Ratchet & Clank technical analysis *spawn

As the video is only 1080p and with high compression, and the fact that there are texture loading bugs on the PC version to boot, not sure why you're so confident. I've posted a comment asking for clarification.

Well, we don't have confirmation this is a texture loading bug. All we know is that textures are acting similarly to Spiderman Miles Morales, which have yet to be "fixed". I also know what highest quality textures are supposed to look like because I've played Rift Apart on PS5. The textures you mentioned above are similar to Alex's textures in his video. I posted this in the ratchet thread -

Screenshot (280).jpg
 
No it hasn't. All of your confusion will be cleared once you realize SSD speed /= I/O speed.

Thanks goodness you're here to educate me. To be clear though, my quote around IO speed was purely related to the drive throughput. I of course understand that there are other factors involved in getting that data in a useable form off the disk and into GPU memory. Primary amongst those would be decompression speed. However I do think that all evidence thus far shows decompression speed to not be the bottleneck.

I'm curious as to where you think the main bottleneck is though?

He got those times with an HDD because he could compensate with higher amount of ram, and ram is faster than even the fastest SSD.

Certainly true in the case of the HDD, but in the overall context of the discussion it is clear that raw data throughput speed from the drive isn't the bottleneck here as 3.5GB/s drives perform identically to 10GB/s drives.

So the bottleneck is somewhere else.
 
Well, we don't have confirmation this is a texture loading bug. All we know is that textures are acting similarly to Spiderman Miles Morales, which have yet to be "fixed". I also know what highest quality textures are supposed to look like because I've played Rift Apart on PS5. The textures you mentioned above are similar to Alex's textures in his video. I posted this in the ratchet thread -

View attachment 9285

I'm curious how you put textures that never load regardless of how long you stop and stare at them down to lower IO performance as opposed to a bug. What's your train of logic there?
 
Thanks goodness you're here to educate me. To be clear though, my quote around IO speed was purely related to the drive throughput.

Thanks for the clarification. Sorry for any trouble it may have caused you to be more specific.... in a technical thread.

I of course understand that there are other factors involved in getting that data in a useable form off the disk and into GPU memory. Primary amongst those would be decompression speed. However I do think that all evidence thus far shows decompression speed to not be the bottleneck.

What evidence are you referring to? I would love to see it.

I'm curious as to where you think the main bottleneck is though?

Decompression is my guess for now, considering the widespread inability to load in high res textures despite player preset selections.
 
PS5's data transfer advantage is effectively larger than DF's claims because Alex's setup constantly fails to load in the highest mip levels. The idea that Nixxes overlooked this is so farfetched, considering the game's nature as data management heavy during signature moments. We have seen this texture behavior before with Spiderman Miles Morales. It's starting to look like there's only so much Nixxes can do to try and keep up with ps5 loading performance.

View attachment 9281

View attachment 9282

Let's wait at least 2 patches in before making any solidified claims or answers. I'm pretty sure most of these issues will be resolved on PC.
 

It's worth remembering that the one way bandwidth for PCIe2 16x is only 8GB/s which is barely faster than some Gen4 NMVe drives and so could be very easily saturated by heavy IO plus normal game data going back and forth.

PCIe3 doubles that to 16GB/s though which gives far more headroom. I would expect that if PCI3 is still a bottleneck, and thus 100% saturated (to enable PCIe4 to give a further speed boost) that we'd be seeing a bigger speedup from the doubling of bandwidth between 2.0 and 3.0. I suspect 4.0 won't give much if any speedup over 3.0. That seems born out by the 4060/Ti results here vs the 3060Ti. The 3060Ti being full speed PCIe4.0 and the 4060's being only half speed. Yet neither seems impacted by that.

What evidence are you referring to? I would love to see it.

Decompression is my guess for now, considering the widespread inability to load in high res textures despite player preset selections.

