Xbox Series X full specs revealed

2020 has definitely been a weird year so far, but even with everything going on we’re still looking forward to the release of the next generation of gaming consoles this fall. If nothing else, faster consoles help to raise the low bar for PC games. Today, Microsoft revealed a ton of information about the Xbox Series X console, including the full set of specs for the system.

Microsoft confirmed previously-stated specs with the reveal. The CPU in the Series X is an 8-core custom Zen 2 AMD CPU that runs at 3.8 GHz. The GPU is a custom RDNA 2 AMD GPU. It runs at 1.825 GHz, has 52 compute units, and can handle 12 TFLOPS. The System on a Chip that houses all this is built on a 7-nm process.

Xbox Series X memory & storage

The system offers 16GB of GDDR6 memory running on a 320mb bus. Microsoft says the memory bandwidth is 10GB at 560 GB/s or 6GB at 336 GB/s. The system includes a 1TB custom NVMe SSD. As I conjectured previously, the “mystery slot” on the back of the system is a memory expansion port. Because the system is built on ultra-fast loading, simply plugging in a USB HDD won’t work for gameplay; at least not for Xbox Series X-specific titles.

The system does support external hard drives via USB 3.2, and Microsoft has confirmed that the drive is a 4K UHD Blu-ray drive.

On top of all this, Digital Foundry did a deep-dive video on the system that shows off the system’s internals, puts the Xbox One X and Xbox Series X side-by-side, and gives a better overall idea of the size of the system.

The front of the system is similar in length to that of the Xbox One X, and DF characterized the system’s size by recalling the classic meme for the Nintendo Wii: “two Gamecubes duct-taped together.” Only this time, they’re talking about size, not specs, because if the specs above are any indication, this thing is a beast of a game console.

The whole video is worth watching. DF’s Richard Ledbetter simulates assembling the system, and you can see the new storage card up close and personal.

0 0 vote
Article Rating
Subscribe
Notify of
guest
11 Comments
Newest
Oldest Most Voted
Inline Feedbacks
View all comments
Adnan Baloch
Adnan Baloch
6 months ago

The 7nm CPU and GPU being close together on the SoC will generate a LOT of heat. How are they getting away with not using water cooling?

willmore
willmore
6 months ago
Reply to  Adnan Baloch

1/3 of the volume seems to be cooling. A huge vapor chamber and lots of thin/long vanes. They said it had a 130mm fan at the top. That should be able to move a lot of air. I’d be curious to get Wattage values from it.

pogsnet
pogsnet
6 months ago

The loading is more influenced by NVMe SSD not the CPU >.<

Krogoth
Krogoth
6 months ago
Reply to  pogsnet

CPU is a factor though. NMVe PCIE SSD is overkill for the console unless game developers start going nuts on I/O.

willmore
willmore
6 months ago
Reply to  pogsnet

Unless they’re doing something completely different on consoles, modern games are CPU bound when loading. Just look at the benchmarks when going from HD to SSD to NVME. It hits a plateau because of the CPU. Look at how much faster things load when you increase the CPU speed with fast storage.

Matthew Masters
Matthew Masters
6 months ago
Reply to  pogsnet

True. The performance differences comparing a conventional magnetic hard drive to an NVMe SSD has nothing to do with the CPU. Where do you gamer tech nerds get this stuff from?

jihadjoe
jihadjoe
6 months ago
Reply to  pogsnet

Actually, no.

When Samsung first released their awesome NVMe SSDs someone did a side-by-side video against a relatively cheap BX300 SATA SSD. In a whole bunch of games there was no noticeable difference, and if the NVMe was ahead it was only by a second at most.

The greater majority of level load time in most games with compressed assets is down to CPU.
https://www.youtube.com/watch?v=GKv8cAaJgqs

willmore
willmore
6 months ago

How does that split memory bandwidth thing work?

willmore
willmore
6 months ago
Reply to  willmore

Lookslike there’s 10 memory chips, so I’m guesing 4 of them are 16Gib and 6 are 8Gib density. That means all busses are active for 10GiB of storage and only 6 of them can access the remaining 6GiB.

psuedonymous
psuedonymous
6 months ago
Reply to  willmore

Likely the same as the GTX 970’s split memory pool did: partial crossover of memory PHY, with at least one controller having to pull ‘double duty’ between pools (i.e. only able to service one or the other at any given time). Likely for the same reason: the ability to harvest dies with one dead memory controller. By using that as the baseline spec, they can build devices with either fully working or one-dead-controller binned dies, whereas requiring all controllers to be working would limit them to harvesting just the uppermost bin. As these dies are a custom design for Microsoft… Read more »

CityZen
CityZen
6 months ago
Reply to  willmore

There are 10 memory chips: 4x1GB and 6x2GB. Each chip has 32-bit I/O at 14Gbit/sec per pin, or 56GByte/sec per chip.
For accessing the first 10GB, all 10 chips are used, and thus you get 10x56GB/sec bandwidth.
For accessing the remaining 6GB, only the 2GB chips are used, and thus you get 6x56GB/sec bandwidth.

11
0
Would love your thoughts, please comment.x
()
x

Pin It on Pinterest

Share This