Intel Arc B570 vs NVIDIA GeForce RTX 2070 SUPER Comparison

The Intel Arc B570 and the NVIDIA GeForce RTX 2070 SUPER occupy the same performance tier, with average benchmark scores of 20556 and 20282 respectively. The data reveals a generational split: the Battlemage GPU wins modern API and compute workloads, while the Turing card retains dominance in legacy DirectX and 2D tests. This comparison pits a 5 nm TSMC design with 19,600 million transistors against a 12 nm part with 13,600 million, but raw specifications only tell part of the story.

Where Each One Wins

The Arc B570 secures its victories in the categories that define modern graphics workloads. Its most decisive win comes in 3DMark Steel Nomad DX12, where it scores 2649 against the RTX 2070 SUPER’s 1651 — a commanding 60.4% advantage. This is not a marginal lead but a generational gap in DirectX 12 rasterization performance. The Intel GPU also wins Geekbench Vulkan by 6.8% (96844 vs 90637) and Geekbench OpenCL by a slim 0.2% (83514 vs 83358), indicating comparable compute throughput with a slight edge for the newer architecture. In PassMark DirectX 12, the B570 leads by 7.5% (72 vs 67), though both scores are low in absolute terms.

The RTX 2070 SUPER dominates the legacy API benchmarks, which is expected given its Turing architecture’s maturity with older DirectX versions. Its biggest win is PassMark DirectX 10, where it scores 132 against the B570’s 65 — a 50.8% advantage. The NVIDIA card also leads PassMark DirectX 11 by 21.9% (151 vs 118) and PassMark DirectX 9 by 26.5% (223 vs 164). These are substantial margins in titles that rely on older rendering paths. The RTX 2070 SUPER further wins PassMark G3D by 21.9% (18169 vs 14195) and PassMark G2D by 24.7% (878 vs 661), showing stronger overall 3D performance in the PassMark suite and better 2D throughput. PassMark GPU Compute goes to NVIDIA by 3.7% (7557 vs 7281).

The Verdict

The data supports a clear split based on workload type. For users running modern DirectX 12 or Vulkan titles, the Intel Arc B570 is the superior choice. Its 60.4% lead in 3DMark Steel Nomad and 6.8% lead in Geekbench Vulkan demonstrate that the Xe2-HPG architecture delivers substantially better performance in current-generation rendering APIs. The B570 also edges out the RTX 2070 SUPER in OpenCL compute (0.2% lead), making it a viable option for compute tasks that leverage this API.

For users with legacy software libraries or workloads that depend on DirectX 9, 10, or 11, the RTX 2070 SUPER remains the stronger pick. Its 50.8% lead in DirectX 10 and 26.5% lead in DirectX 9 show that older APIs still favor the Turing architecture. The NVIDIA card also wins PassMark G3D by 21.9%, which aggregates several DirectX tests, and its 3.7% lead in GPU compute is noteworthy for compute-heavy applications that use the PassMark benchmark.

Both cards sit at the 65th percentile vs all GPUs, confirming they are performance peers in the broader market. The choice hinges entirely on API direction: forward-looking workloads favor Intel, backward compatibility favors NVIDIA. The RTX 2070 SUPER is end-of-life with a successor in the GeForce 30 series, while the B570 is active production, suggesting Intel’s architecture is the one with ongoing driver support and future optimization headroom.

Head-to-Head Benchmarks

The most significant delta in the entire comparison is 3DMark Steel Nomad DX12, where the Arc B570 scores 2649 versus 1651 for the RTX 2070 SUPER. This 60.4% margin is the largest of any test and represents the clearest statement of architectural advantage in modern rendering. The Intel GPU’s 18 ray tracing cores and Xe2-HPG design appear to handle DirectX 12 workloads with far greater efficiency than the Turing architecture’s 40 RT cores and 320 tensor cores.

Geekbench Vulkan shows the B570 ahead by 6.8% (96844 vs 90637), reinforcing the pattern that Intel’s newer architecture excels in contemporary APIs. The OpenCL result is nearly a tie at 0.2% (83514 vs 83358), indicating that raw compute throughput is essentially equivalent despite the B570’s lower FP32 rating of 11.52 TFLOPS versus 9.062 TFLOPS for the RTX 2070 SUPER.

The RTX 2070 SUPER’s largest victory is PassMark DirectX 10, where it scores 132 against the B570’s 65. This 50.8% lead is nearly as decisive as Intel’s Steel Nomad win, but in the opposite direction — legacy API performance clearly favors the Turing card. PassMark DirectX 9 shows a 26.5% NVIDIA lead (223 vs 164), and DirectX 11 shows a 21.9% lead (151 vs 118). These results suggest that Intel’s driver stack or hardware design does not prioritize older DirectX paths, while NVIDIA’s long history with these APIs pays dividends.

