Series S has even less. 7.5 GB shared memory for games
To be pedantic: Games have 8 GB of memory on Series S. The OS reservation is smaller for that console.
Series S has even less. 7.5 GB shared memory for games
Series S has even less. 7.5 GB shared memory for games results into just around 4 GB video memory in most games if you compare texture quality, that's the baseline devs have to work with and the reason why entry level GPUs have 4 GB as well.
But don't fret. Software gets more efficient like Virtual Texturing and SFS, thanks to SSDs much less stuff has to be loaded in VRAM compared to before which can be used to increase texture quality for the same amount of VRAM.
Tech demos like Valley of the Ancient and the matrix demo have crisp 8K textures and only use around 6-7 GB VRAM in 4K.
It's not like theres 16gb available to just VRAM allocation generally on say the PS5 either. let's say 3/4GB will go to the OS (if its anything like the PS4), your looking at around 12GB for games?
That 12 is most likely not going to be allocated in full to the GPU, you need ram for game logic/assets etc too. I'd hazard a guess that around 8 to 10gb at most for VRAM would be a very acceptable amount for high-quality AAA games if well optimized, probably closer to 8gb.
PC gpu's are most often assisted by main system memory which adds another 16 or so GB's of DDR4 (and now DDR5). Someone with a 6gb GPU is most likely running a RTX2060 or below, which is when you want to use DLSS anyway (reducing the amount of vram needed) and lowered settings.
I do not see vram to be a problem at all. AMD gpu's generally sport 16gb dedicated vram (6800 and up), these amounts are likely going to increase. Also not to forget that pc gpu's usually dont content with the CPU for BW/latency and generally much faster memory speeds at that.
I think last generation when consoles had 8gb gddr at the same time when 2gb gpu's where a thing differences in vram could have played a role backthen. Although 3gb gpus (7950 and better) and some 4gb NV models, 2gb gpus where the most common. That was a 4x increase looking at gddr in the consoles favour.
Nowadays not so much, were not looking at a 4 times difference.
You are right, textures on the big consoles are in line what PC GPUs with 8 GB use. And the Matrix demo uses textures specifically optimized for the big consoles and that needs around 7 GB at 4K.
In a game like Watch Dogs Legion, you can max it out with Raytracing and the HD texture pack (which looks a lot better) on a 3080 without any issues, while the PS5 uses lower settings without the HD texture pack. That tells you available video memory for the PS5 in this game must be lower than 10 GB, or else the texture pack would be available.
I assume in the future games will be even more demanding game logic wise because they will finally put the Ryzen CPU in good use, meaning even less VRAM available for these consoles as needed DRAM increases.
The statistical term that you're looking for is the numbers are representative of the user base.STEAM number are and can be misleading as it samples a small amount and then projects totals based on what it samples.
The statistical term that you're looking for is the numbers are representative of the user base.
Because the samples are completely random, misrepresentative data or point skews from month to month would manifest themselves is observable trends of wildly adjusting and inconsistent reports of hardware in use. That does not happen because the probability distribution is sound.
The second number is GPU utilization. You're looking at fan speed. The cost of temporal reconstruction is considerably more than 6%.In the 4k vs 1440P display resolution comparison, do you have it locked to 60 fps? Because both show the same framerate. If you do have it locked, the only difference would show in GPU utilization % which is about 6% higher for the one with the 4k display resolution but then maybe that's jumping around too much even in like-for-like scenes.
Not skews, just differences between regions. I am not aware that the SHS is region specific so a non-issue.It also skews from region to region.
I have a ASUS Zephyrus G14 which both integrated and 3060 GPU. When Steam is running, the discrete GPU is active. I'd wager that most laptops that offer dual GPUs are designed to offer both a battery-friendly graphics option and a decent (or semi-decent) performance option which kicks in when Steam is running front and centre. The alternative is trying to track every damn .exe of every game now and forever.That's further complicated by some number of machines with multiple RTX (2-10+) GPUs in one machine with Steam installed.
But also some number of "casual" machines where Steam is installed on a PC that never plays AAA games and hence in some cases just a basic low end PC is more than enough. Heck, this forum is a pretty small community now and even here there are multiple users with multiple RTX cards in their machine.
That sounds dangerously like PCMR talk, man. Like people who don't play AAA games aren't gamers.
I have dual-3080s in my desktop gaming PC, just saying.
This generation is probably leaning towards relatively fewer PCs being at or above console standard due to the crazy GPU situation the last couple of years.
GPUs are still stupidly expensive. Cheapest RTX card I've seen at Scan or Overclockers (in the UK) is £299 for a 6 GB version. That's more than it was 3 years ago. A 3060 Ti costs ~ £100 more than an entire flippin' XSX / PS5.
PS5 and XSX represent insanely good value at their RRP. Even Series S does. I can imagine there are a fair few folks who are simply pushing back PC upgrades and maybe even buying a console in the mean time.
In my estimation there are far fewer high end raytracing cards around in PC land because of the supply and cost issues of the last few years.
The figures and numbers dont match your thoughts.
there's a lot of pent up demand for more affordable, high performance cards. Performance per dollar is worse than before next (current) gen consoles launched and I think that has to hurt.*
That sounds dangerously like PCMR talk, man. Like people who don't play AAA games aren't gamers.
I have dual-3080s in my desktop gaming PC, just saying.
You misinterpret what I'm saying. I mention that to illustrate that the percentage of RTX GPUS doesn't match the percentage of RTX enabled PCs and thus gamers. IE - if 20% of GPUs in the Steam survey are RTX class GPUs, that doesn't mean that 20% of PCs (or PC gamers) in the Steam Survey are equipped with RTX class GPUs. Of course, everyone could argue until they are blue in the face as to whether that represents a small percentage differential or a large one.
That contrasts with say console gaming where 1 console = 1 console capable of playing games. Of course someone could own multiple consoles, but that goes for RTX enabled PC's as well (including ones with 2-10+ RTX class GPUs).
The other bit is just to illustrate that not all PC gamers are interested in AAA gaming and hence don't need a more performant PC nor would they feel any incentive to upgrade their PC even if it's a potato relative to what's hot and new in the PC gaming space and isn't capable of playing PS4/XBO generation much less PS5/XBS generation games.
Regards,
SB
For the latter bit, not all console gamers are intrested in AAA gaming either, heck theres 100m PS4 users out there with a large part of those playing fifa etc.
LoL, you mention PS4 owners not playing AAA games and then mention a large part of those owners playing a AAA game.
Here's the thing with that, regardless of whether a PS4 owner plays a AAA game or not, they will have a machine capable of playing AAA games. If a PC gamer doesn't play AAA games, there's a good chance that they may have a PC that isn't capable of running AAA games.
So during the PS4 generation all 100+ million PS4 consoles were capable of playing AAA games of that generation. OTOH - all 100+ million Steam machines weren't capable of playing AAA games of that generation.
The Steam numbers aren't misleading, they are precise. It's tracks individual GPUs, CPUs, RAM configuration and so on.
I think that's the point.
I didn't see anybody may that point but I could have missed it. I responded to eastmen's post about crappy PCs being a big tail and was just pointing out that the ownership of non-laptop 3080 and 3090 GPUs (all variations) was very small, meaning that the vast majority of the reported hardware is a tail from the top-end of PC gaming GPUs.When I say “misleading”, I mean when people try to just eyeball the data and make quick assumptions about the demographics. People assumed that RTX ownership is relatively small because the percentages for any one sku is tiny (most are between 1-2%).