PPC 970 Apple Motherboards?

[jettredmont]

Note, IMHO, this is completely off-topic and academic here. However, maybe Lanbrown can tell us precisely how Sun's hardware compared to Intel relates to the IBM 970 and especially Apple's motherboards for such a chip.

Originally posted by Lanbrown
Originally posted by ddtlm

IRQ?s are not a non-issue; today there are IRQ issues. Especially on those SFF PC?s that companies seem to like, like the ones from HP/Compaq. Shipped from the factory they are fine; try to add one or two PCI cards, especially some a few years old and viola, IRQ problems.

Are you sure? Have you been doing IT for an Intel-based company?

I'm not IT, but I've done my fair share of upgrading Intel-based hardware. IRQ conflicts ... gosh, I can't remember the last time I had one! I think the last one that I had was adding a Voodoo graphics card (Voodoo 1 ...) to an onboard-video NEC computer, and that one only gave me about ten-fifteen minutes' hassle (had to set the IRQ instead of letting it set itself). That would have been 1996 if I remember correctly, perhaps 1995 (about a month after the Voodoo 2 cards debuted, if anyone feels like looking it up).

I've moved aging network cards and modems and sound cards between computers without having to think about it at all since then. I even installed a combo USB/secondary PP/SP card into that old NEC before retiring it, and it was an insert/start computer/play prospect!

Again, not coming from an IT perspective, but are IRQ problems really that prevalent? I haven't seen them in years. Seems like you're working off old experience and conditioned reflexes instead of facts here.

I guess this is an attack on interchangable components?

In the PC world, compatibility is not in their vocabulary. Each company likes to do things their way and virtually everything in a PC is all based upon specifications and not standards. ATA/133 and USB are two perfect examples. Maxtor and Agere decided to extend ATA from 100 to 133.

Interesting you should use that as an example. I just bought and installed a Maxtor ATA 133 drive in my Windows machine (the Maxtor was cheaper than its competition, so it's not like I paid extra for ATA 133). I have a "standard" ATA66 controller card in there (and even more "standard" ATA33 on the motherboard which drives my optical and Zip drives). How many problems did I have with the installation?

Well, aside from the fact that I found my 3.5" floppy was dead and I couldn't use Maxtor's "recommended" non-fdisk method of partitioning the new drive (I just used fdisk anyways and it worked perfectly), installation went hitch-free. If there is a problem with Maxtor's ATA133 standard extension, I don't see it.

Most of the big players in the PC world are on the USB committee, which there are compatibility problems with USB. Why, because it?s a loose spec. They have used three different controller chips, plus add all the operating systems in, you have too many to do a through test.

Again, are you basing this on fact or superstition? What kinds of problems have you seen with USB? Aside from the "install driver before or after plugging in" per-device confusion in Windows (which is an unfortunate side-effect of MS's architecture), I haven't seen interoperability problems on USB hardware. The only hardware problem I've had with USB is a USB hub that wasn't fully spec-compliant (the last DLink hardware I've bought, I might add!) Everything else works great. And, yes, I have a USB scanner from Epson, a printer from HP, a camera from Intel, a Belkin UPS, a keyboard and mouse from MS, a flash card reader from Dazzle (not sure who actually makes the device and I've seen identical hardware branded otherwise), along with (occasionally) a Kodak digital camera and a Sony video camera. So, the USB buses (two root ports on my PC == two buses) get quite a bit of pounding. Still, no conflicts.

This is supposed to refute my claim about work per clock?

If it isn?t out yet, it?s a mute point. You are comparing their yet to be released chip to already released chips that most have been out for several years.

Actually, he was comparing the work-per-clock of the Opteron (unreleased) with the IBM 970 (also unreleased). He never mentioned the G4; go back and look at the original post.

If you have to ask why a company would use Sun, you obviously don?t see the whole picture.

Again, he didn't ask why a company would be using Sun. He said he can't see why you would use it as a desktop (which was in response to you saying that comparing Suns and P4s was valid because you use a Sun as your desktop computer).

It was the choice of Intel to take a desktop chip, make a few changes to it and call it a server chip. They made the choice; they must live with it.

Let us not forget that Apple calls G4 computers servers too. Anyone shopping for a server should understand the scope of their options quite well, and will probably already have at least a pretty good idea of the answer to the 32/64-bit question.

On the low end of servers (which traditionally do not require 64-bit procs), the Xeon is a very nice processor. One wouldn't compare a $5k Xeon to a $30k Sun system any more than one would compare a $4k Mac to an Itanium2. Keep your comparisons realistic, in other words, or at the very least acknowledge that a Cray and a Xeon have a slight price difference between them.

What does cost have to do with it? Sun sells a 2-way server and the PC companies do as well. Since Itanium servers are hard to come by and may not have the software one wants, a comparison between those two is totally legitimate.

Well, first of all "what does cost have to do with it"? Everything. Companies are in business to make money, and often making money correlates to not wasting money. If a $5k server is just slightly less fitting for an enterprise than a $30k server, the cost difference will drive the decision. On the other hand, if the difference is measurable and will affect profit-making ability, it may be wiser to go with the more expensive hardware.

So comparing year-old PCs to Suns which are essentially identical to brand-new top-end $16000 desktops is fair?

Let?s see, I also included two servers with the same number of processors. If Intel wants to use desktop processors in a server, then so be it. They all followed the same rules; there are no rules.

Correct. There are no rules. Except, of course, for the company actually paying for the hardware, which is buying said hardware today, not a year ago. Today, if you have $16k to spend on a server you'll not artificially restrict yourself to last year's model just so that Sun hardware looks better.

And now this thread has deteriorated into you defending Sun. Noone here cares about Sun.

I used Sun as an example. You are defending Intel.

So what, precisely, was the point? How does Sun's rockin' upcoming UltraSparq lineup relate to the IBM 970?