The prototype of the unreleased GeForce RTX 3070 Ti graphics card with 16 GB of memory showed up in the photo

NVIDIA was planning to release another model - GeForce RTX 3070 Ti with 16 GB of memory as part of its GeForce RTX 3000 series of graphics cards.
This gas pedal for a long time appeared in various databases of regulators, but eventually did not appear on sale.
Insider MEGAsizeGPU shared photos and characteristics of unsuccessful reference graphics card.Image source: Twitter / MEGAsizeGPUIn appearance GeForce RTX 3070 Ti 16GB was an exact copy of the model with 8 GB of memory, which eventually went on sale.
However, the characteristics of cards are different.
In spite of the fact that the 16GB version was supposed to use the same GA104 GPU based on Ampere architecture with 6144 CUDA cores, 192 tensor cores and 48 RT cores as the 8GB version which went on sale, the chip of the \"older\" model operates at slightly higher frequency range.
As you know, the version with 8 GB of memory GPU has a base frequency of 1575 MHz, and under load it overclocks to 1770 MHz.
The 16GB version of the card, which never went on sale, GPU worked at a frequency of 1590 to 1800 MHz.Cards differed not only the volume, but also the standard of video memory.
Unreleased model used less efficient GDDR6 instead of faster GDDR6X memory.
The GeForce RTX 3070 Ti 8GB model uses a 256-bit memory interface and GDDR6X chips with 19Gbps per pin.
This provides a memory bandwidth of 608 GB/s.
The model with 16 GB of memory was supposed to get GDDR6 chips with 16 Gbit/s per pin, 256-bit bus support and 512 GB/s of memory bandwidth.Unfortunately, the insider did not share the performance data of the GeForce RTX 3070 Ti 16GB.
However, the fact that the gas pedal supports the current versions of the NVIDIA graphics driver gives hope that this information will eventually be published.

ASUS introduced GeForce RTX 3060 Ti Dual with GDDR6X memory and new design

ASUS introduced the GeForce RTX 3060 Ti graphics gas pedal with GDDR6X memory in Dual version.
Not only does it get faster memory, but it also has a freshened up look that sets it apart from other Dual performance graphics cards from the same vendor.Image source: ASUSIt's even more compact than the original.
Its dimensions are 230 × 120 × 42 mm.
By comparison, the original GeForce RTX 3060 Ti Dual has dimensions of 269 × 136 × 52 mm.
Recall also that ASUS releases Dual Mini model with dimensions 200 × 123 × 38 mm.New ASUS GeForce RTX 3060 Ti with GDDR6X memoryOriginal ASUS GeForce RTX 3060 Ti Dual (GDDR6)ASUS GeForce RTX 3060 Ti Dual (GDDR6)Updated GeForce RTX 3060 Ti Dual with GDDR6X memory will be offered in two variants: with higher and lower factory overclocking GPU.
The GPU model marked DUAL-RTX3060TI-8GD6X has a declared range of maximum operating frequencies from 1665 (standard) to 1695 MHz (overclocking mode).
The chip card marked DUAL-RTX3060TI-O8GD6X operates in the range from 1680 to 1710 MHz, respectively.At the core of the GeForce RTX 3060 Ti uses NVIDIA GPU GA104 on Ampere architecture with 4864 CUDA cores.
The updated model gets 8 GB of GDDR6X memory at 19 Gbps per pin.
The original version of the GeForce RTX 3060 Ti, released in 2020, uses GDDR6 memory at 14 Gbps per pin.We should also add that the updated gas pedal ASUS GeForce RTX 3060 Ti Dual has lost one video output.
The card has three DisplayPort connectors and one HDMI.
The same set has the Dual Mini model.
The original GeForce RTX 3060 Ti Dual has one more HDMI port.

