Xbox Series S [XBSS] [Release November 10 2020]

The SeriesS definitely won't run OneX games, not because it has lower bandwidth (RDNA2 is more effective than Polaris at using bandwidth), but because it's lacking 2GB of RAM.
It's just not possible for the SeriesS to run OneX code.

Not all Xbox One X games use the 9 GB. Most of the games have the same textures as PS4 Pro, and that console only has one extra GB.
 
Last edited:
It cannot BC the X1X titles due to lack of memory AFAIUnderstand.

it is S -> S
And X -> X
To be clear i expect it to be as you say.

But I've recently theorized on if it could do it even with less memory.
I suspect it could for grames that rely on streaming.
 
Didn't see the advert.
Up until now they have been firmly against selling the digital only console like the 1Sad
It's in a banner on the main page. This isn't really the same though, you can't buy the XSX digital edition, this is a totally different machine with a totally different price.
 
The thing with not having 12GB in a 2*2GB + 2*4GB configuration is that the SeriesS will definitely not be running One X versions of Xbox One games.
BC on the SeriesS will be a very strange thing, or it'll simply be running OneS code with higher framerate and dynamic resolution (but only where the games / game engines allow it to).
Of course it won't be running One X versions, it doesn't support 4k gaming...

4k textures were almost entirely the difference between One S and One X, some games had the option to turn off 4k in favor of framerate, but that was about it. MS has already said they're doing things to enable faster frames on BC titles, so presumably that would let the XSS hit 60/120fps for titles that are compatible with that. After which there's no value in the One X "versions".
 
To be clear i expect it to be as you say.

But I've recently theorized on if it could do it even with less memory.
I suspect it could for grames that rely on streaming.
I do expect them to do some magic on the 1S versions of the games, higher resolution, double framerates, much like they plan for XSX to a degree.
It's in a banner on the main page. This isn't really the same though, you can't buy the XSX digital edition, this is a totally different machine with a totally different price.
I was talking about the XSS which is a digital only machine
Anyway , not a big deal, just surprised.
..
 
Small
https://www.resetera.com/threads/xbox-series-x-and-xbox-series-s-model-unboxing.283934/post-44918879

118988589_364392684586871_5604498251794812379_n.jpg


119049322_810323703040075_1555836848998557716_n.jpg
 
There's really no reason why a console launching in late 2020 with a 4TFLOPs GPU without an optical drive nor a 2.5" SATA HDD wouldn't be pretty small.

Not all Xbox series X games use the 9 GB. Most of the games have the same textures as PS4 Pro, and that console only has one extra GB.
You mean the One X, right?
You think game developers would leave free RAM completely unused in the One X? That would be a pretty bad use of resources.
If you have free memory, you'll at least load part of the assets needed for the next area / level so you can reduce loading times.

So unless Microsoft is actively forcing all game devs that launched a OneX patched game to patch it again for the Series S, I see no way how the SeriesS will be running OneX code.
They could maybe go around the bandwidth deficit with GPU optimizations and getting maybe more GPU time and compensating on less CPU time, but not on having a whole 2GB less memory.
 
I love this little guy. As an engineer I have a thing for small, efficient devices. The idea of dramatically lowering resource and cooling costs by targeting a 1080p (native) resolution is incredibly appealing.

I do worry about whether its ray tracing prowess will be able to keep up with its older sibling, even with the reduced resolution. Ray/box and ray/triangle intersections are largely independent of target resolution, correct? If so, they will consume a much larger fraction of the available hardware resources (flops, bandwidth, and whatever fixed-function RT resources exist) on the XSS relative to the XSX. Thoughts?
the console word and concept makes a LOT more sense with Series S. This could be MS more successful console ever if they did things well -which they didnt- :D , but it is the only console that could appeal to me this generation and I am not sure.

My (distant) future PC is going to be one in a micro case, for sure.
 
They could maybe go around the bandwidth deficit with GPU optimizations and getting maybe more GPU time and compensating on less CPU time, but not on having a whole 2GB less memory.

To be pedantic, it's at most 1.5 GB less if it's 9 GB and 7.5 GB. Right?
 
Looking at the hands on, can't help but wish the height was a little bit smaller than the OneS. The black base on the OneS has a slimming effect. The XSS will probably look a bit fatter in my unit, as the reduced depth is hidden.

(Although still not sure if I'm preordering the S|X|neither. Have a week decide and work on a wife credit)
 
To be pedantic, it's at most 1.5 GB less if it's 9 GB and 7.5 GB. Right?
I don't know how much OS overhead there is on the SeriesS and the OneX. Are they different?
But even if the difference is 1.5GB, it won't change the fact that the OneX ports are definitely making use of those 1.5GB at all times, even if it's just caching stuff that isn't present on the current level / location.
 
I don't know how much OS overhead there is on the SeriesS and the OneX. Are they different?
But even if the difference is 1.5GB, it won't change the fact that the OneX ports are definitely making use of those 1.5GB at all times, even if it's just caching stuff that isn't present on the current level / location.

