Middle Generation Console Upgrade Discussion [Scorpio, 4Pro]

[iroboto]

Erm... which makes even less sense..?


Maybe it depends on what kind of developer we're talking about. From what I can gather, Thomas Mahler is an asset creator / 3d artist.

As I said, I think it may just derive from the fact that Scorpio is a much more revolutionary difference from the Xbone than the Pro is to the PS4.

but you're looking purely at power and TFLOPS, which are often bound by TDP.
I'm not sure if you are putting enough stock into the feature set difference that could be at play here. We've been on Shader model 5.1 as a standard for DX11 for 7 years now. Things may be relatively new on console, but not necessarily new at all for PC. I have good confidence that no GPU in release today is fully SM6.0 compliant; partially yes.

 

[mosen]

Does 52 CUs at 950 MHz = 6.3TF sound more likely than mixing up abbreviations?

I don't know, read it as you want.

Erm... which makes even less sense..?


Maybe it depends on what kind of developer we're talking about. From what I can gather, Thomas Mahler is an asset creator / 3d artist.

As I said, I think it may just derive from the fact that Scorpio is a much more revolutionary difference from the Xbone than the Pro is to the PS4.

Ok, saying that GCN4 has 64 ALUs makes no sense. Sorry, I didn't know that.

 

[Deleted member 13524]

but you're looking purely at power and TFLOPS, which are often bound by TDP.
I'm not sure if you are putting enough stock into the feature set difference that could be at play here. We've been on Shader model 5.1 as a standard for DX11 for 7 years now. Things may be relatively new on console, but not necessarily new at all for PC. I have good confidence that no GPU in release today is fully SM6.0 compliant; partially yes.

Like I said, he probably has no idea about which custom features are in Neo and how they compare to DX12's SM6 (if those actually are in Scorpio at all).
He's a 3D artist for a 1st-party Microsoft developer. Scorpio is a much bigger jump to Xbone, so he claimed the PS4 Pro is a half-assed implementation in comparison.
From a certain perspective, he's right...

Ok, saying that GCN4 has 64 ALUs makes no sense. Sorry, I didn't know that.

Assuming you're being sarcastic, what's the point of saying how many ALUs there are in a CU without saying how many CUs there are?

 

[mpg1]

52 CUs @ 950MHz = 6.3TFlops
52 CUs @ 1050MHz = 6.98TFlops

Which sort of fits with "6.3Tflops up to 7"...but doesn't fit with speculating up to 1200MHz...

 

[mosen]

Assuming you're being sarcastic, what's the point of saying how many ALUs there are in a CU without saying how many CUs there are?

What's the point of saying that Vega is GCN4?

 

[Shifty Geezer]

Well....either that person is just making shit up...or Microsoft is truly "going for it" . For Microsoft's sake I hope it is the latter.

What do you think are the advantages of going full fat next-gen with Scorpio versus a 4K optional XB1? Doesn't seem to me like consolers are champing at the bit for new machines. I suppose the plus side would be an early start on next-gen and an entrenched install base ahead of whatever Sony puts out, coupled with a lower price when competing head-to-head with a rival system that probably isn't significantly different in visuals.

 

[iroboto]

What do you think are the advantages of going full fat next-gen with Scorpio versus a 4K optional XB1? Doesn't seem to me like consolers are champing at the bit for new machines. I suppose the plus side would be an early start on next-gen and an entrenched install base ahead of whatever Sony puts out, coupled with a lower price when competing head-to-head with a rival system that probably isn't significantly different in visuals.

XBO was made with a different goal in mind. Pretty sure MS is doing their best to bind Scorpio with PC as much as possible, moving that baseline feature set up encourages consumers to switch to Win10. Write off XBO and start with new objectives.

 

[mpg1]

What do you think are the advantages of going full fat next-gen with Scorpio versus a 4K optional XB1? Doesn't seem to me like consolers are champing at the bit for new machines. I suppose the plus side would be an early start on next-gen and an entrenched install base ahead of whatever Sony puts out, coupled with a lower price when competing head-to-head with a rival system that probably isn't significantly different in visuals.

I don't see it as it either or...but both. A future proof/versatile console that can just act as an premium priced "4K optional XB1" at first but is still relatively powerful enough down the road to also act's as a "nex gen console" to run games after XB1/PS4 are dropped off. In other words...not the PS4 Pro.

 

[mosen]

I remembered that an Spanish website released the first rumors about Scorpio which was full of technical problems. Considering that new information are also from an Spanish site I tried to find the previous source. Then I found this Eurogamer Italy article and then this website:

In-Depht Xbox Scorpio: system memory and the operating system design

Take it as you want. It seems that new informations are taken from this site.

 

[Shifty Geezer]

But if it's not a clear next-gen at launch, it won't get next-gen sales. And then when a clearly next-gen console does launch, how does MS compete? Let's say we have 2 years of 4K XB1 titles on Scorpio and then PS5 appears with 12 TF and a clear generational advance on PS4. Is MS's message, "ah yes, but Scorpio can do that too. See, it's a next gen console, only it hasn't had any next gen games thus far because XB1 was holding it back."

