Intel is back in the big game of dedicated graphics cards, with the blue team suggesting that it will launch its first graphics solutions Intel Xe in 2020 through a wide selection of platforms, environments, and prices. This will be the first new discrete graphics card from Intel in over twenty years.
The technology behind this GPU, and its performance potential is still shrouded in mystery, but as the time passes, the details come to light. If it proves to be a viable alternative to Nvidia and AMD, this will be one of the most important events in the industry of graphics cards in this century.
This is what we know up to now.
Pricing and availability
Intel has stated that we can expect a debut in the summer of 2020 and that it pointed to a wide selection of graphics options, some aimed at consumer electronics and other to High performance. Intel revealed the first of its GPU-centric business in November 2019. Intel has said that the GPU Ponte Vecchio not be launched until 2021, as it will be based on a second generation of graphics Xe. WCCFTech reported that the first one will be used in the Supercomputer Aurora.
In mid-2019, the executive of Intel , Raja Koduri tweeted a picture of his Tesla, with the name “ThinkXE” in the plate and with a label that shows the month of June and the year 2020. This could be an anticipated announcement of a launch in June of 2020 for the graphics card Xe, perhaps in time for Computex.
In December 2019, however, began to circulate rumors that the development of graphics Xe not progressed fast enough. A user of the forums Chiphell (via WCCFTech), stated that the development of graphics Xe did not work well and that it is possible to not see any GPU Intel before the end of 2020. They also stated that this would delay the Ponte Vecchio to 2022 at the earliest.
— Intel Graphics (@IntelGraphics) August 15, 2018
If true, this would be problematic for a number of reasons, from entry because the graphics Xe are scheduled to appear in the next processors, Tiger Lake Intel, currently scheduled for a release in 2020. This is likely to be a priority for Intel, so that could cause a shortage of GPU dedicated if launched at a similar time.
There is also the rumor that Intel will be the only company that sells their GPUS at launch, as it has struggled to define new relationships with partners of cards that include to create customized variants and with ‘overclock’.
A leak of information on the release of drivers of video from Intel in the summer of 2019 made reference to four discrete graphics cards different. That suggests that for ‘gamers‘ and enthusiasts ‘hardware‘, there will be a selection of relatively wide graphics cards to choose from.
Architecture and performance
When Intel made its official announcement about the new technology of graphics cards that was developing, the company made it clear that he was building a dedicated graphics card. While this suggested that he was building something other than your GPU incorporated existing, these cards will be based on the same architecture generation 12 in the core of their graphics solutions built for the next generations of processor like Tiger Lake, but scaled properly.
Intel previously announced that there will be three different microarchitectures different as part of the range Xe known as the Xe-LP, Xe-HP and Xe-HPC. These will cover graphic chips entry level, graphics card-level enthusiastic, and GPU-based data centers for rendering and supercomputing. However, in each case, Intel will build the brand of “Xe” for all graphics projects in the future, whether integrated or discrete, although there will be big performance differences between them.
The best view of the architecture behind these new cards came in our own report on a slide obtained from Intel that indicated the TDP (Power thermal design, for its acronym in English), and the architecture “mosaic” of these early cards Xe.
According to an internal presentation of Intel of the beginning of 2019, these execution units (EU) are divided into “tiles”, potentially in a manner similar to the chiplets that we have seen used in the processors Ryzen 3000 Zen 2 AMD. When aligned with a leak of information about drivers in July of 2019, it was deduced that these tiles likely contain 128 EU and could be combined with the Bridge interconnect die multiple built-in (EMIB) from Intel.
In another part of the presentation, Intel said to be developing graphics cards with the full spectrum of price and performance, until the end. Includes designs ranging from a TDP of 75W up to 500W. At the upper end, that is double the TDP of the graphics card for games most powerful of Nvidia, the 2080 You. With the above charge, confirming that the GPU will use an input voltage of 48V, that makes it almost certainly a card of high performance.
It is possible that this card is 500W, and four mosaics is the GPU Ponte Vecchio, which Intel announced in November 2019. Although Ponte Vecchio not debut until 2021, Intel called the “first GPU to exascale” and stated that it would use an interface Compute eXpress Link (CXL) that is running on a connection PCIexpress 5.0.
According to reports, the other GPU mosaic single and dual will have a TDP of 75W, 150W and 300W. Although this last is still high, those are TDP much more typical could well be equivalent to the graphics cards of games for desktop computers the conventional. The cards are comparable products that already exist in these levels include the GTX 1650 to 75W, the RX 5600 XT to 160W and the Titan RTX to 280W.
