End of the line for fast consumer CPU means what for gaming?

gongo said:
We, consumers, do not need faster CPU?
Click to expand...
Well, generally speaking, we, consumers, don't. CPU performance has been stagnating for some time, and nobody has really cared much. In fact, rather than people moving to platforms that offer better performance, there has been been an active migration away from higher power platforms, to laptops, mobiles and ultimately cell phones.

I spoke to an Intel CPU engineer 5 or 6 years ago, relatively new at the company. When questioned regarding the need for higher CPU performance, and the challenges of selling new kit based on performance, he put forth an argument of faith - he assumed that important new software would evolve that justified the effort and expense spent on increasing their x86 performance.
He was wrong. It didn't happen.

As we now start to see cell phones being able to connect to the cloud when asked to do serious computational heavy lifting, client (consumer) side CPU performance is destined to mainly evolve towards lowered expense and less power to allow broader applicability and better ergonomics. In fact it already happened. x86 is a dinosaur driven by market inertia. All of this is not to say that there aren't specific areas where CPU computational power is relevant. My own field, for one, can use as much as anyone can supply for the foreseeable future. But consumers? The killer app never materialized.
 
It's interesting that with cell phones and tablets we are in a CPU/GPU performance arms race anew. Adobe Flash, HD video playback and the emerging 3D gaming market there is like the late '90s of PCs but fighting extreme power consumption constraints. It's pretty exciting. The hardware is already beyond what we had on PC in 2003-4 perhaps but using under 2W total or so. It is totally state of the art even if it's not competitive with the latest billion transistor desktop stuff.

I'd also say that the failure of the 3DS is a clear indication of what the cell phone and tablet markets represent. Angry Birds, Minecraft Mobile, etc say hello.
 
It's pretty safe to say the 3DS has been performing well below expectations. Failure is subjective.
 
End of line for CPU, well, console makers can just have smaller and cheaper CPU and have two GPUs instead.

It's like that for gaming PC too isn't ? Instead of going for something like Hex-core i7, for the money you better off getiting a second graphic card for better gaming performance.
 
I.S.T. said:
I think you are prematurely calling the 3DS a failure... It's not even been out a year.
Click to expand...

We will see but the phone and tablet gaming market is only getting stronger and I think it replaces portables for most people. Controls are somewhat of a problem there but it's not clear if that's a hindrance.

I'm not much of a portable game player though. I'm just observing tablet and phone popularity, who plays with them at my local stores and how the 3DS is in the back corner often untouched. I don't see 3DS exploding in popularity anymore like DS did simply because there is this new ultra popular competition.
 
swaaye said:
We will see but the phone and tablet gaming market is only getting stronger and I think it replaces portables for most people. Controls are somewhat of a problem there but it's not clear if that's a hindrance.
Click to expand...
I'd be willing to pay tons for decent portable Civ/Master of Orion/Space Empires game. Not something that's dumbed down to oblivion both in game mechanics and complexity but the real deal.

I wonder if I should try make something myself :p
 
hoho said:
I'd be willing to pay tons for decent portable Civ/Master of Orion/Space Empires game. Not something that's dumbed down to oblivion both in game mechanics and complexity but the real deal.

I wonder if I should try make something myself :p
Click to expand...
I posted over in the handheld forum about how I played Master of Orion in DOSBOX on my Nook Color. The game was easily controllable. That kind of game works very well with touch. A x86 Windows tablet would let you play whatever you want.
 
swaaye said:
A x86 Windows tablet would let you play whatever you want.
Click to expand...
Problem is I want them on my cellphone :)
I've played open transport tycoon on my n900 and it worked pretty well. With tweaks to UI I can easily see all sorts of turn-based games working really well on them.
 
Is it time for someone like ARM to make a competitive processor to the x86/PowerPC stranglehold?

What stranglehold do I speak of? The console and computing markets for the general consumer! What if ARM designed a CPU based on ARMv8 but was 4-issue wide and had FP optimisations specifically designed to aid graphics/physics work and from the offset aiming for a 3-4GHz clockspeed whilst retaining good power consumption? They could design such a gentle beast with a view to selling it to the console manufacturers.

As it is - I believe IBM have it sewn up for the next-gen -their latest PPC architecture is extremely power efficient and they have configurations to address all different types of power/perf/cost envelopes. x86 is expensive both in cost and power it seems. Case in point would be Atom and Brazos compared to PPC 470. I still believe we are going to see a many core CPU in the next gen consoles from MS and Sony, both with IBM designed CPUs and AMD designed GPUs. That combination currently seems to be the best deal.
 
swaaye said:
It's interesting that with cell phones and tablets we are in a CPU/GPU performance arms race anew. Adobe Flash, HD video playback and the emerging 3D gaming market there is like the late '90s of PCs but fighting extreme power consumption constraints. It's pretty exciting. The hardware is already beyond what we had on PC in 2003-4 perhaps but using under 2W total or so. It is totally state of the art even if it's not competitive with the latest billion transistor desktop stuff.

I'd also say that the failure of the 3DS is a clear indication of what the cell phone and tablet markets represent. Angry Birds, Minecraft Mobile, etc say hello.
Click to expand...

