That's the issue, Kraken is for general game data, BCPack is specifically for texture data. According to devs with over 20 years of experience, Kraken cannot compress texture data as well as BCPack and we still haven't gotten clear figures.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
- Status
That's the issue, Kraken is for general game data, BCPack is specifically for texture data. According to devs with over 20 years of experience, Kraken cannot compress texture data as well as BCPack and we still haven't gotten clear figures.
MasterCornholio
Member
So your saying that Sony could save on resources by having good custom I/O technology?
That's a pretty interesting theory. However I believe these consoles are going to have advantages and disadvantages when compared to each other.
Aceofspades
Banned
We're talking about texture compression.
sircaw
Banned
Forgive me for not being very technical, i am kinda simple minded in pc talk.
But those numbers don't look good at all on the xbox compared to those numbers for the ps5 which look Gargantuan.
Am i missing something because that looks very dire to me.
I understand it is not oranges to oranges but that makes it look likes its oranges to watermelons.
Last edited:
ethomaz
Banned
MS Direct3D documentation.
How do you believe we are talking about 50% or more decompression in less than half a second?
Last edited:
Bo_Hazem
Banned
“PlayStation chief Jim Ryan has stressed that the PS5 launch should be simultaneous around the globe, according to people in the company’s supply chain,” the website reports.
PS5's Launch Aiming to Be Simultaneous Worldwide
Jim Ryan has apparently requested itwww.pushsquare.com
I hope this time there are no more $1000-2000 PS5's for early resellers around other parts of the globe.
MasterCornholio
Member
Sony already does that.
ethomaz
Banned
False.
It can't compress data 50% better than Kraken.
It is Kraken 30-40% vs BCPack 50%.
It is more like 10-20% better at best.
Last edited:
ethomaz
Banned
No lossless algorithms will ever reach the compression level of lossy algorithms.
splattered
Member
I don't know why people are still arguing about SSD and stuff in here.. just come to terms with it.
PS5 is 60% performance and 40% gimmick.
This is their Wii moment, let's see if it pays off
PS5 is 60% performance and 40% gimmick.
This is their Wii moment, let's see if it pays off
The Mad Draklor
Member
What do you mean that's the issue? Not all data are textures. That means BCPack can't decompress any non-texture assets such as 3D objects/models and audio (of which Sony even has a specific solution for that via DMA'ing audio data to a specialized CU). And we also have to take into account that BCPack is not a lossless solution. The more you increase the compression ratio, the more lossy the quality. Kraken is a lossless solution and developers can improve the transfer rates above 9 Gb/s if they want to optimize the compression/decompression algorithms.
MisterXDTV
Member
Can't we just stick to the RAW speeds and be done with it?
5.5 vs 2.4 GB/s = PS5 SSD is 129% faster than Series X
Period...
5.5 vs 2.4 GB/s = PS5 SSD is 129% faster than Series X
Period...
Bo_Hazem
Banned
PS5 is better. Japanese cars are the best (especially Nissan Patrol). Apple is overrated. Audio is the next big thing. Talking to them is like talking to religious persons, while it is 100% clear that they are wrong, delusional and making things up from their imagination, it is not possible to talk sense to them as they repeat their faulty logical circle of "god is real because book said it, book is real because god wrote it!"
Had to settle down to a laugh emoji
Last edited:
Imtjnotu
Member
M-V2
Member
What's I'm not consistent in??XSX has ECC function and features on its GDDR6 memory; it may not serve the exact same purpose as the GPU cache scrubbers but conceptually seems like it would line up with what the coherency engines do (which I assume are there for retaining data integrity and keeping data synced), just at a different level of the memory hierarchy stack.
Outside of the GPU cache scrubbers I'd be surprised if XSX doesn't have equivalents for things like the DMA controller, so I agree on that point. What we will see are differing implementations of some of these things at different hierarchies of the system design, for what best serves the particular system. And that's pretty exciting to think about.
So the paper specs matter again?
