PlayStation III possible details...
- Thread starter alexsok
- Start date
- Status
randycat99
Veteran
Yes, I certainly agree that fan noise is becoming an increasingly important issue. I, too, feel the current PS2 is a bit loud for a "well-mannered" living room device. I don't know where things will truly end up since we are baying for these fantastic amounts of processing performance, but we want it to still fit in a little plastic box and be quiet.
zidane1strife
Banned
When many said the ps2 would only be able to sustain 5M polys ingame... the final product surpassed expectations... I think 3Ghz is feasible, I mean if it's 65nm or smaller... doesn't that help?
randycat99
Veteran
I just think a G3 running at 3 Ghz sounds otherworldly when Motorola struggled to get them to just 500 Mhz and IBM only recently got them to 700+...but what do I know? It just seems like something would have to be redesigned in the G3 for it to really run that fast. We're talkin' P3 age architecture, right? Does anybody think a 3 Ghz P3 is a bit outlandish, just by beneficial die scaling alone?
archie4oz said:
Fair enough, all I remember hearing at the time was that the chips for PS2 were having major yield problems - I assumed they meant both.
Fafalada said:It's interesting that even though GS has 4x transistor count, as well as being a new design, while EE is a Mips Core with a few custom units added, people still like to assume EE was most problematic to manufacture.
Anyway, what Archie said - I haven't heard of EE yield problems either (but plenty about those with GS).
Actually, from what I heard .25 micron EE's were already overclockable
I wasn't assuming it was the most problematic to manufacture, I mentioned it because we were talking about CPU cores rather than GPUs. And as stated above, I heard about yield problems in general - I figured because the EE was such a GFLOP powerhouse at the time, they had just as many problems manufacturing them.
Vince:
As stated (even in the bit you quoted) - it's purely my opinion. And my opinion is that they'd be crazy to have a core with so many CPUs on it. I think I already said so, if they do have some crazy amount of cores on there, then sure, clock speed is going to be limited. I'm not sure why you're trying to tell me something I know already.
randycat99:
There are just some limits to CPU design sometimes, I guess due to the nature of their design limiting clock speed. Whether it's possible to redesign the chip so that it can scale higher is iffy - they managed to do it with the Athlon, but it took a fair amount of effort, but even then it's not scaling that well. I think you have to design from the start for the target clock speeds...
But I'm just speculating. I don't really know enough about this aspect of CPU design.
randycat99
Veteran
That's the catch- if you design a CPU from the ground up for brute scaling while retaining some semblence of output, you end up with a giant core P4. I think it is pretty certain that 16 P4 cores on a chip is utterly nonsensical in terms of manufacturing and the sheer heat output. If you were to say that perhaps a smaller processor could be created that also scaled similarly, well that would be betting you could outsmart Intel's best, wouldn't it?
So I maintain (IMO) that the core choice will be based around a simpler, smaller, computationally more efficent CPU architecture, where clock rates are more moderate (and possibly limited ultimately by some ceiling inherent to the architecture). Such a CPU would be relatively "dirt simple", but you wouldn't flinch at all over throwing dozens of them on a single chip to scale the performance.
So I maintain (IMO) that the core choice will be based around a simpler, smaller, computationally more efficent CPU architecture, where clock rates are more moderate (and possibly limited ultimately by some ceiling inherent to the architecture). Such a CPU would be relatively "dirt simple", but you wouldn't flinch at all over throwing dozens of them on a single chip to scale the performance.
zidane1strife
Banned
heheh.... a...m.#.cough.... will....do...it....
randycat99
Veteran
I was going to say, "why don't you edjumacate the guy instead of rolling eyes", but then I see he is a Senior Member. What's up with that? Can he really be serious with that comment? 
Technologies Maturing end of 2004 and in 2005:
With all these technologies coming into place they will be able to do more with less GHz power.
