[OryoN]

From PC to Next-Gen Consoles: Largest Performance Gap In A While?


I'm just the "average gamer" really curious about future console hardware
capabilities, and how they may overcome/reduce current limitations/burdens in the PC area(and current consoles). Granted, I understand limitations will always be present in any piece of hardware, but what major technical advantages can one really expect form next-generation consoles?

More specifically:

What technical factors do you believe will be less of an issue
for the next-generation consoles Vs current (and perhaps future)PC graphics cards? Or, what features, known so far, do you think will really set consoles apart form PCs in terms of performance?

for example:

-- Will fillrate finally be a non-issue for consoles, especially since they will
be displaying much lower resolutions than high-end PC graphics ?

-- How will these consoles benefit by not having to deal with the “slow” AGP bus that high-end PCs
must work around/cope with.

-- What’s the big deal about Xenon’s shaders having access to main RAM? What kind of advantage will this kind of “tightly interconnected” design of consoles have over PC systems?

-- Although today’s PC graphics cards are becoming increasingly powerful, they are unfortunately limited by today’s CPUs. What position does this then leaves next-generation consoles in, with their extremely powerful, yet, “less general purpose” CPU's? Perhaps a performance gap that’ll be much harder to close this time around?

Being less tech-savvy than most here, this is all can think to ask; but I’m sure you guys can think of a number of other points that may be related to this discussion, in addition to sharing some insight to those questions. I’m anxious to hear your thoughts, so please feel absolutely free to post anything along the lines of this topic. I would appreciate it greatly.

PS: I always browse these forums, but I registered just so I could post this thread. Sorry if it has been brought up before.

[Lazy8s]

PCs will have larger allocations of memory which will allow them to work with bigger data sets. Consoles will access and move data more quickly, initially, for greater dynamic scenes and effects.

[Guden Oden]

PC devs need more memory because they tend to use memory inefficiently, squandering it on stuff like 32-bit textures, or in the case of Doom3, uncompressed textures. Or else people run at high resolutions with tons of AA and need gigantic frame buffers, or a combination of all at once.

Console programmers, having a fixed hardware to target, knows how to better spend the memory as they don't have to work with many different hardware setups to get things to look the same on all systems. If they can skimp on memory somewhere and still make it look great, they can do so. Streaming load systems is also MUCH easier to implement. Metroid is a Prime example here (lousy pun intended). It does both; manages to have hugely complex rooms in memory, as well as loading new things in the background.

Also, fillrate won't be much of an issue next gen because we will either have so much of it there's basically no way we can run out of it at TV res, or else devs will take the matter out of our hands by making the choice for us wether AA will be applied or not and if so - how much. On PCs people can force 6xAA on at all times and make their machines bog down completely.

If PS3 rumor specs are even close to being true, it will indeed be the biggest gulf between PC and consoles EVER seen. Not only because PS3 would theoretically outpace PCs to such an extent it would be rather silly, but also because it would require such a paradigm-shift in how games are programmed to leverage that machine's enormous potential and potentially very alien architecture.

[jvd]

If PS3 rumor specs are even close to being true, it will indeed be the biggest gulf between PC and consoles EVER seen. Not only because PS3 would theoretically outpace PCs to such an extent it would be rather silly, but also because it would require such a paradigm-shift in how games are programmed to leverage that machine's enormous potential and potentially very alien architecture.

Ah thats funny.


Next year we will have dual athlon 64s . r520s and nv50s .

The year after that we will have faster dual athlon 64s and finally games that really take advantage of the extra registers . The average pc will have more than a gig of ram and the video cards will have 512 megs of ram or more .

Even if the ps3 is a large leap it will only be a year or two till the pc surpasses the hardware.

another year or so for pc games to catch up to.

Alot of big things are going to happen with pcs in the next year or two.

Longhorn will finaly set a standard for games. So we can more vrom geforce 2 mx cards being the primary dev platform to dx 9 class cards being the minimum .

[Killer-Kris]

Well so far the only next generation console design that to me looks impresive has been Xbox2 and the tightly integrated CPU/GPU design.

As far as Xbox2 goes, it might not necesarily be that the GPU can access main memory, but more that the GPU can access the CPU's caches, or possibly directly talk to the CPUs. This could be used for a few things off the top of my head such as physics calculations on the GPU, hit detection for extremely high polygon models (especially true if there is a primative processor) on the GPU. Or any other situation where you might want to pass data back from the GPU to the CPU (ie. just a lot of interactivity with geometry and what not).

The other advantage to having the CPU and GPU connected in that fashion is you no longer have to deal with the bandwidth and latency issues of either AGP or PCIE. This makes it so you can pass very large amounts of data from the CPU to GPU regularly.

But other than the configuration of the system there is nothing happening in consoles (that I'm aware of) that gives them a raw power advantage over PCs. PC CPUs are soon going to have 2 or even 4 or more cores, and GPUs are showing no sign of slowing down either.

Though the bandwidth/latency advantage that consoles are going to hold over PCs probably won't go away until nvidia and ati start building x86 CPU cores into their graphics chips. And just like the memory capacity advantage of PCs, I don't see either being fixed any time in the near future.

