Dark Posted January 4, 2022 Posted January 4, 2022 Graphics cards are on everyone's lips and both NVIDIA and AMD know it. Therefore, whoever dominates the current generation over the other will take but all the cake, a large part. In a preview of what should be the pre-rumor room, Huang's have accepted practically all of them in an event that will be remembered for being the one that has presented the most GPUs at one time. These are the RTX 3090 Ti, 3080 12 GB, 3070 Ti 16 GB, 3050 8 GB and of course the RTX 3080 Ti Mobile and RTX 3070 Ti Mobile. There have not been too many surprises in terms of the official presentation, because as they say almost all the fish was sold weeks before the event. But it must be emphasized that the company has launched an ordeal that AMD has not been able to respond to and where it seems that Intel is the great enemy. The highlight of course has been the 24 GB RTX 3090 Ti, the fastest GPU on the planet for gaming and mining that breaks with everything established so far. A presentation like no other on GPU NVIDIA-GeForce-RTX-30-Series-RTX-3090-ti_RTX-3080_RTX-3070-2 Never before in history have six GPU models of the same series been presented at the same time, this (series) not being a novelty. We are used to similar launches when a new architecture is presented, but being this already veteran it is totally new. For this reason, NVIDIA has not wanted to miss the opportunity and will fill the gaming market with four models for desktop and two more for laptops, protecting itself as it has never seen against its two rivals. RTX 3090 TI NVIDIA-GeForce-RTX-30-Series-RTX-3090-ti_RTX-3080_RTX-3070-1 It is without a doubt the most powerful gaming graphics card in history and on the planet, the ultimate GPU. More cores, faster memory and higher overall performance as well as excessive power consumption, NVIDIA is going big, of course. As expected it will be a 628mm2 GPU manufactured by Samsung at 8nm and will have 28 billion transistors being the chip fully unlocked and the second largest GPU in size after Turing TU102. What we will have in exchange are 84 SMs with a total of 10,752 cores that will go at a frequency of 1,560 MHz with a Boost of 1,860 MHz. It will of course be a GPU under PCIe 4.0 x16 (no PCIe 5.0 as it was rumored) that will integrate to its credit a 384-bit bus that will link with 24 GB of GDDR6X at 1325 MHz or 21.2 Gbps effective to give an impressive bandwidth of 1,018 GB / s, that is 1 TB / s. This bandwidth will have to feed the 336 TMUs, 112 ROPs, 336 Tensor Cores and 84 RT Cores that will be linked directly with the 6 MB of L2 that Integra. Regarding the general performance, NVIDIA ensures a round figure: 40 TFLOPS, which would mean being around 10% above the current RTX 3090. NVIDIA-GeForce-RTX-30-Series-RTX-3090-ti_RTX-3080_RTX-3070-6 The most impressive thing without a doubt is that its consumption or TGP will go to no less than 450 watts, 100 more to take the Ampere architecture to its limit. Therefore, you need the 12-pin connector developed by the company or if it ends up reaching the assemblers three 8-pin connectors. Its price will be 1,499 dollars and it will arrive in this month of January, specifically on the 27th. RTX 3080 12 GB She will be in charge of standing up and looking over her shoulder at what Intel presents, since it is called to be the queen of the high-end, renewed that yes. There are no surprises here either, since with 70 SM we have a total of 8960 Shaders, where the biggest update is precisely in its GDDR6X memory, which now reaches 12 GB at 19 Gbps, so with its 384-bit bus we will have 912 GB / s representing the largest performance increase over its predecessor in this section (+ 20%). Like its bigger sister, NVIDIA hasn't revealed its MSRP price, although it looks set to arrive in the first quarter of this year. RTX 3070 Ti 16GB Regarding this NVIDIA GPU, it has been very brief, since despite the fact that it has the same chip and features as its original 8 GB version, the card is updated with twice the VRAM capacity and also goes to modules. GDDR6X at 21 Gbps, which with its 256-bit bus will give us no less than 672 GB / s. This should boost your performance by about 5%. Unfortunately there are no official prices, but there is a release date, specifically June 10. RTX 3050 8GB There are also no surprises here compared to what was leaked at the time, since this RTX 3050 will be the company's lowest-end GPU and will have at its heart a GA106-150 chip with 2560 Shaders, 20 RT Cores, 80 Tensor Cores and 8 GB. GDDR6 at 14 Gbps attached to a 128-bit bus. Its TGP is encrypted at 130 watts and will cost $ 249 as MSRP, hitting the market on January 27.
Premium
Recommended Posts