Speng
- Toshiba and Sony: .65nm CMOS Technology (CMOS5) with High Density Embedded Memories for Broadband Microprocessor Applications.
http://www.sony.co.jp/SonyInfo/News/Press/200212/02-1203/
http://www.beyond3d.com/forum/viewtopic.php?t=3426
IBM's Strained-Silicon, Shrinking Circuits:
Completed on .65nm dimensions to allowing increased performance and decreased power consumption.
http://www.siliconstrategies.com/story/OEG20021204S0041
IBM's Double Gated Transistors: Twice the performance of todays transitors
IBM'sCell Processing:
http://www.eet.com/at/news/OEG20021121S0046
With all these technologies coming into place they will be able to do more with less GHz power.
Speng
randycat99
Veteran
That is not a relevant solution (not saying it didn't work in the DC). It doesn't matter much (as far as fan noise is concerned) if you can move heat from one component to another. You have to dissipate the heat to air in the end, just the same.
[Whoah, dude! When did I become a Senior Member?! Yippee!]
[Whoah, dude! When did I become a Senior Member?! Yippee!]
london-boy said:as recently seen on the geforceFX,
The reason for the NV30's heat related problems that necessitated the fan was nVidia's decision to revert from the planned inclusion of low-k dielectrics - to the more reliable and higher yeilding (aswell as sooner ready) 0.13um Cu process.
If all had gone to plan, it most likely wouldn't have required that. But, TSMC's delays and problems forstalled nvidia and caused physical changes in the design.
I never said you weren't correct, just that there are big holes in what your saying. Especially when you factor in what we already know.
Um, may I suggest you do some research into Cellular Computing. It's fasinating stuff, and if you do, you'll see that the ideas behind said computing is very grounded, very sucessful, very effecient, and potentially very powerful.
It's only "crazy" untill society catches up and it's "the way".
randycat99
Veteran
mech said:
That bar needs to be raised then.
It just doesn't feel right that I could be a senior member here (experience-wise)- let alone at a mere 200 posts.What exactly are these holes?
We're just speculating on what we believe the clock speed will be.... we may as well make up a dart board with some clock speeds at it and throw darts at it, because none of us know exactly what route Sony's going to take. I'm making some speculation and stating what my assumptions are for that speculation - you can't be any more right than I am, because the thing's not even finished yet.
Every console which has had multiple CPUs to program for has been deemed a "programming nightmare". Saturn, PS2, Jaguar, all these machines were very difficult to program for because of the multiple cores that required managing.
Instead of trying to win a point against me with an off-hand remark like "go do some research", why not try and explain why you think multiple CPU setups are going to be the way of the future? I still think there are too many problems with this setup, especially the problem of parallelizing many game tasks - several of which are really quite difficult to parallelize effectively. So how are they going to overcome this problem?
for all those wondering about my post its to point out how stupid this thread is. I guess i should have used 8) To get my point across.
IMO the only way to tell how well the cell chip would scale is to figure out the highest the pushed it and on what mircon . The cooling needed , how many cores were on that chip , how many mhz it costs to add more cores (by this i mean how much the clock speed is limited due to power demands and cooling requirements) what process they will be using at that time , how much a 3ghz cell processer would cost and the biggest thing , would it even need a 3ghz processer with 16-32 cores on it to push the graphics and the ai .
Oh and while i may have poor grammer (never was good with that) and I may not be as wise as most members on this board I am not an idiot. I have been viewing this board and others for a long time.
8) To get my point across. IMO the only way to tell how well the cell chip would scale is to figure out the highest the pushed it and on what mircon . The cooling needed , how many cores were on that chip , how many mhz it costs to add more cores (by this i mean how much the clock speed is limited due to power demands and cooling requirements) what process they will be using at that time , how much a 3ghz cell processer would cost and the biggest thing , would it even need a 3ghz processer with 16-32 cores on it to push the graphics and the ai .
Oh and while i may have poor grammer (never was good with that) and I may not be as wise as most members on this board I am not an idiot. I have been viewing this board and others for a long time.
- Status