In the midst of two GPU juggernauts hyping up their new releases, an unknown in the field is attempting to enter the scenario. Even though Intel’s graphics card track record is almost non-existent, its branding is huge in the PC technology field.
However, they are planning to become competitive in the GPU market with their Intel Xe lineup.
For the longest time, Intel has been a force to be reckoned with in the CPU market, but AMD has objectively caught up and given them some stiff competition.
Who knows if Intel decided to try and do the same thing to AMD in the GPU market? Either way now is the perfect time for a third option to enter the fray.
The rivalry between AMD and Nvidia is re-igniting after years of Nvidia domination and things look to be on much equal footing with the releases of their next-gen graphics cards. This makes up for a perfect storm for Intel to enter the market.
- September 20, 2021: Fixed a few information-related errors.
- September 2, 2021: Added new official information from Intel about Alchemist architecture.
- August 3, 2021: Added new benchmark score added of the 128 EU SKU
- June 23, 2021: Added a new leak of a Xe SKU comparing to the RTX 3070 and RX 6700 XT.
- May 25, 2021: Added information regarding the fact that Intel Xe DG1 is showing up in pre-built systems.
- April 7, 2021: Added a few more leaks.
- March 5, 2021: Added leaks of a DG2 Discrete GPU in use by Intel.
As mentioned earlier, Intel didn’t exactly perform well the last time they tried to launch a graphics card. But that was a long time ago; 20 years ago, to be exact. Things have more than dramatically changed since then.
That’s not to say that Intel has no clue how to develop a graphics card. They know what they’re doing – their integrated Intel HD Graphics have been a staple of a vast majority of modern laptops. Because these integrated chips aren’t meant for gaming, Intel has decided to make its own dedicated graphics card.
Seeing how Intel’s Xe has just been announced, we’ll try to dive deeper into the information we already have. We will only discuss mobile GPUs – Xe-LP and desktop GPUs – Xe-HPG.
Table of ContentsShow
Maybe this is the reason why Intel has decided not to release Xe before then. It could also be for a variety of other reasons, including a move from their speculated 7nm node, which they are now calling 10nm SuperFin node.
It’s worth pointing out that the latest information from inside the company suggests that Intel’s 7nm node might not be seen until 2022 or even 2023, which is very different from the initial report of Intel’s chips being built on this year. However, it’s been announced that they will use a third-party foundry for their GPUs.
“The company’s 7nm-based CPU product timing is shifting approximately six months relative to prior expectations. The primary driver is the yield of Intel’s 7nm process, which based on recent data, is now trending approximately twelve months behind the company’s internal target.”Intel’s official press release
Whether they’ll release all SKUs, lower-end or flagship GPUs first, we do not know yet.
Intel Xe was reported to have begun testing in late 2019 with its intended release scheduled for mid-2020. When the long-awaited moment arrived, all we got was an announcement and a delay instead.
On September 2, 2020, we saw Tiger Lake laptops with Intel’s 11th gen CPUs get released, but also with the much anticipated Intel Iris Xe.
A big caveat is that these are integrated mobile graphics cards, far away from what Nvidia and AMD can offer in that market. Another reason why it might’ve flown under the radar is that Nvidia had a big announcement on September 1, 2020, regarding their flagship next-gen GPU, which was released within weeks.
This could also be a tactic to be the first to enter the market as both Nvidia and AMD tend to release their GPU’s mobile version sometime after the release of their flagship cards.
From this perspective, it appears that desktop GPUs in the enthusiast/mid-range are being prepared for 2021. At first Q1 2021 made the most sense and that is what most tech experts expected. However, with several delays and slowdowns, it seems like even 2021 is a reach now.
Best case scenario, we may see a glimpse of Intel Xe at the end of 2021. Alternatively, we’ll have to see what 2022 brings.
One thing that is almost certain is that Intel will have to price their GPUs lower than both AMD and Nvidia. Despite having a good standing name in the technology world, it’s still somewhat of an unknown in the graphics card market. Sure, its HD Graphics series are solid integrated GPUs, but aside from that, there’s just nothing else.
Although when Intel Xe was first announced many expected it to offer budget cards, with mid-range being the highest they could aim for, apparently they intend to aim a bit higher.
