RTX 3060 is Just 10% Faster Than Its Predecessor In Synthetic Tests

RTX 3060 Ti Founders Edition
(Image credit: Nvidia)

The RTX 3060 launch is almost upon us, but we won't have to wait until tomorrow to see benchmarks. Videocardz.com today posted early numbers of how the 3060 runs in synthetic benchmarks like 3DMark and Unigine Superposition. The site attributes these tests to anonymous sources, and although synthetic benchmarks and pre-launch tests can be inaccurate, the results are quite underwhelming.

Nvidia's GeForce RTX 3060 will be the company's new mid-range card for the Ampere generation, featuring 3584 CUDA cores, 12GB of GDDR6 memory, and a $325 MSRP (good luck getting one at that price). The GPU will be launching on February 25th (tomorrow).

Swipe to scroll horizontally
GPUs:Fire Strike ExtremeTime Spy ExtremeSuperposition 1080P Extreme
RTX 30601028441115073
RTX 2060905038104370
RTX 2060 Super1056040705150

We've highlighted some of the benchmarks Videocardz ran above, but overall, the RTX 3060 managed to beat the 2060 by just 10% overall. Some of the synthetic tests like 3DMark Fire Strike have the 3060 beating out its Turing counterpart by 16%, but in others like Time Spy, the results are nearly identical.

However, like with most of these leaks -- performance can be highly skewed due to the use of synthetic benchmarks and the use of a pre-release driver; both of which can significantly alter the card's performance. So take these results with a grain of salt, as it is almost guaranteed that what we're seeing here don't indicate how the card will really perform on launch day, with launch day drivers.

But, if there's a chance these results end up reflecting the 3060's actual power, the 3060 runs the risk of presenting a seriously unappealing value at $325, which is 30% more expensive than the RTX 2060 despite the card giving you just 10% more performance. 

Stay tuned for our review of the RTX 3060 tomorrow, where you'll get a more detailed, more accurate overview of where the 3060 sits in performance against the RTX 2060 and the other best GPUs currently on the market.

Aaron Klotz
Contributing Writer

Aaron Klotz is a contributing writer for Tom’s Hardware, covering news related to computer hardware such as CPUs, and graphics cards.

  • InvalidError
    It will probably be unobtainable for anywhere near the MSRP for the foreseeable future just like the 2060s are now if you want to buy one new, so the disappointment was going to be inevitable regardless of how good or meh it turns out.
    Reply
  • nitrium
    I have an RTX 2060 and was looking to upgrade to the 3060 - especially after experiencing it's very disappointing performance with Ray Tracing set to On in Cyberpunk 2077 (at 1200p it was basically resulted in unplayable framerates in many areas, no matter what other settings I tweaked). 10% more performance wouldn't be nearly enough to make a difference (I estimate drops to as low as 15 fps on my RTX 2060), although hopefully when it comes to RT it's a LOT more than that.
    Reply
  • Makaveli
    Admin said:
    Nvidia's RTX 3060 was tested in Superposition and 3DMark, and was just 10% faster than the RTX 2060.

    10% faster and double the price lol.

    nitrium said:
    I have an RTX 2060 and was looking to upgrade to the 3060 - especially after experiencing it's very disappointing performance with Ray Tracing set to On in Cyberpunk 2077 (at 1200p it was basically resulted in unplayable framerates in many areas, no matter what other settings I tweaked). 10% more performance wouldn't be nearly enough to make a difference (I estimate drops to as low as 15 fps on my RTX 2060), although hopefully when it comes to RT it's a LOT more than that.

