This week at ISSCC (International Solid State Circuits Conference), Microsoft introduced a chat titled ‘Xbox Series X SoC: A Next Generation Gaming Console’, with {hardware} engineer Paul Paternoster presenting. The 30 minute presentation lined quite a bit about Microsoft’s newest console processor, most of which was a repeat about what we noticed at Hot Chips in August final yr, nevertheless there was a brand new aspect on this presentation speaking about how the console design staff balanced acoustics, energy, thermal efficiency, and processor yield, discussing the place the hotspots within the design originate and the place the efficiency/energy targets of the ultimate silicon have been optimized.
Xbox Series X SoC Recap: Project Scarlett
To begin, right here’s a fast refresher on the processor contained in the Xbox Series X, often known as ‘Project Scarlett’. Scarlett is a monolithic system-on-chip constructed on TSMC’s N7 course of node. The design has 15.three billion transistors, measuring 360.Four mm2, with dimensions of 15.831 x 22.765 mm2.
Inside the chip are eight Zen 2 cell cores, configured into two teams of 4 cores, every sharing Four MB of L3 cache much like AMD’s cell Renoir/Lucienne processors. These cores are related via the Scalable Data Fabric to a customized RDNA-based GPU able to 12 TFLOPS of peak efficiency, enabling variable charge shading, ray tracing, and different new graphical options. This GPU is constructed as a 56 compute unit design, nevertheless solely 52 are used within the closing product (extra on this later). There is extra particulars on the GPU options in our protection of the Hot Chips presentation.
The reminiscence system options 16 GB of GDDR6 via 20 x16 channels. 10 GB of this reminiscence is increased efficiency, enabling 560 GB/s of reminiscence bandwidth for video games, whereas the opposite 6 GB is of decrease efficiency, at 336 GB/s of reminiscence bandwidth, for video games or circumstances the place reminiscence just isn’t the limiting issue. This additionally permits a decrease energy state.
The video codec assist permits 4 show planes with 4K/8K AVC encode and decode, HEVC/VP9 HDR decode, and AVC/HDR encode. Three audio co-processors allow compute offload for MOVAD (Opus/Vorbis), CFPU2 (frequency area processing), and Logan IP for MEC (multi-channel echo cancellation) for eradicating background noise from microphones.
The Hardware Security Processor (HSP) permits root of belief in addition to all cryptographic capabilities, reminiscent of safe {hardware} crypto keys. The HSP can also be a part of Microsoft’s Pluton structure, which we’ll see come to fashionable CPUs that use Windows over the subsequent couple of years. The Media Streaming Processor (MSP) permits quick encryption/decryption and hashing of exterior media units via AES, adequate sufficient for the PCIe 4.Zero enabled storage on the platform. Microsoft states that in comparison with the earlier technology 7200 RPM drives, its new Xbox Velocity Architecture (utilizing NAND over NVMe plus the MSP) permits quicker loading instances coupled with 30-35% area financial savings for video games not already distributed in compressed codecs.
You can learn in regards to the Xbox Series X together with efficiency and knowledge testing in our assessment:
For an article devoted to why NVMe storage issues for these units particularly because it pertains to gaming, and a breakdown of a number of the technical particulars, we’ve got this piece for you:
Compared to the earlier technology, Microsoft is selling its new processor as having
- 3x the CPU efficiency,
- 2x the GPU efficiency,
- 1.7x the reminiscence bandwidth,
- 2x the IO bandwidth,
- 0.9x the bodily quantity,
- 2.4x efficiency per watt,
- Same acoustic efficiency,
- Stricter compliance for media playback,
All of this, in keeping with Microsoft, was achieved inside a +15% energy improve in comparison with the 2013 chip.
We imagine the higher restrict of the Xbox…
