PlayStation III possible details...
- Thread starter alexsok
- Start date
- Status
zidane1strife
Banned
PC-Engine said:
Yeah, but still... the ps2 doesn't have the same core as the psx(i think...)... i believe they'll try as hard as they can to make it 3Ghz, to keep the tradition...
randycat99
Veteran
I heard elsewhere it could be G3 cores, but that could just be rabid rumor. Does it sound feasible at all to have 16 G3 cores with dual vector units (each core) running at 1 Ghz on a chip considering die processes in 2005+? This place seems like the best place to kick around the idea, I imagine. 1 Ghz sounds like a conservative figure for me, but do you think it could actually be scaled as far as 3 Ghz, by 2005+?
randycat99
Veteran
I have to admit that 3 GHz sounds pretty lofty to me, as well, given how far it has scaled so far. Of course, I'm not thinking in a 2005+ mindset. So maybe it could happen given that future? I have no idea, really.
You have keep in mind that the P4 was designed scale with higher clocks. Also the P4 doesn't have 16+ cores and eDRAM. The cpu in PS3 supposed contains many cores and a certain amount of eDRAM for each core.
A chip with so many cores running at 3GHz would probably consume a lot of power and get very hot...just my 2 cents and I could very well be wrong as I don't know when PS3 will launch and the CPU configuration.
A chip with so many cores running at 3GHz would probably consume a lot of power and get very hot...just my 2 cents and I could very well be wrong as I don't know when PS3 will launch and the CPU configuration.
While that's true, high clock speed in itself really isn't as important as final performance as architechure and physical implimentation can and will play significant roles. Case in point (although I don't like using Spec CPU) a 1GHz Itanium2 scores better at CPU2000FP than a 3.06GHz P4... Of course 3GHz would certainly not be unrealistice either...
While also true, a lot can depend on what consists of a 'core' as well...
I know P4 was designed to scale well. That's not my point - by the time PS3 comes out, we'll be having 10+Ghz P4s.
But AFAIK there aren't any CPUs on the market (or about to be released) that haven't reached 1Ghz yet, and even GPUs are up to 500Mhz now. We're looking at 2 years down the track, minimum, and people think 1Ghz is a reasonable clock speed to expect? Perhaps if they have many, many CPUs on the one core - but Sony should know, games aren't exactly the easiest thing to convert for parallel operation (unless it's using raytracing
). If Sony comes out with a 1Ghz CPU with 16 cores on it, I can just imagine the groans coming out of development houses the world over...
Oh yeah, and PC-Engine, the EE ran _extremely_ hot when it came out and Sony had atrocious yields. That's just the name of the game.
But AFAIK there aren't any CPUs on the market (or about to be released) that haven't reached 1Ghz yet, and even GPUs are up to 500Mhz now. We're looking at 2 years down the track, minimum, and people think 1Ghz is a reasonable clock speed to expect? Perhaps if they have many, many CPUs on the one core - but Sony should know, games aren't exactly the easiest thing to convert for parallel operation (unless it's using raytracing
). If Sony comes out with a 1Ghz CPU with 16 cores on it, I can just imagine the groans coming out of development houses the world over...Oh yeah, and PC-Engine, the EE ran _extremely_ hot when it came out and Sony had atrocious yields. That's just the name of the game.
Yeah and that was ther first major foray into advance microprocessor design/manufacturing. It was a gamble and they yields suffered because of it. I highly doubt they would want to take the same risks again.
