CD Projekt dev switching lead platform from PC to console *spawn

I agree with many people who consider that PC hardware is again undergoing much more advancements and will never remain same, in fact the top of the line PCs are already more powerful than next gen consoles.
 
But, there is always a but... it is true that other systems will step up because while the PC is not standing still, so aren't consoles.

However, it seems to me that finally for many companies and studios it was about time that at some point the consoles became *good enough* and have sufficient RAM -which, imho, was the biggest difference between consoles and PCs until now- and they can become the lead platform without limiting the possibilities of PCs.

You speak as though this is the first time this has ever happened. Perhaps you're only getting into gaming mid way through this generation but consoles being roughly on par with, or even slightely ahead of the highest end PC's when they launch is nothing new. It happens every time a new generation launches, that's why they call it a cycle. This isn;t some great fundemental shift in the balance of power, its just the normal reset of power balance we come to expect at the start of every console generation. And as with every generation PC's will once again overtake and within a few years be offering possibilities to developers that simply don't exist in the consoles space. The big difference this time is that the consoles are starting from a much lower performance point than they have in previous generations making the PC's job much easier to open up another generational power gap.
 
Now, with 8GB of memory each, next gen consoles are really good at everything. I can see PCs still offering some extra powers, but mostly some extra resolution and not much -maybe SSAA-.

Right now PCs offer more in almost every possible way, from resolution, to AA, to textures, to HDR, to new techniques that are impossible with current generation hardware, etc.

So the same as every other console generation. At the start of this generation PC's will offer little more than resolution improvements over next gen consoles and as the generation ages, PC's will gradually offer more and more graphical features that the consoles lack

Bare in mind though that the relatively low performance of the new consoles this generation opens up possibilities on the PC side that didn't exist in previous generations. For example the power will exist from day one on the PC this time (if you're willing to pay for it) to take a 1080p 30fps console game and run it at 1080p, 2x SSAA, 60fps and full 3D. That's a massive visual difference IMO.

On the other hand, now that next gen consoles are on their way to our living rooms, we know they will be capable of offering the same rendering technologies than future PCs, a little bit of everything, so I can see how studios are considering the idea of using them as their lead development hardware, it seems reasonable.

Consoles have always been in our living rooms, they'll offer the same rendering technologies as contemporary PC's, not future PC's and studios using consoles as the lead development platform is nothing new.

Honestly I think that's the case because both consoles feature custom audio chips and I am pretty sure the GPUs are heavily customized, by getting rid of silicon that isn't necessary on PC, where some of the silicon is there for legacy purposes.

How does a custom audio chip make consoles unlike a PC? PC's have been using custom audio chips for decades. They've switched to software for the most part in recent years but that doesn't mean custom chips aren't still available and if the console audio chips are offering capabilities that are difficult to reproduce on the PC side in software then we may even see a resurgence of them on the PC.

And the GPU's are not heavily customised. They are virtually off the shelf GCN based GPU's. Probably Sea Islands based.

Not to mention the eSRAM, the DMEs and the display planes.

ESRAM and DME's are features of only the weaker of the two next generation consoles and for the most part seem to be a simply alernative memory arragement to allow for lower production costs (vs 8GB GDDDR5). Look at PS4 and you see almost a carbon copy of the upcoming Kaveri (PC) memory setup.

Display planes do seem quite unique though, I'll reserve judgement on what kind of impact they will have.
 
Last edited by a moderator:
Display planes do nothing a PC cannot do. It's just that they allow the OS to have predictable access to the display. If MS wanted Windows to pop up over running games, that could be achieved in the GPU, but presumably at the framerate of the game. The display planes are not going to allow anything special on the consoles that the PC same basic hardware in PC will not be able to achieve. The audio will be nice, but again not a system changer unless your major interest is audio (and even then you can buy a DSP audio card, so apart from being a different model DSP, it's not a non-PC component).

The CPU is x86. It has all the legacy junk of x86 that AMD are putting in Jaguar. The GPU is also a legacy supporting part as much as the PC implementation of Kaveri is. We haven't heard anything to suggest the GPU is highly different. It has a ring-bus or a scheduler variation, perhaps, but it's stull fundamentally a GCN GPU and a Jaguar CPU. These boxes are effectively flavours of typical x86 PC, in stark contrast to prior consoles that used bespoke CPUs and/or GPUs and extra components. That's why they are great for next-gen PC developers, because they provide developers with the same architecture that they'll be writing for on PC. They'll have a piracy free audience to target, with good baseline specs, meaning they can write their x86 and GCN code and port it to PC pretty easily (depending somewhat on the difference between console APIs and DirectX), unlike this gen and others where they had to substantially rewrite the entire game.

As such, leading on consoles seems a no-brainer to me. It won't impede PC development. Maybe, if you are looking for a very broad audience on PC, the consoles would be a bad choice for scaling down to low-end PC. Otherwise consoles will be equivalent to a stable PC target.
 
