives it a massive connectivity advantage.

For practical use, the RX 470 is the only one with display outputs, making it usable for gaming or general desktop tasks. The CMP 70HX is a mining card with no video outputs, making it useless for anything other than compute or mining. The RX 470 also has a much lower power draw at 120 W TDP, while the CMP 70HX's power consumption is unlisted but its PSU suggestion of 200 W is lower than the RX 470's 300 W recommendation.

Where Each One Wins

The AMD Radeon RX 470 wins in the two benchmarks where both cards are tested. In Geekbench OpenCL, it is 25.1% ahead of the CMP 70HX. In Geekbench Vulkan, it is 10.2% ahead. It also has the advantage of display outputs, a PCIe 3.0 x16 interface, and a lower TDP of 120 W. The RX 470 additionally has a Geekbench Metal score of 41,690, which is the highest individual score for either card in any shared or exclusive benchmark.

The NVIDIA CMP 70HX wins on raw specifications. It has more shading units, more ROPs, higher clock speeds, more memory, faster memory bandwidth, and higher compute throughput. Its 8 GB of GDDR6X memory offers 608.3 GB/s bandwidth, nearly three times the RX 470's 211.2 GB/s. It also supports newer API versions, including DirectX 12 Ultimate and Vulkan 1.4. The transistor density of 44.4M per mm² versus 24.6M per mm² reflects a much more advanced manufacturing process.

The CMP 70HX also has a higher average benchmark score overall, suggesting it performs better in aggregate across all tests in the database. Its nearest rivals include the RX 6700 and RX 6800, which are both newer and more powerful AMD cards, indicating the CMP 70HX punches above its mining-oriented design. The RX 470's nearest rivals include the Intel Arc A370M and RX Vega M GH, which are lower-power mobile parts, suggesting the RX 470 competes in a lower performance tier despite its head-to-head wins.