GIGABYTE AORUS GeForce RTX 5090 INFINITY 32G (40th Anniversary)
GIGABYTE introduced the AORUS GeForce RTX 5090 INFINITY 32G in January 2026 as part of its 40th-anniversary lineup. It combines NVIDIA's Blackwell flagship GPU and 32GB of GDDR7 memory with the distinctive WINDFORCE HYPERBURST cooling system and higher factory clocks. This review examines the card's cooling design, independent temperature and noise measurements, power requirements and unusually high reported US retail price.
What makes the AORUS RTX 5090 Infinity different?
The WINDFORCE HYPERBURST cooler uses a dual-flow-through airflow design and an internal Overdrive fan that activates under heavier loads. According to GIGABYTE's official specifications, the card reaches a 2730MHz boost clock, compared with 2407MHz for the reference RTX 5090.
The RTX 5090 Infinity also features a distinctive anniversary design, but its higher clock speed and cooling configuration do not change the underlying Blackwell architecture. The practical differences depend on sustained performance, temperatures and noise rather than specifications alone.
Gaming performance, temperatures and noise
PC Games Hardware tested the AORUS RTX 5090 Infinity in May 2026, recording GPU temperatures of approximately 77°C and memory temperatures of 84°C. The publication also reported relatively high noise compared with some competing RTX 5090 models.
Its independent cooling and thermal review provides the testing conditions. These are third-party measurements, not results obtained by Qayim Li.
For a clearer understanding of what the factory overclock adds, our NVIDIA RTX 5090 review covers the standard Blackwell GPU, including its 32GB GDDR7 memory, 4K gaming capabilities and reference power requirements.
Power requirements, ports and case compatibility
GIGABYTE recommends a 1000W power supply and specifies one 16-pin power connector, three DisplayPort 2.1b outputs and one HDMI 2.1b port.
The card measures 330mm in length and 65mm in thickness, making case clearance particularly important. Buyers should also ensure the power cable is fully seated and avoid sharp bends close to the connector.
Although the card supports NVIDIA's DLSS 4 features and AI-accelerated workloads, these capabilities come from its RTX 5090 architecture rather than the anniversary cooling design.
AORUS 5090 INFINITY price and competing graphics cards
GIGABYTE has not provided a verified US manufacturer MSRP for this exact edition. A Micro Center listing in June 2026 placed the card at $5,299.99, but that was a retailer asking price, not an official launch MSRP or a confirmed price in Egypt, Saudi Arabia or the UAE.
Buyers comparing the GIGABYTE AORUS RTX 5090 INFINITY with other flagship cards should consider measured cooling performance, noise, warranty and the total system cost. Our RTX 5080 vs RTX 5090 comparison provides additional context on the performance and memory differences between NVIDIA's high-end GPUs.
At a substantial premium over conventional RTX 5090 models, the INFINITY edition makes more sense for buyers who specifically value its anniversary styling and cooling design than for those seeking the best performance per dollar.
✅ Pros
- +Factory clock up to 2730MHz
- +32GB GDDR7 memory
- +Distinctive dual-flow-through cooler
- +Modern DP 2.1b and HDMI 2.1b outputs
- +Manufacturer recommends a 1000W PSU
❌ Cons
- −Relatively high tested noise
- −Test sample VRAM reached 84°C
- −330mm length requires clearance
Frequently asked questions
Is this a different RTX 5090 GPU?
How long is the card?
Does it always run cool?
Quick verdict
The AORUS INFINITY is a striking Blackwell flagship with a distinctive cooler and higher advertised clock, but not a different generation of RTX 5090. Independent testing raises practical thermal and noise considerations, and the exceptionally high reported US retail listing is difficult to justify for buyers prioritizing value per frame.
| Manufacturing | |
| Architecture | NVIDIA Blackwell |
|---|---|
| GPU chip | GB202 |
| Process node | 5 nm |
| Transistors | 92.2 B |
| Die size | 750 mm² |
| Launch date | 2026-01-07 |
| Performance | |
| Shading units | 21760 |
|---|---|
| Ray tracing cores | 170 |
| Tensor cores | 680 |
| ROPs | 176 |
| TMUs | 680 |
| Boost clock | 2730 MHz |
| FP32 performance | 118.8 TFLOPS |
| Memory | |
| VRAM | 32 GB |
|---|---|
| Memory type | GDDR7 |
| Memory bus | 512 bit |
| Memory speed | 28 Gbps |
| Memory bandwidth | 1792 GB/s |
| L2 cache | 96 MB |
| Power | |
| TDP | 600 W |
|---|---|
| Recommended PSU | 1000 W |
| Power connectors | 1x 16-pin (12V-2x6) |
| Cooling | WINDFORCE HYPERBURST; dual-flow-through |
| Connectivity & ports | |
| Interface | PCIe 5.0 x16 |
|---|---|
| Display outputs | 3x DisplayPort 2.1b; 1x HDMI 2.1b |
| Max resolution | 7680 x 4320 |
| Max displays | 4 |
| Card length | 330 mm |
| Slot width | 3.5 |
| Extra features | |
| Upscaling | NVIDIA DLSS 4 |
|---|---|
| Ray tracing | YES |
| Video encoders | NVIDIA NVENC, AV1, H.264, HEVC |
| API support | DirectX 12 Ultimate, Vulkan 1.4, OpenGL 4.6 |
| Availability & price | |
| Street price | 5299.99 USD (Micro Center, Jun 2026) |
|---|---|
No photos yet.
No video yet.
No ratings yet. Be the first to rate!
Rate your experience
New reviews, straight to your inbox
Every new review and buying guide, once a week. No spam.
No spam. Leave with one click, any time.