Depends. It‘s a case by case basis.Why the whole game in memory? The game just need to have the level or what you are seeing. Why the game would need to load the last level in the memory while I am playing the first lvl yet?
You can hear this GIF.Sony HQ
Maybe they can load much more of the open world map with better draw distance on memory while still loading the rest when needed. Plus PC can still use a CPU core to help because, unlike consoles, it doesn't need an entire core just for the OS.Depends. It‘s a case by case basis.
Open Worlds for example has no such levels. Maximum speed of the character is dictated by the I/O speed of the system which was shown in the Spiderman presentation for the PS5. What if I fast travel then? I would have to wait by the traditional amount of time it takes for it to load since it’s not entirely loaded in memory.
The reason they want fast I/O is exactly what you said in the first sentence.Maybe they can load much more of the open world map with better draw distance on memory while still loading the rest when needed. Plus PC can still use a CPU core to help because, unlike consoles, it doesn't need an entire core just for the OS.
You still need an I/O fast enough on the PC in order to fill the ram fast enough. The PS5's I/O directly fills the GPU RAM and bypasses the CPU, which it will not be able to do on the PC at the moment. "More RAM" isn't always the answer if it's bogged down by the current I/O extraction layers. You will be waiting for all that RAM to fill up and dump, waiting turns, and asking for extraction layer permission so to speak. All robbing resources from other areas that brute forcing eventually runs into thermal thresholds.All those games if put on PC would need the same speed of the IO all the time or they could just use more memory?
Not really surprising PC is ahead with such a huge lead. Its probably much easier to publish games there and its public tends to be more open to new creative concepts, along with stronger word-of-mouth advertising.
also lol at the title's wording
When we talk about PS5 I/O, we aren't talking about the SSD or the speed (5.5 GB/s) of the SSD. That's not a problem for PC. On the hardware side, PC currently lacks the dedicated decompression unit that consoles (PS5/XSX | S) have. On the software side, its existing file I/O protocols are ancient. DirectStorage is being pushed to overcome this at least on the software side of things.Because pc lacks memory and bandwidth like consoles and because of that need the same speed of the I/O to compensate?
All cool, but I am saying what if devs don´t need to rely on IO/decompression/SSD so much like they will need on consoles because PC does not lack memory and bandwidth like consoles? Maybe they can "brute force" hte use of more memory to load more stuff needed.When we talk about PS5 I/O, we aren't talking about the SSD or the speed (5.5 GB/s) of the SSD. That's not a problem for PC. On the hardware side, PC currently lacks the dedicated decompression unit that consoles (PS5/XSX | S) have. On the software side, its existing file I/O protocols are ancient. DirectStorage is being pushed to overcome this at least on the software side of things.
Decompression with Gen4 SSD speeds can take up to 24 cores with a conventional CPU (Threadripper) according to NVIDIA, 11 to 13 Zen 2 cores according to Sony and MS, respectively. With just 8 cores inside consoles, you can see they can't just rely on the main CPU for decompression of the game files. This is why console manufacturers built a custom decompression block into the I/O unit of the main SoC that offloads those tasks from the main CPU so the CPU can focus on what it's meant to do like processing of the game physics, instructions, preparing draw calls, etc. without having to worry about decompression overhead that much.
And this is where it's a problem for PC. DirectStorage kinda solves the problem on a software level, not on a hardware level. Gaming PCs with 6C/12T CPU like Ryzen 3600 or 5600X, heck even 8 cores are quickly going to get saturated when the decompression with Gen4 SSD is taking place leaving nothing for the games. They just aren't gonna cut it when games demand data to be decompressed and loaded from the NVMe to VRAM as soon as possible.
To solve this hardware problem NVIDIA is proposing a solution called "RTX I/O" where instead of using the CPU cores, with DirectStorage the game will use the GPU's CUDA cores or the SMs for decompressing the data. This is still going to impact gaming perf, IMO, and NVIDIA won't directly tell you or talk about this, at least not now. Because we'll likely see a dedicated decompression unit similar to consoles in the next-gen PC GPUs that will offload this task from the SMs. I mean this is not some wild concept. Remember this chart?
They'll have something like this up for the I/O decompression talk when selling their next-gen GPUs to you and show you how having a dedicated decompression, I/O cores in the GPU speeds up gaming perf instead of relying on the SM or the CPU.
This means gamers with Turing and Ampere GPUs are likely screwed if/when that happens just like GTX users are screwed without RT and AI cores. Think about this, even the console manufacturers could have gone the same route as NVIDIA and used their GPUs for decompression work, but they didn't. They felt the need to add a dedicated block so that it didn't impact gaming perf.
Now, do you understand what we mean when we talk about PS5/XSX I/O unit? Even XSX and XSS have some form of hardware decompression unit in the SoC.
Man, you still don't understand. Even if you want to use "more memory" it still doesn't solve the challenge on the hardware side of decompressing those data and bringing them into the memory.All cool, but I am saying what if devs don´t need to rely on IO/decompression/SSD so much like they will need on consoles because PC does not lack memory and bandwidth like consoles? Maybe they can "brute force" hte use of more memory to load more stuff needed.
