Velocity Architecture - Limited only by asset install sizes

[function]

Here's the exact quote:



I've highlighted a couple of bits that make me wonder if they were maybe using faster hardware in the early days? The interface is PCIe4 and in the timescales they are talking about the fastest PCIe4 drives were around 5 GB/s on a 4x interface which may have been simple off the shelf models plugged into PC's that were spec'd similarly to the final XSX hardware. So if he was testing Direct Storage on early hardware it's conceivable that he was seeing 5GB/s throughput in those early tests.

I can't see how it's possible on the current final hardware though. We know the interface is 2 lanes of PCIe 4.0.thanks to this confirmation from Seagate. That interface maxes out at around 3.75GB/s and so it's literally physically impossible to be getting 5GB/s from SSD to APU without using compression even if the drive itself wee capable of more than 2.4GB/s.

That's a good spot!

MS were targeting ~5GB/s (after decompression) based on their profiling, but the decompression block didn't even work properly until this year. So 'early on' it would probably make sense to attain that speed using raw transfers and a none final setup.

So basically a 5GB/s setup using all four lanes, swapped out for 2.4 + compression on later hardware.

 

[rntongo]

I think you've misunderstood what that link was saying. The Western Digital SN530 is a 2.4GB/s SSD that uses a PCIe 3.0 (4x) interface. The custom ASIC used on the Xbox drive merely allows it to use a PCIe 4.0 (2x) interface instead (likely to make future expansion cards easier to manufacture).

The drive speed is still the same at 2.4GB/s which is a factor of the controller and the memory speed. The controller is capable of 3.75GB/s if you pair it with the fastest compatible memory. The SN530 as uses in the Series X uses a slower tier of memory which gives you the 2.4GB/s. That's why Microsoft advertise it as such.

The 3.983 GB/s is merely the theoretical max throughput of a PCIe 3.0 4x or PCIe 4.0 2x interface. You're still limited by the controller on the XSX to 3.75GB/s and then limited again by the memory connected to that controller to 2.4GB/s.

you are 100% right!! I completely misread it.

 

[scently]

The interface maxes out at 3.983GB/s. The E-19 controller maxes out at 3.75GB/s. The custom E-19 controller in the Series X has an extra asic such that it maxes out at 3.983GB/s. So its a highly custom SSD not like others that share the same controller. And the technical director for Dirt 5 was very clear when he said,

"And that was without the compression in the hardware, that was just raw."

So whatever figures he gave, with the decompression hardware it would perform better.

Here's what I suspect, the SSD in the Series X can sustain 2.4GB/s constantly. But due to the custom controller and custom firmware it can go higher than this up to 3.983GB/s and maybe sometimes lower.

But again, it would have been good if he had been pressed about whether the data was decompressed first or just sent to RAM immediately for testing purposes. If it was using the CPU for decompression, then its even more impressive because the decompression block can definitely do better.

Thats why he gave the following statement:
"it may well be able to do way better than that."

Have you listened to the interview? Can you point me to the exact point in the video where the Dirt 5 dev made the statement that it this is without the decompression block? Thanks.

 

[PSman1700]

I think you've misunderstood what that link was saying. The Western Digital SN530 is a 2.4GB/s SSD that uses a PCIe 3.0 (4x) interface. The custom ASIC used on the Xbox drive merely allows it to use a PCIe 4.0 (2x) interface instead (likely to make future expansion cards easier to manufacture).

The drive speed is still the same at 2.4GB/s which is a factor of the controller and the memory speed. The controller is capable of 3.75GB/s if you pair it with the fastest compatible memory. The SN530 as uses in the Series X uses a slower tier of memory which gives you the 2.4GB/s. That's why Microsoft advertise it as such.

The 3.983 GB/s is merely the theoretical max throughput of a PCIe 3.0 4x or PCIe 4.0 2x interface. You're still limited by the controller on the XSX to 3.75GB/s and then limited again by the memory connected to that controller to 2.4GB/s.

Yes otherwise it would be 7GB/s before decompression even comes into play.

 

[turkey]

Have you listened to the interview? Can you point me to the exact point in the video where the Dirt 5 dev made the statement that it this is without the decompression block? Thanks.