NVIDIA unveiled an improved GeForce RTX 3060 Ti with GDDR6X memory and a degraded GeForce RTX 3060 with 8GB GDDR6

NVIDIA officially unveiled an improved GeForce RTX 3060 Ti with GDDR6X memory as well as a degraded GeForce RTX 3060 with a reduced to 8GB GDDR6 memory.
Characteristics of the cards appeared on the company's website.
Some partners NVIDIA, such as Manli, ASUS and Colorful, have previously provided their versions of these gas pedals.Image source: NVIDIA \"To diversify the choice of gamers and creators of digital content, we present some new options from our network of partners around the world.
They will be available in November.
In addition to the regular GeForce RTX 3060 Ti with GDDR6 memory we are introducing the faster GeForce RTX 3060 Ti with GDDR6X memory.
The second gas pedal is the GeForce RTX 3060 with 8GB of GDDR6 memory.
Support for both products are included in the new version of the graphics driver Game Ready Driver\" - says NVIDIA.
Model GeForce RTX 3060 8GB - this is actually a copy of the original GeForce RTX 3060 with 12 GB of memory, but with reduced to 8 GB of memory GDDR6 and cut to 128 bits memory bus.
As a consequence, the memory bandwidth of this gas pedal decreased from 360 to 240 GB/s.
This could have a negative impact on gaming performance.
At the basis of a new version of the gas pedal is used the former Ampere GA106 graphics processor with 3584 CUDA cores.
The use of faster GDDR6X memory in the GeForce RTX 3060 Ti, in turn, should have a positive impact on the performance of the graphics card in games.
The faster memory provides a higher bandwidth at 608 GB/s.
Recall that the original version of the GeForce RTX 3060 Ti has a GDDR6 memory bandwidth of 448 GB/s.
The new version of the gas pedal has only changed the standard of memory used.
The volume of 8 GB and support for 256-bit memory bus is preserved.
At the basis of this graphics card is used the former GA104 graphics processor with 4864 CUDA cores.

New GeForce RTX 3060 Ti with GDDR6X memory was faster than GeForce RTX 3070 in some games

Galax has provided the first gaming performance data on the new version of the GeForce RTX 3060 Ti with GDDR6X memory, which was introduced a few days ago.
The faster GDDR6X provides up to 35% higher memory bandwidth than GDDR6.
This should have a positive impact on gaming performance, especially at high graphics settings.Image source: GalaxThe Galax company will release a GeForce RTX 3060 Ti Plus graphics card with 8 GB of GDDR6X memory.
The manufacturer compared the new product to the older model GeForce RTX 3070, also equipped with 8 GB of memory, but slower GDDR6.
Recall that the GeForce RTX 3070 uses a GPU with 5888 CUDA cores, while the GeForce RTX 3060 Ti is equipped with a GPU with 4864 CUDA cores.
GDDR6 memory of the former provides a bandwidth of 448 GB/s, the second equipped with GDDR6X memory, this figure is 608 GB/s.
According to the video published by Galax, GeForce RTX 3060 Ti with faster memory provides a slightly higher level of performance compared to the GeForce RTX 3070 in a number of games.
The tests were conducted in three games: Spider-Man Remastered, Cyberpunk 2077 and A Plague Tale: Requiem.
Unfortunately, the manufacturer did not specify what resolution and graphics settings were used.
In the first two games, Galax compared two almost identical scenes.
In Spider-Man Remastered, the GeForce RTX 3060 Ti with GDDR6X memory provided 81 FPS of performance, while the GeForce RTX 3070 provided 75 FPS.
In Cyberpunk 2077, the advantage of the model with GDDR6X memory remained.
It provided an average of 70 frames per second, and the GeForce RTX 3070 model - 61 frames per second.
But in A Plague Tale: Requiem game GeForce RTX 3070 model was faster - 81 vs.
78 FPS of GeForce RTX 3060 Ti with GDDR6X memory.Officially the company has not yet presented GeForce RTX 3060 Ti Plus card.
At the time of writing this text the gas pedal is not listed on its official website.
However, Galax's statement once again confirms that this video card model with GDDR6X memory does exist and will soon go on sale.

