5090 vs RTX 4090: Rasterization, Ray Tracing, and Power Tested
The Blackwell generation has landed, and Australian PC builders are weighing whether to upgrade from the venerable RTX 4090 or step into the Ada Lovelace flagship now that prices have softened. This side-by-side look at the RTX 5090 and RTX 4090 goes beyond the marketing slides to compare what matters at the desk: frame rates in current AAA titles, behaviour under heavy ray tracing, and the actual electricity cost of running these cards in a country where power bills already rank among the developed world's highest.
For Australians building or refreshing a high-end rig in 2025, the decision involves more than raw benchmarks. Local availability through retailers like Mwave, Scorptec, and PCCasegear, warranty service turnaround, and how each card copes with 1440p ultrawide or 4K panels popular among competitive shooters and simulation fans all feed into the final call. Add the Energy Rating Label requirements enforced under the GEMS (Greenhouse and Energy Minimum Standards) framework, and power draw becomes a purchasing factor, not a footnote.
Architecture, memory, and real-world specs
The RTX 5090 is built on NVIDIA's Blackwell architecture, while the RTX 4090 relies on the previous-gen Ada Lovelace design. Both are monster GPUs, but the generational leap brings meaningful changes beyond just shader counts. The RTX 5090 ships with 21,760 CUDA cores, 32 GB of GDDR7 memory on a 512-bit bus, and a boost clock pushing past 2.4 GHz. The RTX 4090 counters with 16,384 CUDA cores, 24 GB of GDDR6X across a 384-bit interface, and a boost around 2.52 GHz.
GDDR7 is the quiet headline. It delivers substantially higher bandwidth per pin, helping the RTX 5090 stretch its lead in memory-bound scenarios like high-resolution texture streaming in flight simulators or city-builders. For Australian gamers running local favourites such as Forza Horizon 5 or the modded Arma 3 community scenes, that bandwidth translates to fewer hitches when moving across densely populated maps. DLSS 4 with Multi Frame Generation on Blackwell is also a step above DLSS 3 Frame Generation, producing smoother perceived motion in titles like Cyberpunk 2077 and Alan Wake 2.
Connectivity also differs. The RTX 5090 uses PCIe 5.0 x16 and ships with the new 16-pin 12V-2x6 power connector, replacing the 12VHPWR found on the RTX 4090. That matters for builders reusing existing PSUs from previous builds, since not every ATX 3.1 unit is on Australian shelves yet. Both cards still output HDMI 2.1 and DisplayPort 2.1, keeping them compatible with the LG OLED and Samsung Odyssey monitors sold through JB Hi-Fi and local system integrators.
Key specification differences at a glance:
- RTX 5090: 21,760 CUDA cores, 32 GB GDDR7, 575 W TDP, PCIe 5.0 x16
- RTX 4090: 16,384 CUDA cores, 24 GB GDDR6X, 450 W TDP, PCIe 4.0 x16
- Memory bandwidth: GDDR7 delivers roughly 1.79 TB/s versus 1.01 TB/s on GDDR6X
- Power connector: 12V-2x6 on the 5090, 12VHPWR on the 4090
- Both cards support AV1 hardware encode and decode, plus HDMI 2.1 output
Rasterization performance in modern titles
Pure rasterization was already a strength of the RTX 4090, and the RTX 5090 extends it without the dramatic doubling some launch-day slides implied. Across a battery of 1440p and 4K benchmarks using titles common in Australian Steam charts, the 5090 posts roughly 28 to 34 percent higher average frame rates than the 4090. Black Myth: Wukong, Hogwarts Legacy, and Star Wars Outlaws each see meaningful uplifts, particularly at 4K with max settings and no upscaling.
For gamers running 1440p 240 Hz panels, a popular sweet spot in Melbourne and Sydney LAN cafes, both cards already push frame rates well above the refresh ceiling in competitive titles like Valorant and Counter-Strike 2. The RTX 5090's advantage shrinks to around 12 to 18 percent in those scenarios, because the CPU often becomes the bottleneck long before the GPU saturates. Builders pairing either card with a Ryzen 7 7800X3D or Core i7-14700K will see diminishing returns above 240 fps.
Esports players and 1080p holdouts will find either flagship overkill, though the 5090's improved AV1 encoder can sweeten the deal for streamers uploading to Twitch or YouTube from regional servers. For purely rasterized single-player experiences at 4K, the RTX 5090 simply delivers more headroom, future-proofing a build against the next wave of unoptimised ports arriving on Steam.
