Intel has awarded TSMC with some big contracts for future manufacturing of next generation GPUs, according to Taiwan's Commercial Times. As previously covered on TPU, the second generation Battlemage graphics processing units will get fabricated via a 4 nm process. According to insider sources at both partnering companies, Intel is eyeing a release date in the second half of 2024 for this Xe2-based architecture. The same sources pointed to the third generation Celestial graphics processing units being ready in time for a second half of 2026 launch window. Arc Celestial, which is based on the Xe3 architecture, is set for manufacture in the coming years courtesy of TSMC's N3X (3 nm) process node.

One of the sources claim that Intel is quietly confident about its future prospects in the GPU sector, despite mixed critical and commercial reactions to the first generation line-up of Arc Alchemist discrete graphics cards. The company is said to be anticipating great demand for more potent versions of its graphics products in the future, and internal restructuring efforts have not dulled the will of a core team of engineers. The restructuring process resulted in the original AXG graphics division being divided into two sub-groups - CCG and DCAI. The pioneer of the entire endeavor, Raja Koduri, departed Intel midway through last month, to pursue new opportunities with an AI-focused startup.

Qualcomm has been working on its Snapdragon SoCs for quite some time now, with massive success in the mobile phone space. However, the company's processors needed to be up to the task regarding laptops. For a user to not look at x86 offerings, the only remaining performant alternatives are Apple's M processors. In 2021 Qualcomm purchased the Nuvia team that was developing massively efficient and high-performance IP for laptops, similar to Apple M processors. Today, according to the insights from Kuba Wojciechowski (@Za_Raczke) on Twitter, we have some potential information about the upcoming Nuvia-powered SoC codenamed Hamoa.

According to the Twitter thread, Qualcomm's Hamoa processors are part of the Snapdragon 8xc Gen 4 compute platform and feature up to eight high-performance P-cores and four low-power E-cores, all based on Nuvia's IP. Allegedly the P-cores are being tested at 3.4 GHz, while the E-cores are tested at 2.5 GHz. The SoC splits CPU cores into blocks, each being a four-core group with 12 MB of shared L2 cache. There is also an 8 MB L3 cache structure; it needs to be clarified whether it is per core block or for the entire SoC. The chip employs 12 MB of system-level cache, with 4 MB of memory for graphics-related tasks handled by iGPU. The iGPU of choice is Adreno 740, with all modern APIs supported. Discrete graphics solutions are supported by the top-end SKUs, which allow eight PCIe 4.0 lanes to be directed toward dGPU, along with an additional four PCIe 4.0 lanes for NVMe SSD. For RAM, the chip uses up to 64 GBs of LPDDR5X eight-channel memory with up to 4.2 GHz speeds. Chip's media engines are structured to support decoding up to 4K120 and encode up to 4K60 with AV1.

SUS IoT, the global AIoT solution provider, today announced PE3000G—the industry's first edge AI system to support Mobile PCI Express Module (MXM) GPUs from both NVIDIA and Intel. Specifically, the all-new industrial PC works seamlessly with NVIDIA Ampere/Turing or Intel Arc A-series MXM GPUs. Powered by a 12th Gen Intel Core processor and up to 64 GB of DDR5 4800 MHz memory, and combining a proven power design, guaranteed and fanless thermal performance, and superior physical and mechanical ruggedness, PE3000G brings unprecedented longevity, computing power, flexibility and reliability to AI computing at the edge—making it an ideal option for scenarios where resilience, longevity and both CPU and GPU scalability are paramount.

"PE3000G is ASUS IoT's response to the burgeoning demand for accelerating AI inference and extreme deployment in industrial settings," commented KuoWei Chao, General Manager of the ASUS IoT business unit. "With robust power, thermal and mechanical design, it pushes versatile edge-AI-inference applications to business-critical applications. PE3000G is an ideal fit to accelerate edge AI inference in SWaP-constrained applications, such as machine vision in factory automation, outdoor surveillance system and AI-inference systems for autonomous vehicles."