Reactors online are guessing that Xe will be somewhere in the range of Nvidia’s 1600 series, but some also believe that they could be on par with an RTX 2070. One of the main reasons for this is that Intel has confirmed that their GPUs will be featuring hardware-accelerated ray-tracing.
Intel Xe Needs To Be The Cheaper Alternative
This is exactly the reason behind the big dilemma of price speculation. We do still anticipate Intel coming out strong with good budget/low-end options, but the announcement of ray-tracing in their cards puts a serious dent into being able to give any sort of precise price prediction.
At the moment, it doesn’t appear that Intel will be able to fully challenge the top-of-the-line Nvidia or AMD graphics cards, but if ray tracing is executed properly, we might see a good mid-range challenger.
Something essential for a large corporation like Intel is that ruthless competitive spirit. It’s fair to expect that whatever product they release, in whatever range, will be slightly cheaper than whatever AMD or Nvidia have to offer in the same area.
Although this section is still very hard to predict for Intel Xe-HPG (now named Alchemist, we do have a pretty good idea of what Xe-LP (now named Iris Xe) was able to pull off.
Disclaimer: Iris Xe is a portable GPU. As such, it can’t deliver a 4K 60 FPS performance like a high-end desktop graphics card, so just keep that in mind before getting underwhelmed.
Intel’s Ryan Shrout posted the following tweet, which shows the Tiger Lake prototype running a resource-heavy Battlefield V.
Despite this being a relatively short clip, it’s quite impressive that the game is running at full HD 1080p at a decent 30 FPS. Although it would be foolish to assume that this signals the end of gaming laptops, it’s still reasonable to say that laptop gaming is getting more and more accessible.
Now that Iris Xe saw the light of day, we can report that that Battlefield V footage was legit and that the laptop is more than capable of giving you a good approach to gaming. Most notably, it was able to achieve 60+FPS on GTA V and perform at competitive standards for eSports games like DotA 2, Rocket League and CS:GO.
The main reason why gamers should be excited about Xe-HPG is the supposed hardware-accelerated ray-tracing capability. With the introduction of RTX 2000 series, we’ve been thrilled with what real-time ray-tracing can look like, so it will be interesting to see how Intel pulls it off.
We’re also excited to see how ray-tracing performs at a more affordable price point.
It is always nice to see new and improved iGPUs for laptops, but we all know that enthusiasts are looking for something more powerful that is going to bring competition against NVIDIA and AMD.
Fortunately, the last couple of Russian leaks show a few new products being tested by Intel. Demo board DG2 Pre PO board kit host (K83287-202) and Discrete Graphics SDP software development platform (DGD22SEH1X) both showed up on portal.eaeunion.org/.
Pretty vague, but the DG2 found in the codename and name inside of the leaks suggests that they are testing a discrete GPU. We know from previous reports that Xe DG1 is targeted towards laptops while Xe DG2 is supposedly for the enthusiast desktop market.
A new leak from June, shows an Intel XE DG2 448 EU SKU in a benchmark, clocking at around 1.8 GHz with performance very similar to Nvidia’s RTX 3070 and AMD’s RX 6700 XT. It’s not the best, but it’s impressive to see that Intel will be able to trade blows with the GPU giants.
Who knows, this particular SKU’s frequency might be pushed even higher, providing better performance.
Another leak in August, 2021 showing a DG2 SKU working the Geekbench benchmark. This is the 128 EU (1024 cores) GPU, making it the second weakest GPU of the Intel XE DG2 family. Still, it’s something to be excited about because this is the first time an XE GPU passed the 2.0 GHz threshold.
However, even though this SKU clocked at 2200 MHz it is still quite slow considering that it got a score of 13710 in the OpenCL benchmark. To give you a baseline to compare, a GTX 1050 card can easily get over 18000.
Of course, the cause for such a low score could be a number of things. It could be the lack of good graphics drivers or it could be that Geekbench isn’t optimized for Intel XE. Or, it may just be that Intel still hasn’t pushed the clock speeds of this SKU to the maximum.
Additionally, we do know that Intel are planning a 512 EU SKU which should be their flagship GPU. With this leak in mind, this flagship could go head to head with a 3080 or even a 3080 Ti.
Weirdly enough, even though Xe DG1 was meant for laptops or other mobile solutions, it has started showing up in various pre-built systems. We’ve seen one in a CyberPowerPC system at BestBuy.
