You mean like here?
This idea that Sony wasn’t going around making those very claims you suggest they didn’t is verifiably false. That’s data.
2 Likes
Lempel doesn’t say what you seem to think he says. He just says R&C as is (i.e. at its existing fidelity) could not be done on PS4, which is true. He does not say anything about it requiring the full SSD speed to max out the PS5’s I/O, which is what the framing you guys are putting forward relies on since you are trying to leverage the DF vid conclusions in your argument.
Also, please try to delineate between Sony talking up the SSD in general use cases for the platform’s future and claims about the full bandwidth being required for R&C to work. Those are two entirely different claims.
You are kind of right and I tried to touch on this by saying “for the most part” we haven’t seen the gap that’s shown on paper. I didn’t mean to say there was no gap, just that it’s not as large or as common as some thought it would be…yet.
Looking at a handful of games with proper new gen updates (so no enhanced BC for example), most are either exactly the same or much closer than the millions of pixels you claim.
F1 2020 targets the same 4K60 or 1440p120 on both platforms.
Plague Tale is the same 1440p60 on both.
Doom in RT mode targets 1800p on both but runs slightly higher on the SX. One scene for example runs at 1680p on the PS5 and 1800p on the SX, so that’s a gap of 743,520 pixels.
120Hz mode in Doom does show a bigger gap at 1800p on the SX while the PS5 tops out at 1584p, so a gap of about 1,299,456 pixels.
Balanced mode holds pretty close to the 4K60 target on both systems.
The interesting thing about Doom is that it’s one of the very few games where load times are twice as fast on the PS5. Doom also uses VRS on the SX, so the resolution gaps are surely helped by using this feature besides the raw performance gap. So the game does a great job highlighting the strengths of both systems.
Chivalry 2 Targets and largely holds 4K60 in quality mode on both.
Both are 1080p in frame rate mode.
Last we have Metro Exodus where the SX will see higher averages, an example being 1600p on the SX while the PS5 is 1440p, so a difference of 864,000 pixels.
The catch with Metro is that is runs slightly better on the PS5, so there could be a case made that the dynamic resolution found on the SX may benefit from being a bit more aggressive.
So yes, we sometimes do see a gap, but it’s certainly not “almost every 3rd party game” nor is the gap millions of pixels wide. Please keep in mind that I wasn’t trying to argue that there isn’t any gap but instead that it’s too early to judge the GPU delta between the two this early in the generation in any definitive way, just like the SSD delta between the two. I think it’s fair to gauge the progress of utilization on a game by game basis and how the averages evolve, but I don’t think it’s accurate to make any definitive conclusions, especially less than a year into the generation. That’s the point I’ve been trying to make.
If I’m giving off the impression that I’m dealing in absolute conclusions, that’s my bad because I’ve been actually posting for the opposite. I’m reading the claims of the SSD delta as mostly marketing as absolute conclusions because that is a conclusion based on limited data. In my view, it’s more accurate to form conclusions of how the software is utilizing the hardware instead of forming conclusions on the hardware itself based on the software, at least at this point in the generation. In no way am I trying to tell people they need to wait until the end of the generation to form opinions but drive by posts focusing on marketing and making definitive statements of any deltas is premature at this point IMO.
I also want to believe no one here thinks the SSD utilization won’t improve but then I’m not sure why they choose to argue with me when that’s the point I’ve been making. Instead it’s talk of marketing and the delta being BS, when I don’t understand how anyone can form that opinion if they knew things will change in time. Makes more sense to wait until studios have more time to properly take advantage of these systems to make those bold claims IMO.
Regarding using the cheaper/slower SSD right now, I’m actually really curious to see how it performs across a number of games. Hopefully someone sees this DF video and expands on the testing just out of curiosity.
The GPU comparison was more so to help convey my point instead of actually proving anything. You’re right that I did limit it to raw performance because it is true that the existing delta will only increase when the XSX/S GPU features are being properly used and raw performance was the easiest way to compare results. I just wanted to drive the point home that it’s too early at this point of the generation to draw any final conclusions. Both systems still have headroom to explore, especially when almost every game releasing now started before final hardware was even available. I think it would be great for us to have healthy conversations and talk about how games are using the hardware over the course of this generation. Drive by posts do nothing to promote this and instead just look bad.
On the whole marketing topic, to be fair with marketing using technology in their interviews, blog posts, PR statements, etc. it will inevitably be part of the conversation. However I don’t think it’s beneficial to put as much weight behind marketing statements because we all should know there are degrees of spin, stretching, exaggerations, or however way you care to classify it in marketing. The whole point of marketing is to spin your product in the most positive light possible. Look at how the focus has shifted to marketing in this thread and how it has added nothing to the conversation.