Have MS confirmed the 4.8 figure is only for BCPack, or is this an assumption on your end? Because I don't recall them specifying what compression tools (or combinations thereof) are providing some of those numbers.
llien
Member
It was obvious from the beginning it would be misinterpreted into obvlivion and so it is.
It's time for people to get over 36CU debacle.
I don't recall it that way, but it doesn't matter.
I doubt textures, which are already compressed will allow that.
DaGwaphics
Member
I never saw that anywhere either. Just 4.8 being the typical result with compression.
The compression tech is important in next-gen, primarily to help in reducing game install sizes. I expect game sizes to inch up a bit further even without the data duplication, that quickly becomes a problem when you only have a TB of storage to work with. An even bigger issue if storage upgrades cost a kidney.
llien
Member
Build a PC with 2070s and 5.5GB/s SSD.
Then another one, with 2080S and 2.4GB/s SSD.
Then argue the former is faster.
Neo Blaster
Member
Don't worry, he'll probably be back in 7 years for the next next gen when everyone will have forgotten the 2020 insider debacle.
RaySoft
Member
It's hard to know what people are actually talking about when they just say one compression tech is better then the other. There are different sides to it. I think it's well established that the decompression block in PS5 is faster than the XSX equivalent, but BCPack's compression ratio might be higher than Kracken, that doesn't mean their decomressor is faster, only that filesizes could be smaller. You have to look at all the sides when comparing tech, and not just be blinded by one aspect of it. We have to be precise when discussing these things to prevent any misinterpretations.Lmao I don’t want to get too much into the fanboy nonesense. But remember when PS5 specs where revealed by Sony, every Xbox fanboy downplayed the SSD claiming it would have no real benefit other than slightly faster load times, they called Cerny a liar for claiming the SSD could do more. I remember PS fanboys even jumped on the bandwagon and started attacking Cerny. Time went on and several developers along with tech geeks such as NX Gamer, Coreteks and others came out and revealed how the PS5 SSD is a game changer and that it will allow for games an experiences which are just not possible on the Series X or the PC. So now that people or should I say Xbox fanboys are actually learning and understanding that the PS5 SSD will have an impact on gameplay, experience and performance, they’re just doing mental gymnastics to explain how the Series X with it’s magical compression is equivalent or better than the PS5’s SSD and compression. Sony fanboys did the same thing when it came to the difference in Teraflops but that’s another subject.
EDIT: I just watched that Moore’sLawIsDead clip and I laughed when Tom completely shut down the Series X compression being better than PS5
Thirty7ven
Banned
Fixed
If there are no IO bottlenecks and ram the same, it becomes an interesting decision to make if games are made with them in mind. I know which one the devs would choose.
Last edited:
SonGoku
Member
The 6GB/s is a theoretical max output from XSX hardware decompressing block just like 22GB/s is a theoretical max throughput of PS5s hardware decompression unit
Yes... hence 2.4GB/s->4.8GB/s
There might be edge cases were it goes over 4.8GB/s just like PS5 might go over 9GB/s but these are the typical figures we are working working with no theoretical peaks
Last edited:
Neo Blaster
Member
Funny how Xbox fanboys are struggling so hard to close the SSD gap using compression. Just accept it, SX got the flops/CUs, PS5 got the SSD/I/O.
Last edited:
FeiRR
Banned
It's double funny because we're talking about a compression technique. Microsoft can have a patent for implementation but they cannot patent a mathematical method used to achieve compression. If they can do it, other people can do the same, in a different way but with the same formulas.
ZywyPL
Banned
That's actually the vibe I get from the PSSD5 force... Let's see:
- but devs wanted an SSD and that's what Sony gave them! - so did MS.
- but PS5 has dedicated decompression chip! - so is XBX.
- but PS5 has dedicated audio chip! - so is XBX.
I could go on and on with examples but I don't wanna waste my time on pathetic kids who fight online over some plastics boxes, it's not my business to change their attitude, nor I can actually achieve that.
