Intel’s elusive Arc platform has been in the pipeline for many years, and for most of that time, it has largely been shrouded in mystery. But with the official release of the new Intel Alchemist GPUs right around the corner, the time for Intel’s official return into the dedicated GPU space is finally here. This is the first dedicated GPU platform Intel has made since the late 1990s — when they launched their first and only GPU, the i740 AGP graphics card. With the release of the Intel Arc A380 (released in June), the first impressions of the Arc lineup haven’t been anything to write home about, but the excitement for Intel’s GPUs remains. This is a testament to the fact that industry watchers are excited about the fact that Intel’s entry into the space might finally end the decade-long duopoly of Nvidia and AMD. So what is Intel’s Arc platform? And can it compete with Nvidia or AMD Radeon GPUs? Let’s find out.

Hype aside, what we want to do through this article is objectively explore the various aspects of the Arc lineup, from its architecture and specifications to its performance and flagship features. The Arc GPUs are touted to be the next step in graphics architecture, and for that alone, they deserve a technical deep-dive. Also, we want to see how Intel Arc graphics compare to Nvidia or AMD Radeon GPUs.

  • Why Did Intel Develop the Arc GPU Platform?A Look at Intel Arc Alchemist Architecture

  • What Is a Xe Core in Intel Arc GPU?Intel Arc GPU Models and SpecificationsIntel Arc Alchemist GPU: New Features

  • Ray Tracing SupportXe Super Sampling: Intel’s Take on DLSS and FSRIntel Arc GPUs Support AV1 EncodingHow Do the Intel Arc GPUs Perform?Intel Arc A380 vs Nvidia GTX 1650: ComparisonIntel Arc’s Three-Tier Strategy: GPU OptimizationIntel ARC Graphics: Release Date (Expected)Launch Confusion: Is Intel Arc Cancelled?Intel ARC Graphics Cards: Pricing (Rumored)

  • What Is a Xe Core in Intel Arc GPU?

  • Ray Tracing SupportXe Super Sampling: Intel’s Take on DLSS and FSRIntel Arc GPUs Support AV1 Encoding

Why Did Intel Develop the Arc GPU Platform?

The first reaction we had when Intel showcased its dedicated GPU all the way back in 2020 was a sense of bewilderment. Like us, most PC enthusiasts wondered why Intel, a leader in the CPU market at the time, would take this radical jump into the discrete GPU space.

If one were to speculate why Intel did this, we could just look to simple economics. Intel, like all capitalist enterprises, is constantly looking to expand, both in terms of size and profits, and what could be a better opportunity than the ever-growing dedicated GPU space? The GPU market size was expected to grow 94% from $18.2 billion in 2018 to $35.3 billion by 2025, and one could argue that Intel wanted a piece of that pie. 

Coming to the current state of things, after years of hard work, Intel has finally released its first set of graphics cards under the Intel Arc lineup (codename Alchemist). Now it would not be wrong to say that Intel is a newcomer to the graphics market, as the new Arc GPU series is the company’s first foray into the segment in decades. But we have to keep in mind that while they are new to dedicated graphics, they are by no means new to making GPUs.

Intel has a storied history in this industry and has been at the forefront of graphics technology many times in the past. Most recently, they released the Xe graphics architecture with their Tiger-lake CPUs, which brought a huge upgrade from the older UHD graphics. And the Xe graphics performance was so impressive that it pushed Intel to the same level as AMD when it came to integrated graphics.

What Is a Xe Core in Intel Arc GPU?

To start, a Xe-core contains 16 Vector Engines (formerly known as Execution units), each of which operates on a 256-bit SMD chunk. Each vector unit, in this case, can process eight FP32 instructions simultaneously, each of which is traditionally called a “GPU core” in AMD and Nvidia architectures.

Then, Intel takes this Xe-core and forms a larger unit called a Render slice. This slice contains four Xe-cores, 64 Vector and Matrix Engines, plus additional hardware. We can see what the architecture looks like in the image above. It’s complicated-looking, for sure, but it’s also an advanced chip that contains many different tools, making for a truly cutting-edge architecture.

