Next Generation Hardware Speculation with a Technical Spin [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
4x more powerful than XBX is a combination of SSD + CPU + GPU + ram. Immediacy of gaming (shorter loadings) is now their top priority.

it's the combination of speed, not just of the SSD but of the processor, the performance of the GPU and RAM...You're playing then suddenly *bloop* a load screen pops up and drops you out. Our goal is to get rid of those things, that's what we're after.

https://www.eurogamer.net/articles/...cross-gen-halo-infinite-and-the-lack-of-fable

Seems to me their marketing strategy will be very different than XBX's 'most powerful console'. That and their conference was basically 'Games, games, games'. This is very good ambition from Microsoft and even if they didn't show much gameplay, I quite like their vision for next gen.
 
One-X has not double rate FP16... So on FP16 they have 24 TF and 12 TF on FP32... Should be this... So still believe Sony is at max 10 TF (FP32)
 
It's marketing. Numbers should never be taken at face value, especially in great big advertisements like E3. Use the biggest true number you have available rather than a realistic benchmarking for a fair comparison. It's just the way it is and no-one should be surprised or irked by it. Just ensure when you get a number, you run it through appropriate reality-checks before accepting it.

This is very true. Regular people tend to like to look at top end numbers since big numbers are great. Take GPUs for instance where talk often centers around the maximum theoretical TFLOPs which often aren't achievable in real life scenarios.

However, consider a moment if MS is talking about the improvements at the lower end of the spectrum. Basically the bottlenecks. There's areas for dramatic performance improvements in the areas which are the most restrictive on performance.

If 40x represents the removal of many of the bottlenecks that restrict overall storage performance that's a significantly bigger win than if 40x was in reference to the average performance of the storage subsystem or the maximum performance of the subsystem.

In that sense the numbers become a bit more believable. Various bottlenecks in the whole storage system pipeline is what makes storage subsystems feel slower than you would expect (especially PCIE SSDs on PC which can hit ridiculously high transfer rates with ridiculously low latencies, yet real world performance doesn't come close to those lofty numbers). If you remove those bottlenecks I could see a potential for a 40x real performance increase versus a 40x stat sheet number increase.

That said, it's still PR, so skepticism is always warranted. I'd like to have more details. But at least it appears that like Sony, Microsoft took a good long look at the storage subsystem (hardware and software) to see how they could improve worst case performance which in turn would increase real world performance.

Regards,
SB
 
40x HDD could be as simple as 100MB/s Vs 4GB/s, or somewhere thereabouts. 4GB/s or higher won't be an issue for PCIe 4 ssds. 40x faster than even a 150 MB/s linear read is entirely feasible.

4x X1X could include the calculating performance of the ray tracing hardware acceleration. If ray tracing is going to be a big thing next gen it's fair to talk about that as you'd be running it via compute otherwise.

I need to look at the relative performance of rtx and gtx Turing when running ray tracing enabled games and demos.

Will be interesting to see if this makes Sony confirm or not confirm 'hardware' RT.
 
In that sense the numbers become a bit more believable. Various bottlenecks in the whole storage system pipeline is what makes storage subsystems feel slower than you would expect (especially PCIE SSDs on PC which can hit ridiculously high transfer rates with ridiculously low latencies, yet real world performance doesn't come close to those lofty numbers). If you remove those bottlenecks I could see a potential for a 40x real performance increase versus a 40x stat sheet number increase.
The main problem there being Sony are doing exactly the same thing and are talking about a 20x speed up. 40x just seems like one-upmanship.

"How much faster are we?"
"A phenomenal twenty times!"
"Crap. Sony already said that. Can't we get faster?"
"Well, we sped up this particular part of the IO process by forty times, although it only contributes to 3% of the total access speed."
"Forty times faster it is!!"

:p
 
40x HDD could be as simple as 100MB/s Vs 4GB/s, or somewhere thereabouts. 4GB/s or higher won't be an issue for PCIe 4 ssds. 40x faster than even a 150 MB/s linear read is entirely feasible.

4x X1X could include the calculating performance of the ray tracing hardware acceleration. If ray tracing is going to be a big thing next gen it's fair to talk about that as you'd be running it via compute otherwise.

I need to look at the relative performance of rtx and gtx Turing when running ray tracing enabled games and demos.

Will be interesting to see if this makes Sony confirm or not confirm 'hardware' RT.
He did say it...

If you let me back I'll be here. Back to Scarlett... you mentioned it has four times the power of Xbox One X, which certainly sounds good. But what does that mean?

Matt Booty: It's a few things - it's the combination of speed, not just of the SSD but of the processor, the performance of the GPU and RAM, but we're also in a world where speed is starting not to matter. You can make RAM faster either by speeding up the way you access it or by adding more access points. Just think, what are all the things right now which take you out of a game? You're playing then suddenly *bloop* a load screen pops up and drops you out. Our goal is to get rid of those things, that's what we're after.

So it's not just 4x GPU.
 
The main problem there being Sony are doing exactly the same thing and are talking about a 20x speed up. 40x just seems like one-upmanship.
It's regrettable for them that they pulled a 'me too', on a stage as large as this. They could have left the me-too multiplicative business for after E3.

Luckily they did concede that they were using Solid sate as virtual ram of the sorts?
Honestly it was too vague AF for me to clearly grasp which bush they were trying to beat around.

Seems like MS did not want to give away anymore than Sony has.
An interesting note for me to tuck away later.
 
The main problem there being Sony are doing exactly the same thing and are talking about a 20x speed up. 40x just seems like one-upmanship.

"How much faster are we?"
"A phenomenal twenty times!"
"Crap. Sony already said that. Can't we get faster?"
"Well, we sped up this particular part of the IO process by forty times, although it only contributes to 3% of the total access speed."
"Forty times faster it is!!"

