GeForce RTX 5090 Efficiency Evaluate: NVIDIA Stays King Of PC Gaming

	NVIDIA GeForce RTX 5090: MSRP $1,999 NVIDIA’s highly effective however expensive GeForce RTX 5090 is the corporate’s newest flagship GPU and our checks present it crushing every part in AI, gaming, and content material creator workloads.
	Quickest GPU Ever Bleeding Edge Options DLSS 4 RTX Nerual Rendering Nice For AI, Creators, And Gaming Smaller Type Issue Than Earlier-Gen 32B Of VRAM	Steep Worth Premium Excessive Energy Consumption Not A Huge Leap Over 4090

Because it was introduced at CES a number of weeks again, the GeForce RTX 50 sequence, the Blackwell GPU structure, and the flagship GeForce RTX 5090 specifically have been the matters of a lot dialogue. Over two years out from the launch of the previous-gen GeForce RTX 4090, fans, avid gamers and creators have been itching to see what NVIDIA had in retailer. As such, firm CEO Jensen Huang revealed total desktop and cell GPU line-ups, together with the GeForce RTX 5070 all the best way on up the top-end GeForce RTX 5090 we’ll be exhibiting you right here.

Earlier than we proceed although, we’ve offered a ton of Blackwell and GeForce RTX 50 sequence protection, which are must-reads to higher perceive what the GeForce RTX 5090 is all about. On the very least, please take a look at our Blackwell GPU structure piece, which particulars the numerous developments coming with the structure, together with RTX Neural Rendering, RTX Mega Geometry, DLSS 4 with Multi-Body Gen, the brand new AI Administration Processor,  and the RTX 50 sequence’ up to date media engine. Particulars on the remainder of the line-up are additionally obtainable right here. We not going to re-hash lots of the deeper particulars obtainable in these earlier articles right here, so please give them a learn if you’d like the total scoop earlier than continuing…

NVIDIA GeForce RTX 5090 Specs

The GeForce RTX 5090 is constructed across the GB202 GPU, which is manufactured on the identical TSMC 4nm course of because the earlier gen. The GB202 options 92 billion transistors, and as configured on the RTX 5090, it has 21,760 CUDA Cores, 170 RT Cores, and 680 Tensor cores. The GB202 relies on the NVIDIA Blackwell structure and is outfitted with new fifth gen Tensor Cores, 4th gen Ray Tracing Cores, and up to date CUDA cores.

Linked to the GPU by way of a large 512-bit interface is 32GB of GDDR7 reminiscence providing as much as 1.8 TB/s of whole reminiscence bandwidth, which is a 78% improve over the earlier gen. The GB202 additionally encompasses a native PCIe Gen5 interface, and its up to date show engine helps DisplayPort 2.1b with UHBR20 and HDMI 2.1b.

In whole, a full GB202 GPU contains 12 Graphics Processing Clusters (GPCs), 96 Texture Processing Clusters (TPCs), 192 Streaming Multiprocessors (SMs) every with 128 CUDA Cores, and a 512-bit reminiscence interface with sixteen 32-bit reminiscence controllers. As such, a full GB202 GPU contains 24,576 CUDA Cores, 192 RT Cores, 768 Tensor Cores, and 768 Texture Items. A full GB202 GPU additionally options 128 MB of whole L2 cache, together with a 256 KB Register File, and 128 KB of L1/Shared cache.

The GB202 just isn’t totally enabled on the GeForce RTX 5090, nevertheless. Whereas the total GB202 GPU has a complete of 12 GPCs, 96 TPCs, and 192 SMs, “solely” 11 GPCs, 85 TPCs, and 170 SMs are enabled on the RTX 5090. As such, though the GB202 is outfitted with 128 MB of whole L2 cache, the RTX 5090 has 96 MB of L2 enabled.

