Alchemist aXXX, Battle Mage bXXX, Celestial cXXX

Intel seems to be in the final stages of launching the first-generation Arc graphics, and the nomenclature is leaking in the right order. Intel has created a usage guide for adding board members. It contained details I had never seen before (@momomo_us via videocardz), Includes the fact that generations are called Intel Arc aXXX, bXXX, cXXX, and dXXX, respectively. Interestingly, Arc is also not capitalized.

Intel Arc a170, a150, and a130 are all potential naming schemes (along with other variations).

I’m not going to lie, I’m a little disappointed to see Ark not capitalized. Imagine NVIDIA being spelled Nvidia. It’s not the same effect, but I’ve got a glimpse of marketing folklore with ALC Maisto, which looks absolutely great, and I hope Box Art will be more than a complement to it.

The leaked design documentation states that first-generation GPUs are referred to as Intel Arca-series and Intel Arc-a-series graphics, not first-generation Intel Arc graphics. Of course this is not the case in the press, but we always call them the first generation of Intel’s most publicly admired attempts to diversify so far (remember 5g?). I’m out.

Now that the nomenclature has been hashed, given that Intel plans exactly three SKUs, we can make informed guesses about what the nomenclature is.

Here are some of our more likely guesses:

  • Intel Arc a300, a500, and a700 (in terms of performance)
  • Intel Arc a130, a150, and a170 (in terms of performance)
  • Intel Arc a800, a600, and a200 (see Number of Xe slices)

Leaked slides showing possible SKUs proposed and their performance goals.

The first iteration of the Intel Xe HPG GPU has eight slices, each consisting of four Xe cores. This constitutes a total vector / matrix count of 512 (8x4x16). Yes, that was exactly the old “EU” count for this GPU. Assuming the basic architecture is the same (it should be), we are still looking at 4096 ALLUs (512 * 8).

In the graphics world, there is an insatiable demand for better performance and more realism. TSMC is excited that Intel has chosen N6 technology for its discrete graphics solution in the Alchemist family.

There are many factors that make a graphics product successful, including semiconductor technology. With the N6, TSMC provides the ideal balance of performance, density and power efficiency for modern GPUs. We are pleased to collaborate with Intel on the Alchemist family of discrete GPUs.

Dr. Kevin Chan,
Senior Vice President of Business Development at TSMC

Since the Intel Xe HPG is manufactured on TSMC’s 6nm process, it should be very good in terms of power efficiency and transistor density. This also means that when they boot, they can actually get some volume from them. Intel’s Xe HPG architecture achieves 1.5 times the clock rate of Xe LP and 1.5 times the performance per watt. This means that we are considering clocks in the 2.1GHz range, given that the XeLP discrete GPU was clocked at 1.4GHz. Performance leaks and roadmap leaks start popping up and you can’t really wait to get them for review in the lab. Alchemist aXXX, Battle Mage bXXX, Celestial cXXX

Back to top button