Man, you still don't understand. Even if you want to use "more memory" it still doesn't solve the challenge on the hardware side of decompressing those data and bringing them into the memory.
If they could just "use more memory" NVIDIA wouldn't have come out with an "RTX I/O" solution:
Timestamped:
Or
MS wouldn't have come up with DirectStorage:
In this Microsoft Stack talk, Andrew Yeung (DirectStorage Engineer) is alluding to future GPUs incorporating dedicated hardware for I/O and that there will be an "exciting evolution to DirectStorage". This is exactly what I was talking about in my prev post.
Ok firstly, consoles aren't relying on "IO/decompression/SSD" because they lack memory and bandwidth. This is your misconception.I did not say "just" use more memory. I did say "what if devs don´t need to rely on IO/decompression/SSD SO MUCH like they will need on consoles". I expect them to make good use of all my 32+10GB of RAMs, while they only have 13GB on consoles to use.
I didn´t. Is it saying it is impossible to use more memory and less of the I/O? If yes, than I have to wait and see what MS, Nvidia and AMD will do.Ok firstly, consoles aren't relying on "IO/decompression/SSD" because they lack memory and bandwidth. This is your misconception.
Second, devs on PC/consoles have to rely on IO/decompression/SSD along with DirectStorage + SFS because this is a requirement due to NVMe speeds, and "just use more memory" isn't the answer to it. You still have to decompress for those data to go into the memory. This is about decompression. Even if consoles had 32GB RAM and more (applies to PC) they'd still require those dedicated I/O, decompression blocks because, as stated by NVIDIA, Sony, MS, a conventional CPU just doesn't cut it for Gen4 SSD's 4-5+ GB/s of speeds. You had HDD with 50-100 MB/s speeds and CPUs with fewer cores were doing ok. But NVMe is like over 100x faster thus requires the need for a dedicated decompression unit or GPU decompression.
I can see that you haven't watched NV's vid that I linked/probably didn't understand.
And it shows. You clearly don't understand how these things work and you aren't even willing to learn when provided with correct info.I didn´t.
On a PC, it is currently impossible to decompress at NVMe speeds with a 6-8 core CPU and have the same perf as a machine with dedicated HW-accelerated decompression or GPU decompression. See below:Is it saying it is impossible to use more memory and less of the I/O?
Nothing for nothing, but 2/3rds of the threads in this forum read like marketing for the PS5 or Series X (Mostly PS5).
Confirmation bias be strong round these parts.
Both systems are cool.
Correct. Hopefully we'll be seeing such implementations with the nVidia 40xx series and whatever AMD calls their next gen GPU (7xxx?)And it shows. You clearly don't understand how these things work and you aren't even willing to learn when provided with correct info.
On a PC, it is currently impossible to decompress at NVMe speeds with a 6-8 core CPU and have the same perf as a machine with dedicated HW-accelerated decompression or GPU decompression. See below:
This is from the DirectStorage presentation vid I linked. "No such solution for PC". i.e. there's no such solution on PC as it is in the PS5/XSX/S. PC is playing catch up relying on GPU for the decompression as a temporary fix until they incorporate a dedicated decompression unit in the GPU silicon.
I think it’s already implemented hardware wise into the RTX 30xx-series, it’s mostly down to DirectStorage.Correct. Hopefully we'll be seeing such implementations with the nVidia 40xx series and whatever AMD calls their next gen GPU (7xxx?)
No, there's no dedicated decompression unit in RTX cards like you see in consoles. RTX IO will make use of GPU's SM (AMD's equivalent of CU) for this task. Maybe next-gen NV GPUs will have a separate silicon dedicated for I/O.I think it’s already implemented hardware wise into the RTX 30xx-series, it’s mostly down to DirectStorage.
Introducing NVIDIA RTX IO: GPU-Accelerated Storage Technology For The Next Generation of Games
Load instantaneously, experience vast worlds with endless views and rich detail, and further improve gameplay by leveraging the power of GeForce RTX 30 Series graphics cards and NVIDIA RTX IO.www.nvidia.com
AMD will also definitely implement something similar in RDNA3, anything else would be surprising.
Ah, my bad. Yeah there is definitely no mention of dedicated silicon specifically for that task.No, there's no dedicated decompression unit in RTX cards like you see in consoles. RTX IO will make use of GPU's SM (AMD's equivalent of CU) for this task. Maybe next-gen NV GPUs will have a separate silicon dedicated for I/O.
PC has no barrier for entry so it should appeal to more developers. No need to get a dev kit or follow TCRs and go through the cert process and have pricing dictated. Just make it and release it. Hell, You can host your own CDN if don’t want to use Steam/Epic/Etc.
in my (professional) experience from least to most friction of development and release its:
PC > Mobile > Console
Consoles have a lot of upside when developing, but not gonna lie and say it’s frictionless.