As you know, the green team is preparing to launch its flagship Geforce RTX 3090 graphics card during the month of September. NVIDIA intends to keep a tight state of secrecy about its upcoming graphic architecture, Ampere, and the company has decided to take the name Ampere after the physicist Andre Marie Ampere, founder of electromagnetic energy. Nvidia is one of the companies that is very keen not to disclose its secrets and modern technologies until the exact date of their launch.
Despite this, many technical sources were able to get their hands on some information about the RTX 3090 card, and this was thanks to a PDF file that was mistakenly published on the official page of Micron, which was a publication entitled “Demand for high-bandwidth memory” which The company deleted it again but it was too late. Most technical websites have already managed to access this file and read its content. As it became clear to everyone, this post was a promotional advertisement for Micron’s new video memories that are being used and sold to official partners. The most important thing in the Micron publication is that the next generation video memory will come from the GDDR6X model, and this is what caused a great media sensation in the world of technology and hardware.
Before these sites delved into the halls of the upcoming architecture, they had many questions and inquiries about why Nvidia called the next graphics card the RTX 3090 model. If we look at the past, we will notice that the company has not used a serial suffix number 90 since the Kepler architecture with the GTX 600 family of cards. The last graphics card that was released with the suffix of serial number 90 was the Geforce GTX 690 card, which contained two GTX 680 graphics processors. If you remember well, that card was a bulldozer in terms of average power consumption and temperatures. But the strange thing is that Nvidia is absolutely certain that the idea of multiple graphics processors in one card is no longer valid and causes many problems for users, whether in terms of performance or because of the lack of efficient software support for this case of the form of uses.
So the RTX 3090 should only come with one GPU. It is possible that the card contains a graphic processor that contains thousands of cores and microcomponents inside, but the card should not come with two graphic pastes at all costs. There are still questions about what the RTX 3090 Ti will look like, but that’s in case Nvidia was planning to roll it out as well. We wonder what we should see from Ampere’s architecture, what are its technical characteristics, new technologies, and what is the expected price for them. All inquiries we will discuss with you in detail during our article, so make yourself a cup of tea and follow us until the end.
The RTX 3090 GA 102 processor cores are expected to range from 5376 to 7552 CUDA Cores. While the RT nuclei responsible for RayTracing range from 84 to 104 nuclei. Hypothetically, the numbers of cores responsible for artificial intelligence, deep learning, and high-performance computing will come in between 336 to 472 Tensor Cores. For tissue staining and mapping units, there are 672TMUs. As for the drip and eject units, they come in 128ROPs. The video memory will come with a 12GB GDDR6X RAM operating at a frequency of 21Gbps, based on what we saw in the RTX 1080 Ti card, which contained an 11GB video memory with a frequency of 11Gbps. This means that the RTX 3090 is the first NVIDIA graphics card to have a bandwidth of 1TBps or more, which in turn makes it exactly equal to the video memory of the Radeon VII card that comes from the HBM2 type.
If the rumors are to be believed and the video memory characteristics of the RTX 3090 card come in the same way that technical sites expect, then this means that this card will be 60% or more better compared to the RTX 2080 Ti, which contains a video memory operating at a bandwidth of 616GBps. Add to this the advantage of the new engineering architecture and its modern technologies that help achieve higher results on the ground in the characteristics of computing and intensive calculations denoted by FLOPS, whose results are often the same as the rate of the speed of the video memory working frequency, in short, the RTX card is expected to come 3090, including up to 21TFLOPS in math-intensive performance.
Combining both the Memory Bandwidth bandwidth results and the TFLOPS math results, we might end up with a graphics card that is 50-75% stronger in overall performance compared to the RTX 2080 Ti. But we can not yet say for sure that these numbers are the real results just by looking at the numbers of the expected cores and the characteristics of the video memory, and the reason is that there are several different parameters that we cannot predict, such as the real number of all types of cores used, such as the real frequency speed and the architectural efficiency. The graphics card could come with lower frequencies and a higher core count than the expected numbers, or it could come with a lower core count, higher frequency speed, and a higher energy consumption rate.