Last month, Intel released its Arc Alchemist lineup for mobile/laptop configurations. As expected, being the first discrete GPU that the company made, there are some hiccups here and there that happen along the way. Today, we have an interesting case of Intel Arc A350M getting a heavy performance boost with Dynamic Tuning Technology (DTT) disabled. The DTT is Intel's solution to automatically and dynamically allocate power between an Intel processor and an Intel Discrete Graphics Card to optimize performance and improve battery life. This is essentially a competing tech for AMD SmartShift and NVIDIA Dynamic Boost implementations. Thanks to a South Korean YouTuber, BullsLab, we have information that disabling DTT in drivers helps Arc 350M GPU reach higher performance targets.

He found when disabling DTT in drivers that the gaming performance improved significantly and that the Arc 350M was outputting 30-80 more frames per second. This is no slight improvement and shows that the drivers are still not yet mature. Creating a discrete graphics card is not an easy task, as noted here; however, we hope to see Intel put out more fixes in the coming weeks and hopefully end this strange behavior.Below, you can see the YouTube video with benchmarks.

From 2020 to 2025, the compound annual growth rate (CAGR) of 12-inch equivalent wafer capacity at the world's top ten foundries will be approximately 10% with the majority of these companies focusing on 12-inch capacity expansion, which will see a CAGR of approximately 13.2%, according to TrendForce's research. In terms of 8-inch wafers, due to factors such as difficult to obtain equipment and whether capacity expansion is cost-effective, most fabs can only expand production slightly by means of capacity optimization, equating to a CAGR of only 3.3%. In terms of demand, the products primarily derived from 8-inch wafers, PMIC and Power Discrete, are driven by demand for electric vehicles, 5G smartphones, and servers. Stocking momentum has not fallen off, resulting in a serious shortage of 8-inch wafer production capacity that has festered since 2H19. Therefore, in order to mitigate competition for 8-inch capacity, a trend of shifting certain products to 12-inch production has gradually emerged. However, if shortages in overall 8-inch capacity is to be effectively alleviated, it is still necessary to wait for a large number of mainstream products to migrate to 12-inch production. The timeframe for this migration is estimated to be close to 2H23 into 2024.

The discrete GPU market has been a duopoly for quite some time, and when Intel announced that the company is rebooting plans for its discrete GPU lineup, another player was about to break that duopoly. Today, that has been changed forever and Intel has officially become the third manufacturer of discrete GPUs, as we can see on the online listing. On BestBuy, CyberPowerPC has listed "Gamer Xtreme Gaming Desktop" powered exclusively by Intel components. When it comes to the CPU choice, Intel's 6C/12T Core i5-11400F CPU model is present without iGPU. Now comes the interesting part. The GPU powering the system is Intel Iris Xe discrete graphics card, which is a DG1 GPU based on Xe-LP SKU.

This model features 80 EUs, resulting in 640 shading units. While this is not any gaming beast, casual 1080p gaming should be just fine on this configuration. The system is listed for 750 US Dollars, and it is sold out, as of the time of writing this. While the performance of this configuration may not be something monumental, it is an important step towards Intel's inclusion in the discrete GPU market. By using OEMs, the GPU will reach a very large market without any major problems. We are waiting to see the first reviews of the system, which will surely take a good look at the card and examine its performance.

It appears that Intel's DG2 refers to a number of HPG (High Performance Graphics) products within the same family, with rumors surfacing around a possible total of six different graphics products based on the company's latest high performance graphics architecture - and its debut on the high performance discrete market. It's been confirmed that Intel's DG2 products will not be manufactured in-house, via Intel's 10 nm SuperFin technology, but with recourse to foundry partner TSMC's 6 nm fabrication technology.

It seems that DG2 is currently slated for launch based on three different chip configurations: the first is the DG2 512EU, which will power the highest-performance, 4096 shading unit, 8 GB / 16 GB GDDR6 and 192-bit bus graphics card. Another chip is the DG12 384EU, estimated to come in at ~190 mm², available in three different shading unit configurations: 3072 shading units, with an accompanying 6/12 GB of GDDR6 memory and 192-bit bus; 2048 shading units, which reduces allotted memory to 4/8 GB configurations and a 128-bit memory bus; and finally, the further cut-down 1536 shading unit configuration, with a maximum of 4 GB of GDDR6 memory over the same 128-bit bus. The final (current) chip in the DG2 family is the DG2 128EU, with both 128EU and 96EU configurations (1024 and 768 shading units, respectively) carrying 4 GB VRAM over a pretty tight 64-bit bus. We'll see if these leaks actually materialize into final Intel products, and if these design choices are the possible best, considering Intel's technology, so as to assail the two-player party that is the discrete, high performance graphics market.