I think there is more than one force at play here.
But one is simply lack of adaptability and imagination. Apart from Apple, neither the hardware manufacturers nor the vendors who want to sell their kit control the software on top of it. So they sell hardware specs. The press, online and print, live off advertising from the same manufacturers and vendors. That's probably the main reason the press focus on the same properties their advertisers try to exploit for making sales. (Another could simply be tech editor inertia.)

It's pitiful to see hardware marketing try to sell new handsets on for instance CPU MHz. Even PC customers stopped caring years ago, and now they think cell phone buyers will bite that hook?

So I'd say a major reason you see that processing power arms race, is that it's the only thing the hardware manufacturers can or know how to do to try to generate consumer interest. I wish they were at least slightly more concerned with battery power, weight, build quality, display properties, camera responsiveness, et cetera. Even with hardware alone, there is so much they could do, and so many niches to address with a changed mind set.
 
swaaye said:
It's interesting that with cell phones and tablets we are in a CPU/GPU performance arms race anew. Adobe Flash, HD video playback and the emerging 3D gaming market there is like the late '90s of PCs but fighting extreme power consumption constraints. It's pretty exciting. The hardware is already beyond what we had on PC in 2003-4 perhaps but using under 2W total or so. It is totally state of the art even if it's not competitive with the latest billion transistor desktop stuff.

I'd also say that the failure of the 3DS is a clear indication of what the cell phone and tablet markets represent. Angry Birds, Minecraft Mobile, etc say hello.
Click to expand...

The 3DS has some major problems to fight. One, it looks like a DS and the games.. at least from what i know.. looks like DS games. The number of DS consoles i see around kids is staggering. If Nintendo is able to move the 3DS into the "must have" of kids minds it will sell like crazy. But it really REALLY needs the software to prove it.

A DS with a much bigger screen, essentially filling the entire top lid with a higher res screen and better graphics would have killed the sales and made life harder for the VITA.

And a final offtopic comment. Super Mario is the king, holy something, the kids are playing that game today and it´s still an awesome experience.
 
I'm not sure if I posted in there, but I don't think I changed my mind. I consider it far more likely, maybe even a near certainty, that we'll get many-core in order stuff in some shape or form, with graphics and physics remaining the most performance demanding ingredients for games in the future.
 
There are very small OoO ARM cores in phones right now, and this will be expanded upon in coming years. OoO was decided against for this gen due to time to market concerns. The pressure is not the same now.

A heterogenous setup of OoO CPU cores and specialized coprocessors or accellerators is now present on either side of the console hardware niche.
If console designers--who wanted OoO originally--want to keep consoles in-order still, they will need to justify it in the face of an improved baseline of acceptibility and the real benefits for overall performance improvement and consistency.
 
Last edited by a moderator:
I found it really worth it to read this thread again especially as time made its work and as some ideas/implementations did not make it or did not prove themselves.
There was also a interesting discussion going in another between Sebbbi and Andrew Lauritzen about multi thread vs single thread performances.
A bit parallel I find interesting what ARm did with their new Cortex- A7.

Hopefully it's not my job but architects of next generation systems must be in front of tough arbitrations.
My understanding is that the "form factor" will drive a lot of their decisions.

If we still facing a system consisting of two chips (a cpu a gpu) they may have room to pack more than 4 OoO cores offering a signigicant jump in performances.

If it's a SOC/APU it's another matter; they will be pressured from every single factor from die size to power consumption/thermal dissipation (ot make room for a beefy GPU). If llano is any clue 4 aggressively OoO cores may not let enough room for a good enough GPU (either in in die space but that could be workable with a bigger chip or in power/thermal characteristic which my be more problematic).

What I got by reading the aforementioned thread is that R&D is really costly and that IBM may (we'll never know for sure) had offer either to MS or Sony Toshiba not new chip but project they had that were hanging around. I especially keep in ming comments about expansive testing a OoO engine can be among other thing. All wrapped together I believe that neither Sony or MS can pay for the development of something in the same time custom (not problematic) and offering good performances (more problematic). So manufacturers may have to go with close derivative of existing chip if not of the shelves parts put together. My point is that IBM doesn't seem to have part that would lay in between Xenon/ppu and aggressively OoO CPU cores, to use CPUs IBM actually produce a part in between PowerA2 and and POWER7.

So I would be really happy to have the pov of people here that either really have advance knowledge on the matter or are coders about this: as manufacturers may not provide you with a cpu cores that strike the good balance between single and multi threaded performances what would be your first though about a system that actually embark both high single thread performance CPU cores and throughput oriented one? A concrete example would be 1 slightly slimmed down power7 cores (just removing really specific execution unit as the decimal unit) and one power A2 module (without an L2) and somewhere between 2 and 4 MB of L3 (actually the L2 for both the power A2 cores and the GPU) all this running at more reasonable clock speed than Xenon/PPU (by looking at power A2 spec ~2.4GHz could be it).
Would you deem such a cpu set-up good enough to be the brain of a next generation system? Or at least what would be your first thoughts if something like this were to be announced?
 
Last edited by a moderator:
liolio said:
A bit parallel I find interesting what ARm did with their new Cortex- A7.
Click to expand...
Can you give more detail on what parallels you see?

What I got by reading the aforementioned thread is that R&D is really costly and that IBM may (we'll never know for sure) had offer either to MS or Sony Toshiba not new chip but project they had that were hanging around.
Click to expand...
Can you point to the statements that indicated this?
 
You must log in or register to reply here.
Back
Top