:p

Sony are very good with hardware but they don't have nearly as much experience as MS has with software, drivers and multi-use storage throughput experience (from data center storage servers all the way down to low power tablets and all the cases in between). Software, firmware and hardware have to be optimized differently for each use case. A drive optimized to be super fast in a data center is often only average or less than average in a Home PC and vice versa, for example. I could see Sony focusing so much on the hardware that they miss areas of optimization within the software and driver stack that MS likely wouldn't miss.

But again. We're over a year away from release of both consoles (most likely), so it's all PR from both companies at this point. We have no idea what they are really benchmarking when they make performance claims, for example.

Basically trying to compare what each company is saying and determine something meaningful from that is useless.

Regards,
SB
 
40x HDD could be as simple as 100MB/s Vs 4GB/s, or somewhere thereabouts. 4GB/s or higher won't be an issue for PCIe 4 ssds. 40x faster than even a 150 MB/s linear read is entirely feasible.

4x X1X could include the calculating performance of the ray tracing hardware acceleration. If ray tracing is going to be a big thing next gen it's fair to talk about that as you'd be running it via compute otherwise.

I need to look at the relative performance of rtx and gtx Turing when running ray tracing enabled games and demos.

Will be interesting to see if this makes Sony confirm or not confirm 'hardware' RT.

I think the key difference is that Sony showed a real world demo. Microsoft quoted a maximum theoretical speedup irrespective of the other limitations and bottlenecks present within the system.
 
What is better: really 20x faster loadings (demoed in a real game) or theoretically 40x faster loadings?

Remember that they also claimed, using their own logic, that Scarlet was theoretically 4x more powerful than XBX. What internal logic are they using for that 40x claim ?
 
IT'S official, real time hardware ray tracing in the next Xbox.


Going back to this after finally making it through the thread.

This makes me feel even more that it's possible that Microsoft approached both NV and AMD about RT support in their GPUs at roughly the same time and that they both started work on it at around the same time. NV delivered "on time" (or at least faster) with Turing while AMD were late with Navi. This is assuming of course that the hardware RT support is in the Navi GPU cores and not something else that was added to the SOC.

The first roadmaps that had Navi on them had it launching in a similar time frame to Turing.

Regards,
SB
 
What is better: really 20x faster loadings (demoed in a real game) or theoretically 40x faster loadings?

Remember that they also claimed, using their own logic, that Scarlet was theoretically 4x more powerful than XBX. What internal logic are they using for that 40x claim ?
i guess probably 40x theoretical. I mean, even at 50% of the claim, you're then back to PS5 levels. So any % over 50% i guess would be better, I mean if you are speaking empirically, even if they managed 40x faster loading, the difference between what PS5 and Nextbox will still be miniscule. 2seconds vs 1seconds etc.

And whenever ray tracing is involved; it's easy for the performance differentials to amp up quickly.
 
I'm super-excited to see how not just an SSD being added improves things but how tailoring the I/O and overall architecture (along with superior compression<>decompression) to minimise latency and maximise throughput really bolsters these consoles. Removing this bottleneck and making consoles so much more cohesive is no doubt going to have massive implications not only for loading but for gameplay possibilities and the effectiveness of the CPU/GPU/RAM etc. in rendering and simulating worlds.

With the two main players committing to SSD's, Custom I/O, RT (hopefully both hardware-accelerated) and combining that with what will no doubt be at least a solid bump in CPU/GPU/RAM, I think the industry as a whole is going to get a pretty revolutionary technical push forward as opposed to the more evolutionary graphics-only push we saw this gen.

I think the right choices are being made.
 
What is better: really 20x faster loadings (demoed in a real game) or theoretically 40x faster loadings?
I wouldn't know because still not that many data points.
Maybe the 40x is inflated, maybe it's not.
Maybe was just one upping Sony, maybe not.
Fact is we don't know yet, so be skeptical but I wouldn't relegate it to be less/worse to Sony's implementation.

What if MS had gone first and said 40x?
We need dev docs to be leaked.
 
Add read and write speedups together!
lol as much as I want to rag on MS here because of their disaster with XBO launch. They delivered mostly every checkbox on X1X. It's still delivering quite well as of this moment.

I believe their goal back then was to make a 4K title for any XBO title. And they managed that mainly with some. It sounds like 4K60fps is the next goal here and fast loading and something something streaming.

So if we judge by their past performance I think even though MS claims are un-intelligible, I wouldn't necessarily correlate that with being just complete marketing gibberish. Most of Scorpio reveal 18 months in advance was marketing gibberish. Most people did not think it would be possible for them to hit native 4K or ~4k in general. Forums were largely adamant that the claims were entirely full of it. And they delivered for the most part. Very few titles well under the ~UHD area.
 
I could see Sony focusing so much on the hardware that they miss areas of optimization within the software and driver stack that MS likely wouldn't miss.
Although Cerny explicitly talked about the software stack, so they gave it some jolly good consideration. I think it's a bit of a long shot that MS managed to get the same thing running twice as fast. How much faster is XBox at loading than PlayStation? As far as I know, they not, and Sony, with their inferior experience of OSes at all levels, have a much faster machine when it comes to booting up and hibernating. I'm not convinced MS's expertise is likely to net them any advantages, and I doubt Sony will have missed something capable of slowing them down 2x if they really investigated. We'll see!

But again. We're over a year away from release of both consoles (most likely), so it's all PR from both companies at this point. We have no idea what they are really benchmarking when they make performance claims, for example.

Basically trying to compare what each company is saying and determine something meaningful from that is useless.
Definitely. It'd just be nice if they could distance themselves from the one-upmanship and talk more about vision and plans without the need to through competitive numbers around.
 
Status
Not open for further replies.
Back
Top