PC system impacts from tech like UE5? [Storage, RAM] *spawn*
- Thread starter MistaPi
- Start date
![the-ps5-features-a-custom-io-unit-and-a-custom-flash-control_6m26[1].jpg](https://forum.beyond3d.com/data/attachments/3/3912-b6f5cff6c969a940c804cd82e43bde47.jpg "the-ps5-features-a-custom-io-unit-and-a-custom-flash-control_6m26[1].jpg")
Last edited by a moderator:
This hardware IO is easier to implement in a fixed console design. But for PCs, which hardware block is going to pay for the transistor budget?
AMD/ Nvidia on the GPU?
Intel/ AMD on the chipset/ CPU?
SSD manufacturers on the SSD itself/ or HW controllers?
Or purely software driven on CPUs?
If the hardware IO ends up somewhere at the expense of other transistor logic, competition can downplay one metric at the expense of another and slowdown adoption, for example. I suppose if DirectX implemented an API, whether it's software or hardware is left open for acceleration, but where will the hardware reside?
depends on what marketing wants.
Ryzen 5000 series now with new hardware compression for IO load times increase x amount , UE5 games get X amount faster.
New PCI-E 4 nvme drive now with hadware compression logic for IO load times , so on and so forth.
Depending on the actual cost to include we could see it everywhere. What if the compression hardware is on the SSD , CPU and Video card ? Would they be able to just keep the textures compressed even in graphics ram ?
Problem in decompressing inside ssd is that it could lead to bottleneck in pci-e bandwidth. I believe most nvme ssd's and slots are 4 lanes wide. fast nvme ssd be it pci-e 3 or 4 can saturate the respective 4 lane implementation. Decompression in ssd might require 8 lanes or even more to connect the ssd. Decompressing in io-chip/gpu/cpu would save significant amount of pci-e bandwidth.
By hardware IO I assume you're talking about the decompression unit? If so then I don't expect to see that in PC's any time soon. How could they choose what formats it will handle for a start? Even the two major consoles are using different formats. And how would software installs on the SSD know whether the system had hardware decompression or not? And without knowing they'd have to be conservative with compression to ensure it was possible in software which makes hardware decompression arguably irrelevant. Although there could still be some benefit of course assuming the correct format was supported.
In terms of who'd pay for the silicon, the SSD doesn't make much sense because that's on the wrong side of the PCIe interface. Gen4 SSD's will already max out PCIe 4x4 this year so adding decompression on the SSD side would gain you nothing in terms of bandwidth like it does on the consoles. It would save disk space of course, but some high end SSD's already do this so that's nothing new. Coupled with PCIe 5.0 it could get interesting though. 14GB/s of uncompressed data over PCIe is more than enough to keep up with the consoles but with hardware compression and decompression on the SSD itself, you'd also get all advantages of the saved disk space.
I also don't expect to see it on the CPU chip/APU any time soon either unless the industry standardises on a single compression format for everything on the disk like the consoles. It's also too specific to high end gaming while these CPU's have to appeal to a much wider market. It might be a great use case for iGPU's though. But given AMD only integrates GPU onto it's low end parts that doesn't seem to fit either.
The GPU sounds like a possibility but then you'd only benefit for data that's being read directly by the GPU and doesn't need any kind of processing on the CPU. Otherwise I guess it would still need to be decompressed on the CPU in software. I've no idea what the split of those two data types are though to know whether that's a worthwhile compromise.
PC gamers will buy hardware to make games run better and there's a lot of empty PCIe slots on most motherboards now that multiple GPUs isn't really a thing anymore. I'd imaging OEMs would be very willing to build and sell DirectStorage cards under their gaming brands. Aside from that, even the mini-ITX motherboard in my secondary PC has 2 m.2 slots, so that's not exactly uncommon.
Unknown Soldier
Veteran
Really? The only people I see buying hardware right away will be hardcore gamers/YT Gamers if they can afford it and pros. Your casual gamer won't necessarily buy such hardware unless it's cheap or at least doesn't hit the wallet that hard. And there's a lot more casual gamers than hardcore/pros. It's why the console is popular, it offers good hardware at a great price at launch.
Yes, really.
Unknown Soldier
Veteran
I'm not convinced. And I won't be unless the hardware is dirt cheap, which it won't be.
Last edited:
The real question of course being - in what volume?
The most popular CPU among those dedicated enough to build their own gaming rigs right now is the AMD 3600. At the German shop Mindfactory it outsells the 3950 by 30:1. The gamers who buy second hand et cetera, or even use laptops don’t even have a chance to register in these statistics. PC gamers are more sensible with their money than the typical tech site material implies.
