AMD Radeon Pro Vega II vs NVIDIA CMP 90HX Comparison

Head-to-Head Benchmarks

The recorded data contains a single direct comparison between the AMD Radeon Pro Vega II and the NVIDIA CMP 90HX: the Geekbench OpenCL test. In this benchmark, the AMD Radeon Pro Vega II posts a score of 99048, while the NVIDIA CMP 90HX reaches 69000. That represents a 43.5% advantage for the AMD part, a decisive margin in raw compute throughput as measured by OpenCL. The database shows the AMD Radeon Pro Vega II as the winner of this head-to-head matchup, with one win and no losses.

Looking beyond the direct comparison, the average benchmark score for the AMD Radeon Pro Vega II is 109617, placing it in the 94th percentile of all GPUs tracked in the database. The NVIDIA CMP 90HX, by contrast, has an average score of 69000 and sits in the 90th percentile. The percentile gap is smaller than the raw score gap might suggest, which indicates that the CMP 90HX is surrounded by a dense cluster of similarly performing cards near the top of the distribution, while the Vega II sits slightly higher with fewer cards above it.

The AMD Radeon Pro Vega II's nearest rivals in the database include the AMD Radeon PRO W7900, which averages 110725 and trails the Vega II by just 1%. The AMD Radeon Pro W6600X sits at 107342, which is 2.1% behind the Vega II, and the AMD Radeon Pro Vega II Duo scores 106750, a 2.7% deficit. On the other side, the NVIDIA RTX A5500 Mobile leads the Vega II by 3.8% with an average score of 113944. These figures show that the Vega II is competitive with modern workstation and mobile flagship parts, landing within a few percentage points of each.

For the NVIDIA CMP 90HX, the nearest rivals cluster much more tightly. The Intel Arc A770 scores 68809, a mere 0.3% behind the CMP 90HX. The AMD Radeon Instinct MI25 is 0.6% behind at 68562. The AMD Radeon Pro WX 8200 leads by 1.2% with a score of 69870, and the NVIDIA Quadro P6000 leads by 1.4% at 69986. The CMP 90HX is effectively in a dead heat with these four cards, separated by less than two percentage points in either direction. This tells a different story than the head-to-head result: while the CMP 90HX loses badly to the Vega II, it is squarely average among its own peer group.

The database also records other benchmark results for the AMD Radeon Pro Vega II that are not available for the NVIDIA CMP 90HX. In Geekbench Metal, the Vega II scores 130183, and in Geekbench Vulkan it scores 99621. These results indicate that the Vega II's compute performance is consistent across multiple APIs, though no comparable data exists for the CMP 90HX in those tests.

Where Each One Wins

The AMD Radeon Pro Vega II wins decisively in the only directly comparable workload, OpenCL compute, with a 43.5% lead. Its average benchmark score of 109617 versus 69000 reinforces that this is not a narrow victory but a substantial gap in general compute throughput. The Vega II also demonstrates strength across additional API tests, with a Metal score of 130183 and a Vulkan score of 99621, both of which are higher than its OpenCL result. This suggests the card maintains strong performance regardless of the software interface used to access it.

The NVIDIA CMP 90HX, however, has no recorded wins in the head-to-head data. Its strength lies not in beating the Vega II but in its competitive position within its own tier. The database shows the CMP 90HX essentially tied with the Intel Arc A770, AMD Radeon Instinct MI25, AMD Radeon Pro WX 8200, and NVIDIA Quadro P6000, all within 1.4% of each other. For a workload that is well-matched to this performance class, the CMP 90HX is a representative performer, but it cannot match the Vega II's raw compute output.

Use-case separation follows from these results. The AMD Radeon Pro Vega II is suited for compute-heavy tasks where OpenCL performance matters most, and its additional Metal and Vulkan scores indicate flexibility across different software stacks. The NVIDIA CMP 90HX, with no display outputs and a PCIe 1.0 x4 bus interface, is clearly not designed for general-purpose graphics or interactive workloads. Its role in the database is defined by its compute scores alone, and in that narrow scope it lands in the 90th percentile, a solid but not elite position.

Architecture Differences

The two cards come from different manufacturers, different design philosophies, and different process technologies. The AMD Radeon Pro Vega II uses the Vega 20 chip built on GCN 5.1 architecture, fabricated by TSMC on a 7 nm process. It contains 13,230 million transistors on a 331 mm² die, giving a transistor density of 40.0 million per square millimeter. The NVIDIA CMP 90HX uses the GA102 chip with Ampere architecture, fabricated by Samsung on an 8 nm process. It packs 28,300 million transistors onto a 628 mm² die, for a transistor density of 45.1 million per square millimeter. The NVIDIA chip is more than twice the physical size and carries more than double the transistor count, while the AMD chip achieves a smaller die through the denser 7 nm process.

The memory architectures are fundamentally different. The Vega II uses 32 GB of HBM2 on a 4096-bit bus, delivering 825.3 GB/s of bandwidth. The CMP 90HX uses 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s. Despite the much wider bus and larger capacity on the AMD side, the bandwidth difference is modest: the Vega II leads by roughly 8.5%, which is small compared to the massive difference in memory capacity and bus width.

Compute resources also differ significantly. The Vega II has 4096 shading units, 256 texture mapping units, and 64 ROPs. The CMP 90HX has 6400 shading units, 200 TMUs, and 80 ROPs. The NVIDIA card has 50 RT cores and 200 tensor cores, while the AMD card has none recorded for either. The Vega II's pixel rate is 110.1 GPixel/s, lower than the CMP 90HX's 136.8 GPixel/s, but its texture rate of 440.3 GTexel/s is higher than the CMP 90HX's 342.0 GTexel/s.