Honestly I think that's the case because both consoles feature custom audio chips and I am pretty sure the GPUs are heavily customized, by getting rid of silicon that isn't necessary on PC, where some of the silicon is there for legacy purposes.

Not to mention the eSRAM, the DMEs and the display planes. Consoles always retain its capability to be heavily tweaked by their makers.

Custom audio chips are nothing new in the PC space. GPUs by definition are heavily customized pieces of hardware.

There is nothing being put into the consoles currently to allow them to do something that PCs cannot do. In other words, unlike Xenon or Cell, there aren't exactly a whole lot of non-x86 microcode that doesn't exist in PCs. Nor do the CPUs or GPUs operate any differently than the PC ones.

Everything that is being put into the consoles are there to facilitate being able to get game performance in the lowest power and heat envelope possible with the smallest silicon budget possible.

PCs have the luxury of not having to possibly operate within a closed entertainment cabinet with limited or zero airflow. Hence they can brute force many things.

You could potentially make an argument that the hardware arrangement in consoles might be more "elegant" than standard PCs, but even that isn't necessarily true.

Basically all the hardware is so similar that it is going to make multiplatform games much easier to develop as porting them from one platform to another is going to be relatively simple (compared to the past). That is going to be a huge boon not only to consoles but to PCs as well as it means developers and publishers can spend less time and money on porting and more time and money on the game itself.

Regards,
SB
 
There is nothing being put into the consoles currently to allow them to do something that PCs cannot do.

What about HSA, very low latency communication between CPU and GPU that allows for more GPGPU effectiveness?
 
What about HSA, very low latency communication between CPU and GPU that allows for more GPGPU effectiveness?

We've had APU's available in the PC space for almost 2 years now. Nothing new there. The only thing new is that the integrated graphics core will be much more potent, which only makes sense since you won't have a dedicated GPU in the machine.

HSA hasn't taken off on PC mainly due to the fact that most machines don't have CPU's featuring integrated GPUs which are capable of compute. Hence, no one is going to bother building a program specifically targeting those at the moment. But it certainly isn't new to the PC space.

Regards,
SB
 
AFAIK, HSA is a lot more than just having a CPU and GPU in the same chip. There's an unified memory address (RAM and L2/L3 cache alike, I think) and some advanced context switching in the hardware, at least.

No current SoC/APU is fully "HSA compliant" and the next-gen consoles might well be the first to come out with a true HSA.


BTW, Intel has only had one generation of compute-enabled GPUs (at least driver-supported on launch). Things could (should) change a lot this next gen.
 
AFAIK, HSA is a lot more than just having a CPU and GPU in the same chip. There's an unified memory address (RAM and L2/L3 cache alike, I think) and some advanced context switching in the hardware, at least.

No current SoC/APU is fully "HSA compliant" and the next-gen consoles might well be the first to come out with a true HSA.


BTW, Intel has only had one generation of compute-enabled GPUs (at least driver-supported on launch). Things could (should) change a lot this next gen.

Sure but that unified memory address only serves to speed things up, it doesn't offer functionality that didn't already exist in a way that currently can't be done. It's isn't like porting something from PPC instructions to x86 instructions for example. Or Cell to x86. And it isn't like that same virtually identical architecture isn't going to exist on PC soon after the consoles launch, if not before. Either way the APU used in Orbis/Durango are direct descendants and logical evolutions of the APU that AMD released almost 2 years ago. Whether Orbis/Durango existed or not, they'd be coming to PC.

And WRT Intel, I'm cautiously hopeful for that. I'm hoping that Intel doesn't continue to marginalize potential GPU compute capabilities as I believe that for the computations involved GPUs can still crunch through those numbers faster than adding on more capability to CPUs currently.

Regards,
SB
 
Kaveri will be the first fully HSA enabled PC APU and will launch towards the end of this year. To all intents and purposes, PC's and consoles will be receiving full HSA capabilities at the same time. That is of course assuming the consoles are actually fully HSA enabled and not more in line with Kabini.

Obviously as noted Kaveri is just an evolutionary development of what was started over a year ago now with Llano.

So strictly speaking consoles aren't getting anything that's not possible on the PC although you'd need a very specific PC setup to match the exact capabilities and even then you may be reliant on developers supporting that HSA implementation in games.
 
Last edited by a moderator:
Kaveri will be the first fully HSA enabled PC APU and will launch towards the end of this year. To all intents and purposes, PC's and consoles will be receiving full HSA capabilities at the same time. That is of course assuming the consoles are actually fully HSA enabled and not more in line with Kabini.

Obviously as noted Kaveri is just an evolutionary development of what was started over a year ago now with Llano.

