Will Microsoft trump CELL by using Proximity Communication in XB720 CPU?
- Thread starter Brimstone
- Start date
At the time the 360 launched, I think x2's, even the lowend ones, cost about as much as the entire system. AMD had no trouble selling its chips at the time, why would it have given any to microsoft when it could charge a premium in the PC market? (I guess they could have farmed out some manufacturing to IBM, but then AMD wouldn't have seen much back out of that)
512KB X2 I believe fits in about the same die size as the 360 xcpu, and doesn't lose much in performance. It's likely even a 256KB X2 (per core) wouldn't lose that much performance compared to the cost savings.
That, or they could have dumped some 65nm Pentium D's into the system, but I think Microsoft learned the problems with that last time with the Xbox using cheaply dumped Pentium 3's (from the last time AMD was kicking intel's ass) as the prices only stay chip for as long as Intel is stuck with the supply.
If Microsoft doesn't go Cell, they'll probably either throw money at IBM to design something that will come out similar to a multicore Power processor or throw money at a cash starved AMD OR throw money at Intel if they're secretly extremely disappointed with the X360 CPU.
inefficient
Veteran
Much ado about nothing
Jeez, I just noticed that article is from Nov 2003.
Jeez, I just noticed that article is from Nov 2003.
Yeah, thanks for pointing out the cache thing. I forgot about that, but it's very true. I was qouting sizes for 512kbX2 but 256kbX2 dont lose much performance and would likely be similar in size to Xcpu.
But, I still dont know if AMD, like Intel, would refuse to let MS own the IP so they can farm out fabbing for cost reductions, and that's critical.
Price of X2 is irrelevant though, the true cost comes down to die size which we've established as similar. If AMD let MS run with the chip fabbing, then it's irrelevant. Plus it would gain AMD publicity, volume, etc. Just like what it gained IBM. I gather IBM makes a pittance on these consoles, doesn't matter it's still worth it to them. Keeps them relevant after losing the Mac contract. Same type of boosts would apply to AMD.
Actually this can produce a massive performance increase, even higher than what is suggested by FLOPS. In the link I gave you, a single SPE outperforms a P4 by a factor of 3. Cell is not a stupid design like many poorly informed detractors insist. There's no reason to believe that Cell FLOPs are not comparable to conventional CPU FLOPs.
And there's no efficiency argument that will save you from a order of magnitude difference in power, at least not with Cell. Cell2 will blow the pants off of any OOOE based CPU in performance with few exceptions.
Except that Athlon 64's are not licensable products, and they are made by a dedicated fab owned by AMD. Getting one of those will cost a fortune. If you limit yourself to only IBM's chips, you're options are far worse off. Not to mention that the Athlon 64 architecture cost something like a billion dollars to design.
I don't think we'll be running AI or meaningful physics on GPUs any time soon. People have pointed out that GPU-based development is even more difficult than Cell-based development.
You've also went from a triple-core to a dual core and about a 33% decrease in clockspeed. It's unclear if performance went up at all.
They won't, and even if they did, they won't perform the same with non-dedicated fabs.
It's not irrelevant. If these chips can be sold at a better profit elsewhere, neither AMD or Intel would even think about giving them to MS. The only reason IBM makes console CPUs is because they can't sell them anywhere else.
Many of the same things are said every generation, hell they were even said 6 years ago when talking about a potential future PS3. And of course PS3 as it is today would be unfathomable back then. Everybody would be screaming about how to cool the thing. Alas it's not built using fab tech from way back in 99 though, that's the difference.
You can still have rather high performance per watt and have high raw performance, remember we're talking about a console which is nearly a decade away from releasing. So while current fab technology today does not allow for Cell based IC's with let's say 10X the overall power over the current BE(not that it's all I'm expecting it's just a bare bones certain leap in power) to be implemented in a PS4 console today it will be possible in the 2013+ timeframe. I just don't see fab technology to be slowing down anytime soon. If Vince were still here I know he would expand further because he is big into fab tech, I have a feeling he's still here though heh.
I am sorry this is not a PS4 thread so I will not sidetrack the discussion anymore, time will tell though.
Won't fab tech be slowing down sometime soon? I thought 32nm was about the time some quantum barrier was reached that required everything to be SOI to go any lower, and much farther after that and you probably can't use silicon anymore.