NVIDIA will talk about the new graphics architecture on September 20

NVIDIA management representatives at the last quarterly reporting event tried to emphasize that the current difficulties in selling finished products in the gaming segment are temporary and the demand for gaming solutions remains fundamentally high.
The head of the company even said that by the end of the year most of the difficulties should be solved, and in the new year NVIDIA will enter with a new gaming architecture.Image source: NVIDIAAfter his words Jensen Huang has promised to tell more details about this architecture at the GTC, which will start on 20th of September.
The speech of the CEO and CFO of NVIDIA at the last event and then interspersed with hints of the approach of the announcement of a new generation of graphics cards.
Colette Kress even literally said that gaming innovations are \"just around the corner\", and the company's founder added that not only will the new generation of calculation gas pedals with the Hopper architecture, but also the new generation of \"computer graphics cards\" will soon enter the market.Analyzing the situation with surplus gaming products, Jensen Huang expressed the hope that a couple of quarters to solve most of the problems, and by the end of this year NVIDIA \"will be in good shape\".
He also explained that in the new year the company will enter with a new gaming architecture, about which he expects to tell the audience next month.As rightly noted the founder of the company, in the past and current quarters NVIDIA will have to reduce the supply of its products to partners, as well as to carry out price adjustments to quickly free up inventory, but at retail sales of gaming graphics cards, even in such circumstances will be 70% higher than before the pandemic.
This indicates a preservation of high demand in the gaming segment in the medium term.Ampere head of the company believes the most successful architecture of graphics processors ever created by NVIDIA.
The new generation of gaming graphics, he said, will be located at a level above the existing one.
Should this be taken as a hint at a higher price level, it is difficult to say unequivocally, because for the implementation of Ampere stock prices for the products of the current generation will have to decrease.
According to Jensen Huang, the average selling price of gaming video cards in the absence of a significant influence of the cryptocurrency factor will also decrease.
He believes that the average price of graphics cards should be comparable to the price of a gaming console, now it is around $ 500 or $ 599.CFO Colette Kress, in turn, told about the average sales revenue from NVIDIA video cards in retail channels.
Although the company will have to reduce their shipments, their partners will still be able to sell $5 billion worth of gaming graphics cards during the second and third quarters of this year.
NVIDIA's gaming revenue in the second quarter barely exceeded $2 billion, and in the third quarter it will still decrease in sequential comparison.
The imbalance between NVIDIA's deliveries and partners' revenues from video card sales is inevitable given the stock overstocking.

NVIDIA promises to introduce Ampere successor in gaming segment in the second half of the year

Since NVIDIA founder Jensen Huang is very tight-lipped about the timing of new products, one of the few public sources of this type of information remains comments by CFO Colette Kress.
Today she didn't let expectations slip and made it very clear that NVIDIA will introduce a new architecture in gaming segment by the end of the year.Image source: NVIDIAThe interesting thing is that the first remark on this topic during the quarterly conference Mrs.
Kress made on her own initiative, without leading questions from the analysts invited to the event.
In discussing the potential for gaming revenue decline in Q2 on a sequential basis, the CFO explained, \"We are preparing for a transition to a new architecture later this year and expect a sequential decline in gaming revenue in Q2.\" As noted in our previous article on the subject, NVIDIA's gaming revenue might decline 10-20% sequentially in Q2, but year over year growth will still be seen.
Lockdowns in China and events in Ukraine will have a significant impact on such dynamics, but these factors deserve a more detailed discussion in a separate publication.At the same time Colette Kress emphasizes that Ampere is a great architecture, which will continue to be in high demand.
The GeForce RTX 30 family of graphics cards has shown the best sales momentum throughout its lifecycle since it was announced in the fall of 2020.
The PC platform has managed to attract about 100 million new gamers over the past two years, according to Newzoo.
The number of games with RTX support has doubled to 250 titles over the year, and the share of GeForce owners with access to this technology is now approaching one-third, according to NVIDIA itself.The main processes involved in the migration to new graphics architectures for NVIDIA will take place in the second half of the year, as Colette Kress explained.
This should be understood in relation to both the gaming and server segments.
For example, the same Hopper in the server segment will appear as early as the third quarter, but the buildup of shipments of carriers of this architecture will not begin until closer to the end of the calendar year.
\"The bulk of the new product launches will be accompanied by a scaling of shipments in the second half of the year,\" Colette Kress added.

GALAX GeForce RTX 3090 Ti HoF gets 4.3-inch display and support in the kit

The first absentee study of the graphics card GeForce RTX 3090 Ti HoF by GALAX (KFA2) revealed that the power system is made in the \"24 + 4 phase\" scheme, and for additional power are responsible for two connectors type 12VHPWR.
The novelty's official description on the vendor's site gives you a better idea of what it is.
Advertisement Image source: GALAX (KFA2) OEM air cooling system includes three fans, two of which have 102 mm diameter impellers and the middle one has 92 mm diameter.
The HOF Panel III display is 480x272 pixels, 4.3\" full color and 4.3\" diagonally mounted via USB, and can be mounted magnetically or on the table by means of a tilting stand.
In conjunction with the display works proprietary utility HOF AI Software, which is now installed and smartphone, allowing you to control overclocking or work RGB-backlight.Image source: GALAX (KFA2)At the graphics processor graphics card provides three preset frequencies: 1860 MHz, 1920 MHz and 1935 MHz.
The latter can only be called from the proprietary utility.
The graphics card has dimensions of 339 x 144 x 68 mm and is equipped with 24 GB of GDDR6X memory.
The package also includes a special adjustable aluminum support (pictured), which allows you to better distribute the weight of the video card after installing it in the system unit.
Moreover the graphics card comes with brand-name gloves and two power adapters which allow connecting three eight-pin power supply connectors to one 12VHPWR socket.
Three DisplayPort 1.4a outputs and one HDMI 2.1 output can be found on the rear panel.