PassMark G3D, which aggregates 3D performance across multiple tests, favors the RTX 2070 SUPER by 21.9% (18169 vs 14195). This is a substantial overall 3D score difference that contradicts the Steel Nomad result, highlighting how benchmark selection dramatically influences conclusions. The PassMark G2D test also goes to NVIDIA by 24.7% (878 vs 661), indicating better 2D acceleration for desktop and productivity tasks. PassMark GPU Compute is the closest NVIDIA win at 3.7% (7557 vs 7281), a minor edge in compute workloads.

FAQ

Q: Which GPU is faster in DirectX 12?

A: The Intel Arc B570 wins PassMark DirectX 12 by 7.5% (72 vs 67) and dominates 3DMark Steel Nomad DX12 by 60.4% (2649 vs 1651), indicating a substantial lead in modern DirectX 12 rendering.

Q: Does the RTX 2070 SUPER have any advantages in modern benchmarks?

A: No. The RTX 2070 SUPER loses all modern API tests (Steel Nomad, Vulkan, OpenCL, DirectX 12) to the Arc B570. Its wins are confined to legacy DirectX 9, 10, 11, and PassMark-specific G2D/G3D/compute tests.

Q: How do their average benchmark scores compare?

A: The Arc B570 has an average benchmark score of 20556, while the RTX 2070 SUPER scores 20282. The B570 leads by 1.3% in average score. Both GPUs sit at the 65th percentile vs all GPUs.

Q: Which card has more memory bandwidth?

A: The RTX 2070 SUPER has 448.0 GB/s bandwidth from its 256-bit bus and 8 GB GDDR6, while the Arc B570 has 380.0 GB/s from a 160-bit bus and 10 GB GDDR6. NVIDIA’s wider bus provides 17.9% more bandwidth.

Q: What is the transistor density difference?

A: The Intel Arc B570 packs 72.1 million transistors per mm² on a 272 mm² die with 19,600 million total transistors. The RTX 2070 SUPER has 25.0 million per mm² on a much larger 545 mm² die with 13,600 million transistors. Intel’s 5 nm process enables nearly 3x higher density.

Q: Does the RTX 2070 SUPER support newer display standards?

A: No. The RTX 2070 SUPER offers 1x HDMI 2.0 and 3x DisplayPort 1.4a, plus 1x USB Type-C. The Arc B570 provides 1x HDMI 2.1a and 3x DisplayPort 2.1, supporting newer display technologies.

Architecture Differences

The Arc B570 is built on Intel’s Xe2-HPG architecture using a 5 nm TSMC process, while the RTX 2070 SUPER uses NVIDIA’s Turing architecture on a 12 nm TSMC node. This process advantage is stark: the B570’s die measures 272 mm² with 19,600 million transistors, achieving a density of 72.1 million transistors per mm². The RTX 2070 SUPER’s die is 545 mm² with 13,600 million transistors, yielding just 25.0 million per mm². The B570 delivers more transistors on half the silicon area.

Core configurations differ substantially. The B570 has 2304 shading units, 144 TMUs, and 80 ROPs, while the RTX 2070 SUPER has 2560 shading units, 160 TMUs, and 64 ROPs. NVIDIA’s card has more shading units and TMUs, but fewer ROPs. The B570 compensates with higher clock speeds — both base and boost at 2500 MHz versus 1605/1770 MHz for the RTX 2070 SUPER. This clock advantage yields a higher pixel rate of 200.0 GPixel/s versus 113.3 GPixel/s, and a higher texture rate of 360.0 GTexel/s versus 283.2 GTexel/s.

Ray tracing and tensor hardware differ in approach. The B570 has 18 RT cores, while the RTX 2070 SUPER has 40 RT cores and 320 tensor cores. The B570 does not list tensor cores, instead relying on Xe2-HPG’s integrated AI capabilities. FP32 throughput favors Intel at 11.52 TFLOPS versus 9.062 TFLOPS, and FP16 similarly favors Intel at 23.04 TFLOPS versus 18.12 TFLOPS (both at 2:1 ratio).

Memory subsystems favor NVIDIA in bandwidth but Intel in capacity. The RTX 2070 SUPER uses 8 GB GDDR6 on a 256-bit bus for 448.0 GB/s bandwidth. The B570 has 10 GB GDDR6 on a 160-bit bus for 380.0 GB/s. The B570’s memory runs at 2375 MHz (19 Gbps effective) versus 1750 MHz (14 Gbps effective) for NVIDIA, but the wider bus gives NVIDIA the bandwidth advantage.

Power and interface differences are notable. The B570 has a 150 W TDP requiring a 450 W PSU and a single 8-pin connector, while the RTX 2070 SUPER draws 215 W needing a 550 W PSU and both 6-pin and 8-pin connectors. The B570 uses PCIe 4.0 x8, while the RTX 2070 SUPER uses PCIe 3.0 x16. Display outputs also differ: Intel offers HDMI 2.1a and DisplayPort 2.1, while NVIDIA provides HDMI 2.0 and DisplayPort 1.4a with an additional USB Type-C port. Both support DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. The B570 is in active production with a 2025 release, while the RTX 2070 SUPER is end-of-life from 2019 with the GeForce 30 series as its successor.