Evidence that points towards decompression likely not being a bottleneck would be:

  • We have other benchmarks showing vastly higher decompression throughput than R&C is using
  • We see no obvious difference in load speeds between different GPU's of wildly different capabilities (albeit this is on different test systems so a proper test may show different results).
  • We can see GPU usage is well below 100% on the 4090 during the portal transitions
None of that is conclusive, but I'm sure we will get conclusive evidence one way or the other soon enough as it's as simple as benchmarking the transition times on the same system with just the GPU swapped out.

Hell, anyone here with the game could do it right now by simply underclocking their GPU.
 
Last edited:
If he was out of VRAM there wouldn't be any increase in performance increasing PCIEX speed.
How do you figure that? If the game runs out of VRAM then it needs to shuffle data back between the GPU and system RAM that it otherwise wouldn't need to. Running on 3.0 means it can write that data back faster than on 2.0.
 
How do you figure that? If the game runs out of VRAM then it needs to shuffle data back between the GPU and system RAM that it otherwise wouldn't need to. Running on 3.0 means it can write that data back faster than on 2.0.

Horizon: Zero Dawn had the same issue with PCIEX bandwidth and that wasn't running out of VRAM.
 
Has anyone determined if Resizable BAR or Smart Access Memory have any benefits or negative affects (on or off) that could potentially influence Ratchet & Clank: Rift Apart PC SSD/IO performance? Just curious...
 
Horizon: Zero Dawn had the same issue with PCIEX bandwidth and that wasn't running out of VRAM.

Well, kinda - it is extremely vram hungry. It 'runs out' by just not displaying the higher-res textures. 6GB cards would have lower quality textures than the PS4 version at anything above 1080p, and even 8GB cards would have issues with some textures at 4k. So it could be attempting to be swapping textures in and out on lower end cards and this streaming system was never fully fixed, or it never bothers, dunno.
 
Dude's just running out of VRAM because he's trying to use RT on an 8GB card. DirectStorage reduces the PCIe load by keeping the data compressed while it transfers.

Bear in mind R&C only use GPU decompression on a subset of textures, which is why it's only enabled at the High and above texture settings. The large textures are compressed using Gdeflate, the others are just .lzh and are sent over the bus uncompressed. Basically on texture settings at Medium and below, these large gdeflate textures are just ignored.

Even on Medium textures though, with High settings otherwise on my 12GB 3060 at 4K DLSS performance, I can see actual in-use VRAM top out at just over 9GB. I would imagine with Very High Textures and especially RT, it's swapping quite a lot back and forth on an 8GB card.
 
Horizon: Zero Dawn had the same issue with PCIEX bandwidth and that wasn't running out of VRAM.
I don't follow your line of thought but that's okay. The reason I'm sure he's running out of VRAM is because my own 11GB 2080 Ti cannot run the game with RT. My PCIe load is completely minor without RT (10-20%) but it then jumps straight to 50%+ as soon as I enable it.

Why should we assume that his issues are caused by DirectStorage anyway? It's not uncompressed data that's being transferred over bus but compressed data.
 

Noticed the same on my 3060. More stuttering even though committed VRAM usage is just above 10GB with very high textures. Just a consistently higher GPU usage overall. 4k DLSS performance, high settings, medium textures - 60fps 90%+ of the time, no stutters, extremely consistent, no traversal hitches or anything like that when on worlds, - even from an HDD. Drops are only from intensive alpha-heavy combat scenes.

Very High textures, SSD or not - I would need to drop down to DLSS Ultra Performance to have enough GPU headroom, and even then there are small stutters present. So I gather a pain point at the moment is async GPU decompression perhaps.
 
I'm curious how you put textures that never load regardless of how long you stop and stare at them down to lower IO performance as opposed to a bug. What's your train of logic there?

I've flipped back and forth between texture settings on my 3060, and at times doing so the Very High would get stuck at muddy textures, and reducing the setting to medium would fix it. It's clearly fubar.
 
Back
Top