AMD Radeon HD 8970M vs NVIDIA GeForce RTX 2070 Comparison

Head-to-Head Benchmarks

The only direct comparison available in the database is a single Geekbench OpenCL run, and the result is decisive. The NVIDIA GeForce RTX 2070 scores 79,966 points, while the AMD Radeon HD 8970M manages 21,237 points. That is a delta of -73.4% for the AMD part, meaning the RTX 2070 delivers roughly 3.8 times the raw compute throughput in this workload. There is no ambiguity here; the RTX 2070 is the clear winner on the recorded metric.

Looking at broader database averages, the gap persists but narrows slightly. The RTX 2070 holds an average benchmark score of 18,789 across all its recorded tests, against the HD 8970M’s single-test average of 21,237. That comparison is skewed by the different test suites, but the Geekbench OpenCL result is the only true head-to-head, and it strongly favors the newer NVIDIA card. The HD 8970M does not win any recorded benchmark against the RTX 2070; the win count stands at 0 for AMD and 1 for NVIDIA.

The RTX 2070’s nearest rivals in the database include the NVIDIA Tesla K80 (-0.4% delta), the AMD Radeon Pro 5700 XT (+0.6% delta), and the AMD Radeon RX 560X (+0.9% delta), placing it in a tight cluster around 18,600 to 19,000 average points. The HD 8970M’s nearest rivals include the NVIDIA RTX A4000 Mobile (-0.7% delta) and the NVIDIA Quadro RTX 5000 (-1.8% delta), with the AMD Radeon RX Vega M GL (+0.4% delta) and NVIDIA GeForce RTX 5050 (+1% delta) also nearby. The HD 8970M’s percentile ranking is 66, slightly above the RTX 2070’s 63, which reflects the older card’s position among all GPUs in the database, not its performance against this specific competitor.

For compute-heavy tasks, the RTX 2070 also shows its strength in other recorded tests. Its Passmark GPU Compute score is 6,411, and its Geekbench Vulkan score is 82,521. The HD 8970M has no Vulkan or compute scores beyond the single OpenCL result, so the database offers no further direct comparison. The data that exists points to a massive performance advantage for the RTX 2070 in general-purpose compute, and that is the headline number.

The Verdict

The choice is straightforward based on the recorded data. Anyone needing raw OpenCL compute performance should pick the NVIDIA GeForce RTX 2070. Its score of 79,966 is 73.4% higher than the HD 8970M’s 21,237, and that is not a marginal difference; it is a generational leap. The RTX 2070 also brings a much higher average benchmark score across its test suite (18,789), although that figure mixes different workloads and should not be overinterpreted.

The AMD Radeon HD 8970M, however, holds a higher percentile ranking (66th versus 63rd), which indicates that relative to the entire database of GPUs, it sits in a slightly better position than the RTX 2070. That is likely because the database includes many low-end mobile and integrated parts, and the HD 8970M, despite being old, still outranks a substantial portion of them. But that percentile does not help it against the RTX 2070 specifically.

For gaming and DirectX workloads, the RTX 2070 has more evidence in the database. It records Passmark scores of 16,094 in G3D, 129 in DirectX 11, 111 in DirectX 10, 60 in DirectX 12, and 211 in DirectX 9. The HD 8970M has none of those results recorded, so any gaming comparison would be speculative. The only factual conclusion is that the RTX 2070 wins the one benchmark both cards share, and it does so by a wide margin. The verdict is for NVIDIA unless the workload is strictly tied to the HD 8970M’s specific niche, which the data does not define.

Architecture Differences

The two GPUs come from different architectural eras. The AMD Radeon HD 8970M is built on GCN 1.0, the original Graphics Core Next design, using the Neptune chip. The NVIDIA GeForce RTX 2070 is built on Turing, using the TU106 chip. The process nodes differ substantially: the HD 8970M uses a 28 nm process, while the RTX 2070 uses a 12 nm process, both fabricated by TSMC. That node shrink allows the RTX 2070 to pack far more transistors into a similar power envelope.

Transistor counts tell the story. The HD 8970M contains 2,800 million transistors on a 212 mm² die, giving a density of 13.2 million transistors per square millimeter. The RTX 2070 contains 10,800 million transistors on a 445 mm² die, with a density of 24.3 million per square millimeter. The RTX 2070 has nearly four times the transistor count and nearly double the die area, with almost double the density.

The RTX 2070 also introduces hardware features that the HD 8970M lacks entirely. It has 36 dedicated ray tracing cores and 288 tensor cores, which support DirectX 12 Ultimate (12_2) and accelerated AI workloads. The HD 8970M has no such units and only supports DirectX 12 (11_1). The Vulkan support also differs: the RTX 2070 reports version 1.4, while the HD 8970M reports 1.2.170. Both cards support OpenGL 4.6, so that is a point of parity.