HBM and PCIexpress 4.0
In an interview (already removed) with the head of development for Intel graphics, and Raja Koduri, the leader of Radeon Technology Group from a long time ago, suggested that these new graphics cards could not use the GDDR6 more typical for the memory. Instead, it was suggested that these GPUS could use memory of higher bandwidth (HBM), which is much more expensive.
That was downplayed and discredited, mostly by Intel, but the announcement of the Ponte Vecchio GPU itself showed the memory HBM that is used in the design. The internal documents that Digital Trends obtained in February of 2020 were also aimed at the use of memory for high bandwidth in your GPU-Xe, specifically HBM2e, which is the more efficient generation and more rapid HBM. According to reports, are stacked in 3D directly on the GPU using the technology Foveros Intel.
That does not mean that we will see a memory so expensive used in the graphs of consumption, but it is an interesting note that whatever Intel is working, you should be able to take advantage of HBM. It is rumored that the GPU “Big Navi” next-generation AMD has an option of memory HBM along with GDDR6, so that it is possible that Intel offers something similar with its GPU, which offers the choice of high range HBM and the options of low-end GDDR6.
Most of the graphics cards of consumption adhere to GDDR6, including everything from high-end as RTX 2080 Ti Nvidia. The only recent exception has been the Fury-X from AMD, Vega, 56, 64, and Radeon VII. Their graphics cards RX 5700 latest left HBM2 in favor of GDDR6.
Another note on internal documents from Intel that we have seen was a mention of the support PCIexpress 4.0. Although the current graphics cards do not even come close to the slots saturated PCIexpress 3.0 x16, compatibility with PCIe 4.0 refers to the potential performance of the new Intel cards. It could also be that they are best optimized to require less lanes, enabling more lanes for storage units and other additional cards.
Nvidia has been dedicated to ray-tracing in real time, so that is a key feature of their graphics cards RTX today. Despite its slow start, the technology has the potential to become the new most important feature in computer graphics in the next few years. The problem is that the increase of the lighting and shadows realistic can be costly in terms of performance. AMD has hesitated more about plunge into the world of ray-tracing for that exact reason, although there are plans to support it in the future, especially in consoles like the Playstation 5.
Long before the launch of the Xe, Intel already came out of the door to support the tracking of rays in its future GPU. Jim Jeffers, principal engineer senior of Intel (and senior director of rendering and advanced display), made the following statement about it: “The ‘roadmap’ of the Intel architecture Xe for rendering optimized data center includes support for hardware acceleration of ray tracing to the family of Intel Rendering Framework APIs, and libraries”. Still do not know what means this declaration for the ray tracing in games, but, if you have implemented hardware acceleration, it would surprise us if Intel does not also give it to the players.
Drivers and software
Both Nvidia as AMD have their respective software packages for drivers that do more than help the GPU to correctly communicate with the system in general. The characteristics such as the sharpness of the image, filters, shading, recording of games, the entries of low latency and the setting of dynamic resolution, have improved the offerings of the major manufacturers of GPU. Intel will want to do something similar when they launch their graphics cards Xe and has already begun to lay the groundwork for that.
In march of 2019, Intel presented its Command Center Graphics. At this time, only works with the integrated graphics solutions from Intel, but includes options for launch and optimization of games and the possibility of modifying the options of GPU global in all applications. It is pretty basic for now, and offers a basic functionality to the users of GPU-Intel integrated, but they have laid the foundations for a software package GPU-Intel more full when Xe debut in the future.
Along with the development of hardware, it is reported that Intel is spending a lot of time and energy to their development of drivers, and WCCFTech informs that need a lot of optimization before you see the light of day.
The students of AMD are helping to achieve this
Intel has not released a discrete graphics card in 20 years. He developed what became a coprocessor, Larabee, at the end of the decade of 2000, but it turned out to be far from being competitive with the modern graphics cards, even found some use cases intriguing in its own right. To develop its architecture graphic into something worthy of a dedicated graphics card, Intel has hired some industry experts, especially Raja Koduri. He was hired directly from AMD, where he spent several years as chief architect of the Radeon Technology Group, leading the development of architectures Vega and Navi from AMD.
He has been there for over a year and even in the mid of 2018 were joined by Jim Keller, the chief architect of the architecture Zen of AMD. Directs the development of silicon from Intel and according to Intel, will help to “change the way Intel built silicon”. That could be considered as further evidence of the momentum of Intel towards a viable production of 10 nm.
Other former employees of AMD that Intel chose during the past year include the former global marketing director of products for AMD, Chris Hook, who spent 17 years working in the company, and Darren McPhee, who now directs the marketing of products from Intel to discrete graphics.
Recommendations of the editor