Next Generation Hardware Speculation with a Technical Spin [pre E3 2019]

Status
Not open for further replies.
Fresh Cerny patent (May 28 2019) about Real-time adjustment of application-specific operating parameters for backwards compatibility

Performance of a new system may be optimized to run a legacy application written for a legacy system. Performance information for the legacy application is recorded or derived while running the legacy application on the new system. Performance characteristics for the legacy application running on the new system are determined by analyzing the performance information. The performance characteristics include one or more key performance metrics and other performance information. The key performance metrics must be met when the legacy application is run on the new system. The other performance information is useful for adjusting one or more operating parameters of the new system when running the legacy application on the new system. The one or more operating parameters are adjusted so that the one or more key performance metrics are met.

The method of claim 1, wherein the one or more key performance metrics include to frames per second...wherein the legacy system and new system are video game systems...A Key Performance Metric is one that must be met when the application is run on the new hardware. Examples of Key Performance Metrics include, but are not limited to frames per second (e.g., in the case of video intensive applications, such as video games) and instructions per cycle (IPC) binned to a program counter (PC) range.

http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=/netahtml/PTO/search-adv.htm&r=1&f=G&l=50&d=PTXT&p=1&S1=(cerny+AND+sony)&OS=cerny+AND+sony&RS=(cerny+AND+sony)

Could they uncap the default framerate on some PS4 games using the BC emulator?
 
Fresh Cerny patent (May 28 2019) about Real-time adjustment of application-specific operating parameters for backwards compatibility




http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=/netahtml/PTO/search-adv.htm&r=1&f=G&l=50&d=PTXT&p=1&S1=(cerny+AND+sony)&OS=cerny+AND+sony&RS=(cerny+AND+sony)

Could they uncap the default framerate on some PS4 games using the BC emulator?

It reads to me like that patent uses Machine Learning to learn how to better process and render a legacy game.
 
Fresh Cerny patent (May 28 2019) about Real-time adjustment of application-specific operating parameters for backwards compatibility





http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&u=/netahtml/PTO/search-adv.htm&r=1&f=G&l=50&d=PTXT&p=1&S1=(cerny+AND+sony)&OS=cerny+AND+sony&RS=(cerny+AND+sony)

Could they uncap the default framerate on some PS4 games using the BC emulator?

Kind of reads like it's the other way around, that performance while running the old game will be monitored and system parameters will be modified to match the timings of the original system exactly. Maybe this could allow for higher rendering resolution/upscaling/insert your improved graphical presentation here to be patched in with an assurance you won't break anything due to timing issues.
 
Kind of reads like it's the other way around, that performance while running the old game will be monitored and system parameters will be modified to match the timings of the original system exactly. Maybe this could allow for higher rendering resolution/upscaling/insert your improved graphical presentation here to be patched in with an assurance you won't break anything due to timing issues.
Well, then they could also uncap the framerate with a patch + the stuff in the patent the way you described it.
 
Precisely. There's no need for patentable tech in running emulators above normal speeds.

Yeah, as expected this is spelled out exactly in the patent introduction...

Differences in performance of the hardware components of a new device and a legacy device can cause errors in synchronization on the new device, which may cause a legacy application to crash or produce incorrect output when running on a new device architecture. Such differences in performance can arise, e.g., from differences in hardware architecture between the new and legacy devices. It is within this context that aspects of the present disclosure arise.

The performance of an application on a new device may be closely matched to the performance of that same application on the legacy device by tuning the operating parameters of the new device.
This is all about forcing emulation parity to improve compatibility. It'd probably be used to fix fragile games, where games that can run full tilt are left alone to.
 
Precisely. There's no need for patentable tech in running emulators above normal speeds.

Not for a company like Microsoft, who have decades of experience of working with layers of abstraction to aid future legacy support but Sony have never really had to think about this so who knows what janky mess is lurking behind the PS4's interface. I reckon they have an option to disable boost mode on Pro for a reason and not just in case Pro owners want to run software a bit more shit than it otherwise would ;)

Why they'd bother to patent such a method is anybody's guess but patenting new methods for anything is the norm nowadays.
 
