NVIDIA GeForce RTX 3090 Ti vs NVIDIA GeForce RTX 5090 Comparison

Architecture Differences

The NVIDIA GeForce RTX 3090 Ti and the NVIDIA GeForce RTX 5090 represent two distinct generations of NVIDIA graphics architecture, with the former built on Ampere and the latter on Blackwell 2.0. The RTX 3090 Ti uses the GA102 chip fabricated on Samsung's 8 nm process, while the RTX 5090 uses the GB202 chip built on TSMC's 5 nm node. This process shrink is accompanied by a dramatic increase in transistor count: the RTX 3090 Ti packs 28,300 million transistors on a 628 mm² die, yielding a density of 45.1M transistors per mm². The RTX 5090, by contrast, contains 92,200 million transistors on a 750 mm² die, achieving 122.9M transistors per mm². That density difference means the newer chip crams more than 3.2 times the transistors into a die that is only about 19% larger by area.

The memory subsystems also differ fundamentally. The RTX 3090 Ti ships with 24 GB of GDDR6X on a 384-bit bus, delivering 1.01 TB/s of bandwidth. The RTX 5090 moves to 32 GB of GDDR7 on a 512-bit bus, more than doubling bandwidth to 1.79 TB/s. Clock speeds scale upward as well: the RTX 3090 Ti operates at a 1560 MHz base and 1860 MHz boost, while the RTX 5090 runs at 2017 MHz base and 2407 MHz boost. Effective memory speed rises from 21 Gbps on the older card to 28 Gbps on the newer one.

Compute resources expand across every category. The RTX 3090 Ti has 10,752 shading units, 336 texture mapping units, 112 ROPs, 84 RT cores, and 336 tensor cores. The RTX 5090 nearly doubles each of these: 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. Pixel rate climbs from 208.3 GPixel/s to 423.6 GPixel/s, and texture rate jumps from 625.0 GTexel/s to 1,636.8 GTexel/s. FP32 throughput rises from 40.00 TFLOPS to 104.8 TFLOPS, with FP16 following the same 1:1 ratio on both cards.

Connectivity and physical design change as well. The RTX 3090 Ti uses PCIe 4.0 x16, while the RTX 5090 uses PCIe 5.0 x16. Display outputs differ: the older card provides 1x HDMI 2.1 and 3x DisplayPort 1.4a, whereas the newer card offers 1x HDMI 2.1b and 3x DisplayPort 2.1b. The RTX 3090 Ti is a triple-slot card measuring 336 mm long, 140 mm tall, and 61 mm wide. The RTX 5090 is a dual-slot design at 304 mm long, 137 mm tall, and 40 mm wide, making it notably more compact. Both require a single 16-pin power connector, but the suggested PSU rating rises from 850 W to 950 W.

Head-to-Head Benchmarks

The recorded head-to-head data contains three benchmark comparisons, and the RTX 5090 wins all three. The largest margin appears in the 3DMark Steel Nomad DX12 test, where the RTX 5090 scores 18,355 against the RTX 3090 Ti's 5,741. That is a delta of -68.7% from the perspective of the older card, meaning the RTX 5090 delivers roughly 3.2 times the performance in this workload. In Geekbench OpenCL, the RTX 5090 scores 334,370 versus 174,441 for the RTX 3090 Ti, a -47.8% delta that translates to about 1.9 times the performance. Geekbench Vulkan shows a similar gap: 376,728 for the RTX 5090 against 215,633 for the RTX 3090 Ti, a -42.8% delta, or roughly 1.7 times the performance.

The average benchmark score in the database tells a more nuanced story. The RTX 3090 Ti carries an average benchmark score of 131,938, while the RTX 5090's average is 79,842. This apparent contradiction stems from the different benchmark suites recorded for each card. The RTX 5090 has ten recorded benchmark entries, including several PassMark tests with low absolute scores that pull its average down. The RTX 3090 Ti has only three recorded entries, all from higher-scoring tests. The head-to-head comparisons, which use identical tests, are the more reliable indicator of relative performance, and they uniformly favor the RTX 5090.

Percentile rankings also differ. The RTX 3090 Ti sits at the 95th percentile among all GPUs in the database, while the RTX 5090 sits at the 92nd percentile. This is likely an artifact of the RTX 5090's lower average score dragging its percentile down despite its superior per-test results. The head-to-head deltas are unambiguous: the RTX 5090 leads by 42.8% to 68.7% depending on the workload.

The Verdict

The data points to a clear generational leap. The RTX 5090 outperforms the RTX 3090 Ti in every directly comparable benchmark, with margins ranging from 42.8% in Geekbench Vulkan to 68.7% in 3DMark Steel Nomad DX12. The newer card also offers 8 GB more memory, nearly double the bandwidth, and more than double the shading units, RT cores, and tensor cores. Its smaller physical footprint (dual-slot versus triple-slot, 304 mm versus 336 mm long) makes it easier to fit into cases despite its higher 575 W TDP versus 450 W.

The RTX 3090 Ti still holds its own in the database's percentile ranking at 95 versus 92, and its nearest rivals are similar in average score: the NVIDIA L4 is 0.7% behind, while the RTX 4000 Ada Generation, A10M, and Radeon PRO W6800 are each 2.4% to 2.6% behind. The RTX 5090's nearest rivals cluster much closer, with the Tesla P100 PCIe 16 GB within 0.3%, the Tesla P100 PCIe 12 GB within 0.6%, and the RX 6850M XT within 1.1%. This suggests the RTX 5090's recorded average score is depressed by its PassMark entries, not by a lack of capability.