During the EGX 2020 event, Nvidia announced the A100 accelerated graphics processor, the first graphics processor based on the amp architecture for high-performance computing, data analysis, artificial intelligence, and advanced scientific research. The A100 GPU contains the third-generation Tensor Cores cores, which were able to bring with it four times the power of the first and second generation Tensor Cores cores for both Volta and Turing architectures, respectively, thanks to the 7nm manufacturing accuracy.
Since the Geforce RTX 3090 card is the first graphics card for office devices and gaming computers based on its architecture on the architecture of Ampere, its containment of the third generation of Tensor Cores will greatly help it in terms of performance in intensive computing and deep learning.
What we cannot predict is the performance form of the RT cores responsible for ray tracing technology, the A100 accelerated processor did not come with any of them, but it is expected that the second generation of RT cores will meet the needs of users of RayTracing technology. The only question on my mind and mind is whether these cores will come with four times the performance power of the first generation RT cores contained in the RTX 20`s series cards or will they come even double the performance, especially since the current generation cards suffer a lot when RayTracing is activated in games.
Until this point, ray tracing technology is not even suitable for higher-end cards and cannot be enjoyed widely. While the middle-tier cards are completely deprived. Therefore, we expect that Nvidia has a comprehensive idea about this problem, and it should place its bet on the ability of the second generation of RT cores to deal with ray tracing technology and provide the necessary means for game developers to benefit from it in a saturated manner that is not limited to the effects of reflections and shadows.
Some of the leaks that came out of the Chinese technology sites indicated the arrival of the next generation RTX 30`s cards with a new standard in power-pass ports and rates of consumption. Meanwhile, the new 12pin PCIe Connector was announced via a Chinese website called FCPowerUp at the same time as RTX 30`s cards leaked online. This only means one thing, that graphics ampere cards will have a new standard in the form of power outlets and average consumption.
If these leaks are true, the results may be very bad for me and for you, as most of the power supplies that users have only contain PCIe + 12V cables with 8pin and 6Pin connectors only. So we should buy new power supplies, or in the worst case, buy new standardized switching connectors, whether this or that except that both cases are worse than each other.
But misfortunes do not come individually, as the other bad news came from Igor`s Lap Laboratories responsible for the final stages in the manufacture of electronic computing processors, as it indicated to several sources close to it that the RTX 3090 card will come with an average energy consumption of up to 350W of TDP.
Many tech websites are not able to believe, and in fact they have good reasons for this.If we look away at AMD and the Zen 2 architecture, we will find that the consumption rates and energy efficiency were 30% better than the Zen + architecture thanks to the transition from 14nm to 7nm manufacturing accuracy. Could the Ampere 7nm architecture follow the transition from Turing 12nm without any significant improvement in energy efficiency and consumption rates? The whole thing could be incorrect rumors that will be denied and ignored during the official announcement of the RTX 3090, we have to be patient and not believe everything until we see with our eyes. But until that happens, we will remain in a state of extreme concern about upcoming graphics cards’ power consumption rates.
The RTX 3090 card is expected to be announced during the Gamescom event on September 21, and the RTX 3090 and RTX 3080 Ti may be announced together, but as we mentioned to you previously, Nvidia never declares the future except on time, and until then we cannot Nothing to stress. Another thing about the names, it is expected that the next-generation cards will not come with additional word extensions such as Super, but there are rumors that suggest that it may come with the suffix Ultimate or maybe something else entirely.
As for the price of next-generation cards, they are likely to come at astronomical prices, and there is more than one reason to expect this. The first reason is that each Nvidia architecture is more expensive than its predecessor, another reason is that the next generation cards will certainly achieve a quantum leap in the level of performance within games and will raise RayTracing technology to advanced levels. Add to this the intense demand for 7nm manufacturing accuracy and everyone being at the mercy of the Taiwanese company TSMC responsible for developing and providing adequate quantitative production of new processor chips for NVIDIA, AMD and Apple. Huawei’s exit from this list helped TSMC in creating a very huge space that allows new partners to take advantage of the 7nm manufacturing precision and perhaps here we mean Intel.
But in the context of this matter, you find that all the leading companies in the industry fall to TSMC alone and at its mercy, which in turn makes us believe that a higher price for next-generation cards and processors is inevitable. We must not forget, there is Corona, and therefore factories do not operate at full production capacity and there is a severe deficit in the proportion of working hands to limit the spread of the epidemic among people, and this matter alone causes the system of supply and demand to upset.