A particular feature that we are excited about is the new ray tracing unit (RTU) of the Xe-cores. The fact that Intel is going to include this cutting-edge feature in its first attempt – a feat that not even AMD could achieve – is commendable. For now, Intel has claimed that its RTU can do up to 12 rays per box, which is a bit higher than both Intel’s RT cores and AMD RDNA 2 Ray Accelerator. Theoretically, that means that a single RTU core should pack more ray tracing power than a single RDNA or RT core. But the fact that we don’t have any hard evidence means that we should take this claim with a pinch of salt.

So now that we have looked into the details of the Intel Alchemist architecture, let us start with a general overview of the entire GPU lineup. Until now, Intel has said that their Arc lineup will cover two different GPU dies (different underlying chip architecture) that will be further divided into three different product families – Arc 700 series, 500 series, and 300 series. And as expected, they will be divided on the basis of price and performance. Let’s look at how these cards are differentiated below:

SpecificationsArc A770Arc A750Arc A580Arc A380ArchitectureACM-G10ACM-G10ACM-G10ACM-G11Process TechnologyTSMC N6 (6nm)TSMC N6 (6nm)TSMC N6 (6nm)TSMC N6 (6nm)Transistors21.7 billion21.7 billion21.7 billion7.2 billionXe-cores3228248Clock speeds2100MHz2050MHz1700MHz2000MHzVRAM (GDDR6)16/8886XMX Engine units512448384128VRAM bus-width25625625696GPU bandwidth (Gbps)560512512186TDP (in watts)225W225W175W75WRelease Date (rumored)Oct 2022Oct 2022Oct 2022June 2022

The Arc GPU model that Intel has showcased the most till now is the A750 limited edition, as it has been used in most of the performance showcase videos. The “limited edition” moniker here doesn’t entail any performance change, as it mostly shares the same components as the non-limited cards. The basic board and fan design and 225W power remain the same, but it does have a first-party cooler setup, much like Nvidia’s Founder edition cards. Also, in the case of the Intel Arc A770, you get built-in LEDs for that added gamer look.

Intel Arc Alchemist GPU: New Features

Ray Tracing Support

The premier feature that everyone is excited about, as we previously mentioned, is the ray-tracing capability of the new Intel Arc GPUs. As Intel has told us many times, the Arc graphics cards have been constructed from the bottom-up around the DirectX 12 feature set, and ray-tracing was a feature they had thought about since the beginning. But the list of new features doesn’t end here, as the Intel Arc GPUs also come with support for other important DirectX features like variable-rate shading, mesh shader, and sampler feedback. Let’s look at other new Arc GPU features:

Since we have already talked about Arc’s ray-tracing cores in our earlier section, let’s move to the next most talked about feature of Intel’s Arc graphics – XeSS.

The process is quite complicated, but to put it simply, XeSS uses information from spatial and temporal data, i.e. both image data and motion data from previous frames, and leverages AI to further improve image quality.

If you wish to read about how XeSS works in more detail, you can check out the recent blog post shared by Intel. One important detail that Intel has emphasized is that XeSS doesn’t just act as an upscale model but also as an anti-aliasing solution like Nvidia’s DLSS. This means that on top of upscaling the game’s resolution, the GPU also removes sharp edges from pixels by using a smoothing technique.

Furthermore, just like DLSS and FSR, XeSS does not run natively in a video game and requires the developers to actively implement support for it. That means not all games will support this upscaling technique at the moment. This is indicated in the number of games that actively support this technology, as revealed by Intel. So far there are only 14 games listed on Intel’s website. But like all the other AI-based upscaling (DLSS and FSR) techniques, the support for the technology is limited at launch but will most likely grow in the future.

Intel Arc GPUs Support AV1 Encoding

Another interesting new feature included with the new Alchemist platform is AV1 encoding. We will look at what it is specifically later in the article, but in brief, it’s a video encoding format that is not available on most Nvidia or AMD GPUs. Only the recently launched Nvidia GeForce RTX 4000-series GPUs come with AV1 support, as we mentioned in our RTX 4090 vs RTX 3090 Ti comparison.

