AMD Radeon HD 7790 vs NVIDIA GeForce RTX 5090 D V2 Comparison

FAQ

Q: How does the AMD Radeon HD 7790 compare to the NVIDIA GeForce RTX 5090 D V2 in benchmark scores?

A: The two cards have nearly identical average benchmark scores, but they were tested in completely different workloads. The HD 7790 scores 17,666 in Geekbench Metal, while the RTX 5090 D V2 scores 16,504 in 3DMark Steel Nomad DX12. These tests measure different aspects of GPU performance, so a direct cross-test comparison is not meaningful.

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

A: The HD 7790 sits at the 61st percentile, while the RTX 5090 D V2 is at the 59th percentile. Despite the massive generational and architectural gap, the HD 7790 ranks slightly higher in its benchmark pool, reflecting how different test suites weight various GPU capabilities.

Q: What are the nearest rivals for each card based on average scores?

A: For the HD 7790, the nearest rivals are the NVIDIA GeForce RTX 4060 (0.2% ahead), AMD Radeon 780M (0.4% ahead), AMD Radeon Pro 560 (0.7% ahead), and NVIDIA Quadro RTX 4000 (0.7% behind). For the RTX 5090 D V2, the closest competitors are the NVIDIA T400 (0% difference), AMD Radeon PRO W7500 (0.5% ahead), NVIDIA RTX PRO 6000 Blackwell (0.6% ahead), and AMD Radeon RX 5700 XT (0.9% ahead).

Q: What is the transistor density difference between the two GPUs?

A: The RTX 5090 D V2 packs 122.9 million transistors per square millimeter on a 5 nm process, compared to the HD 7790's 13.0 million per square millimeter on a 28 nm process. That is roughly a 9.5x density advantage for the newer card.

Q: Which card has more memory bandwidth?

A: The RTX 5090 D V2 offers 1.34 TB/s of bandwidth from its 24 GB GDDR7 memory on a 384-bit bus. The HD 7790 provides 96.00 GB/s from 1 GB of GDDR5 on a 128-bit bus — a difference of over 13,900% in raw bandwidth.

Q: What is the production status of each card?

A: The HD 7790 is end-of-life, having launched on March 21, 2013, with its successor being Sea Islands. The RTX 5090 D V2 is active production, launched on August 14, 2025, following the GeForce 40 series and preceding the GeForce 60 series.

Architecture Differences

The architectural gap between these two GPUs is enormous, spanning over a decade of GPU design evolution. The HD 7790 uses the Bonaire chip based on GCN 2.0 architecture, belonging to the Southern Islands generation (HD 7700 family). In contrast, the RTX 5090 D V2 employs the GB202 chip built on Blackwell 2.0 architecture from the GeForce 50 series.

Manufacturing technology represents one of the most significant divides. The HD 7790 is fabricated on a 28 nm process at TSMC, containing 2,080 million transistors on a 160 mm² die. The RTX 5090 D V2 moves to a 5 nm process — also TSMC — with 92,200 million transistors on a 750 mm² die. This represents a 44x increase in transistor count and a 4.7x increase in die area, enabled by the transition to a much finer process node.

Compute resources differ dramatically in both quantity and type. The HD 7790 features 896 shading units, 56 texture mapping units, and 16 ROPs. The RTX 5090 D V2 scales this up to 21,760 shading units, 680 TMUs, and 176 ROPs — increases of 24x, 12x, and 11x respectively. Critically, the RTX 5090 D V2 also introduces dedicated hardware that the HD 7790 completely lacks: 170 ray tracing cores and 680 tensor cores, enabling hardware-accelerated ray tracing and AI processing that was nonexistent in the GCN 2.0 era.

Memory architecture reflects the generational leap. The HD 7790 uses 1 GB of GDDR5 across a 128-bit bus, while the RTX 5090 D V2 uses 24 GB of GDDR7 across a 384-bit bus. The memory clock speeds similarly advanced from 1500 MHz (6 Gbps effective) to 1750 MHz (28 Gbps effective), contributing to the massive bandwidth disparity.

Rendering capabilities show a similar progression. The HD 7790 achieves 16.00 GPixel/s pixel fill rate and 56.00 GTexel/s texture fill rate, with FP32 compute of 1.792 TFLOPS. The RTX 5090 D V2 delivers 423.6 GPixel/s, 1,636.8 GTexel/s, and 104.8 TFLOPS FP32 — improvements of 26x, 29x, and 58x respectively. The RTX 5090 D V2 also offers FP16 at 104.8 TFLOPS (1:1 ratio with FP32), a capability the HD 7790 does not expose.

API support reflects the newer card's modern feature set. The HD 7790 supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The RTX 5090 D V2 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 — the higher DirectX feature level and newer Vulkan version enabling advanced rendering techniques unavailable to the older architecture.

Where Each One Wins

The HD 7790's advantage lies in its efficiency profile and the specific benchmark in which it was tested. Its Geekbench Metal score of 17,666 places it at the 61st percentile, and it matches or slightly exceeds several much newer GPUs in that particular test. The data shows the HD 7790 outperforms the NVIDIA Quadro RTX 4000 by 0.7% and comes within 0.2% of the RTX 4060 in the Metal workload. This suggests that for legacy applications, particularly those leveraging Apple's Metal API, the older card remains surprisingly competitive.