All in all, both consoles are extremely close to each other, very similar (they are both build on the same AMD tech after all), and the biggest difference so far is not within the consoles themselves but their presentations/reveals, where Sony focused on the hardware side (SSD, TE) and MS on the software side (API features). I personally don't expect much, if any differences in multi-platform games.
But again, the PSSD5 fanboys are starting to make me wish the consoles still used HDDs, I'd gladly take long loading screens instead of hundreds insecure threads and thousands of delusional posts, and we are just a month after the reveal...
LeVictorien
Banned
When talking about SSDs speed raw numbers are important... when talking about any other thing like FLOPS, CUs, TMUs, CPUs speed etc then it's not so important or relevant... make up your minds people lol.
Thirty7ven
Banned
The argument is number of units and clock frequency matter. Math says there’s a 20% difference.
Now imagine Sony fans started talking about software and how Sony has special software that will close that gap. This has actually happened before, with Xbox One and DirectX. Maybe there is a pattern here?
SonGoku
Member
A quality SSD will outlive a HDD no doubt
Its like comparing a chinease/american car (no offense) to a Japanese car, the latter will be far more resilient
Typical numbers delta is not that bad, XSX solution is pretty dang good still, multiplatforms might just design around XSX SSD spec and PS5 SSD might just bruteforce for less pop in, smoother LOD transitions and shorter loading. Theoretically it could free more memory too but its up for discussion if devs will optimize for that, I'd say just wait for games to see what 3rd party devs do with each solution
I was talking about texture data which is what comprises the largest amount of game data.
Bo_Hazem
Banned
Wasnt the X4 roasted by Top Gear UK against the Land Rover?
What is an SUV anyway? Those need to man-up and return to the 4x4 naming.
My 2016 Nissan Patrol shorty 5-Speed MANual transmission, 2x solid axles with rear OEM diff-lock, wired throttle, 4.8L inline-6, 2" lift kit and 35" muddies
Last edited:
SonGoku
Member
Tell that to the guy (which i was replying to) that was using XSX decompressing hardware block theoretical peak output as a typical figure
Last edited:
Physiognomonics
Member
BattleScar
Member
Doesn't matter if you don't need to use any texture compression and still have higher bandwidth.
5.5GB/s >>> 4.8GB/s
Last edited:
Thanks for the clarification but you got some numbers wrong.
It's 20-30% on Kraken and minimum 50% using BCPack according to @richgel999(Richard Geildrich). So it's more than 20% at a minimum. XBTC(Xbox Texture Compression) is being optimized as we speak, so the
more from him:
Although he mentions this, he further qualifies by saying:
So Sony can offer something comparable to BCPack by using RDO alongside Kraken. On the other hand, It seems likely that BCPack will increase the effective throughput of the XSX SSD and perform better than the PS5 in terms of texture compression. This is before considering SFS.
https://twitter.com/richgel999
Bo_Hazem
Banned
But do we really need downsampling? Isn't it like upscaling in reverse? Maybe a 1080p-1440p mode would make better since. I personally think we need to really move on from 1080p, and give 1440p maximum of 4 years from now before ditching it. Higher resolution = less needed anti-aliasing and much sharper image.
PS4 Pro already supports supersampling.
Last edited:
SonGoku
Member
30-40% going by Cerny typical figures (8-9GB/s)
You could compress further but that would reduce the quality of the textures
ethomaz
Banned
Nope... I did not get the numbers wrong.
8GB/s is 30% compression over 5.5GB/s.
9GB/s is 40% compression over 5.5GB/s.
30-40% and not 20-30% like that guy tried to say.
4.8GB/s is 50% compression over 2.4GB/s.
Doesn't matter if BCPack can do 50% or a bit more in compression it is yet pretty big difference because the difference in compression is really lower than 20% between Kraken and BCPack and the issue is BCPack is lossy while Kraken is lossless... if MS needs lossless compression it will be way lower than these 30-40% of Kraken with ZLib.
That is wrong... BCPack won't even make Xbox SSD perform better than PS5's SSD.
That is just impossible.
It is at best case scenario be near to match the RAW PS5's SSD performance.