So strictly speaking consoles aren't getting anything that's not possible on the PC although you'd need a very specific PC setup to match the exact capabilities and even then you may be reliant on developers supporting that HSA implementation in games.

If AMD is clever and want HSA to be successful i hope they have been willing to give consoles the full HSA with context switching and graphics preemption.I hope that when David Baumann says the semi custom apus leverage the 2013 blocks these are the fully HSA ones or they willl have lose the best opportunity to make its tech shine.In fact reading the HSA pdf the ACEs without context switching take to heavily underused resources in the GPU,and dont forget in PS4 they have settled one ACE per cpu core, so stalling CUs when there are many compute threas running in the GPU will only make things worse.
 
If AMD is clever and want HSA to be successful i hope they have been willing to give consoles the full HSA with context switching and graphics preemption.I hope that when David Baumann says the semi custom apus leverage the 2013 blocks these are the fully HSA ones or they willl have lose the best opportunity to make its tech shine.

Context switching and graphics preemption are 2014 features that will launch with Kaveri's successor, not Kaveri itself as far as I'm aware. So I highly doubt the consoles will be getting this.
 
Context switching and graphics preemption are 2014 features that will launch with Kaveri's successor, not Kaveri itself as far as I'm aware. So I highly doubt the consoles will be getting this.

But do you agree that wouldnt be very clever from AMD not to have made the effort to have the consoles with all the weapons out?.If this tech sing in console gaming PC gamers will start to ask for these APUS.Or they still have not a clue of how to get it and is still in design...that is possible, as getting context switching working must be not easy.

Anyway AMD roadmap is a little convoluted nowadays.That powerpoint saying Kavery is a new architecture could mean anything.
 
Last edited by a moderator:
And as with every generation PC's will once again overtake and within a few years be offering possibilities to developers that simply don't exist in the consoles space. The big difference this time is that the consoles are starting from a much lower performance point than they have in previous generations making the PC's job much easier to open up another generational power gap.

The laws of deminishing returns and the possible end to moores law fly in the face of this rule somewhat. We've seen deminishing returns play a role somewhat in the current gen - the only saving grace for PC has been the extended duration of the generation.

If the next gen is only =<5 years then I think PC's will have a problem surpassing them enough for it to matter. Sure, the hardware will be a lot better despite the possible end of moores law (for now), but you can bet graphics will be "good enough" for 95% of people for at least 5 years, and due to the relative abundance of RAM, simply what is purely possible in games won't become such a limiting factor either.
 
If diminishing returns are going to be so apparent then there's no way at all this generation will last only 5 years. More likely it will be the last console generation. And diminishing returns goes both ways. If people wont notice the visual difference between consoles and pcs with 4x their power, nor will they notice the difference between consoles and pc's half their power. That effectively puts a vastly greater proportion of the pc market on a level with consoles than ever before. Effectively were talking about everything from top end apus and entry level gaming cards being equivilent to next generation consoles from day 1 of launch. It took pcs years to get into that position this generation and that spells good things for pc developer support. Meanwhile high end gamers can at least enjoy the obvious benefits of 60fps and 3d which isnt effected by diminishing returns.
 
There are and there aren't diminishing returns. There's still so much that can be improved that lead to hugely obvious increases in visual IQ. Off the top of my head...

Texture quality. Although it may require a game having 200-400 GB of data in order to fully realize it to a very noticeable degree. So in a way I suppose that has hit diminishing returns.

Tesselated geometry and to a lesser extent POM. Barely been touched, but it also is hugely impressive and noticeable.

New technologies like TressFX. I can't wait until the day hardware is powerful enough that we can have realistic hair on all NPCs as well as the main character. TressFX compared to the best hair that has been featured in games up until now is like Night and Day difference. And just like tessellated geometry, it's difficult for me to play a game that doesn't have it anymore because the improvement is so huge.

It's like back when textured polygons were first introduced into a playable game. It was less pleasant to play games that only used gouraud shaded polygons.

As well, having access to more than 2 GB of VA memory (32 bit PC) opens up a whole world of possibilities. Unfortunately, developers will now have to deal with how to handle dealing with loading the assets from HDD without incurring long load times or heavy texture/detail pop in for streaming games. The HDD is going to be a relatively large bottleneck for graphics as there will have to be compromises made to accommodate how slowly it takes to load in multiple GBs of data.

And so much more.

Lighting can still be improved, but for most people the improvements won't be noticeable. People might be able to say one looks better than the other but might not be able to say why. So it's entered into the area of diminishing returns with regards to how your average gamer perceives it. I'll bet most gamers that played Rage and Crysis 2 on PC didn't even notice just how much the lighting differed.

So lighting is an example of where we're getting into dimishing returns where it requires significantly more processing power to implement noticeable improvements. There's others obviously.

