Could PlayStation 4 breathe new life into Software Based Rendering?
- Thread starter onQ
- Start date
- Status
AFAICS DSoup's post was just a joke mixing together titles that refer to all the qualities you wanted to see in a game.
As for being right, people can read through this thread at your points and agree or not (
I disagree 
D
Deleted member 11852
Guest
Yup, hence the emoticon.
AFAICS DSoup's post was just a joke mixing together titles that refer to all the qualities you wanted to see in a game.
As for being right, people can read through this thread at your points and agree or not (). The only real software renderer I see at the moment is MM's Dreams, and it's not using the APU is a magical new HSA solution as you were suggesting PS4 is uniquely positioned to do. It's using GPU compute. Then we have the likes of Sebbbi's work moving the rendering of traditional triangles fully over the to GPU, having it create the drawing jobs, which again doesn't require PS4 nor HSA. Other points raised like compute-based lighting aren't software rendering. At the root of the disagreements in this thread is a difference in view of what Software Rendering is which isn't worth discussing again.I disagree
The PS4 APU is a product of HSA, the reason that games like Dreams is happening is because it's design made it easier for the devs to take advantage of compute.
Signed distance field rendering is possible on nVidia GPUs. It isn't advanced by HSA. The only reason MM are pushing that boundary is because they are a fully supported 1st party company given the freedom to explore (three years funded experimentation with no need to release a product), with a product vision that required lots of experimental techniques to be tried before MM could find a solution. That included producing tools to enable creation of assets, where everyone else is building engines around the existing tools (triangle based modellers, Maya, ZBrush).
Plenty of other devs on other hardware could implement the same sorts of compute-based solutions given the time and low-pressure funding and the same product requirements. PS4 hardware designs aren't an enabler in this case. I don't recall a single reference to HSA in Alex's Siggraph presentation on Dreams.
Plenty of other devs on other hardware could implement the same sorts of compute-based solutions given the time and low-pressure funding and the same product requirements. PS4 hardware designs aren't an enabler in this case. I don't recall a single reference to HSA in Alex's Siggraph presentation on Dreams.
Last edited:
+1 and some of the R&D is very interesting and could be achieve in other titles or for refinement renderer on a more powerful hardware.
Some of the research was pretty old. One reference for realtime geometry was a technique developed in 2002 running on GeForce 3!
Where do I say that this is only achievable on PS4 with HSA? I said that PS4 APU is a product of HSA & Games like Dreams are happening because the design of the PS4 APU made it easier for devs to take advantage of compute.
You brought up my talk of the APU design to use as a loophole & I explain that it is a product of HSA & it's design did help this to happen.
Your words
There is no future in it yet we are here looking right at it.
Last edited:
That has nothing to do with HSA and everything to do with compute.
The 'future' you are talking of is the "some fringe games" I refer to elsewhere in this thread (and Grall even suggests in post 2) but can't be bothered to go searching for. Yes, some novel engines will explore completely new ways of doing things with specific aesthetics that match the engine (just as they always have, even with 2D hardware), but that'll be enabled by the advance of compute and programmable GPUs and has squat to do with PS4's design. It'll happen with or without PS4. For everything else there'll be triangle rasterisation making use of the tools to create triangle meshes and the hardware built into the GPU designed to draw triangles, and PS4 won't make a dent into that standard.
Still trying to jump through that loophole, this has nothing to do with HSA but the PS4 APU is a product of HSA & the APU was designed in a way that made it better at computing. I simply asked what could be done using the PS4 for software based rendering seeing as it was designed with better GPGPU computing in mind.
Michellstar
Regular
the apus in xbox or ps4, or socs in mobiles are integrated for costs reasons, it's the main driving force.
like fpus before, or memory controllers, now gpus tomorrow even memory
like fpus before, or memory controllers, now gpus tomorrow even memory
No you weren't. You were talking about the close-knit relationship between CPU and GPU enabling better software rendering. See post 9...
And post 14...
If all you're interested in is how GPGPU/compute can be used in software rendering, the title of this thread should be, "could next-gen GPUs and compute breathe new life into software rendering?" and it'd be platform agnostic and only mention HSA in passing as a question over if that brings any advantages or not. And then the discussion should be held in the Algorithms forum and would talk not about software rendering but about programmable GPU techniques for solving (parts of) image drawing, whether using triangle rasterisation or not.
I asked about the PS4 because it has a 8 x 8 MIMD design with shared memory in a closed box.
The PS4 GPU was modified so it could have 8 Compute-only pipelines & each pipeline has 8 queues.
Shared memory doesn't make any difference if you're not talking about CPU<>GPU (HSA) interaction. For GPU compute, it works from its RAM pool.
The ACE's don't enable compute. They only improve efficiency of CU utilisation. There's nothing PS4's CUs can do that larger discrete PC parts can't (and much better when those parts are high end). It's certainly not an '8x8 MIMD' design, whatever that is. It's a GPU with 18 CUs (processors) and 8 compute instruction dispatch engines with 8 instruction queues each. A single compute instruction dispatch engine would be able to do everything PS4's GPU can, just with having to do the jobs serially so less efficiently. But the actual novel rendering algorithms you're talking of don't require multiple ACEs and only require advancement of the core GPU processor architecture to support more varied code.
See number 4? The CPU & GPU is formed together as a MIMD processor.
"As for the "supercharged" parts, or the parts that SCE extended, he said, "There are many, but four of them are representative." They are (1) a structure that realizes high-speed data transmission between the CPU and GPU, (2) a structure that reduces the number of times that data is written back from the cache memory in the GPU, (3) a structure that enables to set priorities in multiple layers in regard to arithmetic and graphics processing and (4) a function to make the CPU take over the pre-processing to be conducted by the GPU."
hesido
Regular
That's not what Shifty and HTupolev are talking about though. You need to eloborate that 8x8 MIMD term. What I understand from the fourth is results of GPU work can easily be handed off to the CPU. I'm guessing memory mapping changes here without any copy operation. Which is great and possible on consoles only.
Last edited:
- Status
Similar threads