The progressive platform is definitely a strong selling point, but probably not when a new, exciting next-gen launches. It's the sort of thing that needs to build up over time, if it can even happen. IMO. So I don't see the value in a next-gen power Scorpio that's not ushering in a next-gen of software. There is such a thing as being too early.

 

[iroboto]

But if it's not a clear next-gen at launch, it won't get next-gen sales. And then when a clearly next-gen console does launch, how does MS compete? Let's say we have 2 years of 4K XB1 titles on Scorpio and then PS5 appears with 12 TF and a clear generational advance on PS4. Is MS's message, "ah yes, but Scorpio can do that too. See, it's a next gen console, only it hasn't had any next gen games thus far because XB1 was holding it back."

The progressive platform is definitely a strong selling point, but probably not when a new, exciting next-gen launches. It's the sort of thing that needs to build up over time, if it can even happen. IMO. So I don't see the value in a next-gen power Scorpio that's not ushering in a next-gen of software. There is such a thing as being too early.

And you might be right, but developers that invest in the full 6.0 will have years of advantages over developers who stick to 5.1.

Imagine say 3-4 more years of this generation, that's a lot of time another developer will have a leg up on you in terms of engine design.
And we all know that engine design certainly takes much longer to develop than hardware power.
eg. look at how long it took for Tiled resources to show up in games. Look at the state of compute shaders vs what we have today, and so forth.

Setting that baseline is nice, because there will be feature parity if you decide to release another console with just increased power.
And, PC gamers will benefit from SM6.0 games as well. So I don't necessarily believe it's that niche if you combined the two populations together.

We've got CR and a bunch of other neat features that really represent the future of graphics. I think giving developers that head start now can't be negative.

 

[Shifty Geezer]

I wouldn't say it's a negative, but I can't see that being a positive for MS unless devs actually do that - Scorpio is allowed exclusives and sells gangbusters at launch, which is unlikely unless sold as a new generation machine. A game made in 2018 will target the full range of PC, Scorpio, PS4 Pro, PS4 and XB1. I'm guessing it'll be a very small proportion of devs who create next-gen engines for Scorpio and high-end PCs alongside their current gen games.

 

[iroboto]

I wouldn't say it's a negative, but I can't see that being a positive for MS unless devs actually do that - Scorpio is allowed exclusives and sells gangbusters at launch, which is unlikely unless sold as a new generation machine. A game made in 2018 will target the full range of PC, Scorpio, PS4 Pro, PS4 and XB1. I'm guessing it'll be a very small proportion of devs who create next-gen engines for Scorpio and high-end PCs alongside their current gen games.

It depends.
We've seen from history that even when all of the features are available, they only have enough time to implement so many at a time.

We still haven't implemented all the features for PS4 and XB1 yet.
And the same goes with this I think. Maxwells + Pascals are a pretty big population, add Vega and Scorpio and you have a reason to start working with some of those features.
According to steam survey last month:
72.24% of hardware was DX12
http://store.steampowered.com/hwsurvey/videocard/

900 and 1000 series are doing very well, like they make up close to 18% of all GPUs on steam.
And then you look at OS survey.
51.14% are W10 64
30.30% are W7 64

That's a lot of people you'd probably want to convert.
Anyway, I get the reason to be pessimistic about it, it's easy to associate everything with dreams and excitement, but I do see somewhat of a business case here.

While Sony is indeed Xbox's biggest threat, Vulcan is a threat to MS core OS business.
If I were MS, I'd try to make sure that everyone convert to W10 as soon as possible, and you need games with these high end features to convince people to move (first party games).

 

[mpg1]

But if it's not a clear next-gen at launch, it won't get next-gen sales. And then when a clearly next-gen console does launch, how does MS compete? Let's say we have 2 years of 4K XB1 titles on Scorpio and then PS5 appears with 12 TF and a clear generational advance on PS4. Is MS's message, "ah yes, but Scorpio can do that too. See, it's a next gen console, only it hasn't had any next gen games thus far because XB1 was holding it back."

The progressive platform is definitely a strong selling point, but probably not when a new, exciting next-gen launches. It's the sort of thing that needs to build up over time, if it can even happen. IMO. So I don't see the value in a next-gen power Scorpio that's not ushering in a next-gen of software. There is such a thing as being too early.

"exciting next-gen launches" days are over in my opinion. It's way to complex and expensive to develop modern AAA engines and games in general combined with diminishing returns from the power of the hardware required to run it to have these large leaps like we used to. Scorpio will blunt any "nex gen" feel PS5 will have unless Sony decides to wait many many years....

 

[turkey]

Prepare to smack me down as pure naive ramblings follow.

I find the multiple Tf count interesting as it fits with some earlier whimsical idea I got from when discussing the CPU and the TDP constraints.

