Information about Xe GPU cards and Intel’s new graphic architecture

When Intel announced its intention to launch a graphics card to compete in the flaming graphics market, many followers and those interested in technology, in general, were puzzled. The reason behind these questions was logical of course, there are a lot of questions that roam the ideas of these followers, such as what architecture the company will use, how it will be built, what foundries will be responsible for the manufacturing process, or even what the accuracy of the targeted manufacturing and the techniques that will be enjoyed It will support x-ray tracking, the markets that the company will target and many other confusing questions for followers.

Over time, information began to gradually continue, such as the label for cards. We have already talked about these graphic cards from the company in the past and we have covered the information that was available at the time, but now that more information has emerged and we have entered in 2020, we had to make a new theme in which we are covering the information we currently have about Xe GPU landing cards. The expected 100.

In fact, Intel manufactures a number of cards and not a single card, according to the company’s statements, cards will be available for the category of games and professionals but also servers. So we are faced with a range of new products and not a single product, which either changes the nature of competition in the graphics card market forever, or it is one of the other projects that the company fails to do and then folds its pages. But in both cases, it is a new experience that the company deserves to be thanked for.

On November 8, 2017, Mr. Raja Kaddouri resigned from AMD and joined Intel’s ranks to become vice president of the visual and general computing departments (graphics and laptop sections), and then began to bring his old team and AMD colleagues into Intel’s ranks. Then in June of the following year, then-CEO Bryn Krzanich revealed to investors their design for the engineering of separate graphics cards that the company has been working on for years, and they plan to release them in 2020.

Intel’s news and statements on graphics cards then went to the bottom of intel’s graphics card announcements until Gregory Bryant, vice president of computing for customers, confirmed at the CES conference in 2019 that intel’s first generation of graphics processors would arrive with a 10 nm manufacturing contract (the manufacturing accuracy that the company currently operates in 2019. Also with its central processors). Jim Jeffers, chief engineer and director of rendering and visualization, announced that Intel Xe cards would support radiation tracking at the FMX19 event.

A period of calm followed by a new news to change the course of things and expectations built by many about the new Intel card, specifically on the 17th of November 2019, where Raja Kudori revealed that xe cards will come with three flavors or copies, high-performance cards, and low-energy cards High-performance cards for computing operations. He also stated that the graphics processing units in the latter category will come by the code name Ponte Vecchio, and will come in 2021 with 7nm manufacturing technology.

Then, at the beginning of this year 2020, a new image of the DG1 SDV card (an abbreviation of the strangest name of a graphics card is: Discrete Graphics One Software Development Vehicle), this image showed us a small card with RGB lights to help developers develop engineering software Xe cards. Finally, next March, 17 March 2020: Chief Developer Relations Engineer Antoine Cohed will “provide a detailed tour of the architectural nature” and xe’s “performance implications” at the GDC.

Technical specifications

Although there is talk that intel Xe GPU cards being built will depend on a new architecture from the ground up, Intel’s Gregory Bryant explained that the company’s first separate graphics cards will be built on the company’s current IP graphics architectures. This means we expect Xe cards to be primarily based on the Gen11 GPU architecture of the new Ice Lake processors.

The way intel’s current graphics card silicon units are designed means that they consist of a number of execution units (EUs) arrayed in groups of 8 related units. This group is called sub-slides or subslices, and eight of these substratum are grouped together to form a whole slide, sharing l3 cache and point data background. This 8 x 8 full-chip group gives the new generation Gen11 GPU from Ice Lake the number of execution units that have (64 units).

DG1 Low Power-Consuming Cards

The DG1 SDV or DG 1 LP as is customary in the first separate card which will be a low-power version, the card was displayed at CES2020 and is running Destiny 2 with the distinctive RGB lighting. The card will mostly come with the same basic installation we mentioned with Ponte Vecchio cards. This means (according to the available information that we spoke of at the top) 96 EU or an execution unit, as described in the CompuBench database list.

These cards are referred to by Intel itself as cards (5 to 20 watts) in terms of energy consumption, or from (5 to 50 watts) with other cards (these two types are mostly for mobile devices and compressed devices), and separate cards for desktops will come with consumption of (20) To 50 watts.

Results then appeared on intel’s GeekBench platform integrated with the Rocket Lake processor with 768 1.5 GHz cores with a 2.3 TFLOPs computation capacity, meaning that we have a card built into the GTX 1650’s performance almost in a computational way, of course, assuming that these frequencies will remain the same without increases. In addition to this information, there is another information that has emerged stating that the company intends to work on activating the work of the card integrated with the external cards to get better performance with less energy consumption (and cost of course).

DG 2 High-Performance Cards

DG2 cards are other separate cards but are named differently because they are for high-performance desktops. These cards are referred to as Xe HP, which Raja Kaddouri referred to in a tweet he launched a while ago to mention that this card is one of the largest silicon chips currently in place. We also got some other information at CES 2020 on these cards as we are in front of three high-performance Intel XeHP graphics cards. The final number at the end of the code name can be considered the total number of units or execution units and we see in front of us 128, 256 and 512, so given a large number of execution units we can expect a distinct performance of the cards as we have already indicated if they come at high frequencies.

Ponte Vecchio High-Performance Computing Cards

Ponte Vecchio or Xe HPC are the third category of graphic cards that symbolize HPC characters as “high-performance computing” to be used as graphics cards for data centers and workstations. This category of cards, due to its large power consumption beyond known limits, claims to be the primary engine of high-performance supercomputers and computing, to be able to operate supercomputers such as aurora supercomputer sought by the United States Get to him. I’m not going to talk about these complex technical details here, but we’re going to talk about it in another article so I don’t take any longer.

Xe architecture performance from Intel

As we pointed out above there are some results that have appeared on geekbench for a compact graphical segment of Intel’s new graphics generation. But the most distinctive is the 3D Mark results that emerged through the _rogame member, which reviews the performance of separate DG 1 cards with the same category cards integrated into Tiger Lake-U processors versus the performance of the new RYZEN 4000 Renoir APU processors.

The results were extremely impressive as the separate card was able to overcome all the other cards in this competition, but the RYZEN 4800 U processor was able to get close to these results. We can see that intel’s external card delivers up to 40% better test performance than the built-in AMD graphical processor except in one test, the first test that relies heavily on the overall lighting process and shadows. The reason here may be the lack of improvements in these points, which can be improved over time.