Last edited:
Not really. You need to look at the effect of the compression ratio. But first, BCPack isn't finalized because once we have actual figures we'll know the actual throughput of the SSD. Or how much higher it is than 4.8GB/s. The PS5 throughput will always be higher but if all they can achieve is 9GB/s while Xbox reaches 6GB/s then they've closed the gap. If we factor in SFS, it's clear that any game designed for the PS5 can also be designed for the XSX. For data that can be compressed to 22GB/s I don't see how RAM would use that so maybe the processors directly accessing this data?
This doesn't mean the PS5 won't always have a higher effective throughput, but the question is will it make a significant difference relative the XSX? So far that's not the case from what we're seeing.
Now this is my point, and why the compression ratio matters. If the PS5 can fill up RAM with 8GB worth of data in ~1 second(2gb per quarter second). Then XSX could do so in ~2 seconds(1gb per quarter second).
On the other end, if XSX is able to achieve a higher compression ratio and hit 6GB/s(which is what MSFT is working on now with XBTC) then they can reduce this time to below ~2 seconds.
The 4.8GB/s doesn't factor in BCPack!! They're still working on it to increase the throughput.
You cannot use BCPack compression ratios on
Noooooo!
Why are you using 50% yet BCPack is being optimized to go higher than that? And also you'd have to use BCPack and Zlib combined to calculate the throughput.
Richard was talking about Kraken's ability to compress texture data relative to BCPack not all data like you're doing. So you're using wrong figures. You're comparing Kraken to Zlib which are both for general game data. Which is correct but if we factor in BCPack, you'll have a higher throughput than the 4.8GB/s for the Series X due to a higher compression ratio.
RaZoR No1
Member
But did they solve the issue with the write limits on SSDs?
Afaik the blocks can be rewriten around 1 billion times (this is a made up number) or something like that and after that, the blocks cannot be changed anymore. I know that the companies use a "smarter" controller so that not always the same block is rewriten and to expand the lifespan.
Is this still a issue with todays SSDs?
Will the Virtual RAM / SSD RAM not shorten the lifespans with the constant write and access to the SSD?
Unfortunately I am not up to date regarding SSDs so I dont know how good they are now.
My concern is, with the custom SSDs in the consoles, once the SSDs died, we cannot replace them probably or wont get the same speeds even with the expansion port ( for example PS5 int. 6 to ext. 2 priorities , I dont know if there is any performance impact for the XSX Expansion Port)
I am really curious, how this will be handled and turn out for everyone.
Ascend
Member
How can you claim Sony went with cheaper RAM and at the same time claim that MS 'sacrificed' RAM?
MS didn't sacrifice RAM at all. It would have been cheaper and easier for them to go the 8 x 2GB way of the PS5, giving them the same 448GB/s bandwidth. Splitting the ram chip configuration to 6 x 2GB + 4 x 1GB was done for a bandwidth advantage. They are not stupid. Why would they invest more money in something that will perform worse? Why this configuration costs more money you ask? Because you need additional lanes and space on your PCB, since you need to install more RAM chips, which would also cost more.
They did say they sacrificed the RAM, but the sacrifice they were talking about was from 16 x 1GB to 6 x 2GB + 4 x 1GB, not from 8 x 2GB to 6 x 2GB + 4 x 1GB. The less RAM chips the cheaper. The amount of RAM chips determines your bandwidth. They couldn't have somehow gone with less chips and achieve the 560GB/s that they do now.
Ass of Can Whooping
Gold Member
You don't need to be concerned about the write limits of an SSD for gaming
Bladed Thesis
Member
This thread has taken a turn for the worse. A long time ago to be honest.
Let's just put 'er down like Old Yeller and wash our hands of it?
Let's just put 'er down like Old Yeller and wash our hands of it?
