NVIDIA GeForce RTX 3060 12 GB Review with a custom card. Is the MSI RTX 3060 Gaming X Trio the new Ampere entry-level drug?

Game performance

In the following result overviews I have on the one hand cumulated the average FPS of all cards and in all games and then calculated and compared this still in percent. I tested a total of 10 games and in QHD-HD the action is still good. After all, we must never forget that this is not cherry-picking, but in any situation there are games in the selection that can easily break the fair weather front. In any case, the GeForce RTX 2060 Super, which was the target, has been clearly put in its place.

If we now look at the percentages, we end up with an average lead of 9 to 10 percentage points of the GeForce RTX 3060 over the GeForce RTX 2060 Super, depending on the resolutions, and with DXR it’s even almost 12 percent. The gap to the GeForce RTX 3060 Ti on the other hand is significant. This renders almost 27 percentage points faster in Full HD, and then about 28 to 32 percentage points faster with DXR and in WQHD. Nvidia really has the courage to fill the gap here, because the Radeon RX 6700XT and 6700 will surely follow soon.

 

In summary: Especially at lower resolutions up to WQHD, this card is a safe bet, because even if the gap to the GeForce RTX 3060 Ti is a bit more significant than expected. But is it enough to achieve the desired playability still good. In some situations, you’ll have to turn up the quality slider a bit in WQHD, but only if you need extremely high FPS numbers and the game doesn’t support DLSS 2.0. It only gets a bit tighter in higher resolutions, but NVIDIA doesn’t advertise the card for high resolutions.

DLSS in particular has never been more valuable for this battle dwarf, because you can, if you rather focus on performance and dampen the other settings a bit as well as do without the DXR effects completely, even still go with the flow in Ultra-HD in the appropriate games quite well. Otherwise, DLSS in full HD also allows DXR, which was probably the intention. If you’ve also looked at the variance page, you’ll also have seen that there’s quite a round image gradient here, especially in Full HD enough. The FPS or percentiles are still much too rough intervals to be able to depict this very subjective impression well.

Application performance and studio

As long as you don’t rely on pure raster graphics, such a RTX 3060 is actually not a bad solution for the somewhat smaller budget, since the card almost always acts faster than many older Turing cards, especially in the designated studio applications, and still remains affordable (if the prices don’t explode again). Because whenever it comes to FP32 calculations and you think you’re getting a bit closer to retirement every time you do a final render, for example, the RTX 3060 is a nasty little killjoy for Turing, Pascal and the Radeons.

Sometimes, even a Quadro RTX 5000 is reached or just beaten, which is of course a good thing to take in suitable applications. NvEnc has been making life easier since Turing, but if all the post-processing filters get a good workout now, you can still be reasonably happy with the GeForce RTX 3060. With Decode and Encode on many 4K clips in the Time Line anyway. A constant 24 FPS is nothing to sneeze at, unfortunately the competitor is (still) at a disadvantage.

And while we were on the subject of moving images, the new Ampere cards also natively support AV1 in the decoder (Intel’s Xe will probably do the same). But why AV1? The codec is up to 50% more efficient than H.264, which means you only need half the internet bandwidth to transmit the same quality – including HDR and also 10-bit encoding. That would go up to 8K – theoretically. Unfortunately, the practice even with the cable providers a Germany looks rather lukewarm. But the beginnings are made and the will counts somehow, even if I miss the 8K TV.

In summary: Yes, you can still work quite fast with the GeForce RTX 3060, even if the efficiency is a bit worse than with the GeForce RTX 3060 Ti. So the non-gamers only will be happy because the wallet will be noticeably relieved. Whether it’s Octane, Arnold or Blender, the RTX 3060 is still a real powerhouse (within the limits of its technical capabilities), and with OptiX, the RTX 3060 can gallantly cover up some of its shortcomings. Always provided that the application supports these features at all.

Which of course also applies to video editing and all those workstation or CAD applications that don’t need certified and optimized drivers to sprint. In general, the RTX 3060 does not cut a bad figure, even if it lacks power for higher resolutions. Whereby one really wonders about the sense of the comparatively lush memory expansion. However, if you believe the performance data of a possibly upcoming Radeon RX 6700, then it will only come with 6 GB memory, which would be the first time that NVIDIA has an advantage in a similar performance class. Hard to believe, but if that’s the case, it’s actually relatively smart and tricky to want to beat AMD at its own game for once.

MSI RTX 3060 Gaming X Trio 12 GB

Visually and haptically, the part is convincing and since it has inherited the genes of the RTX 3070 and 3060 Ti almost 1:1 in the cooler construction and board design, it is almost oversized. Top cooling, virtually no operating noise – that’s how a graphics card should look (and sound), even if it pushes the price up a bit. But there are also much simpler models that you can be happy with, such as the GeForce RTX 3060 EX from KFA2, which I will introduce soon.

Conclusion and final remarks

In general, the GeForce RTX 3060 looks quite ok, because it positions itself very exactly where I had predicted it to be a while ago. It’s faster than a GeForce RTX 2060 Super (as a replacement for the GTX 1060 with 6GB anyway), costs less (MSRP) than its counterpart back then and has become much more performant and efficient. For a final assessment, including the market positioning, we will have to wait for the launch of the new Radeon RX 6700 to be fair. I already wrote that NVIDIA’s feature set, besides Reflex and various latency improvements, ranges from the usual RTX components like ray tracing and DLSS 2.0, to various RTX software (video, voice) for the end customer and all the studio and workstation applications. AMD will have to step up its game here, because there is currently still a big gap in the portfolio.

Especially in the semi-professional areas, AMD is currently a bit behind and we will have to wait and see what else will be launched in the future besides the new hardware. So everyone will have to set their premises and ask themselves what value which feature and which use case really has for them (or not). A review like this can’t take this decision away from anyone, that’s something everyone has to decide for themselves.

The MSRP mentioned by NVIDIA as the starting price for the most basic models of 329 euros is certainly an incentive, but there is also initial information on the board partner cards that these should turn out to be significantly more expensive. A rumored MSRP of up to 400 Euros for the more elaborate cards is a bit of a tough call, unfortunately. And then there is the completely crazy market, which currently drives the prices to astronomical heights, which have nothing to do with the RRP. If the cards are currently available in the shops at all.

Whether NVIDIA’s trick with the mining brake can be effective at all for a few mining applications remains to be seen. Let’s hope so, because this is exactly what will strongly influence the customers’ opinion about the new card and NVIDIA will have to measure itself against the statements made in the run-up to the launch. After all, what good is an empty box in the shop window whose contents you couldn’t pay for anyway? Nothing. But as we all know, hope dies last.

The card was provided to igorsLAB by MSI under NDA for testing with the condition not to fall below the specified release date for the case of the NDA. There was no possible influence of the manufacturer on the test and the results, nor was there any obligation to publish them.