Floating-point performance tells a more complex story. The Vega II delivers 14.09 TFLOPS of FP32 and 28.18 TFLOPS of FP16, with the FP16 figure achieved at a 2:1 ratio. The CMP 90HX delivers 21.89 TFLOPS of FP32 and 21.89 TFLOPS of FP16, at a 1:1 ratio. In FP32, the NVIDIA card leads by a significant margin, but in FP16 the AMD card leads. The Vega II's higher texture rate and FP16 throughput suggest it was designed with compute workloads that benefit from packed math, while the CMP 90HX favors conventional FP32 throughput.

API support also differs. The Vega II supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.3. The CMP 90HX supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Both cards are end-of-life in production status, with the Vega II released in 2019 and the CMP 90HX in 2021.

Specification Differences

The two cards differ in nearly every measurable specification. The Vega II has a base clock of 1574 MHz and a boost clock of 1720 MHz, while the CMP 90HX has a base clock of 1500 MHz and a boost clock of 1710 MHz. The AMD card runs slightly faster in both metrics. Memory clocks differ as well: the Vega II's memory runs at 806 MHz with 1612 Mbps effective, while the CMP 90HX's memory runs at 1188 MHz with 19 Gbps effective.

Memory capacity is a major differentiator: 32 GB for the Vega II versus 10 GB for the CMP 90HX. Memory type, bus width, and bandwidth all follow the same pattern: HBM2 with a 4096-bit bus and 825.3 GB/s for AMD, GDDR6X with a 320-bit bus and 760.3 GB/s for NVIDIA. The Vega II has 4096 shading units, 256 TMUs, and 64 ROPs, while the CMP 90HX has 6400 shading units, 200 TMUs, and 80 ROPs. The CMP 90HX also has 50 RT cores and 200 tensor cores, which the Vega II lacks entirely.

Power and physical specifications diverge sharply. The Vega II has a TDP of 475 W and requires a suggested 850 W power supply, while the CMP 90HX has a TDP of 320 W and a suggested 700 W PSU. The Vega II occupies a quad-slot form factor, while the CMP 90HX is dual-slot. The CMP 90HX uses two 8-pin power connectors, while no power connector data is recorded for the Vega II. The Vega II uses an Apple MPX bus interface and has display outputs (1x HDMI 2.0b and 4x Thunderbolt), while the CMP 90HX uses PCIe 1.0 x4 and has no display outputs. The CMP 90HX is 285 mm long and 112 mm tall; the Vega II has no recorded dimensions. The Vega II had a launch MSRP of 2,199 USD; no launch MSRP is recorded for the CMP 90HX.

FAQ

Q: Which card wins in OpenCL performance?

A: The AMD Radeon Pro Vega II scores 99048 in Geekbench OpenCL, which is 43.5% higher than the NVIDIA CMP 90HX's score of 69000.

Q: How much memory does each card have?

A: The AMD Radeon Pro Vega II has 32 GB of HBM2 on a 4096-bit bus, while the NVIDIA CMP 90HX has 10 GB of GDDR6X on a 320-bit bus.

Q: Which card has higher FP32 throughput?

A: The NVIDIA CMP 90HX delivers 21.89 TFLOPS of FP32, compared to 14.09 TFLOPS for the AMD Radeon Pro Vega II.

Q: Which card has higher FP16 throughput?

A: The AMD Radeon Pro Vega II delivers 28.18 TFLOPS of FP16 at a 2:1 ratio, while the NVIDIA CMP 90HX delivers 21.89 TFLOPS at a 1:1 ratio.

Q: What is the power draw of each card?

A: The AMD Radeon Pro Vega II has a TDP of 475 W and a suggested 850 W power supply, while the NVIDIA CMP 90HX has a TDP of 320 W and a suggested 700 W power supply.

Q: Do both cards have display outputs?

A: No. The AMD Radeon Pro Vega II has 1x HDMI 2.0b and 4x Thunderbolt outputs, while the NVIDIA CMP 90HX has no display outputs.

The Verdict

The AMD Radeon Pro Vega II is the stronger card in the only direct benchmark comparison available. Its 43.5% lead in OpenCL, higher average score, and 94th percentile ranking versus the CMP 90HX's 90th percentile make it the clear choice for compute workloads that run on OpenCL, Metal, or Vulkan. Its 32 GB of HBM2 memory, wider 4096-bit bus, and higher FP16 throughput further support its position as a compute-oriented workstation card. The Vega II also offers display outputs, which the CMP 90HX lacks entirely, making it the only one of the two that can drive a monitor.

The NVIDIA CMP 90HX is not without merit, but its strengths do not translate into a win against the Vega II. It delivers higher FP32 throughput at 21.89 TFLOPS and has a higher pixel rate at 136.8 GPixel/s, along with RT cores and tensor cores that the Vega II does not have. It is also more power-efficient on paper, with a 320 W TDP versus 475 W, and it occupies only a dual-slot form factor. However, its 10 GB memory capacity, narrower 320-bit bus, and lack of display outputs limit its usefulness to a narrow set of compute tasks. Its nearest rivals in the database are all within 1.4% of its average score, which means it is a typical performer in its class rather than a standout.

For users who need a card for general compute with strong API coverage, large memory capacity, and display connectivity, the data points to the AMD Radeon Pro Vega II. For users who prioritize FP32 throughput and do not need display output or large memory capacity, the NVIDIA CMP 90HX offers a competitive option within its own performance tier, but it cannot match the Vega II in the direct comparison. The benchmark results are unambiguous: the Vega II wins the head-to-head, and the CMP 90HX settles for a respectable but secondary position.