Current Generation Games Analysis Technical Discussion [2020-2021] [XBSX|S, PS5, PC]

Status
Not open for further replies.
No, just expecting it to be faster than PC since it shouldn’t be bothered by the I/O stack. There is sufficient reason to believe the Xbox version is likely crippled.
Just like PS5 requires explicit filesystem setup to take advantage of its speed, I’m sure XSX has to have something somewhat similar as well.
faster than pc with faster ssd and cpu ?
 
The speed of the drive itself is largely irrelevant (past a certain point) when the bottleneck is elsewhere. I think it's pretty obvious in this case that the VA isn't being used to it's full potential or else the XSX should easily be able to exceed the performance of the PC drive which is hobbled by an ancient IO software stack. Things are going to get really interesting once Direct Storage starts getting used on the PC.

Do you have a benchmark? We don't know this. Here what is sure the XSX SSD is slower than the PC one and is able to load as fast. It proves something is done on XSX side not done on the PC Drive and decompression is not a problem on PC for loading data from a save. This is not streaming, if it use multiple CPU core or all CPU Power this is not a problem.

There is a moment you need to stop being delusional. The PS5 SSD is faster than the Xbox Series X one. From a streaming perspective there is a coherency engine with cache scrubber present in PS5 not on Xbox Series X same of the work done on Xbox Series X CPU is done fully inside the I/O Complex on PS5.

This is like thinking Xbox Series X will not be faster on raytracing than PS5. On DMC 5 and RE8 raytracing is a bit faster on XSX than PS5.

And PS5 is able to do virtual texturing and Partial resident texture too in hardware this is available on AMD GPU since GCN 1 since 2011. SFS means if the mip level needed is not inside RAM, the XSX can load a lower mip level present in memory, on PS5 it will be a big texture pop in but the SSD and streaming hardware being faster on PS5 side less chance this problem arrive.

https://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/6

Partially Resident Textures: Not Your Father’s Megatexture

Among the features added to Graphics Core Next that were explicitly for gaming, the final feature was Partially Resident Textures, which many of you are probably more familiar with in concept as Carmack’s MegaTexture technology. The concept behind PRT/Megatexture is that rather than being treated as singular entities, due to their size textures should be broken down into smaller tiles, and then the tiles can be used as necessary. If a complete texture isn’t needed, then rather than loading the entire texture only the relevant tiles can be loaded while the irrelevant tiles can be skipped or loaded at a low quality. Ultimately this technology is designed to improve texture streaming by streaming tiles instead of whole textures, reducing the amount of unnecessary texture data that is streamed.

And in Direct X it is called hardware tiled resource
https://docs.microsoft.com/en-us/windows/win32/direct3d11/tiled-resources

Tiled resources can be thought of as large logical resources that use small amounts of physical memory.

This section describes why tiled resources are needed and how you create and use tiled resources.

https://docs.microsoft.com/en-us/windows/win32/direct3d11/why-are-tiled-resources-needed-

Tiled resources are needed so less graphics processing unit (GPU) memory is wasted storing regions of surfaces that the application knows will not be accessed, and the hardware can understand how to filter across adjacent tiles.

In a graphics system (that is, the operating system, display driver, and graphics hardware) without tiled resource support, the graphics system manages all Direct3D memory allocations at subresource granularity. For a Buffer, the entire Buffer is the subresource. For a Texture (for example, Texture2D), each mip level is a subresource; for a texture array (for example, Texture2DArray), each mip level at a given array slice is a subresource. The graphics system only exposes the ability to manage the mapping of allocations at this subresource granularity. In the context of tiled resources, "mapping" refers to making data visible to the GPU.

Suppose an application knows that a particular rendering operation only needs to access a small portion of an image mipmap chain (perhaps not even the full area of a given mipmap). Ideally, the app could inform the graphics system about this need. The graphics system would then only bother to ensure that the needed memory is mapped on the GPU without paging in too much memory. In reality, without tiled resource support, the graphics system can only be informed about the memory that needs to be mapped on the GPU at subresource granularity (for example, a range of full mipmap levels that could be accessed). There is no demand faulting in the graphics system either, so potentially a lot of excess GPU memory must be used to make full subresources mapped before a rendering command that references any part of the memory is executed. This is just one issue that makes the use of large memory allocations difficult in Direct3D without tiled resource support.

After some devs don't use it because it needs a call to CPU


 
Last edited:
The pc version is basically loading as fast as the 'optimized ps5 version', while not supporting direct storage yet. Either the XSX IO isnt really fast orthe PC nvme is just that much faster then both ps5 and xbox nvme setups. If pc is already loading that close to ps5 with the alleged 'heavy IO software etc bottleneck', imagine 7gb/s drives with RTX IO/DS.
 
Can someone post the videos here? I can’t follow the evidence. Is it just the one above?
It’s just 5s v 2.5s??

isn’t there some other video where Xbox is 10s and Ps5 is 1.5s?
 
I thought load time said by DigitalFoundry was 1.5s vs 8.5s, for a 5.666x relative difference. That relative difference is far more than the spec differences between the two SSDs we have been given.