Furthermore, Intel claims that we can use this feature along with Intel’s new Deep link technology, a software that helps GPU and CPU share power/ resources to get even better performance in activities such as streaming and video rendering. But how do the new Arc GPUs perform? Let’s look at this in our Intel Arc GPU performance section below. 

While Intel has been showing both the Arc A770 and A750 GPUs in their various media briefings, only the latter has had proper gaming benchmark results revealed. In a performance showcase video, the Arc A750 was shown running Control at 1440p/High at around 60 FPS. Now, this level of performance for a mid-range card is not bad, but the thing that we’re skeptical about is whether this level of performance is something exclusive only to Intel-optimized games (more on this later).

In another one of Intel’s performance showcases, we saw the Arc A750 Limited Edition compete with the Nvidia GeForce RTX 3060 across many different games like Dirt 5, Cyberpunk 2077, Hitman 3, and Fortnite. The fact that Intel chose the RTX 3060 for comparison is quite interesting and shows that Intel doesn’t want to compete for the performance title with its higher-end GPUs, but rather wants to invade the more popular mainstream graphics market.

Intel Arc A380 vs Nvidia GTX 1650: Comparison

Now coming to the GPU that has actually been tested by many of the hardware publications, the Intel Arc A380. The GPU is mostly aimed at entry-level systems, but even if we compare it to those, the level of performance is not that impressive. Toms Hardware did an extensive breakdown of the A380’s performance, and it shows that while the A380 does come with the latest feature set, including AI upscaling and ray-tracing, its classic rasterization performance leaves a lot to be desired. It can only compete with the 4-year-old GTX 1650 in Nvidia’s lineup and the not-so-great AMD RX 6400.

In a recent YouTube video, Hardware Unboxed also shared a detailed breakdown of a Gunnir-made Arc A380 with a top-of-the-line Ryzen CPU. In this review, the A380 was seen keeping pace with AMD’s RX 6400 GPU in most games. In some games like Tom Clancy’s Rainbow Six Siege, the Intel Arc 380 saw an average frame rate of 93 FPS compared to RX 6400’s 95 FPS. But, in some other games like Assassin’s Creed Valhalla, the Arc GPU struggled to even keep pace with Nvidia’s 1050 Ti, a GPU that was released almost half a decade ago.

Most newer GPUs like, Nvidia’s RTX card and AMD’s RX 6000 cards, all have AV1 decoding abilities, which means individuals with these cards will be able to consume AV1-coded files, but the Arc lineup will be one of the first GPUs to feature a proper AV1 encoder (alongside Nvidia 4000-series), making it vastly more performant for video creators.

Now you must be wondering — why is the performance of the new Intel Arc GPUs not up to the mark? It should be comparable to its Nvidia and AMD counterparts based on its specification, right? Yeah, theoretically it should. But there is a fundamental issue here. Let’s look at all the details.

Tier I: This tier includes games that support the latest APIs and have been optimized for Intel’s Arc graphics architecture. Intel, for now, has put newer games like Cyberpunk 2077 and Shadow of the Tomb Raider in this tier but has argued that the number of tier 1 games will increase in the future. The games in this tier use all the architectural benefits that Arc provides and are said to perform better than their immediate completion (A750 vs RTX 3060).Tier II: Games in this tier have not been optimized for Intel’s Arc architecture as of now, but are based on newer APIs — DirectX 12 and Vulkan. This means they can take advantage of Intel’s architecture to some extent.Tier III: This tier includes games based on older APIs like DirectX9 and will most probably run far worse on Intel’s GPU compared to the graphics cards from Nvidia or AMD in the same price range.

Now coming to the question that everyone is asking at the moment- when will the card be available to buy? This is the million-dollar question, and as much as we would like to tell you the exact release date of the GPUs, the information is just not there. The only set of clues we have come from the cryptic press releases or tweets that Intel posts from time to time.

Hi! Laptops with Intel ARC will be available for purchase by the end of the second quarter of 2022. We suggest you follow our newsroom at https://t.co/AlNpzzMu5P for all the updates.— Intel Support (@IntelSupport) April 6, 2022

The latest news we have comes from Intel, which says we will get further updates “next quarter“. Now, what does this mean? If we are to take Intel quite literally, the start of the next quarter, which is Q4, would be October. This seems to be in line with the new information that has been made available online about the embargo guidelines of the new Intel Arc A770 graphics card.