It’s not exactly clear why Intel decided to start selling this dedicated GPU in pre-built systems since it doesn’t offer any significant power, but it is definitely a sign that they are slowly creeping into the GPU market.
The current best guess is that Intel wants to delve deeper into the small business market. SMBs usually require low-powered, quiet, but effective systems for work, so DG1 is the perfect contender. There’s a lot of money in this market, so it makes sense that a company such as Intel would want to become a part of it.
We currently don’t know the architectural and design differences between Intel’s mobile and discrete GPUs, which means it is impossible to predict any performance.
Nevertheless, these leaks tell us that Xe is continuously in the works and that Intel is aiming to deliver its brand-new products as soon as possible.
With the release of Iris Xe, we only got a glimpse of what Intel Xe is all about, but at least we have a good starting point. There are some rumors and speculations regarding the discrete GPUs from Intel and we’ll cover that too.
Intel Architecture Day 2021
We’ve had a lot of rumors, leaks, suggestions, and assumptions regarding discrete GPUs falling under the Intel Xe tree. Finally, though, we get a deeper and official look at this architecture with Intel’s Architecture Day 2021.
First things first, we finally have a name to refer to Xe-HPG. It’s going to be called Alchemist. What’s even more interesting is that Intel also shared their future plan (named Intel ARC) for their next GPU generations including Battlemage, Celestial, and Druid, released respectively.
This is important for Alchemist because it shows just how dedicated Intel is to become part of the world of gaming GPUs. So, even if Alchemist isn’t competitive with AMD’s RDNA 3 or Nvidia’s Ada Lovelace, they’ll surely be with the next generation (Battlemage).
With that, let’s focus a bit more on Alchemist, what you are all here for.
Based on their roadmap Alchemist SKUs are destined to release during the first quarter of 2022. Those are refreshing news as most of us expected Intel’s competitive GPUs much later.
With a year-long silence on Alchemist, it was unexpected to get such a deep insight into how these GPUs work, specifically Intel’s compute building block, Xe-Core.
Below these next few paragraphs, you will find a bunch of leaks and rumors speculating about the total EUs or Execution Units of Intel Xe GPUs. However, with this press release, it seems like Intel completely is throwing that naming scheme out, so we won’t be hearing about Execution Units anymore.
Every Xe-core includes 16 Vector Engines (VE) and 16 Matrix Engines (referred to as XMX). Each of those Vector Engines processes 256 bits per cycle. Based on Intel’s explanation, each VE has 8 ALUs. That leads to a total of 128 ALUs in one Xe-Core. The core also has its own L1 cache.
But, one of these cores does not make a GPU. At least, not a powerful one. Intel combines four of these cores to create a Render Slice. Now, we’ve jumped up to 512 ALUs and 64 XMX.
Further, there are four Ray Tracing Units in each Render Slice which will help with the promised ray tracing performance. There are also Samplers, Pixel Backends, and engines for Geometry, Rasterization, and HiZ.
However, to get the true potential out of these cores and Render Slices, Intel had to scale it further. Thus their flagship GPU holds eight Render Slices. That is 32 Xe-Cores, 512 Vector and Matrix Engines, and a total of 4096 ALUs.
Naturally, there will probably be weaker GPUs configured with fewer Render Slices.
The GPU’s power will be coming from its Vector Engines, but what is Matrix Engines for considering how much space 512 of them will take on one GPU die.
Intel XeSS Upscaling
Those Matrix Engines have a similar role to Nvidia’s Tensor Cores. Tensor Cores use mixed-precision computing to further accelerate certain workloads. Specifically workloads such as training AI/machine learning models. Because of these Tensor Cores, DLSS 2.0 is so effective at supersampling.
Intel’s XeSS upscaling can utilize XMX to provide just as good upscaling or maybe even more effectively than Nvidia.
As seen in the demo above, XeSS rendered at 4k looks identical and sometimes even better than 4k native while providing up to 2x FPS. At least, that’s what Intel’s claiming. We’ll need to see this with our own eyes before we believe it. We all know how DLSS launch went.
Either way, this seems very promising and could bring serious competition to Nvidia’s DLSS. AMD on the other hand will have to think of a much better solution than FSR.
