It rolls a lot with Nvidia's latest graphics cards: the manufacturer's GeForce RTX 30 series has been launched, from which the company promises significant performance improvements, plus (apart from the top model) quite friendly recommended end-user pricing. The series, for which detailed specifications have not yet been released by the manufacturer, will debut in three models, in addition to the GeForce RTX 3070 and 3080, the top-of-the-line RTX 3090 has been introduced – while players can be happy with the former two.
Unlike its predecessors, which built on Turing, the 30 Series already rests on the manufacturer's Ampere architecture, and is made with Samsung's 8-nanometer manufacturing process. Thanks to the new manufacturing technology, the cards work with significantly smaller transistor sizes than before – the chip accommodates about 28 billion transistors, well above Turing’s $ 19.8 billion. Introduced this spring, the Ampere brings much more muscular tensor cores to GPUs than before, more than doubling the performance produced by the previous architecture, according to the manufacturer.
Of course, the company also sparked the raytracing that gave the RTX series its name. the performance of the RT cores has also become twice as muscular – although the exact details have not yet been elaborated by the manufacturer. In any case, the latter could be especially beneficial to gamers as more realistic ambient lighting with raytracing finds its way into more and more titles – and the feature, although supported on the company’s previous generation cards, still sweated them heavily.
they come with PCIe 4.0 support as expected – providing almost 32 GBps of I / O bandwidth for the RTX 30 series in a full-fledged PCIe x16 socket, doubling the value seen in the previous generation. Of course, you will also need a suitable motherboard to take advantage of the latter, although it is still questionable whether it will slow down performance if the cards are placed on a PCIe 3.0 motherboard. SLI support doesn't go anywhere, devices can be connected via an NVLink connector if multiple cards are housed in a single machine.
An important innovation in the series is the introduction of new I / O features under the RTX IO name – with AnandTech highlighting the manufacturer is believed to be embedding the Microsoft DirectStorage API, which allows asynchronous data streaming from storage to the GPU, largely omitting the CPU from the process, thus reducing latency. The DirectStorage API is expected to debut on Microsoft’s in-tube Xbox Seriex X console. This also opens the door for Ampere GPUs to directly decompress game content, taking the burden off the CPU's shoulders – which can greatly improve game load time later on.
SURPRISING ATTRACTIVE PRICE SHEETS
Turning to the cards themselves, the line is opened by the RTX 3070, where Nvidia mentions performance in excess of the GeForce RTX 2080 Ti – although the company hasn’t talked about exactly how much extra can be expected. However, the promise is still extremely encouraging, especially since the price of the top-of-the-line RTX 2080 Ti card has exceeded $ 1,200 net, while the RTX 3070 will hit stores with less than half the net price tag of $ 500.
Progress is even more drastic for the RTX 3080, with the mid-range model starting at $ 700 net – in return, it will bring about double the performance of the said RTX 2080 Ti model, according to the company. The friendly pricing is no longer followed by the RTX 3090, which tops the series, so it will have to ring a rather brutal net $ 1,500, but in return the card will run games at 60 FPS with Nvidia-enabled raytracing – at 8K resolution.
As for the GDDR6X memory, the RTX 3080 has 10 gigabytes, while the 3090 has a massive 24 gigabytes – the RTX 3070 has 8 gigabytes of field GDDR6 memory. This is paired with 5888 CUDA cores on the latter card, 8704 on the RTX 3080 and 10496 CUDA cores on the RTX 3090 – in comparison, the RTX 2080 works with Ti 4352 CUDA cores. With the RTX 3090, the manufacturer talks about a 384-bit memory bus, a 1.695 GHz peak, and 19.5 Gbps memory bandwidth, all with a chunky 350-watt TDP. The RTX 3080's values are a few more restrained, the device has a 320-bit memory bus, 19 Gbps bandwidth, and at the same time the TDP has been reduced to 320 watts, which of course far exceeds the RTX 2080 Ti's 250-watt appetite – but the peak signal is here 1 , 71 gigahertz. The situation with the RTX 3070 is different, with 220 watts of TDP, 256-bit memory bus and 16 Gbps bandwidth, and a maximum clock speed of 1.73 gigahertz.
With the addition of new cards, HDMI 2.1 is finally coming, which, thanks to its percussive 48 Gbps bandwidth, paves the way for 8K and 165 Hz 4K displays – although it is not yet known whether the standard will be available on the new RTX models. the total possible bandwidth provided by Of course, the technology also supports Variable Refresh Rate (VRR).
The GeForce RTX 3080 will be the first of the new models to debut from September 17, the RTX 3090. top card will follow on September 24, and the most wallet-friendly RTX 3070 will hit stores in October.