Xbox Series X [XBSX] [Release November 10 2020]

Well I don't see it doing too much if it requires 30 secs+ to load a game.

Because its not being used on legacy software.
https://news.xbox.com/en-us/2020/05/28/xbox-series-x-next-generation-backward-compatibility/

Backwards compatible titles also see significant reductions in in-game load times from the massive leap in performance from our custom NVME SSD which powers the Xbox Velocity Architecture.


https://news.xbox.com/en-us/2020/06/10/everything-you-need-to-know-about-the-future-of-xbox/

Backwards compatible titles also see significant reductions in in-game load times from the massive leap in performance from our custom NVME SSD which powers the Xbox Velocity Architecture.


From official site they say backwards compatible titles also use NVME SSD which powers the Velocity Architecture.
 
https://news.xbox.com/en-us/2020/05/28/xbox-series-x-next-generation-backward-compatibility/

Backwards compatible titles also see significant reductions in in-game load times from the massive leap in performance from our custom NVME SSD which powers the Xbox Velocity Architecture.


https://news.xbox.com/en-us/2020/06/10/everything-you-need-to-know-about-the-future-of-xbox/

Backwards compatible titles also see significant reductions in in-game load times from the massive leap in performance from our custom NVME SSD which powers the Xbox Velocity Architecture.


From official site they say backwards compatible titles also use NVME SSD which powers the Velocity Architecture.
The nvme ssd is part of VA.
By definition it will be faster running from it.

When people say its not using VA they are talking about the other aspects which isn't just the nvme drive. Example the new direct storage api, etc.

Similar if someone said BC games not making use of the gpu, obvously getting the raw performance upgrade but it's not making use of SFS, RT, recompiled to make use of other efficiencies etc.
 
All titles will use the SSD, it's just the older titles have specific requests for the CPU to perform the decompression. Xbox One games in BC will continue to use the CPU and that will bottleneck the performance of the SSD.

Xbox Series X games will send the textures to the decompression IO, which would significantly improve loading time.

We'll have to wait and see how much quicker it is in practice.
 
Last edited by a moderator:
A little bit more on the various external storage testing, basically no need or benefit of spending more on USB-NVME when USB-SSD performs equally. Both are significantly ahead of USB-HDD for file copying.

4:24.83 vs 4:33.10
2:22.35 vs 2:18.53

 
A little bit more on the various external storage testing, basically no need or benefit of spending more on USB-NVME when USB-SSD performs equally. Both are significantly ahead of USB-HDD for file copying.

4:24.83 vs 4:33.10
2:22.35 vs 2:18.53

Now, the real question is, how do relative "cheap" USB sticks work with BC games. Worked quite well on xb360 even though it was limited to USB 2.0. 1TB flashpens are really cheap nowadays. Would be great if those would work with BC games.
 
A little bit more on the various external storage testing, basically no need or benefit of spending more on USB-NVME when USB-SSD performs equally. Both are significantly ahead of USB-HDD for file copying.
Ars Technica also did some testing. The results are predictably unpredictable! This is pretty much what I expected bit the external SSD vs HDD performance is in that awkward spot where I'm not sure I really want to pay that much more for an SSD over an HDD.

series_x_load_revise.001-1440x1080.jpeg

The performance differential between RDR2 and Borderlands 2 in terms of One X SSD vs Series X NVMe demonstrates how situational loading improvements will be. There is no getting around that some games are not going to be boosted that much, so I/O is not the barrier in these cases. I don't expect PS5 to be any different. I would really like to understand how RDR2 is building it's world because my reading of these tests is that it's neither particularly I/O or CPU bound. :???:

edit: neither, not either!
 
Last edited by a moderator:
The results are predictably unpredictable! This is pretty much what I expected bit the external SSD vs HDD performance is in that awkward spot where I'm not sure I really want to pay that much more for an SSD over an HDD.
It's not showing difference between external usb ssd compared to hdd for xsx.

Or are you talking about the custom cartridge which will perform same as internal?

Hooking up ssd usb drive i would expect closer to the nvme than the hdd.

