[ad_1]
After months of speculation, Nvidia has finally unveiled Gamescom's next-generation graphics architecture, which CEO Jensen Huang has seen as the biggest breakthrough in GPU technology since the introduction of CUDA cores in 2006. lighting techniques that we think are still far from being possible in the coming years and reshapes the now classic design of the Nvidia reference cooler. But all this has a price.
Next-generation graphics cards are much more expensive than their predecessors, even considering the recent price problems faced by graphic card buyers around the world. Are all the new features of the RTX 2000 series worth it?
The cards
Nvidia had three cards to show at Gamescom: the RTX 2070, the RTX 2080 and the RTX 2080 Ti. It was a surprise, because with past generations, Nvidia has spread the showcase and the output of these cards over a longer period. While the 10 series saw the 1080 release, followed by the 1070 one month later and the 1080 Ti one year later, the Ti 2080 and 2080 were originally scheduled to go on sale from September 20, while the 2070 has a possible release date of October.
We have bad news regarding these release dates. RTX 2080 delivery has been confirmed for September 20th. However, the 2080 Ti will be delayed until the 27th. In a message on the forum, a representative of Nvidia said that the company was waiting for pre-orders between the 20th and the 27th.
"The general availability of the GeForce RTX 2080 Ti has risen to September 27, a delay of a week," said a representative of Nvidia. "We expect pre-orders to arrive between September 20th and September 27th."
These cards have different prices for each one, Huang claiming that third-party reference cards (which are not overclocked or have custom cooling solutions) would be cheaper and slightly less clocked than Founders Edition cards. The Nvidia FE GPUs will be launched at $ 1,200 for the RTX 2080 Ti, $ 800 for the RTX 2080 and $ 600 for the RTX 2070. The reference models will be significantly cheaper at $ 1,000, $ 700 and $ 500 respectively, but Remains much more expensive than previous generation cards have been. The equivalent 10-series GPUs debuted at $ 200-300 less.
As a general rule, we expect more mid-range cards at more reasonable prices to be launched a few months after the flagship GPUs. A hypothetical RTX 2060 and RTX 2050 could appear before the end of the year, but Nvidia has not yet made an official announcement to that effect.
The raw numbers are not as big as one might think
Although the majority of Nvidia Gamescom conferences have been devoted to discussing new technologies supported by next-generation graphics cards, the numbers are likely to interest gamers and hardware enthusiasts alike. Nvidia followed the presentation a few days later with a slide, showing that in the actual gameplay performance, the RTX 2080 behaves at about 1.5 times the speed of the GTX 1080. Nvidia has used a multitude of next generation games as Epic Infiltrator, Wolfenstein II, and Shadow of war.
Here are the specifications of what we are looking at:
RTX 2080 Ti | RTX 2080 | RTX 2070 | GTX 1080 Ti |
GTX 1080 | GTX 1070 | |
CUDA cores: | 4,352 | 2,944 | 2,304 | 3,584 | 2,560 | 1,920 |
Basic speed: | 1,350 MHz | 1.515 MHz | 1.410 MHz | 1.480 MHz | 1.607 MHz | 1.506 MHz |
Speed Boost: | 1.545 MHz | 1.710 MHz | 1,620 MHz | 1.582 MHz | 1,733 MHz | 1.683 MHz |
Memory |
11 GB of GDDR6 | 8GB GDDR6 | 8GB GDDR6 | 11 GB of GDDR5X | 8GB GDDR5X | 8 GB of GDDR5 |
Speed of memory: | 14 Gbps | 14 Gbps | 14 Gbps | 11 Gbps | 10 Gbps | 8 Gbps |
Memory interface: | 352 bits | 256 bits | 256 bits | 352 bits | 256 bits | 256 bits |
Memory bandwidth: | 616 GB / s | 448 GB / s | 448 GB / s | 484 GB / s | 352 GB / s | 256 GB / s |
Power: | 250 watts | 215 watts | 185 watts | 250 watts | 180 watts | 150 watts |
Note: Nvidia Founders Edition models will be launched with price tags, power requirements and clock speeds slightly higher than the reference models, which will be overclocked and modified by third parties.