Focusing this much on marketing is a slippery slope, especially when every company is guilty of embellishing their product for the sake of marketing. There have been Xbox blog posts about the SSD allowing larger, more dynamic worlds, but that hasn’t played out yet either. Are we just not calling them out because they haven’t named any specific games like Sony? Ultimately, I hope people can disconnect the conversations about technology from the marketing because at the end of the day, marketing is there to sell something to us, not to really inform users on technical capabilities. I very much prefer conversations like we’re having right now instead of drive by posts about marketing being BS because yeah no shit, it almost always is a spin to some extent. Why focus more on marketing people instead of what the actual developers say in a thread about game technology.
Sorry for the wordy response, obviously I enjoy talking tech, and I’m hoping that I’m able to explain my pov well enough here.
So are you saying that R&C can be done on the PS4 as is? I’m really trying to understand you here.
1 Like
Haven’t gone through your whole post yet, but speaking of seeing the differences, wasn’t Control a good test? The photo mode with uncapped framerate.
It’s been a while since I’ve seen the DF video, but from what I remember it was about a 30 percent increase in frames on the series x. Which fits the expected difference on paper.
4 Likes
Yup, that is a super special case though. Was really interesting/clever for DF to pick up on that as a test scenario. Mike’s broader point was that we still aren’t positioned to really see the gains XSX has in store over PS5 and both machines will be pushed WAY beyond anything we are seeing now. SSD’s are a huge deal for game design and visuals and R&C is not really showcasing what can be done there (nothing is yet).
2 Likes
Says the guy who I would guess doesn’t do any actual meaningful data analysis.
In the real world there are many times when important decisions need to be made based on a limited data set. In which case you simply use the context and experience to interpret the data as accurately as possible.
Sometimes you can’t just wait and see.
In this case it’s a binary outcome either the slower SSD can run the game as expected or it can’t. No need for more data there. It can. Conclusion the claimed super fast SSD speeds are not needed to run R&C. All the marketing done for it in the lead up to release suggested they were.
So yea, big oof lol I guess. What ever that is supposed to imply.
1 Like
Yeah I have little doubt that we haven’t seen even close to the full potential for both machines.
That being said, I don’t think the increased ssd speeds of the ps5 will ever be a major factor this generation. The jump from hdd speeds to nvme is much more important than an nvme and a faster nvme.
Though I do expect near instant load times for any fp ps5 games, but that’s not a big factor when the alternate is just a few seconds.
2 Likes
Yeah the gap ranged from 3% and 11% up to what they described as outlier at 36%. I wasn’t making a point about there being no gap though. I’ve been trying to say it’s too early to draw any final conclusions on the hardware this early in the generation. Basically we don’t know enough about the hardware limits to say what is and isn’t marketing BS.
Thing is we didn’t need this DF video to tell us that. Insomniac already said R&C can run on the Series X. I think the disconnect is when they talked about the game using the SSD, some took it as a swipe against Xbox when the Xbox SSD is also super fast.
No one is claiming the gap is going to make a major difference in the end. As the generation goes on, devs will continue to better utilize the speed of the PS5 SSD while also leveraging features on Xbox like direct storage and SFS to help make up any potential differences. It’s really two different approaches to the same problem. Somehow the conversation has twisted into Sony’s messaging being marketing BS (when all marketing is to an extent) and this DF somehow proving that when we already knew how this test would go before the game even launched months ago. The raw speeds of both SSDs haven’t even been tapped into yet, let alone any of the unique strengths in IO throughput both systems have. Any definitive claims about raw performance now are premature.
100%. When you move from near-minute and minute+ loading times to few seconds, the difference between a few seconds and a couple of seconds less seems pretty trivial.
When devs are fully utilising the architectures of the two machines, uncompressed, the time to fill the full 16GB of RAM with data from the SSD will be:
- 16 (GB) / 2.4 (GB/s transfer) = 6.6 seconds
- 16 (GB) / 5.5 (GB/s transfer) = 2.9 seconds
When you consider not all that ram is available for games, and that textures will be compressed/decompressed, I expect we will be looking at near-instant loads on PS5, and three seconds at most on SX.
These figures are irrelevant as there are lot parameters involved in actual games for ex: SFS, BCPACK compression, RAM bandwidth etc. There’s a reason why R&C Rift Apart can run on much slower SSD. I believe even the last game of this generation will work like that.
That’s just a basic calculation of how long it takes to fill the memory from the SSD.
I said in the post that there are other factors (decompression, the fact that not all 16GB is being used) that would make it faster.
There are also factors (other processes being hidden behind loading screens) that could add time back on.
But I believe the conclusion is right that once developers are working directly to the new consoles and making full use of the API, we will see near-instant loading on PS5 (fade to black and back in) and a few seconds at most on Series – we already saw this with RE8.
I think it’s logical to infer that Ratchet and Clank can run on a much slower SSD because it is not actually using real-time loading as was implied in the marketing. They were busted in a clip on twitter near release where a bug led Ratched to get sucked from a pocket level into the main level (or vice versa) and one was just floating above the other out of sight – it’s trickery.