Will repeat myself, shame doesn't have usb c, and also gen 2 usb.
 
Ars Technica also did some testing. The results are predictably unpredictable! This is pretty much what I expected bit the external SSD vs HDD performance is in that awkward spot where I'm not sure I really want to pay that much more for an SSD over an HDD.

series_x_load_revise.001-1440x1080.jpeg

The performance differential between RDR2 and Borderlands 2 in terms of One X SSD vs Series X NVMe demonstrates how situational loading improvements will be. There is no getting around that some games are not going to be boosted that much, so I/O is not the barrier in these cases. I don't expect PS5 to be any different. I would really like to understand how RDR2 is building it's world because my reading of these tests is that it's either particularly I/O or CPU bound. :???:
yeah, what many people don't get, that for current gen games IO is not really the biggest limit all the time. Most times it is the management of the data (decompression, algorithms to create things, ...) is the bottleneck there. IO only has an effect up to a specific level and from there it is pure CPU & RAM performance that limits.

Well, the hardware decompression will help, but also only up to a specific level. Some things still need to be done by the CPU ^^ (as some people especially on neogaf don't seem to know).

But what this shows, there is at least one bottleneck less, now :)
 
yeah, what many people don't get, that for current gen games IO is not really the biggest limit all the time. Most times it is the management of the data (decompression, algorithms to create things, ...) is the bottleneck there. IO only has an effect up to a specific level and from there it is pure CPU & RAM performance that limits.

Well, the hardware decompression will help, but also only up to a specific level. Some things still need to be done by the CPU ^^ (as some people especially on neogaf don't seem to know).

But what this shows, there is at least one bottleneck less, now :)
Oh was expecting the nvme to be a lot faster loading.
This is where initial comparisons to PS5 will be favourable.
What happened to it being so much faster, when most of the games load in around the same time.
This will be the case for BC comparisons, and possibly launch titles etc
 
Oh was expecting the nvme to be a lot faster loading.
This is where initial comparisons to PS5 will be favourable.
What happened to it being so much faster, when most of the games load in around the same time.
This will be the case for BC comparisons, and possibly launch titles etc

The nvme can't load faster. It is not the bottleneck. Other tests (conducted by Jeff Grubb e.g. video a few posts above ) even the normal external SSD or an external nvme aren't that much slower. It is only .x seconds slower. So not even USB 3.1 (3.2) is limiting here. I guess if you connect a much slower SSD (like 100MB/s) you would start to see that the internal SSD is still just a bit faster but than in a few seconds range. I really hope someone makes such a test with slower external SSDs.
And yes, those results are only valid for BC games. But PS5 won't change those numbers.
I expect even slower loading times on PS5 because the CPU is clocked a bit lower but for BC titles sony wants to use profiles with lower frequencies (+ optional boost mode). As both current gen consoles are jaguar based the compression algorithms should be quite the same on both. So multiplatform games should use the same compression techs (whatever the developer decided would be "optimal" in the situation).

But well, those are just BC games. Patched games or next-gen games might use completely other loading strategies. Only with those new loading strategies, the PS5 might have a loading-time edge. But it really depends if the already high IO throughput will be a bottleneck at any point.

I would be totally ok with 10s loading screens, even in new games. The stuff made in Ratched & Clank is somehow cool (but I doubt they really need that much IO throughput) but I really can't see how this would change a whole game. From time to time this is impressive, but if your whole game would be based on this it would get really chaotic quite fast.
 
I was surprised that some games was still cpu limited.
I assume that it is running with SMT disabled, that's a lot of comparative horsepower.

Wonder if there's still tweeks to be made with the BC, wouldn't have thought so for XO games. But I guess even a small percentage improvement could make huge differenceces here.
 
I was surprised that some games was still cpu limited.
I assume that it is running with SMT disabled, that's a lot of comparative horsepower.