This is not for all the loading time. This is from save, it depend of the save. It seems between 5 and 8 second on XSX like on PC with a better SSD and 1,5 to 2 seconds on PS5.
 
Okay. So that particular loading from save game thing; needs examining. I do wonder if that is an Xbox platform hit. Syncing save games on Xbox cloud. Which may not be I/O related then.
 
Are the SSDs consistent between XSS and XSX? You'd think we'd see quicker loading on the XSS, since it's got a lot less RAM. I can't follow why the loading is consistently faster on the XSX.

I assume they're using the same decompression hardware too?

These anomalies signify that something's up with the loading on the new Xbox consoles. Hopefully it'll be fixed as the generation progresses.
 
For some late game scene which switching between characters in different places, it's 3-4s vs 12-15s.
Oof. Yea that requires a bit more investigation.
It could be as simple as Xbox save cloud sync which is slow. Or it could be I/O. But seeing how some load times are coming in just after PS5 within reason of their hardware spec difference, I may just look at other things Xbox is doing possibly during that load sequence.
 
Are the SSDs consistent between XSS and XSX? You'd think we'd see quicker loading on the XSS, since it's got a lot less RAM. I can't follow why the loading is consistently faster on the XSX.

I assume they're using the same decompression hardware too?

These anomalies signify that something's up with the loading on the new Xbox consoles. Hopefully it'll be fixed as the generation progresses.
SSDs are the same speed on both. If they are doing a cloud sync right at that junction, they can’t fix that.

It’s can’t be IO. 15 seconds of loading is like 36+ GB of memory loading into 10GB of memory. It can’t be CPU either. If PS5 is finishing in 3-4 seconds, CPU must be % of that, even 1-2 seconds let’s say. That cannot be 13 seconds of IO on Xbox. You’d be filling at less than 1GB/sec. well below spec.
So I can only point it at doing network checks or something
 
Do you have a benchmark? We don't know this. Here what is sure the XSX SSD is slower than the PC one and is able to load as fast. It proves something is done on XSX side not done on the PC Drive and decompression is not a problem on PC for loading data from a save. This is not streaming, if it use multiple CPU core or all CPU Power this is not a problem.

Yes.

https://www.techspot.com/review/2116-storage-speed-game-loading/

This shows 5GB/s+ drives being barely faster than 550MB/s drives. Because they're bottlenecked by other elements of the IO stack.

There is a moment you need to stop being delusional. The PS5 SSD is faster than the Xbox Series X one
.

No-ones making that claim. Everyone knows the PS5 drive is faster than the XSX one. But there seems to be an argument going on here that the PS5 also has a major advantage over the XSX in the rest of the IO stack - to the tune of it being twice as efficient in some of the claims about (5x faster than the XSX).
 
Again first loading time is more than loading data. Second SSD speed is not all. PS5 has 1 MB if SRAM cache into the I/O Complex and it seems some 2GB DDR4 on SSD controller, The XSX SSD is cache less.

https://www.resetera.com/threads/pl...-technical-discussion-ot.231757/post-48562453

Qfg1uHi.jpg


dtrum3zqn5t51.png


https://www.resetera.com/threads/pl...-technical-discussion-ot.231757/post-48565207

It’s DRAM cache for the SSD. 2GB, which is very large (980 Pro only has 1GB).
 
Last edited:
  • Like
Reactions: snc
Yes.

https://www.techspot.com/review/2116-storage-speed-game-loading/

This shows 5GB/s+ drives being barely faster than 550MB/s drives. Because they're bottlenecked by other elements of the IO stack.

.

No-ones making that claim. Everyone knows the PS5 drive is faster than the XSX one. But there seems to be an argument going on here that the PS5 also has a major advantage over the XSX in the rest of the IO stack - to the tune of it being twice as efficient in some of the claims about (5x faster than the XSX).

This is non optimized game for SSD there is other things in HDD games than the OS I/O stack slowing down the loading like monothreaded loading and so on.

The advantage of Direct Storage is less latency with no copy of data from main RAM to VRAM and GPU decompression.

The PS5 has other advantage on the I/O stack like the coherency Engine, the I/O complex has 1MB of SRAM use for table translation and as cache and the SSD controller has DRAM cache it seems 2Gb more than a 980 PRo Samsung SSD.

And the CPU is not involved at all in I/O process in PS5 out of asking for data.

The XSX SSD is Cache less.
 
Last edited:
The one inside PS5 should be 4Gb(512MB).

This is sure 2GB seems too much and Fabian Giesen from RAD tools give a usage for SRAM into I/O Complex. It serve as cache during the decomrpession of data to use less memory bandwith.

This is 512 MB find somewhere the guy on era did an error.
 
Last edited:
I don't think we should exactly be worried for Xbox, it's not like games are pushing the steaming technology right now. RE runs on last-gen slow HDDs.

It does go to show how significantly things have still yet to progress this generation. I hope the next GTA makes better use of the hardwares' streaming capabilities. I'm certain everything will be resolved by then.
 
Status
Not open for further replies.
Back
Top