NVIDIA RTX 3090 Review: the meeting point between gaming and productivity

An exaggerated video card in all. The RTX 3090 Founders Edition is an uncompromising GPU, in performance, size and price. It is a mistake to believe that it is meant for gaming, or at least, just for gaming, because the RTX 3090 is the perfect meeting point between a gaming GPU and one built for work.
The 24 GB of RAM installed inside cannot be explained otherwise, which can make a difference by moving to professional applications. The successor of the Titan, also in our tests, has been confirmed the most powerful consumer video card you can buy, for a product aimed at enthusiasts, and professionals, willing to spend the 1549 euros of the list price.

Ampere on steroids

The RTX 3090 is the largest video card we’ve ever tried. 313 mm long, 138 mm wide but above all 54 mm thick and weighing almost 2.2 kg, NVIDIA’s top-of-the-range is simply huge and occupies three expansion slots. The fixing to the case must be perfect to avoid problems, some custom versions even offer supports to take some weight off the PCIe slot. This is not the case with the Founders variant, which once installed normally stays in place without the need for additional help.
The design incorporates that seen on the 3080, with the due differences in proportions, and the constricting quality is very high, thanks to the use of only metal for the whole body.

The 12-pin power connector is placed at the top, a convenient location but the adapter makes it more difficult to hide the cables in the case, complicating the setup for those who want an installation as clean as possible. In this tab the NVLink connector returns, which allows you to install two RTX 3090s in parallel, to obtain simply impressive computing power. At the bottom is the PCIe 4.0 connection, while on the back there is an HDMI 2.1, with which to exploit the 8K at 60 Hz or the 4K at 120 Hz, and three DisplayPort 1.4.

The dissipation system is identical, but larger than that of the RTX 3080.
The two fans are placed opposite each other on the PCB, fresh air is channeled from the bottom up and towards the rear of the GPU, thus being expelled from the case. Inside the heatsink there is also a vapor chamber system to allow better heat dissipation. During our tests the RTX 3090 reached 74 degrees of peak, remaining instead around 40 degrees in idle, lower values ​​than the RTX 3080, but not surprising given the size of the 3090’s dissipation system.

The noise generated by the card is slightly higher than that of the 3080, but we are still on very good acoustic comfort levels, especially when you think about the power generated by this GPU.

NVIDIA recommends a 750 W power supply for this card, the same one we used during the test, but the consumption is actually not much higher than that of the RTX 3080, which in our tests stood at around 320 W. In this case it goes up to about 350 W, despite the greater number of calculation units, the increase is not so marked.

Specifications and benchmarks

To discover all the news of the Ampere architecture you can read the review of the RTX 3080 Founders Edition, in this article we will focus only on the performance of the RTX 3090, a GPU with truly extreme characteristics.
The GA102 chip is the same as the RTX 3080, made at 8 nm and with 28 billion transistors, on a 628 mm2 Die. In this case though CUDA Cores rise to 10496 (8704 in the 3080), but the Tensor Cores and the RT Cores also increase, reaching respectively 328 and 82 compute units (272/68 on the 3080). The increase also brings with it a small disadvantage, in fact the Base Clock drops to 1400 MHz against the 1440 MHz of the 3080, while the Boost Clock reaches 1700 MHz, against the 1710 of the younger sister.
In reality, the RTX 3090 clock is about 1820 MHz, slightly lower than the values ​​found on the 3080.
On the memory side we find 24 GB of GDDR6X RAM, with a 384 bit bus (320 bit in the 3080) and a clock of 19.5 Gbps, for a bandwidth of 936 GB / s, also higher than the 3080. A configuration of this type promises incredible performance in every game, and indeed it is.

The RTX 3090 excels in every benchmark, from Full HD to 4K, but as you can see it cannot clearly detach the RTX 3080, gaming alone cannot justify the huge price difference between the two models.

Already with 3DMark Time Spy you can see how the leap in performance is palpable but not revolutionary. However, this figure does not reflect the trend we have observed in game benchmarks, where the increase in performance varies a lot by title. For example, Shadow of the Tomb Raider gains very few frames with the 3090, while with Gears 5 in 4K the step change is more marked. When Ray Tracing is activated, at least in our test titles, it seems that the potential of the RTX 3090 comes to the surface, but even in this case all the games we tried were already perfectly usable with the RTX 3080, albeit with a lower frame rate.

Is the RTX 3090 a disappointment then? Absolutely not, because NVIDIA’s goal was not to sharply knock out the 3080 in frame rate. The RTX 3090 was created to combine the world of work with that of gaming, it is in complex tasks such as rendering that you see the difference between the two models. A field where more power means more processing capacity, and consequently less time to complete a job. The difference in performance is evident in the two benchmarks we did with V-Ray Next and IndigoBench, which show how the RTX 3090 can give even professionals an edge over the 3080.
With V-Ray Next, a benchmark that measures rendering performance, the 3090 reached 777 points, against the 628 of the RTX 3080, while with IndigoBenchmark, also designed to test performance in the same field, the 3090 reached altitude 41.151 msamples / s, against 35.920 of the 3080. Numbers that, translated into real life, indicate a precious time saving for graphics professionals.