Originally posted by: babyjocko
Originally posted by: VirtualLarry
...but there are enough reports from otherwise-experienced builders, that can't seem to get things to work right, or find strange incompatibilities that shouldn't exist
It has always been that way. In fact, it used to be MUCH worse back in the day. Back then, there was no such thing as Plug n Play or USB. And, incompatablity ruled the day. Not only did you have to configure each and every (ISA) card, but there was no such thing as IRQ sharing and only a few IRQs available. Then, you had to hope there were no memory space conflicts. There were no standards. And, that, along with sparce, poorly translated documentation made it much more difficult than it is today.
I don't know about that. Nearly every I/O card and mobo that I purchased, had a decent-sized manual with it. Sure, some of them contained rather cryptic and deplorable emulations of the english language, but generally they had plenty of diagrams so it didn't matter all that much. The thickness of the manual would generally correspond with the quality of the product, and the reputation of the mfg.
I kind of agree with you, in that there were possibly more "gotchas", in terms of wierd obscure incompatibilities that gurus with years of experience in systems-integration would know, but any new purchaser of components shouldn't need to know. However at the same time, the number of components on the market now is likely greater than before, so even though there may be fewer problems as a percentage of component models on the market, that number is still likely larger today. (But that's a bit besides the point that we were discussing.)
I kind of liked non-PnP ISA cards though, at least I had full manual control over what IRQs went where. Too bad that most PCI cards don't have a manual jumper to select which PCI INTR that a card should use; there are very few that require more than one. The way that most mobos are wired, various PCI slots have their INTRs wire-OR'ed, in a kind of staggered pattern. Given the mobos physical INTR-line mapping between PCI slots, and the ability to choose the INTR used by the device in the slot, would allow you much greater manual control over which devices physically share INTR signals, and could well alleviate the sorts of contentions that we see today. The solution to choose which PCI INTR line is mapped to what software IRQ isn't good enough, because that affects all of the devices hooked to that INTR line, and doesn't allow the devices on a per-slot basis to choose which INTR line that they are physically wired to. (There are some very ancient, likely 5v PCI cards that I've seen, that do have IRQ jumpers. Not too many though.)
Originally posted by: babyjocko
Furthermore, the industry was very young. So, what you may see today as "beta-test products sold on the open market because of competitive pressure, eventually leading the entire market to be in "permanent beta" phase" is much better today than it used to be because even though technology is rapidly developing as we speak, it is far more mature than it was a decade or more ago.
I think that you are contradicting yourself slightly. The reason that I stated that the market is in "permanent beta-test" mode, is because the products
are not mature, and they
do not work correctly, out of the box. How in the world is that a good thing? "New" technology does no good for me, if I can't get it to work correctly. (On the other hand, putting the name of that new technology on the mobo or component's box and marketing, can certainly help to sell new hardware. AGP 8x, anyone?)
Originally posted by: babyjocko
A new product with new inovative technology made by a reputable, experienced manufacturer will be many times more reliable than a product with new, inovative technology made by a company that was born yesterday. Back then, everything was evolving; Every manufacturer was new; Every product was extremely inferior when compared to today's standards.
Hmm. As far as technology immaturity back then, I kind of agree, but I'm not sure that I agree on quality. There used to be a lot more profit margin on components back then. Rampant cost-cutting has decreased any possible margin of added quality to nearly nothing.
Originally posted by: babyjocko
So, if you haven't done this stuff 15 years ago, then you will never know the differences between the product quality of today and yesterday. You can only go by what I or someone else who was there tells you. And, I'm telling you that product quality today is thousands of times better than it ever used to be. Therefore, 9 times out of 10, it all comes down to the end user not having the clue to get it right.
I agree that the *engineering* quality of most products is better, due to better tools, but I don't think that I agree in terms of *mfg* quality - look at most mobos back then, from "reputable" mfgs, and then look at them today. Where are the fields of tantalum caps on the board, like the days of yesteryear? Why are today's boards dying due to "bad cap" syndrome? Why are so many PSUs of the day dying, and taking components with them? (Ok, granted, power demands of modern components are definately higher, but still - if the quality of mfg of today's components were higher, or even at least equal, then we would see the requisite quality inherent in the PSUs too, and they wouldn't be blowing up.)
I remember earlier days, the days of AST computers, Leading Edge, before it was "Dell or the highway", like it seemingly is today. And those PCs back then, were built like tanks, more or less. I have a 486 server box, that weighs.. I have no idea, but a heck of a lot. It's heavier than modern full-tower "server" ATX cases are. (That's another measure of the relative quality of systems, in terms of cutting-costs - old AT-style cases were often made of far more metal on the whole, and a lot more structurally-stable and better-cooled, if a bit bland in design.)
I do agree that the engineering quality, more or less, is slightly better. Back then, it was entirely possible to buy two different "IDE" drives, and not have them work together properly on the same IDE channel. OTOH, even today, you can buy two different DIMMs, and not have them work together either, although RAM was always a touchy thing, and it was always best to match RAM if possible.