Well, it’s all this confusion that has made PC enthusiasts question whether the Intel Arc graphics cards are being canceled or not. While something like this might sound outrageous to a layman, many tech insiders argue that, unlike previous rumors that have seen the light of day, this one seems to have some backing of facts.

But are these claims true? Yes, to an extent. We have seen firsthand the disastrous way in which Intel has handled this launch so far. After resorting to only a small regional launch (China) in the February launch window, and then missing the summer release window, the past few months have seen Intel continually struggle to kick a single graphics card out the door. Furthermore, leaks from graphics card makers and PC system builders appear to tell a story that Arc cards are already made and waiting in warehouses, but it’s the software and drivers that aren’t up to scratch. The delay has gotten so bad that Intel’s own promotional Arc GPU giveaway in March 2022 had to give winners CPUs rather than the Arc graphics cards they had won.

This corresponds to what Intel had declared in their official release last year at Intel Architecture Day 2021. At the conference, they not only told us that the official codename of the first Xe-HPG was Alchemist, but also gave a clear roadmap/ naming scheme for three other future GPU generations, namely Battlemage (Xe2 HPG), Celestial (Xe3 HPG), and Druid (Xe4 HPG).

These all fall under Intel Arc, the new architecture, and are set to release in the coming years. So looking at both the claims and statements from Intel, we could argue that internally they have committed to sticking with Intel Arc GPUs up until 2025 or later. Now, will they stick to this plan? Guess we will find out.

Finally, coming to one of the most important aspects, i.e. the pricing of Intel Arc graphics cards. This, again, has been an elusive aspect of the Intel Arc GPUs. Intel CEO Pat Gelsinger has spoken numerous times about the topic of Arc pricing but has never really divulged any concrete info. Most of the time, his speeches have revolved around the question of market monopoly, about how Nvidia’s immense success could be attributed to the lack of competition. However, Gelsinger does emphasize the fact that Intel has to be very competitive in terms of pricing to dethrone AMD and Nvidia.

We have seen some good-quality leaks, though, like the aforementioned Wccftech report, which suggests that Intel has a price structure in mind. Again, these might be unreliable and are subject to change, but the leaked image does seem legitimate and is on the same tangent as Intel’s claims. So using the leaked image as a clue this is what we can expect – the Arc A770 will most probably be in the $350-$399 price range, with the A750 just below at about $290-$340. At this price point, the cards will trade blows against cards like RX 6600 and Nvidia RTX 3060, which is what Intel compared the performance of their cards to in their August reveal.

So, the final pricing of the GPUs, at least relying on the leaked estimates for the prices of the Intel Arc GPUs, could look something like this:

How does this stack up with the competition? We would argue that it’s competitive, as Intel intended. Looking at A380’s performance, it is a competitive card in its price bracket at $150, but the question remains whether consumers will see it that way, especially with Nvidia and AMD offerings in the market. Going with a first-gen Intel discrete GPU has a variety of compromises at the moment like the lack of support and bad optimization for older titles. But for what it’s worth, if Intel can stick to the rumored prices and work on their game optimizations, the GPU market could turn spicy.

The bottom line is that Intel is going to have a rough ride if the Arc release continues in the same manner as it has been going. It may be a titan when it comes to the CPU world, but in the GPU space, it’s a newcomer, and till now, has had a shaky start. As the graphics underdog, if Intel wants to make a mark in this market, the company needs to come out with not just aggressive pricing but also strong performance and a rapid pace of improvement. But looking at the performance metrics from both Intel’s own performance numbers and other hardware reviewers, the level of performance required from a current-gen GPU is just not there. Some people argue that if Intel Arc GPUs release in good condition (in terms of driver performance), then Intel could find some footing in the market. However, as we move further into 2022, the situation doesn’t look as rosy. GPU prices for both AMD and Nvidia cards are finally returning to normal, and in the case of higher-end cards, well below launch prices. This makes the graphics card market particularly hostile to any newcomer, let alone someone like Intel that suffers from severe performance and driver issues.