After the press ban for Nvidia’s own “Founders Edition” version of the GeForce RTX 3080 fell on September 16, the graphics card manufacturers’ own variants are now on sale – and the press can report. For this first test we got a GeForce RTX 3080 TUF Gaming (from 809 €) from Asus. This three-slot card has three axial fans and can be switched from the factory-set quiet mode to the louder, but somewhat faster performance mode using a slide switch. With a recommended price of 719 euros, the Asus card is not much more expensive than Nvidia’s Founders Edition.
As with all RTX 3080, the core is the GA102 GPU, which is manufactured using 8 nanometer technology. It consists of around 28 billion transistors and comprises 8704 shader cores. That is almost three times as many as with the GeForce RTX 2080 and exactly twice as many as with the GeForce RTX 2080 Ti, which costs a good 1200 euros.
The number of ray tracing cores has also increased, from 46 for the RTX 2080 to 68 for the RTX 3080. And although it has fewer tensor cores, namely 272 instead of 368, the new card is supposed to achieve higher performance. 10 GB of GDDR6X memory are attached to a 320-bit wide memory interface and transfer 760 GB / s. Nvidia specifies the theoretical FP32 computing power as 30 TFlops; the direct predecessor had 10.6 TFlops. Further technical data of the GeForce RTX 3000 can be found in our message on the architecture of the “Ampere” GPU.
High performance …
The new Ampere GPU does not lack 3D performance, the 3DMark Time Spy gives a first impression: In this DirectX 12 test, the GeForce RTX 3080 scores 15,217 points, leaving the ROG Strix RTX 2080, also from Asus, behind Ti behind (13,154 points). The lead in the Port Royal ray tracing test is just as clear, with 11,653 to 8993 points. The Gainward GeForce RTX 2080 Phoenix GS, which started at a similar price for the Turing launch two years ago, remains significantly behind both.
The RTX 3080 also cuts a very good figure in games: In the action adventure Shadow of the Tomb Raider, it calculates an average of 84 frames per second (fps) even in 4K resolution and with ultra preset. That is a full 20 fps more than the RTX 2080 Ti manages and 35 more than the RTX 2080. With ray tracing shadows on ultra setting, the average frame rate drops to 55 fps, but is well above the 41 fps of the 2080 Ti.
The ray tracing performance in the Blender rendering program, which can use both Nvidia’s CUDA and the newer Optix interface, is impressive. With CUDA, the RTX 3080 calculated the classroom scene in 79 seconds, with Optix in just under a minute. The RTX 2080 Ti takes almost twice as long in both cases.
… high power consumption
However, Nvidia bought the performance jump with a strong increase in power consumption. Regardless of whether in quiet or performance mode: the card draws around 350 watts under full load, and individual power peaks even reach 475 watts. It is significantly more frugal in idle and only requires around 13 watts when it drives a Full HD display. With a 4K display, we measured 15 watts. However, if two monitors are attached to the card, the power consumption increases to 26 watts.
This ampere card delivers the leap in performance that was actually hoped for from the Turing generation. The first impression suggests that the GeForce RTX 3080 can display current games with the highest level of detail in 4K resolution smoothly, even with activated ray tracing effects. It consistently outperforms the GeForce RTX 2080 Ti, and at a significantly lower price.
It is already clear that the GeForce RTX 3090, which is to follow soon, is raising the bar – not only in terms of performance, but also in terms of price. The more exciting question is how the cheaper models RTX 3070 and 3060 will fare.
In addition, AMD is also in the starting blocks with the RDNA2 architecture in the form of the Radeon RX 6000, which should not only compete with the ampere cards in terms of 3D performance, but should also be able to accelerate ray tracing effects. So it promises to be another hot autumn for graphics cards.
|Nvidia GeForce RTX 3080 in comparison|
|GPU||Turing (TU104)||Turing (TU102)||Ampere (GA102)|
|GPU / Turbo clock||1515/1800 MHz||1350/1635 MHz||1440/1710 MHz|
|theoretical computing power FP32||10.6 TFlops||14.2 T flops||30 TFlops|
|Ray tracing / tensor cores||46/368||68/544||68/272|
|3DMark Fire Strike Extreme / Time Spy / Port Royal||12303/10294/6322 points||16295/13154/8993 points||18900/15217/11653 points|
|Shadow of the Tomb Raider 4K (without / with RTX shadow)||49/29 fps||64/41 fps||84/55 fps|
|Blender Classroom (CUDA / Optix)||185/137 s||150/103 s||79/58 s|
|Power consumption idle / 3D full load||11/230 watts||13/267 watts||13/350 watts|
|price||approx. 830 euros||1260 euros||719 euros (RRP)|
More cards with RTX3080
Nvidia’s new graphics chip is not only used in the Asus graphics card we tested, but also in various other models that are currently available for pre-order. The following price comparison always shows the ten cheapest offers.