Guessing that it would be running at 3GHz to keep it inline with tradition isn't a very valid reason IMO especially when:
1. We don't know when it's launching
2. We don't know the processor configuration
3. We don't know the performance target
randycat99
Veteran
More speculation to add to the mix- isn't the CPU industry going through a sort of revisiting where brute clockrate will not be the foremost goal, instead pushing forth with more moderate clocks and advancing throughput via functional units and efficiency tactics? Intel seems to be the sole element to push Mhz to its extremes (and we have seen firsthand what the scale-at-any-cost design approach gets you), and even they have come around to considering a more moderate clock with greater throughput achieved elsewhere. This is not to say that Intel won't continue to push the P4 ever higher and higher. They have a lot invested there, and they have to follow through to make it worth it in the end. The rest of the industry will not necessarily follow suit (to the extreme of beating Intel Mhz, that is), instead opting for Mhz increases that seem to come naturally with die scalings and such. I guess where I was going with all of this, is perhaps we may observe the CPU industry go through a Mhz rise slowdown in the coming years?
megadrive0088
Regular
the EE3 might have more than just 16 cores, AFAIK. the basic CELL has 16-32 cores. but the EE3 version might be many more, with many Cell Chips together. who knows.
As for clock speed, I would love to see a 3 Ghz Super CELL with 64-128 cores and 6 TFLOPs or more of performance, plus a 1.5 Ghz Graphics Synthesizer 3 with 256 pipelines, each better than DX9 feature-wise but I'm sure that reality will be much more down to earth.
As for clock speed, I would love to see a 3 Ghz Super CELL with 64-128 cores and 6 TFLOPs or more of performance, plus a 1.5 Ghz Graphics Synthesizer 3 with 256 pipelines, each better than DX9 feature-wise but I'm sure that reality will be much more down to earth.
Well both the EE and GS were hot on their initial launch rule somewhat due to they're being released at higher clocks than the original design goals. However I don't every remember the EE having any major yeild problems though. The GS on the other hand was quite problematic (in terms of manufacturing).
Actually the PSX was their major foray into major large-scale VLSI design (being a massive jump in their IC design capabilities). The EE is just a logical evolution from that decision. The real big deal or gamble was the decision to take on the manufacturing themselves rather than outsource.
Actually the design can be completed pretty late and still meet production demands. The real need for a system to be logically complete early is to give developers more time on the system to actually have something at launch.
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
mech said:
Since many of you are taking a page out of the How to Lie using Statistsics book, I'll join in:
When the EE launched, it launched at 300MHz. The Intel equivalent was the origional PIII core (starts with a 'k') @ 600Mhz and the newly launched K7 at the same speeds. I think it's fair to say that the EE rips both a new asshole.
So, since your all looking for a 'historical' or 'trend' or 'tradition' - I'd say this is alot better than the trends your using. (Which isn't saying much)
Also, the target year is 2005. Thats basically 3 years away, or 36 months. Moore's Law states that tranistsor densities double every 18 months. Thus, 2 cycles will occur. I'll use Intel's Prescott architecture, which has around 100M tranistsors, as a starting point in 2003. Thus, according to Moore's Law, Intel can only field ~400M transistors in 2005.
Ben's speculation (Whether tongue-in-cheek we shall never know
) puts 900M on Cell IIRC, My estimates (Which I think is way too optimistic) puts 700M yeilding 1TFLOP. How you expect a chip to have such densities (almost 2X) and retain the ability to scale to >3Ghz is beyond me. I doubt SOI will be have that big of an impact.
1-1.5GHz is likely IMHO. With the range extending upto 2GHz, but with little chance.
what i am most worried about is...
how loud will ps3 or any of the next gen consoles be??
i men, FAN-wise...
as recently seen on the geforceFX, engineers will need more and more *cooling power* the more complex the architectures get...
and my PS2 is already loud (to be the *living-room-do-everything-thing* sony wants it to be)....
next thing you know, next gen consoles wars will be fought on *cooling power* rather than *polygon-pushing power* or whatever else
how loud will ps3 or any of the next gen consoles be??
i men, FAN-wise...
as recently seen on the geforceFX,
engineers will need more and more *cooling power* the more complex the architectures get... and my PS2 is already loud (to be the *living-room-do-everything-thing* sony wants it to be)....
next thing you know, next gen consoles wars will be fought on *cooling power* rather than *polygon-pushing power* or whatever else
- Status