Presentation is loading. Please wait.

Presentation is loading. Please wait.

Inside Xbox One Martin Fuller Xbox Advanced Technology Group AMD AND MICROSOFT GAME DEVELOPER DAY - June 2 2014, STOCKHOLM.

Similar presentations


Presentation on theme: "Inside Xbox One Martin Fuller Xbox Advanced Technology Group AMD AND MICROSOFT GAME DEVELOPER DAY - June 2 2014, STOCKHOLM."— Presentation transcript:

1 Inside Xbox One Martin Fuller Xbox Advanced Technology Group AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM

2 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 2 This is a non-NDA event That means there is a limit to how much I can say, go easy! NDA

3 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 3 AMD Jaguar (x64) - 8-cores arranged in 2x clusters of 4 cores each  1.75 GHz  Dual issue  Out of order execution  Speculative execution  Store-to-load forwarding  SSE4.2 and AVX  (Dot product!)  16 x 256-bit wide floating point registers  Hardware pre-fetch CPU

4 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 4 8 GiB of DDR3 at 68 GiB/s  Low latency  Not enough bandwidth to touch all of memory a frame, RAM as a super fast cache 48-bit virtual address space  256 terabytes  Tricky to fragment!  Synced between CPU and GPU 4 MiB of L2 cache  2 MiB per cluster  MOESI protocol for cache coherency  16-way set associative  Per core, up to eight cache requests in flight at once Memory

5 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 5 1. Store to load forwarding saves the dreaded LHS stall  But not spilling out registers is even better 2. The branch predictor is not a crystal ball  Branchless tricks learnt in Xbox 360 era can still apply 3. Hardware data pre-fetch is awesome  Only works with arrays 4. Avoid aliasing load/stores on 2KiB alignments  This causes a false positive that delays load execution 5. Go wide with SSE and leverage all cores  No brainer CPU – Recommendations

6 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 6 AMD GCN 768-SPU 853 MHz 32 MiB of ESRAM at 109 GiB/s 4 Move Engines 3 hardware display planes  Resolution independent  Frame rate independent Exact sRGB this time!  (oh, and its free) Hardware video encode and decode HDMI 1.4a in and out GPU

7 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 7 More than just DMA copy  Memory set  Texture swizzle  JPEG decompress  LZ compress and decompress Move Engines

8 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 8 32MiB of general purpose RAM  Not like EDRAM on Xbox 360  109 GiB/s  Sometimes faster in practice!  Zero contention  Not shared with CPU, SRA’s or video out  ESRAM makes everything better  Render targets  Textures  Geometry  Compute tasks ESRAM

9 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 9 ESRAM can handle concurrent read/writes: Increasing effective bandwidth above 109 GiB/s Operations that can take advantage of this: 1. Read modify write operations 1. Depth buffer / HTILE update 2. Alpha blending 2. Oh, and concurrently DMA’ing resources in/out of ESRAM while also rendering How much effective bandwidth can titles achieve? 1. The current record holder achieved 141 GiB/s from ESRAM (this is a post processing pass in a real title) 2. Of course all titles combine ESRAM’s >= 109 GiB/s with DRAM’s 68 GiB/s ESRAM – Sometimes faster in practice?

10 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM Statically allocate a small number of render targets in ESRAM 2. Alias the same memory for re-use later 3. Partial residency  Put the top strip of render targets (sky) in DRAM, the rest in ESRAM 4. Asynchronously DMA resources in/out of ESRAM Launch titles were at nd wave of titles are now starting to tackle points 3 and/or 4 3 rd + wave will get really good at this! ESRAM – The Four Stages of Adoption

11 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 11 It’s like 8 bit days all over again! (Sort of) Plan the asynchronous moves  Move resources in/out asynchronous while also rendering  New memory map at each stage of the render pipeline Don’t forget, swizzle textures on DMA ESRAM – Memory Maps!

