Wasnt there an IN and OUT HDMI so you could get the TV channels into the XBOX?
Maybe MS havent abandoned the idea completely....or maybe....we should expect some other interaction between games and TV? Or maybe.....its something else
Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]
- Thread starter BRiT
- Start date
- Status
Maybe MS havent abandoned the idea completely....or maybe....we should expect some other interaction between games and TV? Or maybe.....its something else
An HDMI port per GPU.
XD
My mind started thinking they could support VR after all, although it would make more sense to use Virtual Link now. Plus the port should be on the front.
Yes. Pass through for cable.
Fiscal?
Tkumpathenurpahl
Veteran
Cool. Although it might be a bit expensive for a PS5 accessory...
I'm kidding. Please don't take this as a starting point for some fanboy nonsense.
Wrote the script a week ago?
Edit:
Lisa Su just confirmed both consoles launching 2020
Sidenote: I know what I'm purchasing in February.... *drool*
64 core Threadripper in Xbox Series X at $399 launch price confirmed BOOM
cheapchips
Veteran
I know it's just a render but there's no vents on the sides. There's just the holes in the top and a thin rim around the base. Hmm. Not sure I wanted to write that sentence.
D
Deleted member 13524
Guest
Great to know Microsoft is going for the TV TV SPORTS TV crowd.
That went so well last time..
That went so well last time..
They'll just reduce the cache from 288MB to around 16MB. Problem solved.
My guesses is 2nd screen for VR in the future, or for streaming. They have a huge investment in mixer and it's pretty hard for streamers to interact with their community from a console without a complicated setup. A 2nd screen directly on the xbox for the dashboard would be a big help. Maybe the system can keep an app open on the 2nd screen, like netflix, a web browser or whatever. But I'd be more inclined to suspect the streaming setup and VR being the primary reasons.
Haha 10/10!
Stuff like that I'd have thought would be better off handled in customised hardware, such as having a PS4 execution mode that maps a whole load of stuff differently. Even if so, the wavefronts are executed completely differently on RDNA than GCN so the code flow will be different. If the wrong number of CUs is enough to break stuff, wouldn't the change to RDNA be even more impactful with code working at that low a level? And what are the GCN-compatilbility features of RDNA and are they removed in RDNA2??
Wave64 mode looks like it provides a partial foundation for handling the change in width.
It's not sufficient to carry the architecture all the way to transparent backwards compatibility, since RDNA issues two instructions per single original GCN op.
However, some operations in RDNA had small changes made to allow them to behave differently in 32 and 64 lane scenarios. The read and write lane operations can apparently be given larger register ID values to cover for the larger number of lanes.
Other cross-lane operations introduced in later GCN generations are more restricted in how they can read from lanes in RDNA, making the change in width explicit. Sea Islands compatibility isn't affected since it predates those operations. A theoretical enhanced compatibility mode might be able to create a decode method that would extend instruction behavior automatically to wave64 mode without requiring the second instruction issue. There's a direct path for the hardware's decision making at the point of instruction decode in that mode.
There are other issues with RDNA that would require workarounds, such as missing instructions and some bugs that will likely need fixing.
A change to CU count, if PS4 code was allowed lower-level access, might leave the hardware hamstrung if the code can set or read back CU counts or CU bit masks. The hardware wouldn't be able to analyze the overall shader to know whether it can safely run past the 18-CU limit, treat it as if a barrier of some kind existed there, or if it needs to map a given command to a given CU or a second CU in the PS4 Pro. A query of CU values (activity mask, hardware ID) leaves the hardware stuck with no means of communicating higher CU numbers to code that isn't expecting higher values.
There are ways of catching unsupported operations, or ways of implementing extra paths to support them. That's not the same thing as supporting them in a cycle-exact manner.
One difference I made note of for RDNA is its 25% longer VALU latency, which doesn't affect instruction support while having a realistic chance of changing performance. Re-engineering the SIMDs to have a register file and ALU pipeline of equal latency to GCN is one approach, although more difficult to implement if the idea is the hardware can also support RDNA execution at RDNA+ clocks.
Some removed instructions/modes include VSKIP and the fork/join instructions for complex control flow. The loss of the former seems to indicate that this part of RDNA is quite different from GCN, and the loss of the latter seems to make sense since the branch-handling stack has baked-in assumptions about how deep it needs to be with 64-wide waves. RDNA's caches are different and may have corner cases concerning changes in alignment and preferred granularity, which can effect timings.
I'm not discounting some undisclosed way to make RDNA behave that differently from its new pipeline arrangement in hardware, but there are other methods that aren't slavishly beholden to cycle-exact behavior. Sony has patents for being able to cleanly support changing parameters like clock and timings on different hardware. So in that regard, I think there is some kind of context missing for the backwards compatibility modes under discussion that have numbers that don't make allowances for such things.
I don't want to believe they're launching XSX next month Who’s in Europe and staying up hoping that Sony has a big PS5 reveal, which probably won’t happen cause it’s CES, and will get very annoyed that I knew it was going to be a waste of time, why didn’t I just go to sleep silly cow?
Just me?
Just me?
- Status
Similar threads
- Locked
