Dr Evil said:
Acert93 said:
CELL's biggest advantage due to design is scalability. A 4-way 4th gen CELL (4:32 PPE:SPE) or more at 6GHz+ is scalable power and emphasizes continuaty and tools. I like that. I am curious how MS will respond.
I think it has to do much more than 4:32 considering the leap from PS2 to PS3 I don't know if that's doable though.
That is why I added the "or more" and "6GHz+". We really do not know what the next 5-6 years will hold for chips because we are beginning to hit the physical limitations. Even if we look at high end developments, CELL itself shows that for yield, heat, and power consumption issues you can only push so hard on the wall (e.g. CELL can do a full 1:8 configuration AND run almost 5GHz, but for yields and other facts they went with the safe 3.2GHz 1:7). The future is all about manufacturing process.
65nm is coming mainstream in 2006, give or take. I assume, considering the trends the industry follows (or have been set by market leaders like Intel and IBM) that we will see 45nm in 2008, give or take.
From there who knows. They are approaching the physical limits of silicone as Phil mentioned. Lets say in 2011 they do get to 32nm, Sony holds out until 2012 for the PS4 (that is a big assumption consider MS will do a 6year cycle also) so they can hit 32nm.
Throw in the fact the transition from 300mm^2 wafers to 400mm^2 wafers, and there is some hope of a larger CELL than 4:32. But some other reasons to stop and consider the issues:
1. We have no clue how things will react when we hit 45nm and 32nm. There could be so much leakage that it is hard to work with.
2. Yields. The smaller you get, yet keeping a large chip, results in less and less yields. There are ways to help alleviate some of this (CELL disabling a SPE and R500 having redundant caches and ALUs are examples), but the problem is a very real one.
3. Ok, so Sony gets a 8:64 CELL on PS4 on a 32nm process. Now what? That would be a HUGE chip--and with almost nowhere to go in size. It may get another process shrink, but you are talking about a HUGE chip. Basically the machine would be expensive, longer.
4. CELL redisigns. I am guessing future CELL variants will have beefier PPC cores, SPEs with more cache, and so forth.
With quantum computer, Diamond based chips (instead of silicone), and other competing technologies they wont stand still forever. BUT, from what I have seen they are looking at the 2015-2020 range for such a switch over. Basically whenever they hit the wall on the mult-Billion dollar investment they all currently have they will begin to test other paths.
A lot of unknowns at this point. CELL is a good idea of doing things smarter, not just harder. But it requires a rethinking of how we program applications. I think it will be great for the PS4, although we *are* hitting the place where art and budget are more important than just raw power (at least in my opinion). By then it may be more of an issue of who is providing the best tools to expand our concept of "game developer". Allowing young bright kids with ideas and rampant imaginations to work with high end tools to make the games they dream of is something I want to see. Kind of like PC mods, but more so. Fighter Maker was a good example, UE tools past and definately present/future are another.
So having a powerful platform married with even more powerful creative tools is really important. And that is where I think Sony has really got an ace in the hole. CELL will give a really nice transistion to PS4 for developers with hardware ideas they are familiar with and tools they are pros at using. The power is just the icing on the cake! (Btw, not that MS is shabby in tools...)
