Gamasutra GameDev Breakdown of Microsoft Xbox Scorpio [2017: 04-12, 04-14, 04-17]

Ike Turner

Veteran
Heads up for those who don't know yet: Gamasutra will have an exclusive in-depth breakdown of Project Scorpio this Wednesday (this time from a software/gamedev perspective). They were invited by MS like Richard so it will be brand new stuff not rehash of last week's info.
https://twitter.com/gamasutra/status/851557447416168448

It's up:
http://www.gamasutra.com/view/news/..._Project_Scorpio_and_its_brandnew_dev_kit.php

Q&A with Spencer:
http://www.gamasutra.com/view/news/..._Phil_Spencer_looks_to_the_future_of_Xbox.php
 
Last edited:
Remember back in 2013 when multiple developers spoke in videos and in articles and were hyping up unified memory on PS4/X1 talking ONLY about the upsides and absolutely not mentioning the loss of memory bandwidth via contention downsides. Just because they are developers doesn't mean they can't be biased. or feel obligated to give some bias towards the hardware since they had exclusive privilege of being invited to examine Scorpio.
 
Hope they ask more infos about GPU (polaris, vega features...). Specially about Rapid Math, double rate FP16... really hope MS has integrated it.
 
Don't gets your hopes high, on either.
hum. I wondered if maybe they didn't want it for compatibility reasons, but it'd seem shortsighted.

And daft.

Then again, if they at least grabbed FP16 support (GCN3/Tonga), there'd still be wins with register usage & bandwidth vs compute speed.
 
Suggesting it won't have double rate FP16 is crazy talk :LOL:, considering they started developing the SoC one full year closer to Vega's launch than Sony did.

Unless it's causing some weird compromise elsewhere, like limiting clock or something.
 
Suggesting it won't have double rate FP16 is crazy talk :LOL:, considering they started developing the SoC one full year closer to Vega's launch than Sony did.

Unless it's causing some weird compromise elsewhere, like limiting clock or something.
Or MS thinks it was not worth the investment.
 
Indeed.

There are certainly use cases for non-gaming applications as seen with deep learning and mobile application (power consumption), but I'm not clear on the double throughput advantages for high-end gaming; again, there is certainly a boon when it comes to register space and internal memory bandwidth consumption (since GCN3/Tonga), but we shouldn't conflate that with rapid packed math (Vega).

Of course, I'm not sure what the silicon cost or caveats are to its implementation that they would forego it anyway.
 
The scorpio SoC is pretty large just looking at the numbers. If they couldn't even fit the packed math shaders, what are they even doing with all that die space?
 
There are certainly use cases for non-gaming applications as seen with deep learning and mobile application (power consumption), but I'm not clear on the double throughput advantages for high-end gaming; again, there is certainly a boon when it comes to register space and internal memory bandwidth consumption (since GCN3/Tonga), but we shouldn't conflate that with rapid packed math (Vega).
FP16 is a 30% boost in Frostbite checkerboard resolve passes. They didn't specify whether reduced register pressure or double rate math was the main cause for the improvement (likely a bit of both).

Page 82: http://www.frostbite.com/2017/03/4k-checkerboard-in-battlefield-1-and-mass-effect-andromeda/
 
would be nice to get more spec details even minor ones, but that's not what I'm expecting from this.

hoping to be wrong though.
still looking forward to details on tools and maybe xdk, possibly some uwp thrown in.
 
Not the details I was expecting. Didn't go into much depth about the actual software platform. Thought they'd talk uwp, and game feature parity between X1/Scorpio.
 
Disappointing article with regards to specs.
It does reiterate that it will be multiple devs are getting X1 games running on Scorpio quickly.
 
Back
Top