What if

The apu has more potential TDP than it should have or at least the capacity and ability to scale both the CPU and GPU beyond the target envelope.
If Vega and Zen (this would fit for Jag but less beneficial ) are geared to higher frequency then they should have headroom in a console to run well at higher clocks without really loosing efficiency vs TDP. Now imagine like the optional CPU core Xbox one titles can request, the new SDK allows game makers to choose from various CPU / GPU clock speeds. Does your game require a bit more CPU then dock a bit off the GPU and have it. Fixed platform but tweakable. An evolution of what they have now.

I shall now get more wine and be quiet

 

[iroboto]

Prepare to smack me down as pure naive ramblings follow.

I find the multiple Tf count interesting as it fits with some earlier whimsical idea I got from when discussing the CPU and the TDP constraints.

What if

The apu has more potential TDP than it should have or at least the capacity and ability to scale both the CPU and GPU beyond the target envelope.
If Vega and Zen (this would fit for Jag but less beneficial ) are geared to higher frequency then they should have headroom in a console to run well at higher clocks without really loosing efficiency vs TDP. Now imagine like the optional CPU core Xbox one titles can request, the new SDK allows game makers to choose from various CPU / GPU clock speeds. Does your game require a bit more CPU then dock a bit off the GPU and have it. Fixed platform but tweakable. An evolution of what they have now.

I shall now get more wine and be quiet

Clockspeeds determine stability, developers would never have enough test units to test varying clock speeds on how far they can push the silicon. This is MS role here, you'll unlikely to ever see changing clock rates.

 

[turkey]

Clockspeeds determine stability, developers would never have enough test units to test varying clock speeds on how far they can push the silicon. This is MS role here, you'll unlikely to ever see changing clock rates.

It's certainly out there as an idea but having 2 or 3 possible clock combinations would that be so difficult to test and deploy given Microsoft have hardware in some form now and it does not release for over 6 months. This is to be tested and specified by Microsoft, Devs would not get to make it up themselves, merely select the predefined profile you require.

I was just thinking how do you get value from the hardware, you might want a strong CPU for VR or if pushing for 60fps on a 30fps Xbox one title, but a 60fps Xbox one title would leave excessive CPU wasted.

 

[Shifty Geezer]

"exciting next-gen launches" days are over in my opinion. It's way to complex and expensive to develop modern AAA engines and games in general combined with diminishing returns from the power of the hardware required to run it to have these large leaps like we used to. Scorpio will blunt any "nex gen" feel PS5 will have unless Sony decides to wait many many years....

This depends entirely on what appears on Scorpio. If all we see on Scorpio are 4K XB1 games, Sony will be able to release a new console that feels all new and exciting. If Scorpio is sufficiently beefy and allowed to stretch its legs (notably creating optimised 1080p60 games), then PS5 would just look like a 4K Scorpio. And IMO the only way we'll see that is if MS treat Scorpio as a new generation with Scorpio exclusives, blockbuster launch titles that don't run on XB1, and just have XB1 games as a 'BC feature'.

 

[mpg1]

This depends entirely on what appears on Scorpio. If all we see on Scorpio are 4K XB1 games, Sony will be able to release a new console that feels all new and exciting. If Scorpio is sufficiently beefy and allowed to stretch its legs (notably creating optimised 1080p60 games), then PS5 would just look like a 4K Scorpio. And IMO the only way we'll see that is if MS treat Scorpio as a new generation with Scorpio exclusives, blockbuster launch titles that don't run on XB1, and just have XB1 games as a 'BC feature'.

Not sure I agree. Presumably PS5 would run games at 4K standard....which means it would take A LOT of GPU power to make those games look and feel next gen over even PS4...like probably 25-30TFlops. So yeah if Sony decides to wait and release a PS5 in 5-6 years.

 

[iroboto]

It's certainly out there as an idea but having 2 or 3 possible clock combinations would that be so difficult to test and deploy given Microsoft have hardware in some form now and it does not release for over 6 months. This is to be tested and specified by Microsoft, Devs would not get to make it up themselves, merely select the predefined profile you require.

I was just thinking how do you get value from the hardware, you might want a strong CPU for VR or if pushing for 60fps on a 30fps Xbox one title, but a 60fps Xbox one title would leave excessive CPU wasted.

Well today, what happens is that developers program the game so that it runs.
Then they look at bottlenecks and when performance is bad, they optimize it. So assuming it was bad GPU code, they optimize the heck of out of it until the bottleneck is pushed back onto the CPU.
Then the CPU is the bottleneck because it can't feed the GPU enough, so they optimize the CPU code and it shifts the bottleneck back to the GPU.

And they'll keep doing this until they hit the end, either they run out of time, or they can no longer optimize without changing a critical piece of their code, which gets shelved for the next title.
Having said that, we have a great deal of _many_ methods of moving the workload to the GPU, so the setup of having a weak CPU coupled with a strong GPU is fine, as long as you continue to move work over to the GPU.
It's rare if ever that a strong CPU paired with a weak GPU would be considered awesome. I think PS3 may be the only case of it working and it took a _lot_ of time to get it to that point.