The RTX 5090 D V2 wins decisively in virtually every raw specification category. Its 104.8 TFLOPS FP32 compute, 1.34 TB/s memory bandwidth, and 423.6 GPixel/s pixel rate place it in an entirely different performance class. The inclusion of 170 RT cores and 680 tensor cores makes it the clear choice for any workload involving ray tracing, AI inference, or machine learning. Its 24 GB of GDDR7 memory dwarfs the HD 7790's 1 GB allocation, enabling workloads that would be impossible on the older card.

Benchmark results indicate the RTX 5090 D V2's nearest rivals in its test pool are the NVIDIA T400 (0% delta), AMD Radeon PRO W7500 (0.5% ahead), and RTX PRO 6000 Blackwell (0.6% ahead). The HD 7790's rival set includes the RTX 4060 and Radeon 780M, showing both cards compete within their respective performance tiers. For modern gaming at high resolutions, the RTX 5090 D V2 is the only viable option — the HD 7790's 1 GB frame buffer and lack of dedicated ray tracing hardware make it unsuitable for contemporary titles.

Specification Differences

The two cards differ in nearly every measurable specification. The HD 7790 uses a 28 nm process with 2,080 million transistors on a 160 mm² die, while the RTX 5090 D V2 uses 5 nm with 92,200 million transistors on a 750 mm² die. Transistor density increases from 13.0M per mm² to 122.9M per mm².

Memory capacity differs by 23 GB: the HD 7790 has 1 GB GDDR5, the RTX 5090 D V2 has 24 GB GDDR7. Bus width expands from 128-bit to 384-bit, and bandwidth grows from 96.00 GB/s to 1.34 TB/s. Memory clock rises from 1500 MHz (6 Gbps effective) to 1750 MHz (28 Gbps effective).

Compute resources scale dramatically: shading units from 896 to 21,760, TMUs from 56 to 680, ROPs from 16 to 176. The RTX 5090 D V2 adds 170 RT cores and 680 tensor cores; the HD 7790 has none. Pixel rate increases from 16.00 to 423.6 GPixel/s, texture rate from 56.00 to 1,636.8 GTexel/s, and FP32 from 1.792 to 104.8 TFLOPS. The RTX 5090 D V2 also lists FP16 at 104.8 TFLOPS; the HD 7790 lists no FP16 figure.

Power requirements differ substantially: 85 W TDP with a 250 W suggested PSU and 1x 6-pin connector for the HD 7790, versus 575 W TDP with a 950 W suggested PSU and 1x 16-pin connector for the RTX 5090 D V2. Both are dual-slot cards, but physical dimensions diverge — the HD 7790 measures 183 mm in length, while the RTX 5090 D V2 is 304 mm long, 137 mm tall, and 48 mm wide.

Bus interface advances from PCIe 3.0 x16 to PCIe 5.0 x16. Display outputs change from 1x DVI, 1x HDMI 1.4a, and 2x mini-DisplayPort 1.2 to 1x HDMI 2.1b and 3x DisplayPort 2.1b. DirectX support improves from 12 (12_0) to 12 Ultimate (12_2), and Vulkan from 1.2.170 to 1.4. OpenGL remains 4.6 for both. The launch MSRP for the HD 7790 is 149 USD, and for the RTX 5090 D V2 it is 2,299 USD.

Head-to-Head Benchmarks

Direct head-to-head benchmark data between these two cards is unavailable, as the FACT PACK lists no shared test results. Each card was evaluated with a different benchmark: the HD 7790 with Geekbench Metal (score 17,666) and the RTX 5090 D V2 with 3DMark Steel Nomad DX12 (score 16,504). These tests are not directly comparable, so any cross-card comparison must rely on the relative positioning within each card's own performance pool.

The HD 7790's nearest rival data shows it is 0.2% behind the RTX 4060, 0.4% behind the Radeon 780M, 0.7% behind the Radeon Pro 560, and 0.7% ahead of the Quadro RTX 4000. This clustering indicates the HD 7790's Metal performance is tightly grouped with a range of much newer GPUs, suggesting the GCN 2.0 architecture retains strong performance in Metal workloads despite its age.

The RTX 5090 D V2's rival data shows a 0% delta with the NVIDIA T400, 0.5% behind the Radeon PRO W7500, 0.6% behind the RTX PRO 6000 Blackwell, and 0.9% ahead of the RX 5700 XT. The near-identical scores across this group — all within 0.9% of each other — indicate that the 3DMark Steel Nomad DX12 test produces a tightly clustered result for these particular GPUs, despite their very different architectures and specifications.

What the data makes clear is that raw specification superiority does not automatically translate to benchmark dominance within a single test. The RTX 5090 D V2 has roughly 58x the FP32 compute of the HD 7790, yet its percentile ranking (59th) is slightly lower than the HD 7790's (61st). This reflects the different benchmark pools and the fact that Steel Nomad DX12 tests specific rendering features that may not scale perfectly with raw compute resources. Conversely, the HD 7790's strong Metal showing demonstrates that API-specific optimization can keep older hardware competitive in certain workloads.

The most significant interpretation from the available data is that these cards occupy entirely different performance universes despite similar percentile placements. The HD 7790's benchmark score is achieved with 1 GB of memory and 1.792 TFLOPS, while the RTX 5090 D V2 achieves its score with 24 GB and 104.8 TFLOPS. The benchmarks themselves measure different capabilities — Metal is a graphics API test, while 3DMark Steel Nomad DX12 is a specific DirectX 12 rendering scenario. Users should interpret each card's score within the context of its own test suite rather than attempting direct score-to-score comparison.