The GB202 might not be totally enabled on the GeForce RTX 5090, however it’s nonetheless a step ahead over the AD102 powering the RTX 4090. The variety of texture models has elevated from 512 on the RTX 4090 to 680 on the RTX 5090, which ends up in greater bilinear-filtered texel fill charges. The RTX 5090 delivers 1636.76 Gigatexels/sec, in comparison with 1290.2 Gigatexels per second on RTX 4090, although the RTX 5090’s pixel fill fee is definitely a bit decrease than the 4090’s.

The NVIDIA GeForce RTX 5090 Founders Version

What’s inside that field is a very redesigned GPU, which options plenty of progressive firsts. Though it’s far more highly effective than its predecessor, the GeForce RTX 5090 is far thinner. The RTX 5090 measures 304mm in size, 137mm in top, and it’s actually two-slots (40mm) huge – it’s not only a 2-slot huge case bracket. It is a dense package deal for positive, and feels premium through-and-through.

NVIDIA was in a position to shrink the GeForce RTX 5090 by growing what it calls a “Double Circulate By means of” cooler design, which permits each cooling followers to blow air straight by means of the heatsink, for optimum efficiency.

To realize this, the GeForce RTX 5090 has a multi-part PCB setup. The central PCB homes the GPU, reminiscence, and energy circuitry. It rests within the middle of the cardboard, with solely small parts protruding beneath every fan. A separate daughterboard attaches to the central PCB for the PCIe x16 connector, and a 3rd, versatile board runs perpendicular alongside the underside with high-speed signaling connections from the central PCB to the show outputs.

The quantity of engineering that went into the GeForce RTX 5090’s design is nothing wanting spectacular. Not solely is the cooler much more succesful, however NVIDIA needed to resolve plenty of issues to make sure correct signaling over the adjoining PCBs and their connectors. NVIDIA discusses a lot of the design course of in a video obtainable right here in case you’re eager about a number of the nitty-gritty particulars. 

On the high of the cardboard, you will discover the acquainted 12VHPWR 16-pin connector, additionally used on the RTX 40 sequence. On the RTX 5090 although, the connector is angled off the again on the PCB, and recessed barely within the shroud. This configuration ought to permit for simpler cable administration and reduce the necessity to bend the facility feed in shallower PC instances. The included energy adapter requires 4 PCIe 8-pin feeds, just like the RTX 4090 did, however the cabling is longer and much more versatile. The connector on the adapter can also be beefier and emits a offered ‘click on’ when pushed totally into place. Hopefully, these tweaks and updates to the facility connector and adapter assist forestall a number of the points that affected the RTX 4090.

Sitting atop the GPU, reminiscence and energy circuitry is a brand new, unified 3D Vapor Chamber that enables vapor to journey on to warmth pipes to rising thermal efficiency, which is linked to a mix of straight and curved fins within the heatsink meeting. designed to cut back again strain and optimizes airflow by means of the fins. Directional retailers on the facet direct warmth up-and-out to fight recirculation of heat air as effectively.

To make sure optimum warmth switch from the GPU to the vapor chamber, NVIDIA can also be utilizing a really excessive efficiency liquid steel thermal interface materials, just like what’s within the PS5. Liquid steel TIM has a lot decrease thermal resistance than even at this time’s greatest thermal pastes, which permits extra warmth to switch extra rapidly into the vapor chamber, in the end reducing GPU temperatures, however there’s a catch. Liquid steel is electrically conductive, and if not secured in place, it might probably run like several liquid and doubtlessly trigger shorts. NVIDIA sealed the GPU to positive the liquid steel stays in place although, with the cardboard put in in any orientation. Simply remember that “re-pasting” the RTX 5090 down the highway would require additional care to stop potential issues.

Outputs within the GeForce RTX 5090 embody a trio of DisplayPorts (2.1b) and a single HDMI port (2.1b). Their orientations have been reversed versus previous-gen playing cards although, and the brand new case bracket — which does not require venting because of the twin flow-though cooler design — additionally options an anti-fingerprint coating.

And with all of that out of the best way, let’s get to some benchmarks…