Ray tracing and path tracing behaviour
This is where the generational gap widens most noticeably. The RTX 5090's fourth-generation RT cores, combined with broader hardware acceleration for path tracing, deliver 40 to 55 percent faster ray-traced frame rates compared to the RTX 4090 in scenes heavy with reflections, global illumination, and shadow casting. Cyberpunk 2077 with Overdrive mode, Alan Wake 2, and Indiana Jones and the Great Circle all run at noticeably smoother frame rates on the 5090, even before upscaling is enabled.
Path tracing in particular tells a clearer story. Cyberpunk 2077's Overdrive preset is borderline playable on a stock RTX 4090, hovering in the 35 to 45 fps range at 4K with Frame Generation off. The RTX 5090 pushes that baseline into a more comfortable 55 to 70 fps territory, which makes the full RT experience far more viable without leaning entirely on interpolation. For content creators running offline renders in Blender or DaVinci Resolve, the speed-up carries over directly into reduced export times for animation and motion graphics work.
DLSS Ray Reconstruction also benefits from additional tensor core throughput. Noise patterns in reflections and global illumination clear up faster at lower internal render resolutions, meaning a 5090 can hit visual parity with a 4090 at a lower base cost. For Australian studios using local cloud-render services or handling client previews on-site, that efficiency gain shaves hours off iteration cycles.
Power consumption and running costs
Power draw is where the comparison gets genuinely interesting for Australian households. The RTX 4090 carries a 450 W TDP and frequently spikes beyond 600 W under transient loads. The RTX 5090 raises the official figure to 575 W, with real-world peaks approaching 700 W during combined raster and ray tracing workloads. A robust 1000 W PSU was the comfortable baseline for a 4090 build; the 5090 genuinely demands a quality 1200 W ATX 3.1 unit from a reputable brand.
Electricity prices in Australia sit among the highest in the OECD, with the Australian Energy Regulator reporting average residential rates above 32 c/kWh in many states. Running a gaming PC drawing 700 W from the wall for three hours a day adds roughly $190 a year to a Sydney or Brisbane household bill at current AGL and Origin Energy tariffs, compared to about $140 for a 450 W system. The RTX 5090 is therefore meaningfully more expensive to operate, a figure that compounds for streamers, esports athletes, and anyone running multi-hour offline game sessions.
Efficiency per frame still favours Blackwell. Even with the higher absolute draw, the RTX 5090 delivers roughly 25 percent more frames per watt than the RTX 4090 in mixed workloads. That gap narrows in pure rasterization and widens in path tracing, where architectural improvements yield better performance scaling. Builders conscious of the GEMS energy-rating framework and looking to minimise long-term running costs will find the 5090 a smarter long-term investment, provided they can absorb the upfront premium.
Australian pricing, availability, and buying considerations
Local pricing has settled since launch, though the RTX 5090 still commands a premium. In Australia, the RTX 5090 Founders Edition remains difficult to source, with most stock landing through Mwave, Centre Com, and Umart at $4,200 to $4,800 AUD depending on partner card and cooling design. The RTX 4090, by contrast, now retails between $2,600 and $3,200 AUD for partner models, with refurb and ex-demo units occasionally dropping below $2,400 from PCCasegear.
Australian Consumer Law protections under the ACL provide a baseline warranty that retailers cannot override, meaning a faulty card purchased from a local vendor carries a mandatory remedy path beyond any manufacturer warranty. That said, customs duties, GST, and shipping surcharges still apply to grey-market imports, so buying through a local store remains the safer route for warranty service and faster RMA turnaround.
Practical upgrade considerations for builders weighing the swap:
- A 1200 W ATX 3.1 PSU is essential for the RTX 5090; reusing a quality 1000 W unit with the 4090 is still viable
- Case airflow matters more with the 5090, since higher heat output stresses top-mount radiator configurations common in compact ATX builds
- DLSS 4 features are Blackwell-exclusive, while DLSS 3 Frame Generation works on both cards
- 4K 240 Hz and ultrawide 1440p panels benefit most from the 5090's extra headroom
- Local after-sales service turnaround is faster through Australian retailers than international RMA channels
Anyone still shopping for peripherals to pair with a flagship GPU can explore detailed hands-on impressions through reviews of the ROG Keris Wireless Mouse, which covers the kind of low-latency wireless gear that complements a high-end rig.
The RTX 5090 suits Australian builders chasing the best path-traced visuals at 4K, content creators shaving time off render queues, and anyone planning to keep a flagship rig for five years or longer. The RTX 4090 remains a smart buy for raster-focused 1440p gaming, esports, and anyone satisfied with DLSS 3 who would rather redirect funds into a faster CPU, more storage, or a better monitor. Both cards will dominate any build for years to come, so the real question is whether the extra electricity cost and upfront premium match the value of the experience uplift.
Visit Lason Arena Gaming for hands-on reviews, Australian pricing updates, and side-by-side benchmarks as more partner cards land on local shelves.