[Vince]

Originally Posted by Killer-Kris

Well so far the only console design that to me looks impresive has been Xbox2 and the tightly integrated CPU/GPU design

*scratches head*

[zidane1strife]

Longhorn will finaly set a standard for games. So we can more vrom geforce 2 mx cards being the primary dev platform to dx 9 class cards being the minimum .

I've not heard of any other announced consumer cpu in the next 3-5yrs, that would reach Tflop class performance... and if the rumors of 45nm are true...

PS as for longhorn, that's not an advantage....

edited

[jvd]

Originally Posted by Vince

*scratches head*

yea the cell tech is impressive. So i don't know why he thinks that.

But its no quantum leap that pcs wont catch up to quickly

[Killer-Kris]

Originally Posted by jvd

yea the cell tech is impressive. So i don't know why he thinks that.

Am I mistaken about it being Xbox2 that has the 3 Power4 derivative CPUs and the ATI GPU all on one chip?

I haven't heard much about Cell that sounds ground breaking, then again I haven't been following it closely. From my understanding it's a specification for allowing easy integration of what would be multiple ICs into a single chip with high speed interconnects, while we haven't seen much like that in the consumer world it's far from revolutionary.

[Killer-Kris]

Originally Posted by Vince

*scratches head*

So what's confusing?

[jvd]

Originally Posted by Killer-Kris

Am I mistaken about it being Xbox2 that has the 3 Power4 derivative CPUs and the ATI GPU all on one chip?

I haven't heard much about Cell that sounds ground breaking, then again I haven't been following it closely. From my understanding it's a specification for allowing easy integration of what would be multiple ICs into a single chip with high speed interconnects, while we haven't seen much like that in the consumer world it's far from revolutionary.

a quick search of the forums here will give u a good idea.

Sony claims 1tflop performance.

We will most likely see it peak at 256gflops or less.

Still infintly faster than any x86 chip . It should even soundly beat a power pc 4 x 3 set up

The gpu on the other hand will be another story

[Killer-Kris]

Originally Posted by jvd

a quick search of the forums here will give u a good idea.

Sony claims 1tflop performance.

We will most likely see it peak at 256gflops or less.

Still infintly faster than any x86 chip . It should even soundly beat a power pc 4 x 3 set up

The gpu on the other hand will be another story

Ok, from what I've managed to find so far I'm still not impressed. Theoretical performance really doesn't mean a whole lot. If that were the case we'll all be running Sun Niagra CPU's on our desktops in a year or two, but in reality we'll still be running an Athlon or Pentium of some sort because you can actually get decent performance out of it.

Here's the best thread here on B3D I've found discussing the architecture of PS3 http://www.beyond3d.com/forum/viewto...ht=playstation

I didn't read the whole thread but if the diagram several posts in is at all accurate PS3 is going to be very hard to get decent performance out of. That is unless Sony has created a very good threading library and compiler for the developers.

Though for the most part a CPU setup like that is going to need alot of work on behalf of the game developers to make good use of. And frankly 3 very decent at single threaded processing CPUs will be alot easier to use, than 4 psuedo cores with many functional units (I'm assuming there's multithreading going on there or else you'll never make use of all those functional units) .

Granted I would absolutely love to be proved wrong and see developers really start making heavy use of many characters, better AI, better physics, better sound, etc, etc, etc... all on screen at the same time in order to fully utilize the system.

But yes, in short, from the little bit I've found cell still doesn't excite me, it actually sounds like it has more in common with Sun's upcoming processors than useful CPUs like Athlon, P4, PM, Itanium, and Power. I don't doubt one bit that the theoretical performance is 1 TFLOP, but odds are that will never be realized in practice.

[-tkf-]

Tech is one thing.. nah, what is real interesting is when PC's are gonna catch up to the AAA game titles

In any case, i have the impression that thanks to compromises made in order to target as wide an audience as possible, we are rarely seing the best graphics the PCs can deliver. Even Doom3 that looks better than anything else runs "fair" on older hardware, alot of resources were spent on that. Those resources could have been used on other aspects of the game, for example a constant update of the engine as the release schedule were slipping

[Killer-Kris]

By the way, I don't mean to be saying that large multicore CPUs won't be useful. They most certainly will be, it's just that it will be a while before they can be effectively used for games, and even longer for general use.

[j^aws]

My 2 cents and IMO,

1) Bang per Buck,

PS3>Xenon>Revolution>PC

2) Realworld efficiency,

Revolution>Xenon>PS3>PC

3) The most fun? Ignore 1 and 2 and choose the platform that has the games and franchises you want to play!

Btw, Killer-Kris, an old powerpoint presentation summarizing Cell archtecture...

[Killer-Kris]

Originally Posted by Jaws

Btw, Killer-Kris, an old powerpoint presentation summarizing Cell archtecture...

Thank you very much, that was the exact sort of information I was looking for.

And after reading through that quickly, I really am even less impressed with PS3 if it's using that architecture.

