Microsoft provides more details Xbox Series X architecture

Published: 17th Aug 2020, 17:38 GMT   Comments

Microsoft reveals detailed specifications of Xbox Series X SoC

At HotChips 2020, Microsoft has provided more details on the upcoming Xbox gaming console. The System on a Chip (SoC), codenamed Project Scarlett, was manufactured in TSMC’s N7 Enhanced fabrication process. The SoC with a die area of 360.4 mm2 is equipped with 8-cores and 16-thread AMD Zen2-based processing units. Additionally, it features 52 active Compute Units based on upcoming AMD RDNA2 architecture, which are also expected on the upcoming Radeon Navi 2X series.

This chip features 15.3 billion transistors, which is 1.7B more than 545mm2 NVIDIA Turing TU104 GPU (manufactured on 12nmFF). The transistor density has more than doubled compared to the Xbox One X 367mm2 chip, which only had 6.6B of transistors.

Xbox Series X SoC Overview, Source: Microsoft

Microsoft has confirmed that the full silicon features 28 Dual Compute Units, meaning that two of them are disabled for the final retail product (4 CUs are inactive). The GPU accounts for roughly 47.5% of the die area, Tom’s Hardware estimates.

The block diagram for the RDNA2 GPU confirms that the chip will have dedicated Ray Accelerators, which will be responsible for either four texture or four ray operations per clock.

Xbox Series X GPU Block Diagram, Source: Microsoft

The presentation also focuses on increasing the difficulty of next-generation silicon manufacturing. Manufacturing costs have increased and so did the complexity of the modern microprocessors. Microsoft claims that the answer to these problems are patented innovations.  The slide titled “GPU Evolution” illustrates that since 2013, Xbox gaming consoles have seen almost 10x more GLFOPs performance at 1x the power consumption (the slide does not provide a detailed TDP value).

Xbox Series X SoC Overview, Source: Microsoft

Xbox Series X console ray tracing acceleration performance is estimated at a 380G/sec ray-box peak and 96G/sec ray-tri peak. These are DirectX Ray Tracing (DXR) values. The DXR is expected to become a mainstream technology for the upcoming games as DirectX12 Ultimate compatible hardware becomes more popular.

Sadly neither Microsoft nor NVIDIA have provided their performance metrics for DXR performance, so we are unable to compare both architectures. NVIDIA only confirmed that their RTX 20 series can offer 11 Gigarays/sec, and this value is likely to increase with the upcoming Ampere series.

Microsoft’s approach to ray tracing is no different than NVIDIA RTX implementation, as it does not fully replace ray tracing, it’s a hybrid, a simplified version that provides more visual fidelity at reasonable performance, or as Microsoft calls it an ‘economical upgrade to traditional rendering’.

Interestingly the slides also mention Machine Learning (ML) inference acceleration. However, it’s not explained if the hardware features anything like NVIDIA Tensor cores. Thus, the acceleration is likely based on INT4 and INT8 operations running on Stream Cores, as explained by Eurogamer back in March.

Microsoft Ray Tracing and Machine Learning Acceleration

Other slides:

Source: Tom’s Hardware




Comment Policy
  • Comments must be written in English.
  • Comments deemed to be spam or solely promotional in nature will be deleted. Including a link to relevant content is permitted, but comments should be relevant to the post topic.
  • Comments containing language or concepts that could be deemed offensive will be deleted. Note this may include abusive, threatening, pornographic, offensive, misleading or libelous language.
  • A failure to comply with these rules will result in a warning and, in extreme cases, a ban.
  • Please note that comments that attack or harass an individual directly will be deleted and such comments will result in a ban.
  • VideoCardz Moderating Team reserves the right to edit or delete any comments submitted to the site without notice.
  • If you have any questions about the commenting policy, please let us know through the Contact Page.
Hide Comment Policy
Comments