Not sure why but I trawled through for the "forum" aka racing BRiT

1:hr24m

 

[rntongo]

That's a good spot!

MS were targeting ~5GB/s (after decompression) based on their profiling, but the decompression block didn't even work properly until this year. So 'early on' it would probably make sense to attain that speed using raw transfers and a none final setup.

So basically a 5GB/s setup using all four lanes, swapped out for 2.4 + compression on later hardware.

Again, you're absolutely right and I misread it. Otherwise it's still unclear how this 10GB in 2 seconds could be achieved. I remember seeing a linkedin screen grab of someone that had worked on Project Scarlett and the SSD was x4 lanes IIRC but then it later came out to be x2 lanes. So your hypothesis is possible that early devkits had faster SSDs. But that would mean the dirt 5 dev was misleading. If true he should have stated that clearly. Either way he claims the system can do more than 10GB in 2 seconds which is good because the chances of filling up 10GB in less than 2 seconds can mean having a larger active set of RAM just like Cerny showed for the PS5.

 

[scently]

Not sure why but I trawled through for the "forum" aka racing BRiT

1:24

Huh. Thanks. I heard the part where and when he said the hardware was capable of 10gb/2s and I thought that was what was referred to in the article and he never said anything about the decompression block. But having listened to the part you noted he actually made that statement.
I am of the mind to believe him. The reason being he is not just a "dev" with that annotation attached to him just because he works in the game industry. He is an actual dev. A technical director who has had hands-on with the hardware early on so he knows what he is talking about. Also, one of the Xbox hardware dev did state that the 2.4gb/s was conservative and it could go higher than that but they just want to state the sustained speed.

 

[pjbliverpool]

Also, one of the Xbox hardware dev did state that the 2.4gb/s was conservative and it could go higher than that but they just want to state the sustained speed.

How do you explain that this claim would see the Series X significantly exceeding the theoretical limit of the PCIe 4.0 2x link which we know connects the SSD to the APU?

 

[Deleted member 11852]

How do you explain that this claim would see the Series X significantly exceeding the theoretical limit of the PCIe 4.0 2x link which we know connects the SSD to the APU?

Yeah. It's interesting because having made the 10GB/sec statement, one of the other guys on the stream challenges him on why Microsoft are marketing it at half that and he said something odd like you shouldn't be loading 1 texture at a time, you should be loading 500, which doesn't answer the question at all. I think he's just made a mistake, maybe he is mis-remembering the exact figure, maybe he's mixing up GB and Gb or maybe the between his "early testing" and production something changed. ¯\_(ツ)_/¯

It's possible that Microsoft have something additional on the I/O controller side that can boost bandwidth in certain situations but given how open Microsoft are when discussing their hardware, it is odd that they would not have mentioned it. Series X's system architect disused the NVME and I/O solution with Digital Foundry and there is no suggestion of anything more.

 

[Johnny Awesome]

No secret sauce. Xbox SSD is 2.4 GB/s. Period. SFS is going to help make up the difference, but Sony will always have the edge on I/O speed IMO.

Despite being a proud XSX owner. I have to stay true to my belief that neither company has secret sauce.

That's not to say that tools and dev skill won't greatly improve. They will.

 

[Deleted member 11852]

That's not to say that tools and dev skill won't greatly improve. They will.

Perhaps not so much skill as experience. This console gen will represent a paradigm shift for devs in rethinking data; formatting data, storing data, and loading data - if they want to minimise load times as much as possible. It's not as easy as it sounds, hence why its around six seconds to get into Spider-Man but 20+ seconds to get into Demon Souls, a PS5 exclusive game. Decades of conventional wisdom needs throwing out. Hopefully Insomniac will share their experiences at GDC, as will other devs when they begin to wrestle with this.