This is a Sony dev though. I know there was told of reverting to VLIW with Super SIMD but that seems far-fetched.

Maybe Sony knows that porting/remastering PS3 games is kinda stupid when 360 and PC versions are going to be more straightfoward to port? Why convert VLIW code if you can run it native.......lots of big games optimized for the Terascale architectures.
 
Not for a company like Microsoft, who have decades of experience of working with layers of abstraction to aid future legacy support but Sony have never really had to think about this so who knows what janky mess is lurking behind the PS4's interface. I reckon they have an option to disable boost mode on Pro for a reason and not just in case Pro owners want to run software a bit more shit than it otherwise would ;)

Why they'd bother to patent such a method is anybody's guess but patenting new methods for anything is the norm nowadays.

So... where does the patent fit when compared to existing BC implementations? Even MS implemented a legacy/performance mode for the enchanted 360 titles running on Scorpio.

(haven't had time to read through)
 
The patent is about collating metrics, and not what to do with them other than to aim to get the same metrics on the target machine to ensure games run the same. I don't think it fits in any other emulator strategy. If it does, Sony won't be the first one to try this!
 
So... where does the patent fit when compared to existing BC implementations?

I'm assuming this most likely applies directly to the new AMD GPUs having SFUs while the PS4 GPU has them as runtime macros or instructions, so the runtime characteristics between the two are vastly different. Everything old is new again, Navi having SFUs just like the old VLIW5 hardware had them.

Are the new SFUs instructions identified? Any performance given on how much of a difference between SFU and the older runtime macros or instructions? Any SFU instructions useful for RealTime RayTracing, like intersection tests?
 
So... where does the patent fit when compared to existing BC implementations? Even MS implemented a legacy/performance mode for the enchanted 360 titles running on Scorpio.

(haven't had time to read through)

I read it the patent overview twice and I'm unclear about what this method is doing specifically. Legacy code support is in Microsoft's DNA, Windows would never have got so widespread without it and it's difficult to imagine Microsoft designing anything new without leveraging existing methods, all of which are predicated on techniques that include legacy code support but I very much doubt that when Sony were architecting PS4's software stack that they were thinking about whether PS4 games would have to work on PS5 and when you don't think like that, making that happen x years down the line becomes much harder when things change substantially.
 
Someone please help me understand this particular part within this patent. I know it's PSVR (possibly PSVR 2) and motion tracking related, but the hardware or console they're describing connecting the device doesn't sound like PS4.

Attention-based rendering and fidelity
FIG. 1 is a block diagram of an exemplary electronic entertainment system 100. The entertainment system 100 includes a main memory 102, a central processing unit (CPU) 104, at least one vector unit 106, a graphics processing unit 108, an input/output (I/O) processor 110, an I/O processor memory 112, a controller interface 114, a memory card 116, a Universal Serial Bus (USB) interface 118, and an IEEE 1394 interface 120, an auxiliary (AUX) interface 122 for connecting a tracking device 124, although other bus standards and interfaces may be utilized. The entertainment system 100 further includes an operating system read-only memory (OS ROM) 126, a sound processing unit 128, an optical disc control unit 130, and a hard disc drive 132, which are connected via a bus 134 to the I/O processor 110. The entertainment system 100 further includes at least one tracking device 124.

The CPU 104, the vector unit 106, the graphics processing unit 108, and the I/O processor 110 communicate via a system bus 136. Further, the CPU 104 communicates with the main memory 102 via a dedicated bus 138, while the vector unit 106 and the graphics processing unit 108 may communicate through a dedicated bus 140. The CPU 104 executes programs stored in the OS ROM 126 and the main memory 102. The main memory 102 may contain pre-stored programs and programs transferred through the I/O Processor 110 from a CD-ROM, DVD-ROM, or other optical disc (not shown) using the optical disc control unit 132. The I/O processor 110 primarily controls data exchanges between the various devices of the entertainment system 100 including the CPU 104, the vector unit 106, the graphics processing unit 108, and the controller interface 114.

