1. I don't know exactly what you don't agree about the SSD throughput and what I said, but you got some things/facts wrong (intentionally or unintentionally).
I would like for you to quote those bits, given you're literally saying the same things I said in my post above (above the one you read).
The BCPack compression is equivalent or better to what Kraken does, because it's specialized for texture compression, while Kraken isn't.
Based on this, it's wrong to estimate compression is worse because throughput from the SSD is higher. Stick to that layer of the pipeline.
2. PS5's SSD is a custom PCIe gen 4.0 nvme with a 12-channel memory controller. Xbox and PC's (even latest pro consumer ones that have higher raw speed) have 4 channels only and they don't have the decompressor chip. Only server side ones are matching it or surpassing it and it will take some time until PC's get comsumer SSDs as fast/good.
12 R/W channels to NAND ICs, yes, but that's not the only way to get proper performance on the SSD. You could at the release time of the consoles and now get 6/8 channel controllers today that output higher sequential performance. The NVMe custom controller is mostly here to add finer grain priorities over standard NVMe (3 weighted priorities against 6 with their approach).
https://nvmexpress.org/wp-content/uploads/NVM-Express-1_4-2019.06.10-Ratified.pdf (see pages 91-93).
The decompressor chip isn't on the controller though, and general purpose hardware can definitely get the same results with BCn data using XPU cores, not mentionning the extra pools of DRAM that massively reduce the overhead. It's been the approach for a very long time, similarly on consoles for texture decompression.
developer.nvidia.com
The new units designed to work on SSD data on the fly, are new though.
The hardware on PC isn't constrained by price, thermals of memory capacity, so there is more than enough spare performance to enable this.
3. The part about Kraken being worse is not true. Here's a quote: "According to Oodle, the previous 8-9GB/s IO bandwidth figure was provided by Sony by multiplying the 5.5GB/s peak bandwidth of the PS5 SSD for the standard compression ratio of 1.45/1 or 1.64/1. However, when factoring in both Oodle Kraken and Oodle Texture, the developers of the technology have seen an impressive compression ratio of 3.16/1 for a texture set in a recent game; this would translate into an IO bandwidth peak of 17.38GB/s." And yes the 21-22GB/s is theoretical max peak, that's taken from a different source.
Oodle isn't Kraken. For textures, BCPack is superior, since it is directly reusing the same BCn texture data as Oodle, simply rearranged differently to fit their approach. It's not unusual, Microsoft already claimed higher compression ratios:
https://www.anandtech.com/show/15848/storage-matters-xbox-ps5-new-era-of-gaming/2
At least with 2.4GB/s going above 6GB/s, and presented during their HC32 presentation as well.
Kraken + Oodle, that's perhaps equivalent, since its own creator described it as a 3.16:1 ratio here:
https://cbloomrants.blogspot.com/2020/09/how-oodle-kraken-and-oodle-texture.html
and they themselves don't take into account their big ratio outliers listed here
http://www.radgametools.com/oodlekraken.htm
let's see how long it takes for the 1st game using this new tech that PS5 games have (again see Ratchet & Clank Rift Apart, coming next month, June 2021) to come to Xbox and PC.
If it comes this year then yes, PS5's SSD+IO+Kraken is nothing special, if it takes years to come, then this is the benchmark, that it took that long for the rest of them to catch up to this PS5 feature.
I don't see the point, PS games haven't come to PC for a dozen years at least, this doesn't tie to any hw limitations. As for the UE engine demo, everybody at Epic said they would be identical experiences regardless of the platform..
Sony purposedly imposes 6 months to 1 year exclusives to most of their studios, and they only recently started the PC releases because they see most of the japanese customers and worldwide boost in PC sales, and they also revealed some of their partnerships were conditioned on them delivering PC versions of their games (at least on the EGS).
Funnily enough, Steam and Sony have ported their controller support to PC, so you get full features in games like Metro Exodus. Pretty cool.
There is no debate here, time will determine the nature of a win or a loss/tie on this aspect only, I'm not talking about other things.
The fact that you're looking for a win/loss here is worrying, I think. We're having a technical discussion, not looking to say one or the other is in the wrong.
99% of the posts I've seen online on new console hw makes me sick to my stomach as a hardware engineer. Some people stumbled on the Cerny talk and watched a very light description of their console. When it's the first time many people looked at a technical presentation, it's bound to cause issues.
I've seen people saying the PS5 is better because it has a DMA controller and the other doesn't (real quote). Cerny was also misleading about features we know now, all consoles have (audio, decomp, coherency) and the usual lack of transparency in Sony's hw reinforces this. I think his bit on GPU performance was appaling.
In the end, all it's used for is saying "we're the only ones who can do such things" when we all know it's not the case. So people please, stop acting like that.
5. Diminishing the importance of compression and how the same game has a much lower disk size on PS5 vs other mediums while highlighting that Xbox has a bigger SSD size, as more important is really low. If we are to be fair, one has better compression and one has a bigger SSD (until now). And it's not one game only, but quite a few and probably more to come. I have nothing more to say about this because for me it's obvious.
Nobody diminished the importance of compression, but you do get my point though ? Saying there's more storage space because compression shows some games are slightly smaller on a smaller SSD is not really true. Multiple games show opposite results and the overall picture is not really clear, especially looking at recent titles.
I don't think the kraken+oodle addition over bcpack grants any groundbreaking gains "over competition" there.
We've been quoting so many figures for throughput, it's almost random.
- Sony has a 5,5GB/s part, with max theoretical ratios of 4:1 in Kraken, which leads them to quote a unit test result with a dataset that scales well, offering 22GB/s (100% compressed data, 4:1 ideal ratios). It's not a realistic figure because it's 100% SSD utilization, only textures, and at an outlier ratio performance, as you can see on the RAD Kraken site.
- MS guarantees 2GB/s throughput per game, announced average rates at 2,4GB/s, which turns into 4.8GB/s and >6GB/s with >3:1 compression ratios. Add on top the 2.5x of SFS that fetches partial texture data and reduces the overhead by allowing to fit as much more into VRAM, and I don't see any particular disadvantages over texture data. The same isn't true for uncompressed assets, so there you go.
The best bit is both companies say average usage is lower, like the 8-9GB/s for the PS5, but they didn't put forward a figure or showcase real time access to the drive. Performance is bound to be worse because in game access isn't fully sequential.