NVIDIA GeForce RTX 5090 Founders Edition preview: A sleek and compact next-gen powerhouse
Is SFF-compatible graphics cards the way to go?
#nvidia #geforcertx5090 #gaming
When was the last time we saw an NVIDIA flagship GPU with a dual-slot design? (Image: HWZ)
It’s been almost a fortnight since NVIDIA CEO Jensen Huang officially unveiled the next generation GeForce RTX 50 Series at his CES keynote address. But we finally have the GeForce RTX 5090 Founders Edition (FE) in our own lab and while we still can’t talk about benchmarks and performance numbers (you’ll have to come back on 23 Jan, 10pm SGT, when NVIDIA lifts the review embargo), we can show you a closer look at the card as we unbox it.
Much has been touted about the flagship GeForce RTX5090 card’s performance, but the key talking point is the incredible design work that has gone into the Founders Edition models – including the RTX 5080, RTX 5070 Ti and RTX 5070. These GPUs not only showcase NVIDIA’s engineering and design capabilities, but also technical capabilities that balances cutting-edge specifications with practical considerations – NVIDIA’s RTX 50 Series FE cards are the only cards right now that meet NVIDIA’s own SFF-Ready guidelines.
Yup, you can fit one of these babies in a small-form factor PC.
GeForce
Graphics Card | RTX 5090 FE | RTX
4090 FE | RTX
3090 Ti FE | RTX
3090 FE |
GPU | Blackwell (GB202) | Ada Lovelace (AD102) | Ampere
(GA102) | Ampere
(GA102-300) |
Process | 4nm (TSMC) | 4N
(TSMC) | 8nm
(Samsung) | |
Transistors | 92 billion | 76 billion | 28
billion | 28
billion |
Streaming Multi-processors (SM) | 170 | 128 | 84 | 82 |
CUDA cores | 21760 | 16384 | 10752 | 10496 |
Tensor Cores | 680 (Gen 5) | 512
(Gen 4) | 336
(Gen 3) | 328
(Gen 3) |
RT Cores | 170 (Gen 4) | 128
(Gen 3) | 84
(Gen 2) | 82
(Gen 2) |
GPU base / boost clocks (MHz) | 2017 / 2407 | 2230 / 2520 | 1670 /
1860 | 1395 /
1695 |
Memory | 32GB GDDR7 | 24GB GDDR6X | 24GB GDDR6X | 24GB GDDR6X |
Memory bus width | 512-bit | 384-bit | 384-bit | 384-bit |
Memory bandwidth | 1,790GB/s | 1,018GB/s | 1,008GB/s | 936GB/s |
Interface | PCIe 5.0 | PCIe 4.0 | PCIe 4.0 | |
TDP | 575W | 450W | 450W | 350W |
Price | US$1,999 | US$1,599 | US$1,999 | US$1,499 |
As NVIDIA’s new halo part, the RTX 5090 takes over from the RTX 4090, boasting an impressive 32GB of fast GDDR7 memory operating across a 512-bit memory interface, and delivers a staggering 1.8 terabytes per second of bandwidth. With 680 5th Gen Tensor Cores, new Streaming Multiprocessors (that are not present in the RTX 4090) that are optimised for AI-driven graphical tasks like RTX Skin, and 170 4th Gen Ray Tracing Cores (33% more than on the RTX 4090) to support RTX Mega Geometry, the RTX 5090 is a huge generational leap over its predecessor. While numbers like these may seem abstract, their real-world impact is immediately apparent. At the NVIDIA Editor’s Day event, we witnessed games running with path-traced visuals at 4K resolutions with frame rates exceeding 240fps, while creators see substantial gains in applications like DaVinci Resolve and Blender, where rendering times are slashed thanks to the GPU’s neural rendering capabilities.
The Founders Edition cards now support DisplayPort 2.1. (Image: HWZ)
The RTX 5090 FE follows the angular position of the RTX 3090 FE card, which creates lesser tension and stress on the power cables. (Image: HWZ)
The heatsink fins features a very subtle, but very nice, concave design. (Image: HWZ)
The improvements in raw performance are underpinned by a completely reworked thermal design. One of the standout features of the RTX 5090 Founders Edition is its dual airflow system. Traditional GPUs, including the RTX 4090, relied on a single flow-through design where air would enter from one side and exit through small exhaust vents. This design, while effective, had its limitations at higher power levels, often resulting in increased fan noise and thermal inefficiencies. With the RTX 5090 FE, NVIDIA has gone a step further by introducing a dual airflow system with a unique three-piece PCB design. This innovative design allows air to flow directly through both the front and back fans, effectively doubling the cooling capacity. By shortening the PCB (which is really small compared to the RTX 4090), NVIDIA has created an uninterrupted path for airflow, significantly reducing resistance and ensuring optimal thermal performance.
The smaller PCB also enables the integration of a 3D vapor chamber. This chamber distributes heat evenly across the GPU core, transferring it to the fin stacks for efficient dissipation. The result is a thermal solution that NVIDIA says keeps the RTX 5090 FE running at peak performance without the acoustic footprint typically associated with high-power GPUs. Compared to the RTX 4090 FE, the RTX 5090 delivers a quieter and more efficient cooling experience, even under heavy loads.
Just when we thought RTX 5090 FE was going to get bigger (compared to the RTX 4090 and 3090 FE cards below it), we were pleasantly surprised to see a more practical design. (Image: HWZ)
If you're intending to build a super small-form factor PC, the RTX 5090 FE is your only option right now. (Image: HWZ)
Linking up to the main PCB are the remaining two boards; one that links to the display outputs and the other as the PCIe connector. Expectedly, the RTX 5090 supports PCIe 5.0 and DisplayPort 2.1 too.
The RTX 5090 Founders Edition is shaping up to be a standout example of NVIDIA’s focus on not only specifications but also thoughtful hardware design. By completely rethinking the PCB and airflow layout, NVIDIA has delivered an FE card that is not only insanely powerful but also fits into small-form factor builds without compromising cooling or noise levels. While I’ll reserve final judgement until I can dive into the benchmarks, the RTX 5090 FE is a great example of how the madness of huge, thick and heavy graphics cards since the RTX 3090 days can be done away with. But from what I’ve seen at CES 2025, it looks like NVIDIA partners such as ASUS ROG, MSI and Gigabyte aren’t listening – yet.
Bonus picture: Don't have a 12V-2x6 power connector? Well tough luck, you'll have to use this cable adapter. (Image: HWZ)
The NVIDIA GeForce RTX 5090 Founders Edition, along with the RTX 5080 Founders Edition and partners cards will be available from 30 January 2025. The RTX 5070 Ti and 5070 cards will be released in February 2025.
Read more:
- NVIDIA GeForce RTX 50 Series: How Blackwell's Neural Rendering and DLSS 4 are shaping next-gen gaming
- NVIDIA CEO Jensen Huang officially unveils the GeForce RTX 50 series at CES 2025 keynote
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.