Do you have a source for this? Everything I've seen suggests it's heavily reliant on UE4.
Not that I can quote publicly, no.
This comes from the information I got from a couple of friends working as programmers for Splash Damage.
Do you have a source for this? Everything I've seen suggests it's heavily reliant on UE4.
I am not talking about data center SSD. I am saying no one really knows what the speeds on these consoles are. All speculation. You and I have been on many versions of this topic to know how so many guesses, educated or not ended up.This is not speculation, current datacenter SSD with customized filesystem and application goes much faster then "normal" PC SSD...
I posted some link, nothing to do with consoles...
Understood. Thanks.Not that I can quote publicly, no.
This comes from the information I got from a couple of friends working as programmers for Splash Damage.
I am not talking about data center SSD. I am saying no one really knows what the speeds on these consoles are. All speculation. You and I have been on many versions of this topic to know how so many guesses, educated or not ended up.
And most of these when put through some basic reasoning seem farfetched to the extreme.
Block sizes are the determining factor which Part 2 of the tests show you. They reduce latency on access and read. Parallelism can be simulated quite good by the number of threads and queues handling the reading. A data core provider is just the same excluded in its own core/fixed function in HW. Differences even with a console will not look much different under the same circumstances (FAT not in memory, similar data block sizes, same IOPS specs) when we take into account all those tests were made on a 8C/16T system and the load was minimal for those tests.But this is used on PC with a filesystem slowing down the SSD
For example SSD are good accessing randomly data but you need to have the filesystem tailored around that...
and for maximize utilization you need to have Parrallel I/O like datacore provide for datacenter...
Block sizes are the determining factor which Part 2 of the tests show you. They reduce latency on access and read. Parallelism can be simulated quite good by the number of threads and queues handling the reading. A data core provider is just the same excluded in its own core/fixed function in HW. Differences even with a console will not look much different under the same circumstances (FAT not in memory, similar data block sizes, same IOPS specs) when we take into account all those tests were made on a 8C/16T system and the load was minimal for those tests.
Edit: However the main difference to console will be that the customizing there will involve in-memory FAT and maybe direct DMA between the shared main memory and the SSD.
I did not say game systems do not benefit from such capabilities, I just wanted to express a PC can pull off a lot of that too even without the need of having customized HW. So tests there are not useless. However it should be made clear that those results from my tests are indicators and not hard facts.I cant look it up right now, but NVMe has a lot of potentially useful ways game systems can benefit, the ones you said are just a few (DMA baby!!!, deeper/ooo queues, etc etc
I did not say game systems do not benefit from such capabilities, I just wanted to express a PC can pull off a lot of that too even without the need of having customized HW. So tests there are not useless. However it should be made clear that those results from my tests are indicators and not hard facts.
You mean like how No Man's Sky does it? Because that's already short enough of a transition to be cartoonishly unrealistic, so part of the PC sims' longer transit time is just adding some time to make it feel more realistic and give a better sense of scale. On the other hand, No Man's Sky is pretty much the king of pop-in, so assuming SSDs as the baseline could help there.
Some news about no news from TGS probably (related to PS5/Sony):
My secret hope is they all have thunderbolt3, but I am not sure about amd support for it... and with pcie4, it may not be fast enough... time will tellI wonder if we will start to see external ssd's being sold with the same spec as the ps5/scarlett internal ssds
So ppl can expand there storage but still have the custom ssd's benifits.
I don't think Sony reinvented Solid State Storage paradigms like some here. They probably enabled a couple of cool customizations to speed things up in a controlled console environment and MS has probably done the same. Any low cost update you can make to consoles to eek out more performance is a huge win.
I don't think Sony reinvented Solid State Storage paradigms like some here.
I know about it, I am just pointing out using benchmarks for games that dont at all try to utilize nvme are pointless
I don't think Sony reinvented Solid State Storage paradigms like some here. They probably enabled a couple of cool customizations to speed things up in a controlled console environment and MS has probably done the same. Any low cost update you can make to consoles to eek out more performance is a huge win.
High-performance SSDs can perform hundreds of thousands of I/O operations per second. To achieve this performance, drives make use of parallelism and complex flash management techniques to overcome flash device limitations. These characteristics cause SSD performance to depart significantly from that of disk drives under some workloads. This leads to opportunities and pitfalls both in performance and in benchmarking. In this paper we discuss the ways in which high-performance SSDs are different from consumer SSDs and from disk drives, and we set out guidelines for measuring their performance based on worst-case workloads. We use these measurements to evaluate improvements to Linux I/O driver architecture for a prototype high-performance SSD. We demonstrate potential performance improvements in I/O stack architecture and device interrupt handling, and discuss the impact on other areas of Linux system design. As a result of these improvements we are able to reach a significant milestone for single drive performance: over one million random read IOPS with throughput of 1.4GBp
My secret hope is they all have thunderbolt3, but I am not sure about amd support for it... and with pcie4, it may not be fast enough... time will tell
The important thing is how they will used it. Like how they used an outdated crappy APU to do awesome stuff with it.I don't think Sony reinvented Solid State Storage paradigms like some here. They probably enabled a couple of cool customizations to speed things up in a controlled console environment and MS has probably done the same. Any low cost update you can make to consoles to eek out more performance is a huge win.