There are a number of interesting inter-generational changes at play here. CUDA cores have increased the same type of numbers – but no percentages – between the 9oo and 1000 series graphics cards, which should result in a noticeable, even significant, increase in overall performance. The clock speeds have actually decreased, which is not entirely surprising, but it's a bit more when they are associated with an increase in power consumption. These RT and Tensor nuclei may need some juice.
GDDR6 memory provides a solid bandwidth and speed bandwidth for the 2000 series, bringing the two models 2080 and 2070 closer to the GTX 1080 Ti, but not quite.
The biggest thing to remember from these numbers is that for all the advances in ray tracing and intelligent processing capabilities of AI, the RTX 2000 series is much more like a typical graphical generation leap. . We will not know for sure what this means before getting our hands on the material in a few weeks, but that could mean that we are aiming for a much more conservative overall performance increase, rather than the "10 times" numbers that Huang has repeated ad nauseam during his presentation with a focus on ray tracing.
Ray tracing and AI
Although the number of traditional CUDA cores in the new graphics cards has increased at all levels, the most interesting achievement of this new generation, we are told, is the addition of hardware dedicated to ray tracing and IA. The Turing architecture includes RT cores that use clever tricks to accelerate ray tracing, which allows for realistic lighting and reflections in gaming without too much overhead.
These RT cores will work in parallel with Turing Tensor cores, which use supercomputer-driven AI to fill in the blanks using a technique called "denoising" – a new form of advanced anti-aliasing. Huang also mentioned the possibility of rendering fové, which could help make virtual reality titles much less greedy in hardware by concentrating the processing power where the player looks and by making everything in its peripherals at a level of lower detail.
These new technologies at the heart of the 2000 Series architecture allow some games to take advantage of real-time thinking and advanced anti-aliasing techniques like never before. The demonstrations at Gamescom showed us explosions that would not usually be visible to the player, being rendered and reflected in materials such as a car door or eyeball character, which are visible.
It's beautiful. Spear throwing has often been considered as the ultimate goal of digital visuals, making true rays of light in a scene. Nvidia made it clear during his demonstration that his new cards are much more powerful to handle this type of rendering than any card that has come to him. The RTX 2070 would have been faster than the Titan XP. At the ray tracing.
But as with all generations of graphics cards that preceded it, this metric is not the only one to measure graphics cards.
Ports, noise and cooling
Although the hardware inside the 2000 series cards has changed, the exterior has also undergone a major overhaul. After decades of unique fan, reference graphics cards, Nvidia has added a second to its design. He created a much more insane wind tunnel, no doubt to keep the temperatures consistent with the power requirements of the new RTX series graphics cards.
We are assured by Huang that this also has much quieter graphics cards. Even when the clock is completely overclocked, we are told that the temperatures remain constant and the noise at a comfortable level. Providing such claims will require more than a non-powered card on stage, but better cooling of reference cards will be an asset for those who do not want to wait for third-party alternatives.
When you connect your screen, the 2000 series offers some port options. In addition to the more typical DisplayPort 1.4 and HDMI 2.0b connectors, there is also a VirtualLink, USB C-shaped port designed to provide power and video to virtual reality headsets. Although none of the major supports this feature for the moment, this could mean a reduction in cables and cable size for VR in the future.
Is the performance worth the price?
As exciting as all of the above features and specification enhancements are, they must be taken in the context of the costs. The RTX 2000 series cards are the most expensive new Nvidia graphics cards that have been launched long ago. The RTX 2080 Ti will cost $ 1,200 for the Founders Edition and not less than $ 1,000 third. The 2080 costs respectively $ 800 and $ 700, while the 2070 will cost $ 600 and $ 500.
This is a lot of money, especially for what should be a mid-range card in 2070. With its much closer specifications of the RTX 2080, Nvidia could consider offering this card as a high offer range, with more affordable GPUs, as a potential 2060 and 2050 coming later.
Developers are unlikely to adopt ray tracing much before there is a sufficient number of RTX compatible cards and we still do not know how other GPUs such as 10 or AMD Vega cards and RX will be able. manipulating it themselves. These cards may still be in competition with the 2000 series in no-throw games, which will be the majority of years to come.
Updated September 15, 2018: Update with news that the RTX 2080 Ti has been delayed for a week.
Source link