For a buyer choosing strictly between these two, the RTX 5090 is the stronger performer in every measured head-to-head test. The RTX 3090 Ti remains a capable card, but its architecture is two generations older, and the benchmark deltas show no workload where it leads. The RTX 5090's launch MSRP is 1,999 USD, identical to the RTX 3090 Ti's launch MSRP of 1,999 USD, which makes the newer card the more compelling option on raw performance alone.

Specification Differences

The two cards differ across nearly every specification field. The RTX 3090 Ti uses the GA102 chip on Ampere architecture with an 8 nm Samsung process, while the RTX 5090 uses the GB202 chip on Blackwell 2.0 with a 5 nm TSMC process. Transistor count rises from 28,300 million to 92,200 million, die size from 628 mm² to 750 mm², and transistor density from 45.1M per mm² to 122.9M per mm².

Base clock increases from 1560 MHz to 2017 MHz, boost clock from 1860 MHz to 2407 MHz, and effective memory speed from 21 Gbps to 28 Gbps. Memory size grows from 24 GB to 32 GB, type changes from GDDR6X to GDDR7, bus width from 384 bit to 512 bit, and bandwidth from 1.01 TB/s to 1.79 TB/s. Shading units rise from 10,752 to 21,760, TMUs from 336 to 680, ROPs from 112 to 176, RT cores from 84 to 170, and tensor cores from 336 to 680. Pixel rate goes from 208.3 GPixel/s to 423.6 GPixel/s, texture rate from 625.0 GTexel/s to 1,636.8 GTexel/s, and FP32/FP16 from 40.00 TFLOPS to 104.8 TFLOPS.

TDP increases from 450 W to 575 W, slot width shrinks from triple-slot to dual-slot, and suggested PSU rises from 850 W to 950 W. The bus interface changes from PCIe 4.0 x16 to PCIe 5.0 x16, and display outputs shift from HDMI 2.1 with DisplayPort 1.4a to HDMI 2.1b with DisplayPort 2.1b. Dimensions shrink from 336 mm by 140 mm by 61 mm to 304 mm by 137 mm by 40 mm. Production status moves from end-of-life to active, release date from January 2022 to January 2025, and successor from GeForce 40 to GeForce 60. Both cards share the same 1x 16-pin power connector, DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 support.

FAQ

Q: Which card has more memory bandwidth?

A: The RTX 5090 has 1.79 TB/s of bandwidth versus 1.01 TB/s for the RTX 3090 Ti, thanks to its 512-bit GDDR7 memory compared to the 384-bit GDDR6X on the older card.

Q: How much faster is the RTX 5090 in 3DMark Steel Nomad DX12?

A: The RTX 5090 scores 18,355 versus 5,741 for the RTX 3090 Ti, a delta of -68.7% from the older card's perspective, meaning the RTX 5090 is about 3.2 times faster.

Q: Do both cards use the same power connector?

A: Yes, both the RTX 3090 Ti and the RTX 5090 use a single 1x 16-pin power connector, though the RTX 5090 requires a 950 W suggested PSU versus 850 W for the RTX 3090 Ti.

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

A: The RTX 3090 Ti's GA102 has a density of 45.1M transistors per mm², while the RTX 5090's GB202 achieves 122.9M transistors per mm², a substantial increase from the 8 nm Samsung process to the 5 nm TSMC process.

Q: Which card has more RT cores and tensor cores?

A: The RTX 5090 has 170 RT cores and 680 tensor cores, compared to 84 RT cores and 336 tensor cores on the RTX 3090 Ti, roughly doubling both counts.

Q: How do the physical sizes compare?

A: The RTX 3090 Ti is a triple-slot card at 336 mm by 140 mm by 61 mm, while the RTX 5090 is a dual-slot card at 304 mm by 137 mm by 40 mm, making the newer card shorter, narrower, and thinner.

Where Each One Wins

The RTX 5090 wins in every directly comparable benchmark category. In 3DMark Steel Nomad DX12, it leads by 68.7%. In Geekbench OpenCL, it leads by 47.8%. In Geekbench Vulkan, it leads by 42.8%. This pattern holds across general compute and graphics workloads, suggesting the RTX 5090 is the better choice for high-fidelity gaming, ray tracing, and GPU-accelerated compute tasks.

The RTX 3090 Ti's advantages are more situational and relate to its existing ecosystem rather than raw speed. It sits at the 95th percentile among all GPUs in the database, higher than the RTX 5090's 92nd percentile. Its nearest rivals in the database are all within 0.7% to 2.6% of its average score, indicating strong consistency across its recorded benchmarks. The RTX 3090 Ti also has a lower TDP at 450 W versus 575 W, which may matter for systems with tighter power budgets or smaller PSUs, though the suggested PSU rating is still 850 W.

For users prioritizing raw performance in current DirectX 12 and Vulkan workloads, the RTX 5090 is the clear pick based on the head-to-head data. For users who already own an RTX 3090 Ti, the upgrade path is straightforward: the newer card delivers 42.8% to 68.7% more performance depending on the test, with the same launch MSRP of 1,999 USD. However, the RTX 3090 Ti remains a high-percentile card in the broader database, and its end-of-life status does not diminish its recorded benchmark results. The RTX 5090's active production status and newer architecture make it the more future-proof option, particularly given its PCIe 5.0 interface and DisplayPort 2.1b outputs.