Like I said, being pedantic. Series X OS overhead was listed at 2.5 GB, so for worst case for Series S if nothing scales down it's 10 - 2.5 = 7.5 GB for games. So the difference is at most 1.5GB.
 
If I understand what you are asking, how can you saturate a 224 GB/s pathway if you RAM is only 10GB? That's easy. You aren't reading the entirety of you RAM every frame, lots of data is sort of just there being used as needed. The main part of 3d rendering, at least when it comes to bandwidth isn't about what you've got stored in ram, it's writing and reading render targets. And that bus isn't 100% efficient. And the bandwidth usage scales linearly with resolution and framerate. IIRC a 1080p framebuffer at 32bit color and 32bit Zbuffer require the GPU to write about half a gig a second at 60fps. HDR, higher resolutions, and higher framerates are all going to multiply this. And that doesn't account for games that do funky things like rendering Z only before rerendering the scene, or full resolution render to texture, or whatever the requirements of raytracing are. And this all in contention with reading textures and other game assets, and on console there is contention with what the CPU is doing. And all of this talk this generation about loading assets from SSD just in time is going to play a role here, too.

Random fact, the Xbox 360's eDRAM was only 10MB and it has 256 GB/s bandwidth, and it was exclusive to the GPU. More than 1000 times smaller and it still had more bandwidth.

I feel like I'm falling on deaf ears.
If there is no impact and doing this was fine, we'd see a lot more of gtX550 situations and Microsoft would see no reason to explicitly say this at all.
If this is fine there would be simply four banks at 56GB/s each and we would just add them. the fact that they didn't tells.

Hypothetical situation:

chips A B C D
A is 4GB, other three are 2GB.

Situation A: I left 1.25GB in each of A,B,C, and D.
That's 5GB of data I'm trying to read across 4 chips.
time used would be 1.25 units, and my effective bandwidth is 4GB/unit of time

Situation B: I left 2G of data in A, and 1GB in B,C,and D.
That's ALSO 5GB of data across 4 chips I'm trying to read
Time used would be 2 units, and my effective bandwidth is 2.5GB/unit of time

My effective bandwidth here is 62.5% of what I would have in situation A normally.

Of course you can say that I can find work for the other chips to do while I read A but that's, again, explicit planning you'd have to do.

Of course the best solution is to avoid 2GB at all costs.

Any utilization of the 2G at the same time of the 8G will lead to bandwidth contention because physically the channels for the B C D chips won't be able to access chip A, and by creating an unbalanced workload across 4 chips will create extra inefficiency on top of the usual situation.
 
Last edited:
I feel like I'm falling on deaf ears.
If there is no impact and doing this was fine, we'd see a lot more of gtX550 situations and Microsoft would see no reason to explicitly say this at all.
If this is fine there would be simply four banks at 56GB/s each and we would just add them. the fact that they didn't tells.

Hypothetical situation:

chips A B C D
A is 4GB, other three are 2GB.

Situation A: I left 1.25GB in each of A,B,C, and D.
That's 5GB of data I'm trying to read across 4 chips.
time used would be 1.25 units, and my effective bandwidth is 4GB/unit of time

Situation B: I left 2G of data in A, and 1GB in B,C,and D.
That's ALSO 5GB of data across 4 chips I'm trying to read
Time used would be 2 units, and my effective bandwidth is 2.5GB/unit of time

My effective bandwidth here is 62.5% of what I would have in situation A normally.

Of course you can say that I can find work for the other chips to do while I read A but that's, again, explicit planning you'd have to do.

Of course the best solution is to avoid 2GB at all costs.

Any utilization of the 2G at the same time of the 8G will lead to bandwidth contention because physically the channels for the B C D chips won't be able to access chip A, and by creating an unbalanced workload across 4 chips will create extra inefficiency on top of the usual situation.
I think often in this discussion about the pool split, there is mix-up between memory contention and memory access. Memory contention will always occur whether the pool is split or not. That part is unavoidable.
The CPU has access to all 4 chips
the GPU has access to all 4 chips
And the CPU will always have priority over bandwidth.

That maximum bandwidth number will never be 100% because contention will cause bandwidth to drop.

Nor does not mean every single memory location needs to be interleaved over 4 chips. Some of those data will fit onto a single cache line pull of 1 chip.

people need to look at data as being, not continuous. But instead a lot of small fragments making a whole.

Understanding this point, that when the CPU needs stuff, it will access 1 chip for instance, is no different if it's monopolizing all chips, or just the 1 chip.

There is 2GB that can only be accessed on its own, that's where you place memory that is very small in nature that doesn't need to be interleaved over many chips.

This isn't the same as the GPU having it's pool split. Where you are splitting the buffer work over as many memory chips as possible and this remaining 0.5 GB is holding up things making the other chips useless. Because the GPU cannot proceed to the next step until the last step is complete.
 
Last edited:
This is pretty small
It's pretty thicc. As somebody who really loves the aesthetic of the white One S, the black circle/grill of Series S doesn't work for me. And now we have as bunch as stupid memes as we did for PS3.

I didn't think Series S was going to be for me, it's Series X or nothing. But the All Access deal is great.

Well played, Microsoft, well played! :yes:
 
Back
Top