Xbox to have 3x processing power vs. PS3??

<A HREF="http://gamesdomain.yahoo.com/feature/117755" target="_new"> clickable linkeage</A>

Man.. sometimes you know, I just wanna ignore all articles until the actual hardware is released.. so just many opposing stories out there.

-------
<b>It's a man's obligation to stick his boneration in a women's separation; this sort of penetration will increase the population of the younger generation.</b>

Well that is a true point I was about to bring up:
The branch prediction is a key component of x86's success and, well, problems, in modern CPUs. Take the Pentium 4, any wrong prediction creates a deep flushing of the pipeline. As if 20 cycles wasn't enough to refill the P7, the Prescott makes it even worse with 31 cycles, that's over 1.5 times more latency to even START executing code again! Of course they improved the prediction history (well I seem to recall they did) and probably the algorithms, but the point is, you're predicting, since x86 is all based on the CPU's intentions with playing with the code it's given.

So back to my first statement, the branch prediction was needed for x86, however Sony is moving towards a somewhat EPIC-like architecture, where you instruct the CPUs what to do. With such control, prediction becomes much less needed.
I just read a bit more on branch prediction on the Itanium. It seems that while branch prediction can be disastrous, EPIC brings almost half less penalties. So, perhaps I was a bit too hasty claiming no branch predictors are needed. It seems yes they are, but at a much less significance. If anything, I think Sony and IBM chose to make the main core have predictions which, perhaps explaining this choice of disposal of BPs, are split into each available core, so each tests the outcomes and reports the results or stores them, awaiting for the result of the input from the oncoming data which will reveal which of the branches were true. In such case, you can bear having to flush the pipeline because you have 7 other cores available, or at least up to 7. That would be pretty cool actually, because the cores act as slaves that you ask to try things for you, short of doing the wrong choice (sort of like food testers haha!) and when you get the actual result, you know which core has the right answer and simply call it to execute the code resulting of the branch chosen.

Now whether or not the X360 is stronger is not always obvious. The article (which I have only just scanned through) seems quite founded with plausible data, but with facts like "278.7GB/sec" of bandwidth, you must take the time to decode the actual truth behind it: memory bandwidth combined with onboard memory bandwidth. That's like claiming my system has well over 40GB/sec of bandwidth because I decided I measure my bandwidth by adding my memory bandwidth to my cache's bandwidth. EDRAM, while in fact RAM, is a sort of huge cache nonetheless anyways, making my comparison fairly accurate. (ok it's not SRAM, but it's still very fast nonetheless)

--
The <b><A HREF="http://snipurl.com/blsb" target="_new"><font color=red>THGC Photo Album</font color=red></A></b>, send in your pics, get your own webpage and view other members' sites.<P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 05/25/05 00:43 AM.</EM></FONT></P>