Nvidia RTX 3060 12GB brings a new level of performance to the mainstream market. The RTX 3060 was officially released today, with a starting price of only $329. Is it realistic?You will most likely find one at this price, just like you find one RTX 3060 Titanium 399 USD, Tencent 3070 US$499, or Tencent 3080 US$699—maybe not entirely impossible, but extremely unlikely. NVIDIA’s Ampere Architecture Powering many people now Best graphics card, They have all seen from gamers and Cryptocurrency miners. Nvidia added firmware and driver code to detect Ethereum mining, which should help, but when people are willing to pay Extreme scalper pricing on eBay, Even with cards like GTX 1660 Super and RTX 2060, everything we have GPU benchmark The hierarchy is now almost sold out.NVIDIA is even working with partners Bring back the previous generation Turing and Pascal.
All of this will not make it a bad GPU, but we expect the RTX 3060 to be as difficult to obtain as any other modern GPU.Finally, the current Ethereum mining boom Will disappear, but it may take a while A year or more before we see the end of the chip shortageAt this point, this shouldn’t surprise anyone, but if you have been hoping for affordable gaming PC upgrades, then this is a frustrating state of affairs.
Unlike the previous Ampere GPU, Nvidia will not provide RTX 3060 Founders Edition, so we are looking for a third-party graphics card. Nvidia provided us with the EVGA GeForce RTX 3060 XC in this release evaluation, which is a fairly compact and relatively inconspicuous card. There is no metal (or even plastic) backplane, no RGB lighting, and two custom-sized 87mm fans for cooling in a 2.0-slot form factor. The card measures 202x110x38mm and weighs 653g, which is a considerable speed change compared to other third-party Ampere cards we have reviewed so far.
This is of course for a reason. It costs money to create a mainstream card and decorate it with all the bells and whistles. And we think that most gamers who buy value for money can get better service through moderate design with good performance. The RTX 3060 will definitely have extreme variants, some of which will be more expensive than the budget RTX 3060 Ti option. Let’s be clear: in most cases, even the fastest RTX 3060 cannot beat the 3060 Ti — yes, even with 12GB of VRAM. That’s because once it exceeds 8GB, the memory capacity is not an important factor, and because of its wider memory bus, it has more memory bandwidth, which provides a great advantage for the 3060 Ti. In addition, the number of GPU cores for the 3060 Ti has increased by 35%.
|Graphics card||RTX 3060 Titanium||Tencent 3060||RTX 2060 Super||Tencent 2060|
|Process technology||Samsung 8N||Samsung 8N||TSMC 12FFN||TSMC 12FFN|
|Mold size (mm^2)||392.5||276||445||445|
|Basic clock (MHz)||1410||1320||1470||1410|
|Boost clock (MHz)||1665||1777||1650||1680|
|Video memory speed (Gbps)||14||15||14||14|
|Video memory (GB)||8||12||8||6|
|Video memory bus width||256||192||256||192|
|GFLOPS FP32 (boost)||16.2||12.7||7.2||6.5|
|TFLOPS FP16 (Tensor)||65 (130)||51 (102)||57||52|
|Launch date||December 20||February 21||July 19||January 19|
|Launch price||399 USD||USD 329||399 USD||USD 349|
Here is how things are broken down, comparing the RTX 3060 with its closest Ampere brothers and Turing predecessors. RTX 2060 and 2060 Super show the change of -60 suffix cards between Turing and Ampere.Ampere gives you one a lot of More shader cores, which means potentially higher computing performance, and a small improvement in the memory bandwidth of the 12GB card.It also doubles the VRAM capacity (at least until Expected RTX 3060 6GB to appear, But maybe Nvidia may leave it to the RTX 3050 series) and boast of improvements in RT and Tensor cores and memory subsystems, all of which have brought better performance. The power usage remains similar, with 170W TGP (total graphics power), which is much lower than the 220W TGP of the RTX 3060 Ti.
An interesting news is that this is the first time NVIDIA has used 15Gbps GDDR6 memory. Except RTX 2080 Super is equipped with 15.5Gbps VRAM, RTX 20 series cards all use 14Gbps memory. This slightly narrows the bandwidth gap between 3060 and 3060 Ti, although the additional 64-bit interface width still gives the GA104 card a clear advantage. GA106 has no advantage in ROP and rendering output, because it has only 48-the same as RTX 2060.
However, the difference between Turing GPU and Ampere GPU does not always appear in the specification table above. In theory, the FP32 performance of the RTX 3060 is 95% higher than that of the RTX 2060, and the core performance of the FP16 Tensor is 97% higher than that of the RTX 2060. In practice, the actual performance difference is much smaller, and half of the FP32 pipeline shares processing resources with the INT32 pipeline. For gaming purposes, the 3060 should not be slower, but in most cases, it will only be about 20-25% faster.
This is the first desktop card to use the Nvidia GA106 processor. At a high level, there are 3 GPCs (graphics processing clusters), each with up to 10 SMs and 16 ROPs (two blocks, each with 8 blue rectangles, each block is located at the bottom of the GPC) . The complete chip has 30 SMs, and 3060 disabled two and finally got 28 SMs, but everything else was put on hold. (Please note that the mobile RTX 3060 has all 30 SMs enabled, although it is only equipped with 6GB of memory, and its clock frequency is also lower than that of a desktop card.)
Each SM contains 64 dedicated FP32 CUDA cores, plus 64 FP32+INT32 CUDA cores-each cycle can only use FP32 or INT32. SM also contains a second-generation RT core and four third-generation Tensor cores. The performance of each core is twice that of the previous generation core. In the case of sparseness, the speed of the Tensor core may be four times Turing. Times. Finally, there are six 32-bit memory interfaces, each of which is connected to a single 8Gb or 16Gb GDDR6 module-the latter is currently reserved for desktop computers, and 8Gb modules are used in notebook computers.
The complete GA106 chip has 12 billion transistors, which is lower than the 17.4 billion in GA104. This reduces the chip size from 393 square millimeters to only 276 square millimeters, which not only helps reduce chip costs, but also increases the number of chips Nvidia can obtain from a single wafer-if you want to know, GA106 is less GA102 The size is only half of 628.4 square millimeters, with 28.3 billion transistors. It is estimated that Nvidia can obtain approximately 130 chips per wafer (some of which are defective, most of which end up as partially failed chips) using GA104, while the smaller GA106 size allows approximately 200 chips per wafer . More chips means higher output and more graphics cards. This is hope.
more: Best graphics card
more: All graphic content