Regards,
SB
 
I say we've hit diminishing returns for the subject matters that developers carefully learned to adress because they knew were the ones they could do right. A bald space marine in a mostly static industrial environment full of flat squarish opaque non reflective geometry is not gonna look much better. It's when developers do tackle those things we've learned to be impossible in games, and manadge to get it right that players get impressed. Many first psone games looked mildly superior to snes ones because they were concentrating in 2d cartoony sprite based games. But time showed us the system was capable of much more. Xbox360 was called xbox 1.5 in the begining wasn't it?
 
Well, I have been thoroughly thinking about what you said regarding consoles and the fact that they use x86 technology and.... with all due respect, I disagree with willardjuice, pbjliverpool and Shifty.

The fact that they have x86 technology doesn't mean the new consoles are the PCs you want to make them out to be. It's not the same hardware set by any means of the imagination.

Durango for instance has been designed around parallel processing, how does it compare to a PC? A new development strategy has to be thought up in order to take advantage of the console.

On PS4 or Durango you may get 60 fps and the smoothest gameplay you have ever experienced on the system, and on the PC you could get 1 fps because some CPU thread is lagging behind.

On a console everything is the same, you are programming for one machine and that's all that there is to it.

So you can't really compare a 1.6GHz 8 core x86 CPU in a console to say the same thing in a PC.

I mean, just because parts of the hardware could work on a PC it doesn't mean it is a PC at all. :???:
 
Here is an example to illustrate my point.

They say Durango's GPU is based on the Radeon 7770 1GHz Edition, yet the PC GPU operates at 1.15 GHz -350 MHz faster than Durango's GPU and uses GDDR5 memory, yet Durango performs better when it comes to raw numbers.

Why?? You may ask. Specific and specialized console hardware, simple as that. Here are the facts.

http://www.memoryexpress.com/Products/MX39034

http://www.vgleaks.com/durango-gpu-2/

Here is a quote from vgleaks article to again make myself understood.

There is nothing remotely similar to a PC in Durango when the console doesn't even have video memory -VRAM- in the traditional sense.

ESRAM

Durango has no video memory (VRAM) in the traditional sense, but the GPU does contain 32 MB of fast embedded SRAM (ESRAM). ESRAM on Durango is free from many of the restrictions that affect EDRAM on Xbox 360. Durango supports the following scenarios:

•Texturing from ESRAM

•Rendering to surfaces in main RAM

•Read back from render targets without performing a resolve (in certain cases)

The difference in throughput between ESRAM and main RAM is moderate: 102.4 GB/sec versus 68 GB/sec.

The advantages of ESRAM are lower latency and lack of contention from other memory clients—for instance the CPU, I/O, and display output. Low latency is particularly important for sustaining peak performance of the color blocks (CBs) and depth blocks (DBs).
 
Well, I have been thoroughly thinking about what you said regarding consoles and the fact that they use x86 technology and.... with all due respect, I disagree with willardjuice, pbjliverpool and Shifty.

The fact that they have x86 technology doesn't mean the new consoles are the PCs you want to make them out to be.
They are PCs in architecture regarding how developers will address them.

There is nothing remotely similar to a PC in Durango when the console doesn't even have video memory -VRAM- in the traditional sense.
AMD APU based PCs don't have VRAM. They work from DDR3. They have SRAM caches.

A laptop with a 4 core AMD APU at 1.4 GHz and 6 GBs DDR3 shared memory, and a desktop with a quad core i7 and 16 GBs DDR3 and dual nVidia GTX 680s with 4 GBs GDDR5 each, and a nettop with dual core i3 and integrated graphics, and an Intel Mac, and a Linux media box, are all the same from the POV of the developer in how they design their games. You have CPU work, GPU work, all processors running the same code and working the same way in how they deal with threads, accessing memory the same general way. Okay, Durango's memory architecture is going to add a little more complexity, but 'same' can be measured by degree, and when you compare this gen to previous consoles, clearly they aren't anything like as removed from PC as PS1, 2, and 3 and their peers have been. When can argue specifics and perhaps whether a 5-10% variation in system design constitutes a different architecture or not, but the generalisation willardjuice made will still stand. You can write x86 CPU code for PC and it'll port straight over to consoles. You can write GPU code and it'll port straight over to consoles. There are no obvious barriers between developing on console and porting to PC or vice versa.
 
Here is an example to illustrate my point.

They say Durango's GPU is based on the Radeon 7770 1GHz Edition, yet the PC GPU operates at 1.15 GHz -350 MHz faster than Durango's GPU and uses GDDR5 memory, yet Durango performs better when it comes to raw numbers.

Durango's GPU has 12 CU units at 800Mhz. A 7770 has 10 at 1000Mhz.
 
Cyan to be blunt, it's clear you don't quite understand some of the technologies/features you describe in your posts. I recommend that before making sweeping statements, you should at least have a firm grasp of what you're talking about.
 
Back
Top