Nvidia just reported $1.34 Billion in revenue from gaming in the last quarter. That's a lot of people willing to spend money for a better gaming experience.
Don't get me wrong, I'll be first in line to take one of these drives. I don't think it'll be a hard sell to the PC gaming population. We used to clamor for sound cards....
Errr - that’s the Switch breaking records.
Regarding the PC part of that revenue, it is inversely proportional to the price per device.
We know the add in board volume, and we have a fair grasp of product distribution. The lunatic fringe is small. Attractive marks, obviously, but not a large group relatively. (We can quibble about the size, and where to draw lines, but ”PC gamers" as a group are cheap. If you decide not to include the Fortnight/.. volume of players in your target demographics, well, then your target demographics suddenly get one hell of a lot smaller in volume.)
You can't really know because at the moment, hardware changes aren't essential. If it became a case that they had to buy an upgrade or not play PC games any more, would they give up PC gaming, move to consoles, or upgrade?
They attribute it to lock-down people buying new GPUs. Data Centre revenue was up 80%
Consumer GPUs have also been increasing as I understand it. Sure, NSW helps, but it's not the driver here.
While it wouldn't happen over night, I think utilizing the 8x/16x PCIe slots on motherboards for "gaming tailored SSDs" or drives with custom decompression/compression blocks on the PCB is the way to go until M.2 slots on the motherboards get enough bandwidth to be able to support it.
Obviously not everyone is going to have a spare 8x/16x PCIe slot available, however, games should be able to be designed to scale down well enough that they should work on a good quality M.2 NVMe drive as it is. So people who don't have a spare PCIe slot, or don't want to purchase a new drive, can at least play the games, just not at the highest settings. The people who would (and there are tons who would I'm sure) are likely people who already have higher end equipment and spend tons of money on stuff like this in the first place.
I'm not saying it's ideal, but I mean we have GPUs, which were meant to accelerate graphics, Soundcards which were meant to improve audio, Network cards which were meant to improve networking performance... and we also have PCIe SSDs meant to improve storage speeds and capacity.... why not have a card specifically designed to accelerate storage for the purpose of gaming, if it will help?
Obviously not everyone is going to have a spare 8x/16x PCIe slot available, however, games should be able to be designed to scale down well enough that they should work on a good quality M.2 NVMe drive as it is. So people who don't have a spare PCIe slot, or don't want to purchase a new drive, can at least play the games, just not at the highest settings. The people who would (and there are tons who would I'm sure) are likely people who already have higher end equipment and spend tons of money on stuff like this in the first place.
I'm not saying it's ideal, but I mean we have GPUs, which were meant to accelerate graphics, Soundcards which were meant to improve audio, Network cards which were meant to improve networking performance... and we also have PCIe SSDs meant to improve storage speeds and capacity.... why not have a card specifically designed to accelerate storage for the purpose of gaming, if it will help?
They don't at the moment, but when it comes to next gen games, it might be something devs have to take into consideration to allow storing more of the level and assets into RAM so that the typical NVMe drive can keep up.
I'm of the opinion that games should utilize as much of the resources available to them. Fill all my RAM and all my VRAM as much as you can to cache data and reduce the pressure on the I/O as much as possible.
I'm of the opinion that games should utilize as much of the resources available to them. Fill all my RAM and all my VRAM as much as you can to cache data and reduce the pressure on the I/O as much as possible.
One of the "leaks" on YouTube said Nvidia is working on a RAM/SSD caching system. I wouldn't put it past them to put dedicated hardware on the GPU and use their MDF to drive adoption. Also, if studios are going to standardize on Kraken/DirectStorage then GPU companies would be smart to follow.
Last edited:
No CPU has sufficient PCIe lanes to allow a 16x GPU + either and 8x or 16x SSD. The best you could get would be an 8x SSD but then you'd be halving the bandwidth to your GPU.
There's little reason why M.2 NVMe drives can't be fast enough. On a 4x interface they'll be pushing 7GB/s of data by the end of this year. That's already quite a bit more than the XSX and not far off PS5 speed. Also with a relatively low amount of compression applied (although still enough to take out a couple of cores of the average CPU!) that 7GB could reach or even exceed the PS5's compressed throughput.
And that's just with PCIe 4.0. We might see PCIe 5.0 as early as next year in Zen 4 and/or Alder Lake which would likely soon be followed by NVMe drives in the 10-14GB/s range.
The people who would (and there are tons who would I'm sure) are likely people who already have higher end equipment and spend tons of money on stuff like this in the first place.
Last edited:
Similar threads