Wonder if there's still tweeks to be made with the BC, wouldn't have thought so for XO games. But I guess even a small percentage improvement could make huge differenceces here.
It is not only that. The code for the games is optimized for the small jaguar cores and it's instruction set. There may be things in the code that you would do in another way with another processor.
Without patches from developers, the new consoles can only brute force their way through those titles.
E.g. GCN optimized code might not be optimal for RDNA . There might be commands that need more cycles on RDNA than on GCN etc.
Than there is the bandwidth that hasn't been doubled. So double performance is not always given even if the GPU & CPU is more than 2x faster.

In some engines there is even the problem that algorithms work well for the desired target-framerate (e.g. 30fps) but get really inefficient for more than 30 frames.

What really boggles me is, that games with dynamic resolution have still a fixed upper resolution. They might now run in 4k60 but can't because the games caps it at 1744p or something like that. Is there a reason why those upper limits exist?
 
Without patches from developers, the new consoles can only brute force their way through those titles.
Yep, I thought would be enough to brute force it to be honest.
If nothing else through frequency that they run at.
To be fair we are only talking probably a handful of games here, but that is the point of the forum.

It will be interesting to see if there's any improvements to it as time goes on, or if they'll just consider it done and good enough which it probably is.
Some of those games a simple recompile probably would give a big boost.

Edit
What really boggles me is, that games with dynamic resolution have still a fixed upper resolution. They might now run in 4k60 but can't because the games caps it at 1744p or something like that. Is there a reason why those upper limits exist?
only game I can think of that doesn't have upper hard-coded limit is titanfall 2.
Probably has something inherently to do with the way DRS works, I believe it was covered in a DF video now that I'm thinking about it.
What I do have issues with is that it would be easy to just say, running on a xsx|s change upper limit to 4k and 1440p.
Hence why I think xbox should take the lead with doing basic updates at least, for all their games
 
Last edited:
I'm not sure I really want to pay that much more for an SSD over an HDD.

Make sure to read their update (posted in here earlier) that Ars has found external SSD on SeriesX shows nearly the same loading times as internal NVME on Series X for BC games.

Even in RDR2, the difference of HDD vs SSD (Nearly same speed as NVME) is 119 seconds vs 68 seconds on Series X. As to whether that's worth $120ish for a 1TB SSD versus a 4TB HDD or $180-200 for 8TB HDD, yeah, that's a bit of a tougher call.

If you don't already have mass storage for the console, I'd go for that first. Or if you're fine on level loads inbetween the game, then Quick Resume should have you covered on getting back into the game. Or if you're fine with the juggling to and from internal nvme with hdd of about 10 minutes for a 49 GB game then mass storage is the way to go.

If you already have mass storage, and plan to have a few BC games on your play rotation, then getting an external SSD isn't a bad idea if you really want to save a minute off initial loads and every level loading inbetween. Its either half the cost or double the storage of the official NVME Card.
 
yeah, what many people don't get, that for current gen games IO is not really the biggest limit all the time. Most times it is the management of the data (decompression, algorithms to create things, ...) is the bottleneck there. IO only has an effect up to a specific level and from there it is pure CPU & RAM performance that limits.
With RDR2 in particular, I'd expect there to be an I/O bottleneck and/or a CPU bottleneck in terms of loading/generation the world environment and both of these resources are massively boosted on Series X which makes the lack of improvement with RDR2 more pronounced.

I wonder if there is a lot of software-managed I/O, unpacking and decompression in RDR2. It's a real headscratcher.

Make sure to read their update (posted in here earlier) that Ars has found external SSD on SeriesX shows nearly the same loading times as internal NVME on Series X for BC games.
That update was the version of the article I read.
 
yeah, what many people don't get, that for current gen games IO is not really the biggest limit all the time. Most times it is the management of the data (decompression, algorithms to create things, ...) is the bottleneck there. IO only has an effect up to a specific level and from there it is pure CPU & RAM performance that limits.

Well, the hardware decompression will help, but also only up to a specific level. Some things still need to be done by the CPU ^^ (as some people especially on neogaf don't seem to know).

But what this shows, there is at least one bottleneck less, now :)

True, though there is a chance these things only take so long on the CPU because devs never bothered optimizing them since IO was a big bottleneck anyway. No point in making level set-up code run faster than it takes the HDD to feed such code the data.
 
Back
Top