    If you want playable performance with RT you are too far down the product stack I wouldn't look at anything less than the 3080.
    Reply
  • watzupken
    nitrium said:
    I have an RTX 2060 and was looking to upgrade to the 3060 - especially after experiencing it's very disappointing performance with Ray Tracing set to On in Cyberpunk 2077 (at 1200p it was basically resulted in unplayable framerates in many areas, no matter what other settings I tweaked). 10% more performance wouldn't be nearly enough to make a difference (I estimate drops to as low as 15 fps on my RTX 2060), although hopefully when it comes to RT it's a LOT more than that.
    Realistically, if a RTX 3080 is struggling with this game with RT on, I am not very hopeful that the RTX 3060 can do any magic here. Besides the cut in CUDA cores, you need to remember that the RT and Tensor cores are also quite heavily cut down for this card. So RT performance will take a tumble and I suspect that will also be the case for DLSS performance.
    Reply
  • watzupken
    Somehow I feel that the Ampere architecture seems to be very CPU bound or somewhat inefficient, which is why it tends to stretch its legs only when testing it at 4K. As we go down the product stack and with lower resolution, I feel the gains over previous gen may be capped. This is just my opinion based on what I observed so far. In any case, I think its best to wait for official testing results.
    Reply
  • cryoburner
    But, if there's a chance these results end up reflecting the 3060's actual power, the 3060 runs the risk of presenting a seriously unappealing value at $325, which is 30% more expensive than the RTX 2060 despite the card giving you just 10% more performance.
    The article's math seems way off. Since when was the RTX 2060 just $250? It started at $350 at launch, and while a few lower-end models were briefly available for around $300 or so following the 5600 XT's launch a year ago, most remained priced at or above the new 3060's MSRP. The 2060's MSRP was never officially lowered following the 2060 SUPER's launch, they just launched the new card (more accurately a slightly cut-down 2070) at $400, a $50 higher price point compared to the 2060.

    As for the leaked performance numbers, that wouldn't be too surprising. The 3060 Ti offers performance around that of a 2080 SUPER. But despite having a similar name as the 3060 (non-Ti), it is using a different, larger graphics chip with over 35% more cores enabled, and 33% higher memory bandwidth. Since a 2080 SUPER is only around 30% faster than a 2070 in graphically demanding games, and around 35% faster than a 2060 SUPER, I would expect the 3060's performance to be roughly comparable to those cards.

    So, effectively, it is bringing the performance of a ~$400 2060 SUPER down to $325, while adding more VRAM than is probably needed, and likely improving RTX/DLSS performance slightly. Or at least it would be, but the current shortages ensure that it will probably cost close to double its MSRP for months to come, and might not even reach that price before the year is through.

    nitrium said:
    I have an RTX 2060 and was looking to upgrade to the 3060 - especially after experiencing it's very disappointing performance with Ray Tracing set to On in Cyberpunk 2077 (at 1200p it was basically resulted in unplayable framerates in many areas, no matter what other settings I tweaked). 10% more performance wouldn't be nearly enough to make a difference (I estimate drops to as low as 15 fps on my RTX 2060), although hopefully when it comes to RT it's a LOT more than that.
    The 30-series doesn't improve RT performance much more than non-RT performance. However, the 2060's 6GB of VRAM apparently can't handle RT particularly well in that game, and it sees a disproportionately large performance hit with RT enabled as a result, even when using DLSS upscaling at 1080p. See these benchmarks, for example, where the 2060 SUPER with its 8GB of VRAM manages around double the performance of the 2060 with RT enabled (1080p results are near the bottom of the page)...
    https://www.techspot.com/article/2165-cyberpunk-dlss-ray-tracing-performance/So, I would expect a big performance boost in that particular game with RT enabled, but again, similar performance could have already been had with a 2060 SUPER. I wouldn't expect any of these cards to produce a 60fps experience with RT enabled even with DLSS at 1080p, but staying above 30fps (with an average around 45fps) might be possible. Of course, without RT, even your existing card can probably get close to 60fps with DLSS, so you would still be seeing a notable performance hit from enabling it. If you were going to upgrade from a 2060 to improve RT performance, you would probably be better off moving up to at least something like a 3060 Ti. That is, if graphics card prices were not going to be terrible for a while. As it is, it's probably best to just hold out for the time being.
    Reply
  • domaru
    I wouldn't be much surprised if it was around 2060S or 10% faster. In games 3060ti is up to 50% faster than 2060S but also have 40% more cuda cores as compared to 3060 non-ti.

    I say the 3060 performance in games will be only slightly above 2060S unless you enable rtx.

    I also say if you can get 2060S considerably cheaper than 3060 then it's a better deal
    Reply