Captain Hero
The Spoiler Soldier
PS5 fat will be as huge as PS3 fat .. I’d love it
And how are you calculating the figures. The compression ratio on the PS5 SSD is 1.5
Nope the 4.8GB/s figure does not factor in BCPack. The 6GB/s does if I had to guess because that's the throughput they're targeting. So it will be between 4.8-6GB/s
thicc_girls_are_teh_best
Member
They look like smaller numbers in relative comparison, but also keep in mind there are things along the I/O, controller, file system etc. pipeline (plus other features in terms of modifications to certain aspects of the GPU and features within the silicon) that are working alongside those numbers.
This goes for both systems, but in XSX's case I don't think their team would've settled on those performance numbers if they didn't feel them sufficient for data loading and streaming purposes, both on their platform and in general with where the role of SSDs will fall for serving use-case purposes for the consoles. MS and Sony have just seemingly taken some different approaches but, hey, like other things, paper specs don't tell the whole story.
I think something important to keep in mind with the MS numbers, too, is that those are sustained performance figures, if we take their words at face value. They're the numbers they are guaranteeing the SSDs will perform at, at all times. And the reason they have to stress that is because they're also positioning XSX for server markets, where sustained/constant performance is a necessity. If they have peaks notably higher than that, but they're random and inconsistent, then there's no reason for them to mention that
With Sony's numbers, I'm going to assume those are sustained, but all I will say is that they don't necessarily have a market reason to stress if the numbers are sustained or not, since for that type of thing the consumer market has different standards and needs vs. data and business server markets. So is there a chance they could be speaking of peak numbers? Well, the possibility always exists, but I'd personally put it within a margin of error, if not 0%.
One thing I hope Sony clarifies if how the SSD (especially in addition to an optional NVMe drive sitting in the expansion bay) impacts the variable frequency setup in the system. The internal one alone is going to need a decent bit of power, let alone if an approved 3rd-party drive is sitting there alongside it. So if there are instances where power needs to be reduced among other system components to keep the GPU at max frequency, will the SSD(s) factor into that and by how much? Would a 2% frequency drop result in a 2% speed drop for one SSD, both SSDs, or each SSD? Would the speed drop be 2% or greater?
Those kind of questions pop up, because just with the internal drive that is a decent number of high-speed NAND modules and a decent amount of power needed to them and other parts of the SSD, and it multiplies by a factor of 2 when you throw the 3rd-party optional drives in there, as well.
It depends on the quality of the NAND. SLC NAND will remain the best at P/E cycle endurance ratings and speed, while QLC will remain the worst in these aspects. However, wear-leveling technologies and programs have improved a ton since several years ago, and you can get pretty good mileage out of QLC NAND drives now provided you don't need them for servers or environments where write-heavy operations are the norm.
I'd suspect both systems are using either TLC or QLC NAND, and no small pool of SLC or MLC as a cache (i.e SLC and MLC NAND will be absent in them). Neither have DRAM caches, either, tho PS5 will have an SRAM cache (probably around 32 MB - 64 MB; if it's PS-RAM it could be larger while still staying affordable), and XSX will probably use a portion of the reserved 2.5 GB GDDR6 OS memory for caching of its SSD.
For replacements on PS5, you'll need a minimum 7 GB/s drive, to get the same performance as the internal 5.5 GB/s one. On XSX, the expansion storage card just plugs into the back and has similar use-case to a memory card, but offers the exact same performance as the internal drive. We probably don't need to worry about the systems going kaput if the internal storage gets used up beyond further use.
Right. The sacrifice MS made was in regards to going all 2 GB modules or a mix of 2 GB and 1 GB. I expect their RAM to be more expensive though not just because of more modules (10 vs 8), but because MS have added ECC to their memory which to my knowledge Sony isn't doing and they wouldn't have a need tbh since they aren't making PS5 for server markets in addition to the console market.
One thing they could both still do is go with somewhat faster chips, but with it about to be May I doubt any changes that big will come into swing.
Gotta give it to rntongo on this one at least regarding the BCPack optimization; one of the XSX team members mentioned on Twitter in responses that they are still looking to push the compression effectiveness further.
Apparently there's a good deal of room for them on that front but the question is will they actually implement it in the system (and how much further they can push the compression before it's time to wrap up work).
Last edited:
Radical_3d
Member
- Status