In the same way, developers were claiming things done in Miles Morales and Horizon Forbidden West couldn’t possibly be leveraged without the SSD … then those games got announced for PS4. Awkward.
When DF did their tests on file transfers, though, the slower SSD was noticeably slower as you would expect. It’s just that Ratchet, like almost all launch window games, is not really tapping into the full capability.
1 Like
RE8 wasn’t using XVA. But in reality you don’t need to transfer 16GB of data all at once using various optimisations like SFS you only need to load a quarter of all the data at once.
This doesn’t account for the memory bandwidth differences.
I had to look pretty hard to even find a reference to this, do you mean this guy? 'cause he isn’t a developer from insomniac he’s an indie dev giving an opinion.
1 Like
We still don’t have solid confirmation about memory footprints allowed to be used by the games on the PlayStation side. The best we have is a few off the cuff posts from a developer saying 12 GBs for PS5. If that is indeed the actual limit, then having an extra 1.5 GBs usable on Series X (13.5 GBs total) that can be used as extra memory buffer may negate most IO differences. We also have a situation where quoted specs are using different rates, one system provided sustainable rates while the other provided peak rates.
RE8 updated last minute to Ps5 IO apis but didn’t for direct storage on SX that’s why the difference was among the biggest ones. All things equal I don’t think there’s a single game where the difference isn’t within <1s
If a developer was using SFS, surely they would use the technique to make more of the memory – raise LOD distance for example – rather than leaving part of it empty as a result of the savings?
And a game won’t need to fill 16GB of RAM ever, as the whole pool isn’t accessible for gaming. The calculation was just supposed to be a brute example of how long it would take to fill the full memory from the SSD, to demonstrate the real-term difference between the Series and PS5 I/O capabilities - i.e. that it’s not really very significant.
Perhaps someone with more expertise could feed in, but my limited understanding of memory bandwidth was it related to how fast the processor can access the memory. It’s the I/O that determines how fast the memory can be filled from storage. Could be wrong, but I don’t think that should impact load speeds.
Perhaps my memory is playing tricks on me, but I think some of your stats are wrong.
Doom for example from what I recall the smallest difference is sub 1600p versus 1800p on SX (the max for both) but they mention Ps5 drops more often and more severely.
RT mode can also see Ps5 dropping to 50-70% of the target res while SX stayed at it IIRC.
They were actually surprised because in this game the difference was much bigger than what the specs suggest and assumed it was due VRS.
Same for metro, difference was only small when doing nothing in small areas, but in battles and more open areas SX would pull ahead. You mentioned that Ps5 ran slightly better but I recall that only being for one scene, the most demanding one of the game. They showed some scenes where SX ran better despite the higher resolution, but overall was fairly stable on both.
F1 from what I recall is already 4k60 on 1x (reconstructed), so it makes sense neither console would have trouble running it native 4k with increased settings, so perhaps not a good datapoint…
5 Likes
SFS calculates for all the LOD distances that requires to be loaded for a particular scene. Your reasoning assumes that games will always grow in terms of bandwidth requirements but if that’s the case then even PS5’s ultra fast SSD bandwidth will fall short one day. Developers will always want to leave some overhead in the bandwidth and optimise for minimum loading at a time (since they have DirectStorage).
Yes that’s also subjected to how fast the memory can be written.
I do assume that developers, or those working on AAA games that are made to push PCs much more powerful than either console, is going to use every spec of the machine to its maximum.
Why wouldn’t they? They make their top settings for graphics cards with up to 24GB VRAM, in PCs with 32GB system memory.
If SFS frees up memory, then they are going to take that saving and use it to push something else further, like pushing out the distance at which they lower LODs or something else. They aren’t going to just leave it empty.
I suppose the “problem”, if you want to call it that, arises when you have one side going much harder with their marketing than the other. The conversation wasn’t “twisted” about Sony’s marketing approach, but rather they chose to hype up their SSD technology to the moon.
I don’t think it’s fair to directly compare the marketing regarding the SSD of Xbox and Sony. Xbox talked about their velocity architecture yes, a couple videos etc, but they didn’t make even close to as big of a deal as Sony did with their “blazing fast” SSD speeds. It reached the point where people were convinced that the faster SSD would bridge any processing gap between the systems (and yes you still have people who think this.)
Basically, if you talk a big game then you should be prepared to back it up. Whether Insomniac directly said if R&C could only be done with the PS5s SSD or not, they did heavily imply it and let people run with that implication. Of course that’s more on Sony, but still the same thing really. Similar to how Sony made it seem that the Unreal Engine 5 demo was only possible due to their SSD speeds.
I think it’s completely fair to question Sony’s marketing around the SSD, at least for now.
6 Likes