Being surpassed by the RX 5000 and RTX 2000, the NVIDIA RTX 3060 appears in 3DMark leaving a bad taste in the mouth for not measuring up in terms of performance. It was tested with an i9-10900K in the Fire Strike and Time Spy tests, are you noticing the drop in CUDA Cores?
Do not forget that it is a synthetic benchmark on graphics cards, but the low result obtained by an RTX 3060 whose assembler does not appear anywhere is shocking. We do not know if it will have to see the limitation of the SKU by cryptocurrencies, but the source shows us captures in Fire Strike Extreme, Port Royal, Time Spy, and Superposition.
RTX 3060 sinks into 3DMark and its optimal resolution will be 1080p
A few days ago, NVIDIA made official its new range of graphics cards focused on cryptocurrency mining. Additionally, he stated that the new GeForce GPUs that are coming out (RTX 3060, 3070 Ti, 3080 Ti, etc.) will be limited for mining via firmware, BIOS, and, even, in the SKU itself. This was the measure that Jensen Huang's team offered gamers not to run out of their RTX 3000 graphics cards.
That said, what about the 3DMark test? Well, the source has leaked several captures of the NVIDIA RTX 3060 in 3DMark along with an i9-10900K and the results obtained are the following:
- Fire Strike Extreme: 10,089 points.
- Fire Strike Ultra: 5217 points.
- Time Spy: 9276 points.
- Time Spy Extreme: 4367 points.
- Port Royal: 5072 points.
Putting this data in context, we see a lower score than the RTX 2070, 2060 Super, and the AMD RX 5700 XT. Directly, the RTX 3060 Ti would play in another league, being much superior to the mentioned GPUs, and to save you comparisons, we use the source table.
If we look closely at the tests of the NVIDIA RTX 3060, we see what estimated FPS we will see in some games based on the result obtained:
- Battlefield V in 1440p: between 70 and 90 FPS.
- Apex Legends at 1440p: around 90 FPS.
Logically, it speaks of 1440p without Ray Tracing (or so we assume), so we have gone to check what maximum and stable FPS the RX 5700 XT offers in those games together with an i9-10900K. In the same resolution, it offers the following:
- Battlefield V at 1440p: 92 constant FPS and 115.7 maximum FPS.
- Apex at 1440p : constant 87 FPS and 109.7 maximum FPS.
So it looks like the RTX 3060 won't rival the RX 6700 XT, far from it. It will have the same role as the RTX 2060 at the time: an input Ray Tracing GPU that you can't ask for much at 1440p, with 1080p being its best resolution.
On the other hand, they have also offered us the results obtained by the RTX 3060 in the Ungine Superposition benchmark:
- 1080p -> 5073 points.
- 7680 × 4320 -> 2838 points.
Who does not know this benchmark, basically it is an extreme test that uses the UNIGINE 2 Engine graphics engine and that seeks to test the stability of the GPU with a heavy test. To do this, it uses around 900 interactive objects and measures the virtual reality experience.
Why such low performance?
We are left to ask ourselves, why that score? The NVIDIA RTX 3060 is surpassed in these 3DMark benchmarks by graphics cards of previous generations with 8 GB GDDR6, while the protagonist equips 12 GB GDDR6. However, memory is not everything, but the importance of shaders that have integer gains, let's see what specifications each one offers:
| RTX 2060 | RTX 2060 Super | RTX 2070 | RX 5700 XT | RTX 3060 | |
CUDA Cores | 1920 | 2176 | 2304 | 2560 | 3584 |
Tensor Cores | 240 | 272 | 288 | 40 CUs | 112 (3rd generation) |
RT Cores | 30 | 3. 4 | 36 | N / A | 28 (2nd generation) |
Memory | 6 GB GDDR6 | 8GB GDDR6 | 8GB GDDR6 | 8GB GDDR6 | 12 GB GDDR6 |
Turing GPUs come with more RT Cores and Tensor Cores than the NVIDIA RTX 3060, although NVIDIA stresses that the RTX 3000 uses 2nd Gen RT Cores and 3rd Gen Tensor Cores (presumably better). Of course, we see an increase in CUDA Cores thanks, among other things, to Samsung's 8nm process (Turing came with a 12nm process).
On the other hand, we are struck by the doubt of the limitation for mining, will it have affected the performance of the GPU? Well, the debate is served after seeing the score obtained in the tests. That said, we want to emphasize that they are synthetic benchmarks.
What do you think of this performance?
Source: VideoCardz
0 Comments