AMD Radeon Pro Vega 16 vs NVIDIA GeForce RTX 2070 SUPER Comparison

The AMD Radeon Pro Vega 16 and the NVIDIA GeForce RTX 2070 SUPER are both end-of-life GPUs from the same general era, but the recorded data shows they occupy very different performance tiers. The RTX 2070 SUPER wins both shared head-to-head benchmarks by enormous margins, roughly 76 to 78 percent in compute-oriented tests. The Vega 16 counters with a leaner design: an integrated form factor, a 75 W TDP, and HBM2 memory on a wide bus. This page walks through the database entry for each card and explains where the numbers land.

FAQ

Q: Which GPU is faster in the shared benchmarks?

A: The RTX 2070 SUPER, decisively. It won both head-to-head tests: geekbench_opencl (83358 vs 18268) and geekbench_vulkan (90637 vs 21832). That corresponds to deltas of 78.1 percent and 75.9 percent respectively in favor of the NVIDIA card.

Q: How much raw compute does each card deliver?

A: The RTX 2070 SUPER delivers 9.062 TFLOPS FP32 and 18.12 TFLOPS FP16. The Radeon Pro Vega 16 delivers 2.437 TFLOPS FP32 and 4.874 TFLOPS FP16.

Q: Which card has more memory bandwidth?

A: The RTX 2070 SUPER, at 448.0 GB/s over a 256-bit GDDR6 bus. The Vega 16's HBM2 delivers 307.2 GB/s over a 1024-bit bus, which is respectable for its class but still well short of the NVIDIA card.

Q: Do both cards support ray tracing?

A: The RTX 2070 SUPER has dedicated hardware for it: 40 RT cores and 320 tensor cores. The Vega 16's database entry lists no RT cores or tensor cores.

Q: How power-hungry are these two?

A: The Radeon Pro Vega 16 carries a 75 W TDP in an integrated (IGP) form factor. The RTX 2070 SUPER carries a 215 W TDP, a dual-slot card requiring one 6-pin and one 8-pin power connector, with a suggested PSU of 550 W.

Q: Where does each card sit in the overall database rankings?

A: The Radeon Pro Vega 16 sits in the 68th percentile versus all GPUs; the RTX 2070 SUPER sits in the 65th percentile. Note that the head-to-head benchmark data and the average composite scores (23250 for the Vega 16, 20282 for the 2070 SUPER) are drawn from different test mixes, which is why the percentile ordering does not match the direct comparison results.

Architecture Differences

These two cards come from entirely different lineages. The Radeon Pro Vega 16 is built on the Vega 12 chip using AMD's GCN 5.0 architecture, manufactured on GlobalFoundries' 14 nm process. The RTX 2070 SUPER uses NVIDIA's TU104 chip on the Turing architecture, built at TSMC on a 12 nm process with 13,600 million transistors packed into a 545 mm² die at a density of 25.0M per mm². The Vega 16 entry does not record transistor count or die size.

The resource gap is large across the board. The Vega 16 fields 1024 shading units, 64 TMUs, and 32 ROPs. The 2070 SUPER quadruples several of those figures: 2560 shading units, 160 TMUs, and 64 ROPs. That resource difference flows directly into the throughput numbers: 113.3 GPixel/s and 283.2 GTexel/s for the NVIDIA card versus 38.08 GPixel/s and 76.16 GTexel/s for the AMD part, a roughly three-to-fourfold advantage in pixel and texture throughput.

Feature sets diverge as well. The 2070 SUPER includes 40 RT cores and 320 tensor cores, hardware the Vega 16 lacks entirely. API support reflects the gap: the NVIDIA card supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the Vega 16 supports DirectX 12 (12_1) and Vulkan 1.3. Both support OpenGL 4.6.

Clock behavior differs notably. The Vega 16 runs a base clock of 815 MHz with a boost of 1190 MHz, while the 2070 SUPER runs 1605 MHz base and 1770 MHz boost. Combined with its far larger resource counts, this is why the NVIDIA card's compute figures tower over the AMD card's.

Where Each One Wins

The RTX 2070 SUPER wins every direct comparison in the database. Its 8 GB of GDDR6, 448.0 GB/s of bandwidth, dedicated ray tracing and tensor hardware, and much larger shader array make it the clear pick for any workload that scales with raw GPU throughput: the GeekBench OpenCL and Vulkan results show compute advantages of roughly 76 to 78 percent. PassMark results reinforce the picture, with a g3d score of 18169, a gpu compute score of 7557, and DirectX 9 through DirectX 12 passmark results of 223, 132, 151, and 67 respectively, plus a g2d result of 878.

The Radeon Pro Vega 16 wins on physical and power efficiency grounds, not benchmark grounds. It is an IGP-class part with a 75 W TDP and no external power connectors, whereas the 2070 SUPER is a dual-slot discrete card measuring 267 mm long, 116 mm high, and 35 mm wide, needing both a 6-pin and an 8-pin connector. The Vega 16 also brings HBM2 memory on a 1024-bit bus, an unusual memory configuration at this power level. Its GeekBench Metal score of 29650 is the single recorded result where it stands alone, since the 2070 SUPER has no Metal entry in the database.

