Participating in this "last mile communications revolution" is much more about working with sluggish local governments...
Message #137274 from stak at Jun 13, 2001 2:08 AM
"Tim, >>For many of them, it makes little sense to buy a blazingly fast PC if overall throughput for many intended uses is throttled by the dial-up speed. Compression can only go so far (typically not much more than 2x), so a faster CPU is not of much use.<<
"Do you feel Microsoft's .Net will put more demands on the processor and make people feel that it's necessary to upgrade? "
I follow the ".Net" thing through an interest I have in Smalltalk. A guy I respect, Dave Simmons, is a principal developer of Smallscript, a Smalltalk-like core ingredient of .Net. Having said this, I expect most PC users to say "Huh?" to this new thang called ".Net."
Maybe a year from now the average user, and the average corporation, will have heard of it. For now, though, it's a curiosity, barely on the edge of the radar screen.
(How many readers *here* really know what .Net is supposed to deliver? And of those, how many care what a product is _supposed_ to deliver, as opposed to what is actually shipping?)
">>I figure this is why a _lot_ of individual users are shrugging at the prospect of replacing their systems for a marginal improvement in speed, and why many corporations are deciding to save money by not upgrading the systems they installed in droves in 1999 in preparation for Y2K.<<
"I think so too, it's this apathy that's killing sales. There has to be some discernible difference to make individuals fork out. Moving up the GHZ range to 2.0 and beyond will make the improvement even harder to notice. There's not much way around it."
"I wonder if for how many of the corporations it's just a case of having no money left to spend on upgrades?"
Yeah, I agree. For a lot of corps with 500 MHz machines sitting on the desktops of their secretaries (excuse me, "human interface engineers") and purchasing agents and all, it's not a compelling decision that they upgrade all of these machines to 1 GHz and beyond.
">>People who have access to cable modems or ADSL are getting this, but I would expect those people are slightly _more_ likely to get faster CPUs than those who are still stuck with 56K or slower dial-ups.<<
"I would have agreed with you about "those people are slightly _more_ likely to get faster CPUs" in the past. I think those that have just got broadband are pumped up and they want to see what a faster CPU can do for them too. But after awhile( when they have to use dial-up again) many realize that the pipe is more important to them than the CPU. So they(myself included) are tied into 600$/year instead of ~$275, which is around $600 over a 2 year period that could have gone to upgrading to the fastest CPU."
I agree. And this is why I think the Intel focus on "communications" is a misleading _swamp_ to contemplate.
For maybe a hundred million households in America, not to mention Europe and Asia, high bandwidth connectivity is a pipe dream. Known usually as "the last mile," this is a problem of copper cable, fiber optic cables, satellite downlinks and uplinks, whatever.
(Note: I don't count DOCOMO and other cellphone standards as part of this communications revolution. I count my Net connection in this universe, but not my cellphone. Your mileage may vary.)
Participating in this "last mile communications revolution" is much more about working with sluggish local governments and local infrastructure providers than it is about silicon chip production.
"One note: Our area has had blanket high speed availability for 4-5 years and it changes the way of thinking about computer use. Many still use slower PCs because their usability comes from the net, not from their apps on the hard drive. "
Yep, this was my point. Not a lot of reason to have a 1.5 GHz machine if the dominant use is constrained by waiting for Flash downloads... (I concede that a 1.5 GHz machine may be needed for some users: Quake 5, Autocad, movie editing, etc.)
">>In contrast to your point, I'd expect those with slow connections are in no hurry to upgrade to the fastest CPUs. It's probably not a matter of either/or.<<
Yes that could very well be, I'd like to see a survey on what PC users value more nowadays, processor speed or bandwidth speed. And why or why aren't you upgrading to the fastest CPUs. "
I expect someone has done such surveys...and I expect I already know the results.
A leading PC architecture guru friend of mine, who was at the lunch arranged by Paul Engel after the Intel shareholder's meeting, was telling us that he is perfectly happy with his "40 MHz Motorola CPU" machine. (I didn't ask, but I assume this is an older--very old, actually--Macintosh.) His point was that he's not editing movies or designing chips, his downloads are more constrained by bandwidth than CPU power, and that his ancient 40 MHz machine does what he asks it to do.
While I admit that a 40 MHz non-PPC Mac is just too sluggish to me (I'm now working on a 400 MHz PPC Mac, a G4 Tower), his point probably applies to tens of millions or more cases. Once a machine is fast enough to handle all reasonable intended uses, many users have little interest in upgrading.
Machines became fast enough to handle 30 frames per second, HDTV-resolution video a couple of years ago. (Modulo download speeds, of course!) It's not clear that most users will feel the need to upgrade to speeds beyond this.
Except, of course, that anyone buying a new machine will naturally seek the fastest speed for the price. (That is, anyone looking to spend $$1000 on a new system will of course pick the 1.2 GHz Athlon or P3 or P4 over a 600 MHz machine. But how many will decide their 600 MHz machine is "too slow" and decide to spend $1000 and hours of upgrade headaches to upgrade? This is a much different issue.)
Lest you all scream at me that I am a heretic. I'm not arguing against faster CPUs on vague grounds that existing machines are "fast enough." What I'm saying is that all of us, including the bean counters within small businesses and larger corporations, have to decide when it makes sense to upgrade existing systems. For my friend the noted PC architecture consultant, his 40 MHz 68030 or 68040 is apparently "fast enough." For many others, including some SI subscribers I know here, 400-800 MHz Pentium 3s are "fast enough."
Only a very few computer users are really constrained by CPU speed. (I'm not counting servers and other infrastructure providers, where CPU speed determines throughput. Hence the success of the Xeon and multi-CPU market.)
Part of my recent skepticism about Intel's new focus on "communications" is related to the "last mile" problem: the real communications bottleneck is related to our (America's) infratructure. Old copper cable running down country lanes, draped through decaying urban corridors, etc. Until ADSL and cable modems are more widely available, or, better yet, fiber optics, then most PC users will be more constrained by communications speeds than by CPU speeds.
And laying copper or fiber optic cable is not an Intel specialty. Hence my skepticism that "communications" will be a massive market for Intel.
Enough for now.
--Tim May |