Intel has apparently begun shipment of its discrete Iris Xe LP-based DG1 graphics card to OEMs and system integrators, which means we will soon see these graphics cards hitting the market - in a manner of speaking. The quantities aren't yet known, but considering Intel's intentions of only shipping it to OEMs, volume shouldn't be quite significant. It remains to be seen whether DG1-toting systems will even be available to the general public, or if these will be sold primarily to business customers. However, considering that the discrete DG1 only offers entry-level performance due to its 80 EUs (less than even the 96 available through integrated graphics on Intel Tiger Lake CPUs), hopes placed on this particular graphics card as somewhat remedying the current industry ailment of undersupply won't materialize.

One interesting tidbit, however, is that system integrators will have to use specific hardware on the systems they build that carry Intel's DG1, as the blue giant has specified that these graphics cards will only work pending specific firmware updates that enable them to function on certain chipset and processor products. Namely, and according to Intel speaking to Legit Reviews, "The Iris Xe discrete add-in card will be paired with 9th gen (Coffee Lake-S) and 10th gen (Comet Lake-S) Intel Core desktop processors and Intel B460, H410, B365, and H310C chipset-based motherboards and sold as part of pre-built systems. These motherboards require a special BIOS that supports Intel Iris Xe, so the cards won't be compatible with other systems."

Intel in its Q3-2020 earnings release unveiled the Iris Xe MAX brand under which it will launch its first discrete GPUs in over two decades. The company also disclosed that it has started shipping the product (i.e. commenced mass production and started shipping volumes to its partners). The first Iris Xe MAX discrete GPU is expected to be based on the Xe LP architecture, and we expect it to target the mobile market, as the likes of ASUS has already developed ultraportables featuring the discrete GPU. Intel unveiled a proof-of-concept of an asymmetric explicit multi-GPU technology which could see the Iris Xe iGPU of the 11th Gen Core "Tiger Lake" processor work in tandem with a discrete GPU based on the same architecture, which is now turning out to be the Iris Xe MAX. We expect the Iris Xe MAX to make life miserable to entry-level mobile dGPUs from NVIDIA and AMD.

ASUS unveiled a special variant of its 2020 VivoBook Flip 14 convertible notebook, which features the "First Intel Discrete Graphics," expected to be the Iris Xe DG1 discrete GPU based on the Xe LP architecture. Intel appears to be promoting the combination of the Iris Xe dGPU with its 11th Gen Core "Tiger Lake" mobile processors, as a potent ultra portable computing solution with rich graphics and mobile gaming. This eats into the markets of entry-level dGPUs by NVIDIA and AMD. Intel is also developing an iGPU+dGPU Multi-Adapter technology that could leverage asymmetric explicit multi-GPU to accelerate gaming performance. The VivoBook Flip 14 TP470EZ features a 14-inch Full HD (1920 x 1080 pixels) display, and so the combination of the "Tiger Lake" Xe LP iGPU and dGPU could take a swing at Full HD AAA and e-sports gaming.

Jon Peddie Research have released their report on the overall market outlook for GPU shipments for Q4'2019, and the news are great for AMD. Due to the launch of more affordable Navi-based 7 nm graphics cards, the company managed to achieve a growth of 22.6% in shipment volume for the last quarter of 2019, compared to Q3 of the same year. This 22.6% volume increase is pretty significant (and is miles ahead of competitors NVIDIA (whose shipments decreased by -1.9%) and Intel (a 0.2% increase), having increased AMD's overall market share by 3%. This means that AMD now commands 19% of the overall GPU market share, surpassing NVIDIA (which counts with 18%) but both being dwarfed by Intel (with a commanding 63% share). It's important to note here that the numbers include integrated- and discrete-GPUs, and AMD's numbers could be assisted by its mobile processor and APU sales, just as iGPUs make up all of Intel's numbers.

Those numbers are skewed, of course, when we look solely at the discrete GPU market share, with NVIDIA commanding a huge, 73% chunk of the market against AMD's paltry (by comparison) 27%. All in all, Jon Peddie Research reports that the overall PC market increased by 1.99% quarter-to-quarter and increased by 3.54% year-to-year, thus resulting a good performance for these "little" chips.