Specification Differences

• Manufacturer and chip: AMD Vega 12 (GCN 5.0, Radeon Pro Mac generation) vs NVIDIA TU104 (Turing, GeForce 20-series)
• Process node and foundry: 14 nm at GlobalFoundries vs 12 nm at TSMC
• Shading units: 1024 vs 2560
• Texture units: 64 vs 160
• ROPs: 32 vs 64
• RT / tensor cores: none listed vs 40 RT cores and 320 tensor cores
• Clocks: 815 MHz base, 1190 MHz boost vs 1605 MHz base, 1770 MHz boost
• Memory: 4 GB HBM2, 1024-bit, 307.2 GB/s at 2.4 Gbps effective vs 8 GB GDDR6, 256-bit, 448.0 GB/s at 14 Gbps effective
• Throughput: 38.08 GPixel/s and 76.16 GTexel/s vs 113.3 GPixel/s and 283.2 GTexel/s
• Compute: 2.437 TFLOPS FP32 and 4.874 TFLOPS FP16 vs 9.062 TFLOPS FP32 and 18.12 TFLOPS FP16
• Power and form factor: 75 W IGP, no power connectors vs 215 W dual-slot card with 1x 6-pin and 1x 8-pin, suggested PSU 550 W
• Display outputs: Portable Device Dependent vs 1x HDMI 2.0, 3x DisplayPort 1.4a, 1x USB Type-C
• DirectX / Vulkan: 12 (12_1) and 1.3 vs 12 Ultimate (12_2) and 1.4
• Release date: 2018-11-13 vs 2019-07-08
• Product line position: no predecessor or successor listed vs predecessor GeForce 10 and successor GeForce 30
• Launch MSRP: not recorded for the Vega 16; 499 USD launch MSRP for the RTX 2070 SUPER

Both cards share a PCIe 3.0 x16 bus interface, OpenGL 4.6 support, and end-of-life production status.

Head-to-Head Benchmarks

Only two tests appear in both cards' records, and the RTX 2070 SUPER takes both by landslide margins.

GeekBench OpenCL: 83358 for the RTX 2070 SUPER against 18268 for the Radeon Pro Vega 16. The delta of 78.1 percent means the NVIDIA card delivers more than four and a half times the OpenCL compute throughput. This aligns almost exactly with the FP32 spec gap (9.062 vs 2.437 TFLOPS), which indicates the benchmark is saturating each card's shader array without unusual bottlenecks.

GeekBench Vulkan: 90637 for the 2070 SUPER against 21832 for the Vega 16, a 75.9 percent gap. The Vulkan result is the Vega 16's stronger of the two showings, marginally closing the distance relative to OpenCL, but the gap remains overwhelming. The NVIDIA card's Vulkan score also exceeds its own OpenCL score, suggesting the card extracts additional throughput through the lower-overhead API, while the Vega 16's Vulkan result sits between its OpenCL (18268) and Metal (29650) scores.

For additional context from the 2070 SUPER's own record: it posts a 3DMark Steel Nomad DX12 score of 1651 and a PassMark g3d score of 18169. The Vega 16's average composite benchmark score of 23250 sits within a rounding error of its nearest rivals in the database, including the NVIDIA P106-100 (23249, a 0 percent delta), the Radeon RX 6600M (23273, -0.1 percent), the Radeon R9 M290X (23276, -0.1 percent), and the Radeon AI PRO R9700 (23315, -0.3 percent). The 2070 SUPER's composite average of 20282 lands close to the Quadro M4000M (20480, -1 percent), the RTX 3070 Mobile (20534, -1.2 percent), the Arc B570 (20556, -1.3 percent), and the Arc A750 (20582, -1.5 percent). These composite pools are built from different test sets, so they describe each card's neighborhood within the full database rather than a direct matchup between the two.

The Verdict

The data leaves no ambiguity about throughput. For any workload measured in the shared tests, the RTX 2070 SUPER is the pick: it wins both head-to-head benchmarks by roughly 76 to 78 percent, quadruples the Vega 16's FP32 compute, triples its pixel rate, nearly quadruples its texture rate, and doubles its memory capacity and shading resources. Its DirectX 12 Ultimate and Vulkan 1.4 support, plus dedicated RT and tensor cores, extend the lead into feature territory the Vega 16 does not occupy.

The Radeon Pro Vega 16's case rests on constraints, not scores. It fits a 75 W envelope with no power connectors in an integrated format, and it pairs that with HBM2 on a 1024-bit bus delivering 307.2 GB/s. Where power, space, or platform integration rules out a dual-slot 215 W card, the Vega 16 remains the only option of the two. It also holds the sole Metal result (29650) between them, relevant only for that API's ecosystem.

Choose the RTX 2070 SUPER when performance is the criterion. Choose the Radeon Pro Vega 16 when the deployment target demands its power and form factor, accepting the large performance gap the benchmarks document.