Gigabyte to unveil G7, G5 and A7 entry-level notebooks with GeForce RTX 3060 and AMD and Intel chips

Gigabyte's new low-cost laptops are expected to go on sale next month. They will be unveiled in January at CES 2021 right after the announcement of new AMD mobile processors and next-generation NVIDIA graphics cards.

The G5 series & ; are 15-inch laptops with state-of-the-art Intel Core 10th generation processors (Comet Lake-H). The devices will get Core i7-10870H or Core i5-10500H CPUs. In addition, the G5 will offer a 1080p panel at up to 240Hz. This gaming laptop will be equipped with GeForce RTX 3060 Ampere graphics with 6GB of GDDR6 video memory.

The 17-inch G7 model will get a Core i7-10870H processor. Later, Gigabyte may add additional options with other CPUs. The 17.3-inch IPS display will be able to offer a 144Hz refresh rate. The G7 will also use GeForce RTX 3060 mobile graphics.

Finally, the Gigabyte A7 model is quite interesting. It is the first laptop with a Ryzen processor from this manufacturer and one of the first in the world to reportedly get the latest Ryzen 7 5800H Cezanne processor (eight Zen 3 cores, 16 threads and Vega graphics). It is identical in design to the G7 series, so a direct comparison between these devices will be extremely curious. The laptop will also be equipped with a mobile graphics gas pedal GeForce RTX 3060, but, unlike Intel variants, there will be available models with a higher power graphics card Max-P.

NVIDIA will present its AMD Smart Access Memory analog, which will accelerate Ampere operation on any platform

AMD presented Smart Access Memory technology during the announcement of Radeon RX 6000 graphics cards. The point is that while in conventional Windows-based PCs processors can only access part of the graphics memory (VRAM) at the same time, it can be used to expand the data channel, allowing the processor to immediately use the entire video memory array and eliminate potential bottlenecks, achieving improved performance.

Future

AMD has published tests of Radeon RX 6000, in which it demonstrated the performance gain by using Smart Access Memory in conjunction with the latest Ryzen 5000 series processors. But NVIDIA already has something to answer that advantage. According to the company's announcement for Gamer's Nexus website, it will present a similar feature that improves the efficiency of data transfer between the graphics gas pedal and processor. In fact, in the NVIDIA lab, this functionality already works on Ampere graphics cards. In addition, NVIDIA claims that its technology will be versatile and will be able to work equally well with Intel and AMD processors, including through the PCIe 3.0 bus, while AMD's solution requires the AMD Ryzen Series 5000 CPU, X570 chipset motherboard and Radeon RX 6000 graphics gas pedal. The description says that Smart Access Memory allows the CPU and GPU to exchange information over a wider PCIe channel. Technical details are unknown, and AMD simply says that the CPU and GPU are usually limited to «aperture » 256 MB of data transfer, which binds the hands of game developers, and also requires frequent switching between CPU and main memory if the amount of data exceeds this limit. The result is lower efficiency and lower performance. Smart Access Memory, according to AMD, removes this barrier. This description is similar to the BAR size changes in PCIe & ; the standard feature described in the PCIe specifications is considered by NVIDIA. And if the graphics processor supports it, the performance can be improved by simply setting the appropriate parameter in the BIOS of the graphics card and motherboard. NVIDIA said its GPUs support this feature, although it does require activation. As for processors, any Intel or AMD chip working with PCIe bus should be able to enable this technology. AMD may in the future reduce CPU and motherboard requirements if NVIDIA does in fact introduce similar functionality on all platforms. NVIDIA currently claims that its early testing shows similar performance improvements to AMD's Smart Access Memory and that similar functionality will be included in future firmware upgrades. However, the company has not announced the release date of this update.