The cell idea seems perfectly suited for
1.) Graphics
2.) Cluster applications
3.) DSPs
4.) network processors

This thing sounds like it will suck massively for general purpose computing unless Sony makes some really good libraries. And sadly at this stage games still fit much better into the single threaded programming model, than they do into the massively parallel with no data dependancy between threads which is what is going to be required to make use of an architecture like cell.

I think games are heading in that direction, but not yet, at least not that fast.

[j^aws]

Originally Posted by Killer-Kris

.....
This thing sounds like it will suck massively for general purpose computing unless Sony makes some really good libraries. And sadly at this stage games still fit much better into the single threaded programming model, than they do into the massively parallel with no data dependancy between threads which is what is going to be required to make use of an architecture like cell.

I think games are heading in that direction, but not yet, at least not that fast.

Those concerns have also been raised by devs on these forums and we've seen nothing from patents on how it will deal with single threaded, general purpose computing. And, IMO, this compiler would break or make the system and as such would need to be the smartest component of the whole Cell architecture.! :P

[jvd]

) Bang per Buck,

PS3>Xenon>Revolution>PC

what do u mean bang per buck?

our buck or the companys ?


Our buck will be revolution > ps3 > xenon>pc

Companys wil lbe pc < revolution < xenon < ps3

[Killer-Kris]

Originally Posted by Jaws

Those concerns have also been raised by devs on these forums and we've seen nothing from patents on how it will deal with single threaded, general purpose computing. And, IMO, this compiler would break or make the system and as such would need to be the smartest component of the whole Cell architecture.! :P

Absolutely!!! And just think how much money Sony would make by selling that compiler tech to Intel to help with Itanium. (I also suppose it would be of value to other companies who have been working on that particular problem for even longer, or all those researchers in universities who also haven't solved those problems)

So I take it that Sony has been fairly tight lipped about their plans as far as that goes?



So I hope that by now I've sufficiently explained why I'm not impressed with the PS3 design in the least. It will probably have close to 1Tflop theoretical performance but I doubt it will ever come remotely close to that. To Sony this is wasted silicon, and therefore wasted $$$. Currently 2-4 cores are difficult to utilize, let alone 32-64.

My main beef with the design is that it's a complete and utter waste for the intended market. If this was ment to go up against niagra servers, or to be used for some sort of scientific computer I'd be oohing and ahhing over it like most everyone here seems to be doing because you can actually use it there.


Now back on topic... from the reading I've been doing it looks like all the next gen consoles will have the tightly knit CPU/GPU advantage. I guess it was just that I was the most familiar with the next Xbox specs sorry about that.

[j^aws]

Originally Posted by jvd

what do u mean bang per buck?

our buck or the companys ?


Our buck will be revolution > ps3 > xenon>pc

Companys wil lbe pc < revolution < xenon < ps3

Our buck as consumers! and I said IMO!!! and I'm not even going to speculate this from the companies POV!

My assumption was at $300 release price for all consoles and average $1000 for the PC but adjust accordingly depending on what you think the actual price will be...

[j^aws]

Originally Posted by Killer-Kris

So I take it that Sony has been fairly tight lipped about their plans as far as that goes?

Yep...the basic software building block of Cell is the 'software cell' or 'APUlet' and we know very little on how they are compiled.

Fyi, Cell is a collaboration between Sony, Toshiba and IBM (STI) and it's meant to scale from the smallest consumer chip to Peta-FLOP class supercomputers...

Here's an old article on STI plans...

...and some recent plans...

[PiNkY]

The point about brute force approaches (as opposed to brainiac ones) is that you are basically trading resource efficiency for lots of resources. PS3 will probably never approach the real world / theoretical performance ratio of current pc implementations in the "more meaningful" tests but that figure just does not mean a whole lot. Though, I completely agree that the challenges in designing a highly parallel architecture certainly are on the software side. Or, to quote dr. marc tremblay [sun's man in charge for their niagara processor]:
"....I hate to do this, but I have to defend the Itanium design a little bit in the sense that these two chips will be running drastically different applications. So it will be easy for CELL to claim some pretty outrageous numbers, but it will be hard to exploit this while running TPC-C or whatever..."

[Deepak]

I too want to know how much gap will there be this time around between consoles and PC at launch (and afterwards) compared to last time. And will PC find it difficult this time to overtake consoles?

[cybamerc]

Guden Oden:

> Also, fillrate won't be much of an issue next gen because we will either
> have so much of it there's basically no way we can run out of it at TV res,

Sure it will. Next gen games will support HD resolutions.

[jvd]

Our buck as consumers! and I said IMO!!! and I'm not even going to speculate this from the companies POV!

My assumption was at $300 release price for all consoles and average $1000 for the PC but adjust accordingly depending on what you think the actual price will be...

Haha . I would say the reveloution wil lbe the cheapest console to produce and actualy be the most power full in what it actualy puts on the screen.

So in terms of graphics the best bang for the buck would be that. In terms of games well that goes to nintendo again imho. And since its the cheapest to design and make it woudl be the best bang for the buck for nintendo.


Now pc guys make the most money and alot of diffrent companys make alot of money off the pc