The RTX 2070’s FP16 capability is listed at 14.93 TFLOPS (2:1), which indicates half-precision throughput is double its FP32 rate. The HD 8970M lists no FP16 figure, so the database does not confirm it has any half-precision acceleration. The FP32 numbers are also far apart: the RTX 2070 delivers 7.465 TFLOPS, while the HD 8970M delivers 2.304 TFLOPS.

Specification Differences

The memory subsystems are a clear differentiator. The RTX 2070 has 8 GB of GDDR6 memory running at 1750 MHz (14 Gbps effective), while the HD 8970M has 4 GB of GDDR5 at 1200 MHz (4.8 Gbps effective). Both use a 256-bit bus, but the RTX 2070’s bandwidth is 448.0 GB/s versus the HD 8970M’s 153.6 GB/s, a near threefold advantage.

Compute resources favor the NVIDIA card across the board. The RTX 2070 has 2,304 shading units, 144 texture mapping units, and 64 raster output units, compared to the HD 8970M’s 1,280 shading units, 80 TMUs, and 32 ROPs. Pixel rate is 103.7 GPixel/s on the RTX 2070 versus 28.80 GPixel/s on the HD 8970M. Texture rate is 233.3 GTexel/s versus 72.00 GTexel/s.

Clock speeds are higher on the RTX 2070: base 1410 MHz and boost 1620 MHz, versus the HD 8970M’s 850 MHz base and 900 MHz boost. Power consumption scales accordingly: the RTX 2070 has a TDP of 175 W, the HD 8970M a TDP of 100 W. The RTX 2070 is a dual-slot card with a 1x 8-pin power connector and a suggested PSU of 450 W, while the HD 8970M is an MXM module with no listed power connector or PSU suggestion. The RTX 2070’s physical dimensions are 229 mm in length, 113 mm in height, and 35 mm in width; the HD 8970M has no recorded dimensions.

The RTX 2070 offers multiple display outputs: 1x DVI, 1x HDMI 2.0, 2x DisplayPort 1.4a, and 1x USB Type-C. The HD 8970M’s outputs are listed as portable device dependent, meaning they vary by laptop implementation. Both cards use PCIe 3.0 x16.

Release dates are far apart: the HD 8970M launched on May 13, 2013, and the RTX 2070 on October 16, 2018. The RTX 2070 has a recorded launch MSRP of 499 USD. The HD 8970M has no MSRP recorded. Both are end-of-life products.

FAQ

Q: Which GPU wins the only shared benchmark?

A: The NVIDIA GeForce RTX 2070 wins the Geekbench OpenCL test with a score of 79,966, against the AMD Radeon HD 8970M’s 21,237, a delta of -73.4%.

Q: How much faster is the RTX 2070 in memory bandwidth?

A: The RTX 2070 has 448.0 GB/s of bandwidth, while the HD 8970M has 153.6 GB/s, making the RTX 2070 roughly 2.9 times faster.

Q: Does the HD 8970M have ray tracing or tensor cores?

A: No. The HD 8970M has no ray tracing cores and no tensor cores. The RTX 2070 has 36 ray tracing cores and 288 tensor cores.

Q: What is the transistor density difference?

A: The RTX 2070 has a density of 24.3 million transistors per square millimeter, while the HD 8970M has 13.2 million per square millimeter, on a 12 nm versus 28 nm process.

Q: Which card has a higher percentile ranking among all GPUs?

A: The HD 8970M ranks in the 66th percentile, while the RTX 2070 ranks in the 63rd percentile, despite the RTX 2070 winning the head-to-head benchmark.

Q: What is the memory size and type for each card?

A: The RTX 2070 has 8 GB of GDDR6, and the HD 8970M has 4 GB of GDDR5.

Where Each One Wins

The RTX 2070 wins in every recorded category that matters for modern workloads. It dominates the single head-to-head test, has a superior memory subsystem, higher compute throughput (FP32 at 7.465 TFLOPS versus 2.304 TFLOPS), and supports newer APIs including DirectX 12 Ultimate and Vulkan 1.4. For any task involving ray tracing, tensor operations, or high-bandwidth memory access, the RTX 2070 is the only choice between the two. Its Passmark G3D score of 16,094 also suggests it handles legacy DirectX workloads well, though the HD 8970M has no comparable score to challenge it.

The HD 8970M’s only advantage in the database is its percentile ranking. Sitting at the 66th percentile versus the RTX 2070’s 63rd means it outperforms a larger fraction of the GPUs in the database overall. That is a measure of relative standing, not absolute performance, and it likely reflects the fact that the HD 8970M, despite its age, still beats a large pool of weak integrated and entry-level parts. It also consumes less power at 100 W TDP versus 175 W, which could matter in a portable MXM context, though the database does not provide any efficiency benchmarks to confirm that trade-off.

For a practical recommendation: if the workload is confined to older DirectX 11 or OpenGL titles on a laptop with an MXM slot, the HD 8970M might still function acceptably, but the data does not show it winning any test against the RTX 2070. The RTX 2070 is the winner in compute, memory, and feature support. The HD 8970M wins only in the abstract sense of its percentile rank and lower power draw.