12 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM Are you bandwidth limited? 2. Have you maxed out the fixed function hardware? 3. Do you have spare compute resource? Then use async compute! Titles have barely scratched the surface yet: Watch this space! Maxing out the GPU

13 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM Use ESRAM  First for depth / stencil  Then colour targets  Then everything else 2. Sort by state / shader / use hardware instancing  (Batch batch batch!) 3. Always swizzle textures 4. Be wary of using too many general purpose registers  Keep an eye on occupancy in PIX, we normally recommend >= 4 5. Avoid reading DRAM via the CPU-coherent bus 6. There is no hardware integer divide The usual GPU recommendations

14 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 14 DX11 was designed for the desktop (a long time ago, 2008!)  Abstracts a variety of different GPU architectures  Manages VRAM residency for you  Over subscribing VRAM is a serious performance pitfall  Handles hazards  Developers can handle these at a higher level => less cost Xbox One will run vanilla DX11 PC code  Easy port  Extensions available for low level access Graphics API

15 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 15 DX11.X  Some DX12 features available right now on Xbox:  Turn off hazard tracking  Simple fence API  Deferred contexts re-implemented  New resource descriptor model  Draw bundles (Xbox specific, not the DX12 API) Graphics API

16 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 16 The CPU cannot saturate DRAM bandwidth on its own, the GPU can!  Significant performance degradation from DRAM contention  Fancy CPU features don’t help if memory starved 10. Use ESRAM as much as possible 20. Leave DRAM for the CPU and DMA 30. goto 10; DRAM - Contention

17 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM Hardware data cache pre-fetch units are awesome  Manual pre-fetch is near pointless once hardware pre-fetch is spinning  Wasting bandwidth if only operating on small arrays 2. Write combined memory pages and SSE streaming store instructions by-pass the cache  No load - halves the bandwidth consumed by the CPU 3. Pack your data!  Expanding / compressing data is cheap (CPU & GPU)  F16C (half float) CPU instructions  Store to load forwarding avoids LHS stalls 4. Swizzle your textures  Move engines can swizzle on copy DRAM – Love your bandwidth

18 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 18 Custom audio hardware  Very fast  Lots of features  Kinda cool! Nuff said Audio

19 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM ERA  Exclusive Resource Allocation  Only one active at a time  Custom OS  (Games!) 2. SRA  Shared Resource Allocation  Win8 core  (Apps) 3. Hypervisor  SRA and ERA use different virtual address space 3x Operating Systems

20 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 20 ERA can be in one of several states 1. Full screen  Full resources (even with snapped app up) 2. Constrained (Windowed)  Slightly less CPU and GPU resource  No input  Same amount of memory 3. Suspended  Zero CPU and GPU resource  No input  Same amount of memory Limited time to save after receiving a suspend message PLM (Program Lifetime Management)

21 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 21 Hardware:  Higher resolution colour and depth  Better ranges  New – infrared!  Microphone array  No tilt motor Software:  Improved skeletal tracking  Improved biometrics Kinect 2.0

22 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 22 6x Bluray = ~26 MiB/s  To install a 50 GiB Bluray at ~26 MiB/s = ~33 minutes Too long to wait… bored now…  Game must start after an initial payload has been installed.  When running title can hint as to what to install next.  No direct access to Bluray.  Could be digital download  It’s obvious but I’ll say it anyway – compress you assets! Streaming install

23 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 23 Cloud compute: Developer’s code is hosted and executed in Windows Azure Game code execution automatically scales based upon usage Live services: Stats, analytics, matchmaking & storage. Secure! The Cloud

24 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM Is your code 64-bit compliant? 2. Can you scale to 6 cores? 3. Adopt new DX11.X API extensions  Manage your own resource hazards 4. Make sure you use ESRAM effectively 5. Package content for streaming install  Game design considerations 6. Quick save on ERA termination 7. Kinect, Smartglass 8. Cloud services Challenges

25 AMD AND MICROSOFT GAME DEVELOPER DAY - June , STOCKHOLM 25 (That I’m allowed to answer) Thank You! – Questions?

26 © 2014 Microsoft The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. Microsoft makes no warranties, express, implied or statutory, as to the information in this presentation.


Download ppt "Inside Xbox One Martin Fuller Xbox Advanced Technology Group AMD AND MICROSOFT GAME DEVELOPER DAY - June 2 2014, STOCKHOLM."

Similar presentations


Ads by Google