Tech

The Nvidia RTX 50 series “Blackwell” graphics cards have arrived, including the RTX 5090. Priced at $2000 MSRP, it is an expensive offering but promises some enticing features: top-tier gaming performance with the GPU GB202, 32GB of high-speed GDDR7 memory for gaming and content creation, and DLSS 4 with multi-frame generation to maximize refresh rates on nearly all supported gaming monitors. But does it live up to the hype? Many reviewers have tested the RTX 5090 Founders Edition with a new set of demanding titles, examining both ray tracing and rasterization performance at various resolutions. I don’t have the card to make the tests myself so I have searched and compiled information on benchmarks for the GPU, but before that let’s discuss the physical design of the card, and performance expectations based on its specifications and testing approach. It is immediately apparent that the Nvidia Founders Edition is the most compact top-tier card we’ve seen from Team Green in years. Despite consuming 575W which is a lot, its two-slot design fits effectively in small form-factor PCs. This is achieved through a centralized circuit board and two continuous flow fans that draw air through impressively dense heatsink stacks. It’s an astonishing piece of industrial design. Credits to: Nvidia Third-party designs from Nvidia partners are expected to be more conventional, featuring familiar triple-fan designs, making the Founders Edition potentially harder to obtain. The card requires 600W of power via a 12V-2×6 power cable, which can connect directly to compatible ATX 3.0+ PSUs or through four 8-pin PCIe cables using the provided adapter. This adapter is angled to prevent excessive strain on the cable, while an updated connector is designed to avoid the carbonization issues seen on the RTX 4090 FE unit. In addition to a standard HDMI 2.1 port, the card features three DisplayPort 2.1 outputs for connecting HDR 4K 240Hz displays without needing DSC (Display Stream Compression). On the specifications, we have 21,760 CUDA cores, boost clocks of 2.41GHz, and 1792GB/s memory bandwidth, paired with 32GB of GDDR7 across a 512-bit bus. These figures indicate substantial improvements over the RTX 4090 FE, which has only 16,384 CUDA cores and a 384-bit memory interface with 1018GB/s bandwidth, despite having higher boost clocks of 2.52GHz. This results in a 33% increase in GPU cores, a 77% enhancement in memory bandwidth, and a 28% rise in power consumption. Most of Nvidia’s marketing regarding the 5090 and 4090 primarily focuses on comparing multi-frame generation to single-frame generation. Regarding performance gains between the 2 cards, there was a 20% to 35% gain based on the titles, the difference was bigger when using higher resolutions mainly due to the 32GB of GDDR7, also it needs to be mentioned that the drivers for these cards are still very recent and improvements could be made over the lifespan of this generation that could increase the performance. With all that said, below are links to some benchmark results that I won’t be publishing here and information I have been using to make this overview, you should check them before reading further. Eurogamer: Part 1, Part 2, Part 3  Engadget: Part 1 Conclusion This is not a card for everyone and the generation performance jump isn’t so high as NVIDIA was publishing without the fake frames in between if you have a 4090 then you will be more than happy to keep your card since you won’t be losing much on not buying the 5090, especially with the price of MSRP on 2000$ and the 3rd party cards launching prices being much higher than that, at the time I’m writing this the 5090 it’s not available for pre-order but the 5080 is, and on Portugal, this card is starting at 1500€ with higher end models at 2000€ so I would not be surprised to see the 5090 prices starting at 3000€ for base models and the higher end models reaching prices up to 4000€ which is insane, this makes me worried about the RTX 4070 and 4070 TI that may launch with prices that won’t make any sense to recommend. If you are running with an older generation card like a 20 series a 16 series even a 10 series card it can make sense to upgrade to this generation if you use the technology advancements that these cards bring like the AV1 encoder and DLSS4 but I would still wait and see what NVIDIA brings more to the table with the RTX 5060 and 5050, with the right price which at this point seems unlikely it could be a good alternative.