- Release Date: Q1 2022
- Code Name: Alchemist
- Architecture: Xe-HPG
- Process: TSMC 6nm (N6)
- Performance target: Nvidia GeForce RTX 3070 (rumored)
The Intel Alchemist, aka DG2, GPU will be the first gaming graphics card to be split from Intel’s Xe-HPG yarn, which promises to be the basis for many more competitive GPUs from the chipmaking giant. Alchemist also happens to be the first product from Intel’s high-performance graphics brand Arc, spanning hardware, software and services. This will all start early next year, when we’ll see Intel go all out, hoping to turn on Nvidia and AMD’s hydraulic controls on gaming GPUs.
There are two key battles for PC gaming hardware: Intel vs AMD and AMD vs Nvidia. It has been so for thousands of years. Or at least decades. However, Intel’s Alchemist and Xe-HPG architectures could change that. Intel will want to take a slice of the pie between the red and green teams historically, and take a chunk of the profits from gaming GPUs.
Intel is also not entirely new to the GPU market, as it is one of the largest GPU manufacturers in the world. All but a handful of Intel CPUs feature an integrated GPU, and while that doesn’t mean much to gamers, the silicon actually forms the basis of Intel’s Xe-HPG, the architecture that powers Intel’s high-end gaming dreams.
Xe-HPG is derived from Intel’s Xe-LP and Xe-HPC architectures, but with some specialized changes. As Intel explained at its Architecture Day 2021, Xe-HPG is built for gaming first and foremost.
Still, Intel’s entry into the graphics game was welcome even before we delved into the architecture. With GPU prices rising, especially with shortages rampant, it’s time for others to compete with AMD and Nvidia.
Whether you’re a prospect, a hesitant, or someone else’s die-hard, Intel Alchemist and Xe-HPG should get you excited like a Christmas kid.
Intel Alchemist Specs
Intel Alchemist is not a single graphics card, it is actually a series of graphics cards. We used to expect a breakdown by Execution Unit (EU) and memory, although that’s not entirely accurate these days, as Intel has moved from the humble EU to supporting the new, all-encompassing Xe core.
Xe cores are the new fundamental building blocks of Intel’s gaming GPUs.
This doesn’t completely change the status quo, as the Xe core and EU should share the same basic functionality, which is to group the ALU together. However, the Xe cores in the Xe-HPG are different from the previous ones, so our best plan for possible specs is only slightly off.
The new Xe cores in Alchemist GPUs have 16 vector engines and 16 matrix engines, effectively double the number in Xe-HPC. Intel says this is a decision for its Alchemist GPUs to put gaming first, and is necessary to build large gaming graphics cards.
So how many Xe cores and render slices will you find in an Alchemist GPU? That’s a good question, so far we’re still dealing with rumours based on the assumption that these will be the EU.
Previous rumors (non-existent tweets from leakers Komachi_Tohsaka) is highly recommended for Alchemist’s 512 EU and 384 EU configurations, which might still be correct if you just swapped the phrase EU for Vector Engine.
If everything remains the same between Xe-HPG and Xe-LP (the version in Intel’s current 11th Gen Tiger Lake mobile processors), then we can expect a total of 4,096 FP32 cores in the 512 EU chip. This is an 8-wide ALU per EU, capable of 8 floating point operations per clock per ALU.
With the new Xe cores, you seem to be able to replace the 8-wide ALU per vector engine with an 8-wide ALU per EU, so you’ll see more or less a similar composition.
Those matrix engines are also very interesting. You can find them in the diagram above labeled XMX, which stands for Xe Matrix Expansion – as long as you generously give Intel an “X” in “Extended”.
These cores provide acceleration for AI workloads, such as deep learning inference tasks. Think of them like Intel’s version of Nvidia’s Tensor Cores.
Intel plans to use these cores for a variety of AI tasks, but from a gaming standpoint, we’re excited about the acceleration of its Xe Super Sampling (XeSS) upscaler.
Zoom out a bit and things get really interesting for this game architecture. Four Xe cores are combined into a render slice, and up to four ray tracing units can reside within each render slice.
These units accelerate real-time ray tracing workloads, which Intel has confirmed will be supported by its Alchemist graphics cards.
It’s not the only DX12 Ultimate feature on the agenda, either. Intel has attached fixed function units to render slices to support sampler feedback and mesh shading on Alchemist.
Up to 8 render slices will make up the rumored top-of-the-line Alchemist graphics card, which seems like a whole bunch of chips already in gaming from AMD and Nvidia.
At least it has a name to stand up to those behemoths, Intel Arc. This is a start.
