AMD officially announced its new Ryzen AI Max 400 (Gorgon Halo) series processors, which are designed in a structure that is compact enough to fit into small PC-sized systems but offers a full 192 GB of combined memory.

This historic announcement, which coincides with a difficult period in which there are significant difficulties in the global supply of DRAM and memory chips, is of great value, especially for artificial intelligence developers, content producers and researchers.

Built on the strong foundations of the previous generation Strix Halo series, this brand new architecture targets professional users who want to run complex and massive artificial intelligence models safely on the local device.

These new chips, which AMD is once again breaking new ground in the hardware world, are at the center of the technology agenda with both their unprecedented memory capacity and significant frequency increases.

Gorgon Halo Architecture: What’s Different Compared to Strix Halo?

The new generation Ryzen AI Max 400 series, developed by AMD under the code name “Gorgon Halo”, does not include radical architectural changes on the basis of processor cores, but it offers extremely valuable performance improvements at key hardware points.

The new series fully preserves the Zen 5 CPU architecture, RDNA 3.5 graphics processing technology and XDNA 2 neural processing unit (NPU) features reaching 55 TOPS processing power, which were introduced to hardware enthusiasts in the previous generation chips.

The changes made by the engineering team focus mostly on the total memory capacity and operating frequencies of the processors.

Ryzen AI Max + PRO 495, the absolute top model of the series, is equipped with a small but effective overclocking of 100 MHz compared to its predecessor and reaches a frequency of 5.2 GHz in turbo mode. This powerful processor, consisting of 16 cores and 32 threads in total, also includes the high-performance Radeon 8065S integrated graphics unit with 40 computing units (CU).

It seems that a factory overclock was applied on the graphics side and the GPU frequency was increased to 3.0 GHz. The PRO 490 (12-core) and PRO 485 (8-core) models, located in the middle and lower segments of the series, run at a maximum speed of 5.0 GHz and there is no major clock speed update on the CPU side in these models.

Both models are powered by the Radeon 8050S graphics processor with 32 compute units.

In fact, when we look at the general technical picture, it is very clear that AMD’s main focus is on expanding the memory limits of the device rather than the processor cores. The maximum memory capacity limit of 128 GB in the previous generation Strix Halo chips reaches 192 GB in the form of LPDDR5X with the Gorgon Halo architecture.

What Does 192 GB Unified Memory Provide Us?

For a standard end user who only does daily work or plays games, a memory with a capacity of 192 GB may seem extremely unnecessary.

However, for small businesses, independent researchers or software developers working in the field of artificial intelligence, system memory capacity often stands out as one of the biggest performance bottlenecks.

At this point, the huge capacity offered by the AMD Ryzen AI Max 400 series plays a life-saving role. This memory architecture rewrites the rules of the game for professionals who want to run large language models (LLM) on the local device in complete security and privacy, without sending any data to external cloud servers.

According to the official information shared by AMD, Gorgon Halo chips claim the title of being the first x86 client processor in the world that can run a large language model with more than 300 billion parameters directly on the device.

In order to support this great claim, the company allows a huge portion of 160 GB of the total combined memory of 192 GB to be assigned directly by the system as VRAM.

Under normal conditions, in order to run giant artificial intelligence models of this size uninterruptedly, valuable cloud computing services or a large number of huge graphics cards are needed in the system.

AMD engineers claim that developers can save up to $750 per month on equivalent cloud API costs by using these next-generation devices as a single unit.

Availability and Supply Issues

Although the technical specifications offer theoretically enormous numbers, reaching these new generation high-capacity processors is not as easy as it seems. Major computer manufacturers of the technology world, such as Asus, HP and Lenovo, are planned to launch their systems carrying these new processors only in the third quarter of 2026.

On the other hand, pre-orders of special developer boxes called Ryzen AI Halo, which contain AMD’s previous generation Strix Halo processor, are available in June with a high price tag of $ 3,999.

The situation that really worries the industry is the global memory crisis that has shaken the hardware world to its core. Heavy DRAM supply issues in everything from smartphone production to large data centers leave hardware manufacturers in a pretty dire situation.