Leaks And Rumors
As mentioned earlier, Intel will be releasing a graphics card for different market sections and despite that being certainly commendable, we’ll only discuss mobile and desktop versions.
Unfortunately, there simply isn’t enough information about the desktop variant except for the fact that it will employ a GDDR6 memory type.
While that is a good sign, it means that Intel will be lagging from the start considering that the current market leader Nvidia has equipped their flagship GPUs, RTX 3080 and RTX 3090 with brand new memory type Micron’s GDDR6X.
It’s worth noting that this disadvantage relates only to the top cards on the market as RTX 3070 also has GDDR6 as well as AMD’s RDNA 2 cards. Aside from that, Intel can boast about their 96 execution units, which is a good number to compete with both AMD and Nvidia’s mobile solutions.
Another sign that Intel is heading in the right direction is the fact they achieved a core clock of 400Mhz with a boost clock of 1350Mhz. Intel’s last-generation Iris Plus was slightly behind AMD’s Vega 11 but Iris Xe competes more than competently. Still, the mobile integrated GPU market is ever-evolving, so that too can change quickly.
The latest news indicates that Intel’s high-performing graphics card will have 512 EUs (execution units) which will rival AMD’s latest RX 6000 outing. This isn’t the only place where Intel will look to challenge AMD and Nvidia as the rumored numbers for the raw shader power look like they could position Intel’s best card shoulder to shoulder with other GPU giants.
Based on a few new leaks on Twitter it looks like Intel is planning to release 6 discrete GPUs. That is more than anyone expected. Although we should always take Twitter leaks with a bit of grain of salt.
From this Twitter post, it seems like Intel will satisfy the low-end, mid-end, and high-end sectors of the GPU market.
- 512 EU (4096 ALU) / 256bit-bus / 8 or 16G VRAM
- 384 EU (3072 ALU) / 192bit-bus / 6 or 12G VRAM
- 256 EU (2048 ALU) / 128bit-bus / 4 or 8G VRAM
- 192 EU (1536 ALU) / 128bit-bus / 4G VRAM
- 128 EU (1024 ALU) / 64bit-bus / 4G VRAM
- 96 EU (768 ALU) / 64bit-bus / 4G VRAM
This leak is a bit outdated after Architecture Day 2021 and Intel’s decision to switch to a new naming scheme and get rid of EUs, but, it seems that some rumors were indeed correct.
It is true that Intel’s flagship (based on official info) will have 16GB of VRAM and 4096 ALUs. We aren’t certain whether we can translate the performance of 512 EUs to the 512 Vector Engines though, but we will still note that it is the same number as the rumor.
Getting 12GB and 16GB of VRAM on Xe is also a bit hopeful, but it is not impossible, especially if Intel decides to go with GDDR6. GDDR6 is much cheaper to manufacture than GDDR6X.
Take the RX 6000 and RTX 3000 series as an example. The RTX 3060 got 12GB of VRAM while most of AMD’s lineup got 16GB of VRAM. To put that into perspective, the RTX 3080 has just 10GB of VRAM probably because it uses GDDR6X.
In other words, it’s not going to be too expensive for Intel to push Xe to 16GB VRAM.
The 4096 ALUs for the 512 EU variant are just a prediction based on the knowledge people have about the Intel Xe architecture.
However, that prediction could become real. If it does, this means that the 512 EU variant can be put against AMD’s RX 6800. One of the best GPUs on the market right now.
But, it can go against AMD’s graphics card only if it can match it in clock speed. We need to see DG2 reach a core clock speed of at least 1.8 GHz to have comparable performance.
With that in mind, the second-best variant with 384 EU or 3072 ALUs could be put against the RX 6700. Again, only if the core clock speeds can be matched.
Why We Should Be Cautiously Optimistic About Intel Xe
Intel promises to take on NVIDIA and AMD with its Intel Xe-HPG gaming GPUs
This should probably go without saying, but we need to do so, at least for posterity’s sake.
Mainstream Xe is still some time away and, based on the shakiness Intel has experienced lately, we wonder how soon and how good their promises can be.
Still, there is room for optimism. Intel is a company with a long history in technology and even if they’re off to a rocky start, they certainly have the resources and the know-how to compete in this field.
We might not be getting a full-on three-way war between Intel, AMD, and Nvidia, but something is brewing and as gamers and tech enthusiasts, we can’t wait.