RTX 4090 vs RTX 3090: How does the top-of-the-line GPU compare to its predecessor?

The Nvidia Geforce RTX 4090 is the highest-end flagship entry in the Ada Lovelace-based RTX 40 series lineup. It is the third BFGPU (Big Ferocious GPU) designed and sold by the American semiconductor company.

The RTX 4090 is a part of the -90 series lineup that was introduced in 2020 with the RTX 3090 GPU. The latter was the best-in-class consumer-grade rendering machine that money could buy until the "Ti" refresh dropped earlier this year. The 4090 has now replaced the RTX 3090.

The new GPU is based on newer technologies, comes with a wider memory bus that is clocked at a faster speed, and utilizes the improved Ada Lovelace architecture. In video games, the 4090 has been seen to beat the 3090 by a solid 70-80% difference. In some titles, the new graphics card goes on to deliver over twice the performance of its last-gen equivalent.

---

How massive of an upgrade is the RTX 4090?

The Geforce RTX 4090 Founders' Edition graphics card (Image via Nvidia)

Nvidia has made massive performance uplifts with the RTX 40 series GPUs. The new graphics cards can now utilize the full power of the latest CPUs in the market, like the Ryzen 5000 and 7000 series chips, the Intel Alder Lake, and the upcoming Raptor Lake processors. However, this performance gain comes at a massive cost for thermal and power efficiency.

The new RTX 4090 comes with 16,384 CUDA cores, which is a massive bump from the 10,496 found in the RTX 3090. The new GPU packs 512 TMUs (Texture Mapping Units) as compared to the 328 found in the 3090.

The number of ROPs (Render Output Units) has also been bumped to 176 from 112. The memory size remains the same, at 24 GB. Both GPUs use GDDR6X memory, which is based on a 384-bit bus. However, the bandwidth has been boosted to 1,018 GB/s as compared to the 936.2 GB/s memory found on the 3090.

Major advancements have been made in terms of architecture. The new GPU is based on the TSMC 4nm mode. In contrast, the 3090 was developed on Samsung's 8nm technology. The total amount of cache has also been significantly bumped in the new GPUs. While the 3090 had a 6 MB L2 cache, the 4090 packs a whopping 72 MB L2 cache.

The new GPU also comes with significantly higher operating clock speeds. It has a base clock of 2,235 MHz and a boost clock of 2,520 MHz. In comparison, RTX 3090 has a base clock of 1,395 MHz and a boost clock of 1,695 MHz.

All of these improvements translate to a more than 200% increase in theoretical performance. While the RTX 3090 is capable of 35.58 TFLOPS, the new GPU can output 82.58 TFLOPS.

These improvements translate to a massive performance boost, both in video games and in productivity scenarios.

In most games, the RTX 4090 is capable of delivering 4K 120 FPS using the highest settings and ray tracing without DLSS. The list includes AAA titles like Forza Horizon 5, Far Cry 6, Assassin's Creed Valhalla, and more.

In terms of productivity performance, the new GPU almost dwarfs the last-gen RTX 3090 and the 3090 Ti. It also comes with support for AV1 encoding, like the latest Intel Arc Alchemist GPUs. Thus, professional users who want their renders and video exports to get done faster can opt for the latest GPU.

Overall, the RTX 4090 almost demolishes the 3090. When the last-gen Titan-class GPU launched, it was priced at an eye-watering $1,499. The Ti revision was priced ridiculously at $1,999. However, both of these graphics cards can now be bought for around $1,000, thanks to the crypto crash and the successive drop in graphics card demand.

Thus, users have to pay a hefty $600 premium for the latest GPU. Considering that this GPU is almost twice as fast as its last-gen counterpart that consumes similar power, it is a worthy choice.