The graphics processing unit 108 executes graphics instructions received from the CPU 104 and the vector unit 106 to produce images for display on a display device (not shown). For example, the vector unit 106 may transform objects from three-dimensional coordinates to two-dimensional coordinates, and send the two-dimensional coordinates to the graphics processing unit 108. Furthermore, the sound processing unit 130 executes instructions to produce sound signals that are outputted to an audio device such as speakers (not shown).

Why are they describing the vector unit as separate logic outside of the GPU? I'm pretty sure all modern day CPUs/GPUs have some form of integrated vector logic / instructions sets on handling vector arrays. Am I misreading something here? Because GCN has the vector units integrated into them (see pic below). So why is the patent referring to a decoupled vector setup? If not, wouldn't the patent embodiment simply state that GPU is handling "render related task," rather than stating that the vector unit can communicate directly to the GPU through a dedicated bus as if it was something separate? Or is the patent being pedantic on the process of moving data around within the GPU? I'm confused...

44981-image.jpg


-----------------------------------------------------------------------------------------------------------------------

Anyhow, the main part of the patent if anyone is interested.
Methods and systems for attention-based rendering on an entertainment system are provided. A tracking device captures data associated with a user, which is used to determine that a user has reacted (e.g., visually or emotionally) to a particular part of the screen. The processing power is increased in this part of the screen, which increases detail and fidelity of the graphics and/or updating speed. The processing power in the areas of the screen that the user is not paying attention to is decreased and diverted from those areas, resulting in decreased detail and fidelity of the graphics and/or decreased updating speed.
 
Last edited:
Someone please help me understand this particular part within this patent. I know it's PSVR (possibly PSVR 2) and motion tracking related, but the hardware or console they're describing connecting the device doesn't sound like PS4.

Attention-based rendering and fidelity






Why are they describing the vector unit as separate logic outside of the GPU? I'm pretty sure all modern day CPUs/GPUs have some form of integrated vector logic / instructions sets on handling vector arrays. Am I misreading something here? Because GCN has the vector units integrated into them (see pic). So why is the patent referring to a decoupled vector setup? If not, wouldn't the patent embodiment simply state that GPU is handling "render related task," rather than stating that the vector unit can communicate directly to the GPU through a dedicated bus as if it was something separate? Or is the patent being pedantic on the process of moving data around within the GPU? I'm confused...

gcn-7000-series-southern-island-structure.jpg


-----------------------------------------------------------------------------------------------------------------------

Anyhow, the main part of the patent if anyone is interested.

Because it’s like an AMD chiplet design. The first drawing shows everything hooked into an I/O processor.

It should be noted that the first figure looks like an apu that communicates with the i/o processor. The cpu, gpu, vector unit and main memory communicate through a system bus that connects to an I/O processor. The gpu and vector unit has a dedicated bus between them while the cpu and main memory has a dedicated bus. The i/o processor has it own external memory pool.

It doesn’t look like any design I’ve seen from Sony. But I haven’t look at Sony patents in years.

Stripping out the uncore would allow for a smaller apu.
 
Someone please help me understand this particular part within this patent. I know it's PSVR (possibly PSVR 2) and motion tracking related, but the hardware or console they're describing connecting the device doesn't sound like PS4.

Attention-based rendering and fidelity






Why are they describing the vector unit as separate logic outside of the GPU? I'm pretty sure all modern day CPUs/GPUs have some form of integrated vector logic / instructions sets on handling vector arrays. Am I misreading something here? Because GCN has the vector units integrated into them (see pic). So why is the patent referring to a decoupled vector setup? If not, wouldn't the patent embodiment simply state that GPU is handling "render related task," rather than stating that the vector unit can communicate directly to the GPU through a dedicated bus as if it was something separate? Or is the patent being pedantic on the process of moving data around within the GPU? I'm confused...

gcn-7000-series-southern-island-structure.jpg


-----------------------------------------------------------------------------------------------------------------------

Anyhow, the main part of the patent if anyone is interested.

Emotion engine in PSVR2. Or better yet: it's Cell. Someone hold Shifty in place before he gets too excited.
 
Status
Not open for further replies.
Back
Top