Intel says Arc covers everything from hardware to software, so it’s likely we’ll see Intel improve and expand its Command Center app in time for launch, and over the past few years it seems to be gradually improving through the game — — Keep in mind the focused implementation of Xe.
Intel is also promising more capabilities for its software solutions, such as capture and streaming, so Chipzilla’s command center is sure to offer more.
Launching a graphics card is both a hardware challenge and a software challenge – if you ask Intel’s chief architect Raja Koduri, he’ll tell you Software is the hardest part, actually. So it’s just as important, if not more important, for Intel to identify its driver package at launch.
Starting with memory configurations, Intel seems keen to match AMD and top Nvidia in raw memory capacity. Paired with Intel’s Xe-HPG GPU, it’s said to go up to 16GB of GDDR6, running at speeds between 14Gbps and 18Gbps. That’s all the information extracted from Intel’s website in the above leak.
Intel is suspected of using a 256-bit bus for the top-of-the-line 512 EU graphics card, which would put it on par with Nvidia’s RTX 3070 and AMD’s Radeon RX 6800, although the latter does benefit from Infinity Cache.
We have less specific information on power requirements and TGP, but Intel’s choice of foundry and process requires some relatively broad assumptions to be made. It’s not its own, which may once have been considered a blasphemy to Intel engineers. Intel has confirmed that it will outsource the production of Alchemist GPUs to another TSMC for its 6nm N6 process.
A hint that the alchemist’s ultimate power needs came from a YouTuber Moore’s Law is dead, claims to have taken images of Intel DG2 cards during testing that clearly show the 8+6 pin power connector setup. They also suggest a 275W TDP, which is slightly higher than previously expected, but still within the possible range.
So the power has not been confirmed yet, but it is roughly in the whereabouts of 225-275W. It’s also reportedly the right choice for a graphics card targeting Nvidia’s GeForce RTX 3070.
Intel Alchemist Performance
All the rumors suggest Intel’s Alchemist’s top-of-the-line chip is targeting Nvidia’s GeForce RTX 3070.
From specs to the company’s lofty ambitions, it makes sense for Alchemists to see Nvidia’s third card as a worthy opponent. Still, if Intel can pull it off, it’s no mean feat for a first-gen gaming architecture.
Nvidia’s RTX 3070 is a card that’s hard to beat or even impossible to match. According to the following benchmarks. This is a card that matches Nvidia’s previous generation’s best and most expensive GeForce cards.
Sadly, we have little evidence today of Intel’s proficiency, or lack thereof. After the specs were checked, the raw bandwidth and compute power looked good, but building a GPU requires a holistic and generally fine-tuning approach, so one has to wait and see how it gets out of the full picture.
Intel has published information on at least a performance/watt and frequency boost for Xe-HPG over Xe-LP, and it appears to be a 1.5x increase in both.
This could potentially push the Alchemist GPU clock speeds above 2GHz, as the discrete form of Xe-LP, the Iris Xe Max, is capable of 1.65GHz.
Well, at least there’s the possibility of a really competitive card here. Even if it’s not an RTX 3080 Ti, there are relatively few graphics cards of any kind on the market right now, but a new manufacturer with its own dedicated GPU supply could prove vital to gamers trying to grab any discrete GPU in 2021.
However, with TSMC at the helm, a company under enormous pressure to deliver chips, can Intel really turn things around? Fingers crossed.
intel alchemist price
We are now fully in the speculative jungle, and potential pricing is truly the realm of speculation and recklessness. However, from our rough estimates of performance, we can safely assume at least a price range.
Intel needs to compete with AMD and Nvidia to have a chance to enter the gaming GPU market. Unless Intel offers any premium pricing model, which has admittedly priced its own chips into oblivion beforehand (just look at the Core i9 11900K), we should expect a cap of $579 for the AMD RX 6800 and $399 for the low end, Nvidia GeForce RTX 3060 Ti price.
Of course, I’d love to see a cheaper graphics card that puts AMD and Nvidia in a fierce price war, but sadly the GPU, memory, PCB, and hundreds of minor components all add up. There’s a price floor to Intel’s efforts, and I doubt Chipzilla is willing to go below it.
Intel Alchemist release date
It was assumed we’d see the DG2 by the end of the year, but with the announcement of its Arc branding and Alchemist codename, Intel announced that we’ll see the first graphics cards in early 2022.
There is mention of a launch from Xe-HPG this year Intel Graphics Official Twitter Account, but the August 16 Alchemist Bulletin undermined that. Obviously, we will have to wait until the first quarter of 2022. And now, that seems to be a long way off.