As we are approaching the year 2020, when Intel is rumored to launch its discrete graphics cards to the hand of consumers around the world, we are gearing up on the number of leaks about the upcoming products. Thanks to Twitter user @KOMACHI_ENSAKA, who found the latest EEC listing, we have new information regarding Intel's upcoming DG1 discrete graphics solution.

In the leaked EEC listing, the DG1 GPU is being presented as a GPU with 96 execution units, meaning that Intel is planning to take on entry-level graphics cards with this GPU. If the graphics unit is following the same design principle of the previous-generation GPUs, then there should be around 8 shading units per one execution unit, totaling 768 shading units for the whole DG1 GPU. If the 12th Gen Xe design inside the DG1 follows a different approach, then we can expect to see a double amount of shading units, meaning 1536 in total.

According to the sources close to DigiTimes, Intel will unveil its first discrete 10 nm graphics cards named "Xe" very soon, with the first wave of Xe GPUs expected to arrive some time in 2020. Said to launch mid year, around July or August, Intel will start selling initial Xe GPU models of the long awaited product to consumers, in hope of gaining a share in the massive market using GPU for acceleration of all kinds of tasks.

Perhaps one of the most interesting notes DigiTimes reported is that "... Intel's GPUs have already received support from the upstream supply chain and has already been integrated into Intel's CPUs to be used in the datacenter and AI fields.", meaning that AIB partners already have access to first 10 nm graphics chips that are ready for system integration. First generation of Xe graphics cards will cover almost whole GPU market, including PC, datacenter, and AI applications where NVIDIA currently holds the top spot.

During interview with Russian YouTube channel called PRO Hi-Tech, Raja Koduri, Intel's chief architect and senior vice president of architecture, software and graphics, talked about his career, why he left AMD, and where Intel is going with its discrete GPU attempts. However, one of the most notable things Mr Koduri said was regarding upcoming GPU lineup code-named Arctic Sound. He noted that Intel plans to release first GPU as a mid-range model at a price of $200, while enterprise solutions that utilize HBM memory will follow that.

Koduri said that he wants to replicate AMD's strategy of capturing high-volume price-points, such as the $199 Radeon RX 480. The plan here is to bring an affordable, good performing GPU to the masses - "GPUs for everyone" as he calls them. Additionally, he states that Intel's current strategy revolves around price, not performance, providing best possible value to consumers. Intel's approach for the next two or three years is to launch a complete lineup of GPUs, with a common architecture being used for everything from iGPUs found inside consumer CPUs to data-center GPUs.Update: PRO Hi-Tech has posted a snippet of Raja Koduri interview, without the Russian overlay commentary. What he said was actually: "...Eventually our architecture, as publicly said, has to get from mainstream, which is starting at around $100, all the way to data-center class graphics with HBM memory...". This means that the previous speculation about $200 graphics card is false, as he didn't say that. All he said is that Intel wants to enter the "mainstream" GPU market and work its way up to data center.

Jon Peddie Research, the market research firm for the computer graphics industry, has released its quarterly Market Watch report on worldwide GPU shipments used in PCs for Q4'18. Overall GPU shipments decreased -2.65% from last quarter, AMD shipments decreased -6.8% Nvidia decreased -7.6% and Intel's shipments, decreased -0.7%. AMD's market share from last quarter decreased -0.6%, Intel's increased 1.4%, and Nvidia's market share decreased -0.82%. Year-to-year total GPU shipments decreased -3.3%, desktop graphics decreased -20%, notebooks increased 8%.

Although overall GPU shipments declined PC sales saw an uptick of 1.61% which is a positive sign for the market overall. "The channel's demand for add-in boards (AIBs) in early 2018 was out of sync with what was happening in the market," said Dr. Jon Peddie, president and founder of Jon Peddie Research. "As a result, the channel was burdened with too much inventory. That has impacted sales of discrete GPUs in Q4, and will likely be evident in Q1, and Q2'19 as well."

Intel is working to build up hype and awareness around its discrete graphics efforts, of which we're still to see more off besides Intel's continuous acquisition of AMD-based talent for that particular endeavor. It's relatively hard to build up enthusiasm for something other than the fact there is a third competitor entering the high-performance graphics card space; the rest is mostly rumors, speculations, and declared intentions.