Going by those patents, they didn't really. What they did is propose a bunch of optimizations that, when taken together and tuned for a very specific purpose, vastly outperforms off-the-shelf stuff.
But that's to be expected, general purpose will never perform as well in a specific scenario as something purpose built for that scenario. You're trading performance for flexibility. Consoles don't need that flexibility, so it's a smart move to make.
Somthing like that would probably be needed I don't think usb 3's 625MB/s will be enough.
However thunderbolt3 is a intel/apple thing, I don't think Sony or MS would use it.
If branded externals were the only way (which Sony and ms will be happy about making money on them) they will probably use a custom connector.
Thunderbolt3 is now USB4, and is/has been completely openSomthing like that would probably be needed I don't think usb 3's 625MB/s will be enough.
However thunderbolt3 is a intel/apple thing, I don't think Sony or MS would use it.
If branded externals were the only way (which Sony and ms will be happy about making money on them) they will probably use a custom connector.
That's good.
We should see it in next gens then?
Seeing how it's like 8x faster the usb3
Some news about no news from TGS probably (related to PS5/Sony):
Some news about no news from TGS probably (related to PS5/Sony):
On PC devs can simply set a minimum spec. for required SSD performance.
They already set minimum requirements for games now, so nothing really changes.
PC gamers are those most used to having to upgrade hw to enjoy the latest software goodness.
It's no different from them having to upgrade to an RTX card to play games with RT features today.
While I did those test, that was not the conclusion. It really was dependent on certain factors like block size and random and sequential read how big the differences were. Btw all tests I made are threadmarked in this thread as I reposted them early in the thread by adding a 3rd part to it: Link
No need for an apology! There is so much different information out there, it is easy to forget where you have read or seen it. Of course today you cannot see the real benefits in games because those games are not made with SSD in mind, means the throughput requirements are catered to HDDs. Reason I did the tests on different scenarios of access patterns and file system parameters and not with games and a stop watch!Thank you for the link, and yeah, I got your test results mixed up with a video I was watching, which actually did claim an NVMe doesn't add too much for games. Apologies.
All I say is this is a typical Gamingbolt article. I avoid those ...PS5's Zen 2 CPU Is "A Huge Step Up", Will Reduce Development Time "Significantly" – Lost Wing Developer
"It will be a huge step up," he said. "It gets rid of so many restrictions with how many calculations we can do on the fly, we have so many ideas which are very hard, or even impossible to implement in the current generation of consoles. Also it actually reduces development time significantly as less time needs to be spent on optimisation."
Hmm,PC brute-forcing coming to consoles? ;)
No need for an apology! There is so much different information out there, it is easy to forget where you have read or seen it. Of course today you cannot see the real benefits in games because those games are not made with SSD in mind, means the throughput requirements are catered to HDDs. Reason I did the tests on different scenarios of access patterns and file system parameters and not with games ans a stop watch!
All I say is this is a typical Gamingbolt article. I avoid those ...
----
Off-topic:
Sylens is a bitch ...
I think we will see a ps5 announcement event in Feb, and a scarlett one a month or 2 after.
I don't see why they would want to change it up.
It will be sunshine and rainbows until they find how to push its limits and they're back to optimizing.PS5's Zen 2 CPU Is "A Huge Step Up", Will Reduce Development Time "Significantly" – Lost Wing Developer
"It will be a huge step up," he said. "It gets rid of so many restrictions with how many calculations we can do on the fly, we have so many ideas which are very hard, or even impossible to implement in the current generation of consoles. Also it actually reduces development time significantly as less time needs to be spent on optimisation."
Hmm,PC brute-forcing coming to consoles? ;)
It will be sunshine and rainbows until they find how to push its limits and they're back to optimizing.
They don't have to reinvent anything. They just have to play to their strengths of solely being a console manufacturer in the gaming spec. Eg, Ryzen 3 has 24 PCIe lanes Usually a NVMe drive takes up 4 of those lanes. And an NVMe controller reads/write data to those 4 lanes. Nothing stops sony from taking up 8 lanes exclusively for their SSD and thus doubling the bandwidth. Or the could use 2 SSD controllers on their PCB making heir SSD act like a RAID drive and again, doubling the bandwidth.I don't think Sony reinvented Solid State Storage paradigms like some here. They probably enabled a couple of cool customizations to speed things up in a controlled console environment and MS has probably done the same. Any low cost update you can make to consoles to eek out more performance is a huge win.