 

[John Norum]

Yeah. It's interesting because having made the 10GB/sec statement, one of the other guys on the stream challenges him on why Microsoft are marketing it at half that and he said something odd like you shouldn't be loading 1 texture at a time, you should be loading 500, which doesn't answer the question at all. I think he's just made a mistake, maybe he is mis-remembering the exact figure, maybe he's mixing up GB and Gb or maybe the between his "early testing" and production something changed. ¯\_(ツ)_/¯

It's possible that Microsoft have something additional on the I/O controller side that can boost bandwidth in certain situations but given how open Microsoft are when discussing their hardware, it is odd that they would not have mentioned it. Series X's system architect disused the NVME and I/O solution with Digital Foundry and there is no suggestion of anything more.

simply we have not access to real HW data, there could be some customizations on the I/O that we don't know about, I just believe the developer, he is a trusted actual dev working on the hardware, used to talk about data numbers, no need to think he was wrong, I think this is not the case. At the end of the day it is capable of 10 GB tranfert in 2 seconds
Anyway game loading time is not double than ps5, are very comparable, in some cases slower, in some cases even faster, so there's something that we're missing

 

[function]

I'm sure the XSX and XSX are able to read 10GB / 2s because they have a PCI-E 4 interface and the IO stack is streamlined. That doesn't mean that the final consumer units will have SSDs in that can provide that.

The early dev kits he specified didn't have a properly functional hardware decompression block (that came in 2020), so speeds that this early kit hit raw, later kits were able to effectively reach using HW decompression and cheaper, less power hungry drives.

MS could probably put faster drives in the XSX later when they're cheaper and cooler (even the current controller can support faster speeds), but the baseline would always have to be what we have now, and that wouldn't make the IO block and its decompression hardware any faster.

I think the setup has been designed with fast movement of a camera through a detailed gameworld, where the game is ideally using SFS, very much in mind. I expect MS will have done their homework, and that it'll hold up well to this. Most of the time, during gameplay, I think the bottleneck will be somewhere else.

 

[PSman1700]

Was expecting no load times at all after all the hype, or maybe a second or two. The biggest game changer was going to be the CPU DF thought.

 

[Deleted member 11852]

simply we have not access to real HW data, there could be some customizations on the I/O that we don't know about, I just believe the developer, he is a trusted actual dev working on the hardware, used to talk about data numbers, no need to think he was wrong, I think this is not the case.

Who to believe and who to think maybe got it wrong? Game dev or Microsoft technical fellow and Xbox system architect Andrew Goossen?

Hmmm.. tough one.

 

[Metal_Spirit]

Who to believe and who to think maybe got it wrong? Game dev or Microsoft technical fellow and Xbox system architect Andrew Goossen?

Hmmm.. tough one.

I cannot say this for sure, but I'm guessing the Dev made a mistake somewhere.
The SSD has a sustained speed of 2.5 GB/s. Shure, it may go higher on peak transfers, but you cannot count on that.
The controller also claims a maximum of 3.8 GB/s. So if this Dev managed to find a way to break the rules of phisics, he may as well share it with the world, because I would like to know how much a 5.5GB/s sustained SSD can transfer, although I'm guessing 22 GB in 2 seconds using the same math.

 

[pjbliverpool]

I cannot say this for sure, but I'm guessing the Dev made a mistake somewhere.
The SSD has a sustained speed of 2.5 GB/s. Shure, it may go higher on peak transfers, but you cannot count on that.

It's actually a peak transfer speed of 2.4GB/s. With very well managed data it may be able to hit and even sustain that at times, but it's never going to go above it without decompression. Not that that's anything to worry about. That's an insane level of real world speed compared to everything we've seen to date.

 

[AlphaWolf]

It's actually a peak transfer speed of 2.4GB/s. With very well managed data it may be able to hit and even sustain that at times, but it's never going to go above it without decompression. Not that that's anything to worry about. That's an insane level of real world speed compared to everything we've seen to date.

No. Not peak. That's a claimed sustained value.

 

[pjbliverpool]

No. Not peak. That's a claimed sustained value.

Yes and under ideal circumstances the peak can presumably be sustained. The reality is that the peak is determined by the controller capabilities and memory speed though, and Microsoft have based their specification on that (as have Sony).

Let's say the XSX SSD really was operating with the fastest, most expensive memory that it's Phison E19T controller was able to handle meaning that it could transfer 3.75GB/s or 7.5GB/s with BCPACK compression. Do you seriously think that Microsofts multimillion dollar marketing machine wouldn't pick up on that fact and use it? And that they'd rather use the estimated real world throughput when Sony are by contrast using the theoretical maximum throughput?

 

[BRiT]

Or maybe just maybe they went with sustained speed because they know its a major issue with throttling for PC NVME.