Color me skeptical. I don’t recall the 360 HDDs being high performance, but that didn’t stop MS from beating users over the head with its pricing.
Xbox Series X [XBSX] [Release November 10 2020]
- Thread starter Megadrive1988
- Start date
-
- Tags
- microsoft xbox xbox series x xbsx
Color me skeptical. I don’t recall the 360 HDDs being high performance, but that didn’t stop MS from beating users over the head with its pricing.
Which they drastically changed their practice on X360 near the end of that family and never sold expansion units for the Xbox One family. I think Microsoft will only be getting the license fee from the Seagate Expansion products, that is they won't be like the X360 drives at all where it was entirely MS label and all. You didn't know what was inside unless you opened it up.
I think you may be conflating Sampler Feedback, and Sampler Feedback Streaming.
SF tells you which mip-map you need, which mip map you have, and how close to needing a new mip map you are. This feature may or may not be present on PS5.
SFS allows the GPU to act as though it has the needed mip map in memory even if it hasn't, and blends the new new higher detail mip map in the lower res once it's in memory to hide the missed transition.
SFS is cool to be sure, but it's the icing on the SF cake. We still wait to find out if SF is an XSX console exclusive, like VRS.
Most of the resident texture savings MS talked about are due to SF allowing only needed texture tiles to be loaded, and not due to SFS. Though SFS might make cutting things closer, and saving more memory, more realistic my making transitions less obtrusive.
If I'm not mistaken the last official HDD was release in 2014 at 500 GB for $129. Off the shelf multi-TB drives could be had for cheaper prices during that time.
2015? You can afford to be generous for a old gen product when your new product is over a year old.
I am not saying its guaranteed that MS will revert to old pricing practices but I'm not about to expect cheap pricing on a whim.
Megadrive1988
Veteran
The 12 TFLOP XSX APU or does that include Lockhart?
Between the 2 different APUs I could see 4 projects.
XSX
Lockhart
XSX-based xCloud server blade
Lockhart-based xCloud server blade.
It's:
XSX
XSX Cloud
MS Device - wait for more on this one
???? last one is outside my reach. Lockhart does not use Scarlett SoC. It wouldn't be efficient.
Silent_Buddha
Legend
Considering Microsoft's stance on expandable storage starting towards the end of the X360 era and extending throughout the XBO era, as Brit pointed out, I'm doubtful that they'd suddenly reverse course on this.
There are XBOX branded external drives available that carry a small premium over non-XBOX branded external drives.
https://www.amazon.com/Seagate-Game...rds=xbox+external+drive&qid=1592435776&sr=8-3
That's one example made by Seagate. I fully expect something similar for the XBSX. I also wouldn't be surprised if Microsoft allows other SSD makers to make licensed XBSX SSD expansion drives.
After all, as far as we know, the NVME solution will be using a bare bones off the shelf SSD controller with likely minimal components. So making them should be cheap. Even with markup they're likely to still be significantly cheaper than enthusiast level NVME drives.
Cheap controller, fewer channels needed for the speed required means less NAND chips needed, fewer components needed to handle higher speeds, simpler board design, etc. If they allow multiple SSD drive vendors to make officially branded drives like that Seagate one above, that will also keep prices in check through competition.
Regards,
SB
Silent_Buddha
Legend
You missed the other important part. It allows to only load parts of a MIP level. You don't need to load in the entire MIP level. The developer can specify exactly which part of the MIP level to load.
Regards,
SB
Sampler feedback allows that. It's literally part of the DX12U spec for PC:
I'm a bit of a dick, but I'm not a dick who's making this up.
The vram saving part of this is part of the DX12 Ultimate spec. It's in their developer docs. For real!
Sampler Feedback Streaming (on XSX) builds on this by taking the instruction to load a higher detail mip level for a mip tile, but if it can't get it in time it uses the lower detail one. Then when it can get the full res mip tile, it blends over the lower res one for (x) amount of frames to hide the transition.
Sampler feedback is part of the DX12U spec. Nvidia already have it, RDNA2 will bring it for AMD on the PC end.
Does all of this reduce the amount of data being streamed by the SSD or not? According to a graphics engineer working on the XSX, it does. That was my only point before you felt compelled to explain something nobody even asked for.
Sampler feedback resolves case issues that you can experience during streaming virtual textures and other methods. The feedback system is there to provide incite to the developers on what they are sampling before they pull the textures they want. It will help improve what tiles they retrieve. But it comes at a cost.
It is more or less the concept of tiled resources, so yes it can save massive bandwidth. Tiled resources was promoted just before the last console gen, but the HDD in those were the problem, because if you only load small parts of textures (like mini-files) you would cripple the HDD bandwidth. The SSD don't have this problem. You can load minimal files without a noticeable bandwidth impact. To load fast enough of the current gen, games had to load textures etc in packets. This lead to the problem, that you have massive amounts of things you don't really need in memory (and duplicates). The ssd make is possible just to load things you really need in a few frames into memory and with something like tiled resources, this gets much more efficient.
But I guess the PS5 has similar concepts, because the base of tiled resources should be in RDNA 2 for sure.
That may have been your intended point, but what you wrote was something else, and function corrected you with a technical explanation of the difference between SF (the thing you are talking about) and SFS (the thing you mentioned), which is a subtle quality variation on SF.
There's also the implied part of your post that XBSX is doing something unique with data access meaning it can get better performance from its hardware, and function is explaining the part that allows faster streaming (by steaming more efficiently) isn't unique to XBSX.
If you didn't want a technical explanation of SF, you probably shouldn't be posting on B3D...
There's a difference between SF and SFS and we need to use the correct terms here to not muddle conversation. please appreciate the time function put in to clarify the terms and engage in the discussion on a technical level. SF is not SFS, as you so eloquently explained. SF without the texture filters is not very useful since there will always be a delay between determining that a mip is not resident and it being made available. If you are targeting mip change on a nearly per frame transition basis (as I suspect MS is trying to do), then specialised hardware need to be available to smooth transitions or else it will be pop-in galore which defeats the purpose of texture streaming.
I don't think so. You can just prefetch a little earlier instead of later. SF improves the implementation of tiled resources to make it easier and more efficient. SFS helps hide situations where your prefetch has failed, but if your prefetch is good enough, that shouldn't happen that often (if ever!). Potentially, the faster your storage, the less that is a problem. In RAGE for example, SFS would have softened the texture transitions so they were less jarring, as I understand it. But if running from a modern M.2 SSD, that pop-in wouldn't happen in the first place for SFS to help at all.
MS invested in silicon to address situations that will almost never happen... Good to know.
Similar threads
- Locked