The Zen 2 CPU in the console is not the same as the PC part. The lanes are coming from the IO die (IOD) which is probably absent in the console SOC. So we do not even know how they implement high speed SSDs.They don't have to reinvent anything. They just have to play to their strengths of solely being a console manufacturer in the gaming spec. Eg, Ryzen 3 has 24 PCIe lanes Usually a NVMe drive takes up 4 of those lanes. And an NVMe controller reads/write data to those 4 lanes. Nothing stops sony from taking up 8 lanes exclusively for their SSD and thus doubling the bandwidth. Or the could use 2 SSD controllers on their PCB making heir SSD act like a RAID drive and again, doubling the bandwidth.
Point is, what ever they do will become the PS standard and every dev building for it will be guaranteed that kinda performance. That can't happen on PCs cause devs will have to at the very least conform to a PC with a PCIe4 drive at best. Devs can't build on PC with the assumption that PC gamers will have SSD drive in RAID or have them on a riser using 8 PCIe lanes as opposed to 4.
MS will probably fall somewhere in between, as they have t ensure that their games can work just fine n a PC. Or they go with a similar approach to sony and have code that changes the way their game runs depending on what platform its running.
I don't see any platform warring. I see people about a great feature and I see a user questioning if developers will even take advantage of the feature.I'm sure there'll be a default code path in the API to let the system manage the data streaming from the mass storage device, but devs will also be given the option to manage it themselves to better optimise for the performance of their game.
I think some people here arr far too preoccupied with who will have the fastest SSD storage solution that they're missing the forest for the trees.
We should all be rejoicing that consoles are getting NVMe-level SSD mass storage bandwidth at all as standard. The difference between it and a shitty HDD is so huge that every game being able to rely on SSDs means gaming will return to the level of immediacy that we enjoyed back in the cartridge era. It's fucking awesome.
Who cares if Sony's solution is 5, 3, 2 or 1Gb/s faster or slower than MS's? The fact of SSD level performance being the new baseline is what's important and what everyone should be focusing on, instead of contriving new ways to spin the discussion into platform wars.
It will be interesting to see how this influences PC gaming (console ports) regarding CPU,considering next gen base will be 8 core Zen2.
Colbert ,i think your (8 core?) Zen+ will get very sweaty next gen :)
It will be interesting to see how this influences PC gaming (console ports) regarding CPU,considering next gen base will be 8 core Zen2.
Colbert ,i think your (8 core?) Zen+ will get very sweaty next gen :)
While I am not a PC elitist my CPU never will be sweaty because it lives in a swimming pool ;)A lot of people's PC's are going to be sweaty. I can't wait to see the salt from the elitist segment of the PC crowd, it'll be entertaining.
While I am not a PC elitist my CPU never will be sweaty because it lives in a swimming pool ;)
While I am not a PC elitist my CPU never will be sweaty because it lives in a swimming pool ;)
Obviously the APUs in the consoles will also have a I/O portion. Just like how chips in the pre chiplet era had their I/O portion too. The only real argument will b that the consoles APU will have nowhere near 24 PCIe lanes. I am guessing half that at the most or even a third. Consoles don't need that many, and even if they had only 8? Thats 8 times ore than the I/O part of the chip had in the current gen consoles who were even only using a sata bus as opposed to a PCIe bus.The Zen 2 CPU in the console is not the same as the PC part. The lanes are coming from the IO die (IOD) which is probably absent in the console SOC. So we do not even know how they implement high speed SSDs.
Laziness! (joke)Also it actually reduces development time significantly as less time needs to be spent on optimisation."
Laziness! (joke)
It will just shift the goalposts higher and the same limits will apply this time next gen.
But this isn't simple at all. Many won't know how to figure out if their SSD meets this minimum speed, and that's not even getting into the problem of used or low-quality SSDs not meeting their stated speeds. It has to actually be simple, which means either requiring "an SSD" (which forces them to develop around some pretty low speeds) or "an NVMe SSD" (which is rather pricey and restrictive). This isn't like with GPUs (as you can turn the graphics down to ridiculous degrees on PC) or CPUs (virtually nothing in the PC space is as weak as the Jaguars, and even if your CPU actually was you could still play a game at 10-20fps).
I don't see any platform warring. I see people about a great feature and I see a user questioning if developers will even take advantage of the feature.
That's not platform warring.
If you can't stand people talking and comparing speeds, probably best to not set foot in a speculation thread that primarily revolves around technical specifications.
To bring it back to the original discussion, and your first paragraph, whilst I have zero knowledge of game development I don't see why it would be anything other than 'it just works' (again assuming one single storage pool) - I don't know why it would be something that a developer would have to specifically 'enable' or make use of above and beyond their standard method of loading things in.