According to various sources, Huang will unveil the next generation of Ampere GPU at the GTC’s keynote speech on May 14, and he even recorded a funny video for it. But just recently, a YouTube UP owner released the full information on the new card RTX 3080 Ti.
Huang did it again, this time by putting his own new products in the oven.
On May 12, NVIDIA posted a video on Youtube of Huang removing a DGX A100 from the oven and saying “it’s been baking for a while.”
Needless to say, this eye-catching move is a warm-up for the upcoming release of the Ampere GPU.
The DGX A100, which is powered by the A100 GPU, will be shown off at GTC 2020, and the Tesla A100 will be the first Ampere architecture.
Master Huang in NVIDIA GTC 2020 keynote speech, will be in Beijing time on May 14 at 9 PM officially began, the long-awaited new graphics card series, this is really coming.
Youtuber: Reveal the new detailed parameters
But there have been plenty of leaks on the Internet about the new RTX 3000 series graphics card, which, after all, has consumed too much patience.
Yesterday, May 12, Moore’s Law is Dead, a well-known tech blogger on YouTube, revealed the main specs, performance and comparison of the RTX 3080 Ti.
According to his sources, the full range of graphics cards will use RTX (graphics ray tracing technology), the HPC computing card will be released in May, and the game card will not be released until September.
Based on the new Ampere architecture GA102 core RTX 3080 Ti, from the production process to the core architecture and technical parameters, have been greatly updated.
RTX 3080 Ti has built-in 5376 CUDA, adopts 7nm class process, core frequency reaches 2.2ghz, single precision floating point number reaches 21 TLFOPs, video memory matching is 18Gbps GDDR6, bit width 384 bit, total bandwidth 863 GB/s, Power consumption is maintained at 220~230 watts.
In addition to these strong performance parameters, other information has been revealed previously, such as 4x optical catch performance, will support DLSS3.0, and so on.
In terms of performance, Ampere Gpus can improve IPC performance by 10-20%, double the L2 cache, and run more frequently, reaching the baseline level of 1900MHz.
The RTX 3080 Ti will have the same physical appearance as the RTX 2080 Ti, but it will have three fans instead of two.
It is said that these data from NV insiders, if the information is true, Huang jiazhu this time is a big deal.
Up to 70% performance improvement compared to the previous generation
Compared to the RTX 2080 Ti of the previous generation, the 3080 Ti is really much better.
The 3080 Ti Boost frequency is increased by 42%, the single-precision float is increased by 50%, and the bandwidth is increased by 40%.
In a number of 4K game measurement, 3080 Ti than RTX 2080 Ti frame rate increased by 40%, after the transformation can be 50%, N card special optimization game can achieve 70% performance improvement.
Here’s a detailed comparison of the parameters:
| _ | RTX 2080 Ti | RTX 3080 Ti |
| architecture | Turing | Ampere |
| The core model | TU 102 | GA 102 |
| Process technology | 12 nm | 7 nm |
| Single floating point performance | 13.4 TFLOPs | 21 TFLOPs |
| Stream processors | 4352 | 5376 |
| Memory type | GDDR6 | GDDR6 |
| Memory capacity | 11 GB | 18 GB |
| Memory bits wide | 352Bit | 384bit |
| The core frequency | 1350-1545 MHz | 2200 MHz |
| Memory frequency | 14000 MHz | 18000 Mhz |
| Video memory bandwidth | 616 GB/s | 864 GB/s |
| The power consumption of the TDP | 250 W | 220-230 W |
| External power supply | 8 + 8 pin | 8 + 8 pin |
Given that the RTX 3080 Ti has taken the RTX 2080 out of the box, it’s no surprise that the performance is strong.
As for the authenticity of this revelation, it will have to wait until NVIDIA GTC Keynote tomorrow night to reveal, but considering huang has released the warm-up video, this time he is estimated to have a big trick to put.
Beijing time on May 14 at 9 p.m., Master Huang launched the online broadcast, to view the full video can be accessed:
www.nvidia.cn/gtc/keynote…
– the –