Canadian Mac Forums at ehMac banner
1 - 6 of 6 Posts

433 Posts
Discussion Starter · #1 ·
- iPod mini will get a price reduction when shipped worldwide in April due to less than exciting sales in US.

- the iPod will not get a form revamp this year but will get an update adding more functionality (pitch control, etc)

- Apple displays will finally get a new form to match the G5's

- G5 will break 3 ghz

- New iMac

- "iBox" by MWNY. So many other companies are starting to do some form of this now. If apple doesn't get in soon, they won't at all.

well, those are mine! what are yours? i love this stuff...


Premium Member
31,036 Posts
3 gHz has some special problems so I suspect 2.3 then 2.6 and not 3 until next year.
Besides clock is a minor factor.
The architecture of the G5s is excellent and "waiting" is ludicrous. Apple addressed the correct factors getting an amazing balance. Getting a few more mHz is a marketing game they are well out of. Please please someone get clockless on the market.

from The real computer chip speed barrier

"Three-Legged Race

Despite all of this complexity, the core of the problem is simple: The bus is not keeping pace with the core. Specifically, the front-side bus connecting the CPU to memory runs at a mere fraction of the clock speed of the CPU's core (the part of the chip that performs calculations). Core frequency for a modern Pentium 4 chip has passed 3 GHz, but bus speeds are stalled at 533 MHz.

The result is a rather amusing three-legged race between Intel and DRAM vendors, according to Thomas R. Halfhill, a senior analyst with InStat-MDR. "The disconnect between memory and CPU has been the main problem with mainstream PCs for a long time now," he told NewsFactor. He described the problem like this: Intel can hold everyone back because no memory can move data to the processor at speeds faster than that of the front-side bus Intel builds into its chips. On the other hand, even Intel's fastest bus will not boost processing speed if DRAM vendors continue to struggle with the speed of the interface on their chips.

At the end of the day, everyone struggles, Halfhill concluded. "Really, it's just a very tricky technical problem to deal with the physics of sending more and more memory down that bus. It's a very tough problem to design for." Some companies, such as Rambus and Mosys, have built their business on improving the basic design of memory chips, but according to Halfhill, "The most promising [things] on the horizon are the gradual improvements in double- and quad-speed data rate memory" from conventional DRAM makers.

Get Smarter, Not Faster

But those who decide how chips should be designed, known as chip architects, say the emphasis on bus speed is somewhat overblown. David Patterson, Pardee professor of computer science at the University of California, Berkeley and the father of RISC computing, told NewsFactor that it is less the speed mismatch between components that counts, and more the fact that the processor is not designed to use DRAM intelligently.

"DRAM architecture simply has not kept pace with processor design," said Dr. Patterson, leading to high latency. However, "If processors could do more tasks in parallel, rather than trying to do each task at a faster clock rate, more would get done overall." He noted that Intel's attempt at putting multiple virtual processors on a single chip, called hyperthreading, is a step in the right direction. Patterson's own research into what he calls Intelligent RAM uses a computer chip that is almost entirely RAM and a bit of logic. Patterson says this chip would be ideal for PDAs, in which most or all of the instructions and data a processor needs can be kept on chip at all times.

What Color Is Your PC?

Of course, the 64-megabit question is: How do you tell exactly which bottleneck is bugging a user of a particular computer system? There are few good comparisons of overall computing performance. The Standard Performance Evaluation Corporation, or SPEC, based in Warranton, Virginia, has for years provided what has come to be regarded as the standard measure of basic chip performance, by running a variety of sample algorithms against each new processor. But just about everyone in the semiconductor industry knows the tests do not really model what a user might want to do.

How bad is the divide? Halfhill noted that cache memory has grown so large that some processors can hold the entire test suite in on-chip memory, masking any bus speed problems. And John D. McCalpin, a senior scientist with IBM, wrote in a recent comparison of benchmarks and observed system performance that there is virtually no correlation between real-world applications and standard benchmarks. His conclusion: "In order to make this quantitative evaluation, we need to gather and share a lot more data from real applications."

So, are bottlenecks going away anytime soon? Probably not. But the annoyance they cause can be minimized by savvy computer buyers who know which questions to ask. "

Premium Member
1,439 Posts
Anyone have any idea when the dual 3g is due out?

I'm saving for my first mac right now, and hoping to buy when the 3g comes out... not sure how much longer I can wait, though! All I know is that it's supposed to be out "by summer 2004".
1 - 6 of 6 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.