Of course, successfully switching to something else could even speed up fab tech development, such as magnetic semiconductors or just some other semiconductor besides silicon. Maybe carbon nanotubes will come out of left field.
I've heard no such thing. If you're referring to gate leakage, I think that was solved by the introduction of high-k dielectrics. There is a problem with the photolithographic aspect of sub-32nm processes, namely the lack of an effective photolithographic system that engrave patterns at the 22nm and below. Originally, we wanted EUV (extreme ultraviolet) systems to do this, but EUV isn't nearly ready. We're instead talking about stuff like double patterning and double exposure, but these are significantly more expensive methods. I think up to 22nm we're OK, and anything below that is too far in the future to predict.
Not by the timeframe we're talking about.
Silicon still has legs left in it. At 22nm, it isn't an intractable problem, just a nasty and ugly problem. 16nm and under are more of a question mark though.Take a look at what's going to ship in PC CPUs in the next 2 years. That's what kind of technology MS will be using in the X3 (if the X3 will even exist).
Personally, I think MS will try to get Compaq, Dell, etc. to make the hardware and MS will just specify the minimum hardware requirements. The X3 will be a PC with certain guaranteed features and performance.
Personally, I think MS will try to get Compaq, Dell, etc. to make the hardware and MS will just specify the minimum hardware requirements. The X3 will be a PC with certain guaranteed features and performance.
What? They've been going further and further from that route with every xbox. They'd need software unification before they could do that anyway, and they'd doom themselves in the console race if they didn't have something affordable and powerful to sell.
You keep ignoring the main point: The Athlon X2's performance is out-of-the-box, immediately available to everybody. Cell's performance lies in a chest with seven padlocks in the highest tower of the castle, complete with wall, moat and a dragon in front of it.
First and most obvious, they've only made one move with the Xbox hardware. Don't word things like they have a long, well established history of diverging from PC space.
Second, MS has been working towards unified APIs for quite some time.
Third, stop looking at Xbox trends and look at Windows trends. How much more streamlined and integral can gaming get to Windows without PCs being Xboxes than run Windows as well?
Fourth, look at PC price trends compared to console price trends.
Fifth, look at how much top end power you're losing with PC parts over console parts and what difference that power makes in the final game.
Sixth, look at the Home & Entertainment's continued struggle to sell hardware at a price high enough to break even.
There are trends in favor of my hypothesis.
Except that this is a console, and developers have 5-6 years to figure it out. Once they do, it is reasonable to assume that the Cell will come close to its peak performance, leaving the A64 X2 in the dust for most situations.
They will go there, of course, but that doesn't really tell us about what new fundamental technologies they might employ in its design, which I think was almighty's question.
Not quite. You still make the step up to concurrency if you want its full performance, and you can still get your hands dirty with optimization and lower-level code. I'm sure you can get a given level of performance more easily, but I guess the argument would be that while the peak is easier to reach, it is also lower.
Raytheon has already made a monster CPU for DARPA that smokes CELL.
http://www.internetnews.com/dev-news/article.php/3667476
http://insidehpc.com/2007/03/26/raytheons-monarch/
http://www.internetnews.com/dev-news/article.php/3667476
http://insidehpc.com/2007/03/26/raytheons-monarch/
Last edited by a moderator:
But that was a pretty significant move. From using external IP in the form of discrete chips, to owning all the core IP of their console.
And APIs are used to isolate lower level implementation details from higher level systems. Implementation details like architecture.
Right, look and the end price for PS2 compared to the cheapest PCs. Consoles start highish in price and then quickly cost reduce like all other mass market CE devices. With the XBox and PS brands battling for market share cost reduction is absolutely imperative.
The cost-reduction enabler is integration (and economies of scale production).
In the console market you don't win by maintaining a high price on your hardware. You win by lowering cost which enable price reductions which results in massive market share advantages, which in turn results in revenues from game license royalties. In the proces, try not to hemorage to much cash.
Nope
Cheers
Where does it says that?
All they're trumpeting is that they're more power efficient than Cell or Xeon.
Also, a flops rating on its own is a meaningless value. We need to know about the IPC and the supported instruction set and the real cycle cost of every ops.