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

[Shifty Geezer]

Are they really claiming FP16 numbers to get that 4x? Seems way underhanded.

Don't accuse them of being underhanded until we have details. Could mean 4x faster in raytraced scenes for all we know. No-one should point fingers. Just take this numbers under advisement and continue as usual, waiting on better details, discussing what they might or might not mean, without judgement.

 

[BRiT]

"Up to 4x" sounds like FP16 marketing speak.

Sounds like improvements from CPU and GPU combined.

 

[Shifty Geezer]

Except Xbox One X has the fastest drive performance of all current-gen consoles.

I mean in particular games. That is, games that might be super slow on all platforms that show great speed-ups on SSDs in general and PS5 and Scarlett in particular, giving a 40x speed up in Scarlett.

 

[BRiT]

I mean in particular games. That is, games that might be super slow on all platforms that show great speed-ups on SSDs in general and PS5 and Scarlett in particular, giving a 40x speed up in Scarlett.

Ah, yes, especially in situations where there is heavy decompression. Those poor little Jaguar and custom decoder cores just can't keep up.

 

[shiznit]

Is it possible to guess the gddr6 memory configuration from the pictures?

I think I count 12 chips. 384 bit bus would be awesome.

 

[Picao84]

Yup. And now we have this rumor....

https://twitter.com/x/status/1137833936682274816

It would be interesting if Scarlett would have an API, but PS5 would with a discrete GPU. Chances are probably close to none though.

 

[Jay]

Just watched vid and he says 4x processing power than 1X.
So I think it is as I said before, overall processing power. Last time I'm sure they said gpu power.

This time they are purposely, and for good reason being vague about it.

Hopefully dev kits are going out and people are willing to leak after E3.

 

[anexanhume]

Apple used FP16 to advertise that the Mac Pro GPU reaches 56TFLOPS. I thought it was pretty scummy

I thought they had two dual GPU card configs? That's 14TF Vega VII times 4. No deception there.

 

[Adonisds]

I thought they had two dual GPU card configs? That's 14TF Vega VII times 4. No deception there.

You're right, I'm wrong

 

[Jay]

Would be nice if they release a couple vids, not going into detail but showing what SS can do (bit like Sony vid), example of a scene on it, RTRT one also.
Was actually hoping to at least get that during show though.
I think numbers will probably leak before MS put it out there this time around.

 

[Silenti]

What was the die shot? I could swear it looked like the One X shot from a few years ago.

 

[Pixel]

Just watched vid and he says 4x processing power than 1X.
So I think it is as I said before, overall processing power. Last time I'm sure they said gpu power.

This time they are purposely, and for good reason being vague about it.

Hopefully dev kits are going out and people are willing to leak after E3.

Nah overall system performance can't be 4x, since the GPU represents the overwhelming majority of computational performance any gaming system nowdays. Ryzen2 will probably do somewhere around hundred of billion OPS in peak theoretical situations, GPU in Anaconda will likely be in the range of ten of trillion operations per second. You are talking about an order of magnitude difference of 1/100.

Of course it could be conflated oh this could mean CPU limited scenarios on XOX where the XOX cpu cores are at 100% utilization and the XOX gpu is only at ~50% utilization, then you'd get 4x the performance on Anaconda (assuming the double GPU performance ballpark is correct(I think it is) since they are moving onto Ryzen 2 which ceratainly could fall under the 4x the cpu performance of XOX's 2.1ghz mildly customized Jaguar cores.

 

[BRiT]

Nah overall system performance can't be 4x, since the GPU represents the overwhelming majority of computational performance any gaming system nowdays.

You haven't met the Jaguar cores then.

 

[Pixel]

You haven't met the Jaguar cores then.

I hadn't completed my post

 

[Deleted member 91098]

I think I count 12 chips. 384 bit bus would be awesome.

Yeah, me too and it seems to be HC-14 from Samsung which would mean 672 GB/s (unless underlocked).

 

[Jay]

Nah overall system performance can't be 4x, since the GPU represents the overwhelming majority of computational performance any gaming system nowdays. Ryzen2 will probably do somewhere around hundred of billion OPS in peak theoretical situations, GPU in Anaconda will likely be in the range of ten of trillion operations per second. You are talking about an order of magnitude difference of 1/100.

Of course it could be conflated oh this could mean CPU limited scenarios on XOX where the XOX cpu cores are at 100% utilization and the XOX gpu is only at ~50% utilization, then you'd get 4x the performance on Anaconda (assuming the double GPU performance ballpark is correct(I think it is) since they are moving onto Ryzen 2 which ceratainly could fall under the 4x the cpu performance of XOX's 2.1ghz mildly customized Jaguar cores.

Depends how you measure it, if you just add TF and Gflops together then your right the GPU dwaths the overall figure.
But you know that's not how a game works.
Gpu could be 2x
Cpu could be 8x
SSD 40x
But they could give them a weighting as just adding it together wouldn't make sense either.
Bit like a benchmark will give parts different weightings towards overall figure.
That's why giving overall figure like that is hard and I'd hoped they wouldn't do it.

 

[tunafish]

Where do they mention hardware RT?

From E3,
4x XB1X CPU and GPU.
40x XB1X drive performance.

At face value, that's kinda unbelievable. Sony's been talking about a 20x improvement and we've debated how realistic that is, so short of a battery backed up RAM cache, how is 40x possible? And 24 TFs for the GPU?? Only if it releases in 2025 (announced for 2020), or costs $2000.

Those numbers are...marketing. Some sort of arbitrary relative performance. We have quite a ways to go to determine what the real hardware is.

You can easily honestly say that the crappiest SATA SSD you can pick up from the corner CPU part store is more than 40x as fast as the HDDs from past gen, just not on every measure. I wouldn't presume to interpret that as "40 times on the most important measure".

Wow, it's like PS5 and Scarlett will basically be clones of each other.

Hardly surprising, given how there's only one supplier they both need to buy from. Especially given how the one who customized more at the beginning of last gen (XB1 and the ESRAM) ended up regretting it and doing away with the system on the mid-gen upgrade.

 

[MrFox]

They have about twice the memory bandwidth of xb1x.

I don't really see more than 10 chip (3 each side, 4 bottom, if all symmetrical), but it's probably 12x HC14.

 

[BRiT]

We'll have to wait to see how they compare.

Why, when we can continue this discussion in the entirely Baseless Thread? https://forum.beyond3d.com/threads/...-with-no-technical-merits-post-e3-2019.61246/

 

[Pixel]

Depends how you measure it, if you just add TF and Gflops together then your right the GPU dwaths the overall figure.
But you know that's not how a game works.
Gpu could be 2x
Cpu could be 8x
SSD 40x
But they could give them a weighting as just adding it together wouldn't make sense either.
Bit like a benchmark will give parts different weightings towards overall figure.
That's why giving overall figure like that is hard and I'd hoped they wouldn't do it.

True that is how marketing works. Another possibility others in the thread have mentioned is they are comparing theoretical peak 16bit float performance between XOX and Anaconda.