I'm trying to settle a dispute with a friend.
Which has the higher bandwidth?

PCI (64bit 33 or 66MHz)
Firewire.

PCI 32bit/33mhz is usually considered around 133 megabytes per second, or 1064 megabits per second

PCI 64bit/33mhz or 32bit/66mhz is 266MBps or 2128Mbps.

PCI 64bit/66mhz is 532MBps or 4256Mbps.

PCIX is 64bit/133mhz, and is 1064MBps or 8152Mbps.

1394 is 50MBps or 400Mbps

1394 rev b is up to 200MBps or 1600Mbps

Of course, that's all theoretical maximums...

In theory, the USB interface on the car player should be 2mbps faster than Ethernet, but it's not...

How in the world did you get into such an argument? (!?)

Calvin

USB depends on implementation. Different classes carry different speeds. The custom class for the empeg isn't as fast as Mass-Storage class (this info from Hugo, along with the actual average speeds to expect).

I bet the argument might have had something to do with deciding whether or not you would see full performance from using a PCI Firewire card. If so, the answer is still "maybe" Using it for a single hard drive you'll still be limited by the actual ATA disk anyway.

Bruno

We were discussing bandwidth and he was extolling the virtues of firewire. I said it was a fine external bus because of PnP, hotswap etc and somehow he didn't think that PCI was faster.
It's hard to argue with someone who makes non-statements like " but it must be faster because Multimedia professionals use it"
??? He couldn't explain the relevance between the two statements and was happy to get sidetracked on that.

I'll be happy to point him to this thread. coz I was right

Thanks guys!

I'll be happy to point him to this thread.

Point him, too, at the thread on linux-kernel where Jeff Merkey is complaining that his 3ware IDE raid cards start misbehaving when his 220MBytes/sec disc bandwidth starts to saturate PCI64/33...

Nice problem to have, if you ask me...

Peter

and somehow he didn't think that PCI was faster.

And how did he expect the computer to use a fast interface when every implementation of firewire controllers on computers has been via the PCI bus somehow? (Or AGP with the ATI card)

He was probably mislead by the whole megabyte and bit problem. I have always seen PCI numbers in megabyte, and firewire as megabit. And lets see:

Megabyte:
Hard drives
AGP

Megabit:
Networks
USB
Fibre

Internal firewire has always puzzled me. Sure it would be ok for a hard drive, but start introducing two or three fast ones and the limit of the current spec is hit. SCSI is about to hit 320MBps, only because some Ultra 160 drives when operated off an array can tax the bus.

Oh, and PCIX has speeds of 50mhz and 100mhz as well, but I couldn't remember the bandwith calculation. PCIX had some other nice features, but I can't remember what they were.

Keep in mind, though, that the vast majority of PCI busses on PeeCees are 32bit/33MHz (someone correct me if I'm wrong; maybe only the actual pluggable part of the bus is 32/33?), which would mean that you're wrong. Of course, the FireWire adapter is not likely to be connected to the computer anywhere other than to the PCI bus, so it's not like any speed improvement due to FW would be usable.

Yep, for the most part PCs use a 32/33 PCI bus. The biggest problem can be that one device can make the entire bus slow down. But, there are some new PCI controller chips that have 2 busses off one chip. That should help to get a few faster slots in future motherboards. Unfortunatly Intel chose to make an AGP slot instead of pushing better PCI implementations to consumers. The reason I day that is due to the fact that video cards are the only common device in a consumer PC that could push the needs for a faster bus.

1394 reb b though will need a faster bus or risk flooding the PCI bus. Keep in mind Apple is probably going to call this interface "Gigawire". So be ready for another confusing round of 1394 naming schemes.

I don't know much (read ``anything'') about PCIX, but AGP has some nice features beyond what PCI can deliver, like DME and, uuhhh ... some others that I forget right now.

Edit: Ooh! Sideband addressing, which is nice. Not to mention that there's something to be said for being the only device on the bus.

The reason I say that is due to the fact that video cards are the only common device in a consumer PC that could push the needs for a faster bus.

True. The only other thing I can think of are harddisk controllers, more specifically SCSI controllers. The last Adaptec SCSI controllers (29160 and up) all have 64 bit PCI interfaces (though they also work in "normal" PCI slots).
IDE won't make that much of a difference anymore, since the limits of that interface have pretty much been reached. But maybe serial ATA might ? I can tell you, that's one piece of technology I'm looking forward to. I only hope that they develop it properly so it works as promised from the very beginning. Otherwise we might have a second USB débâcle on our hands.

Quick summary PDF on PCI-X

PDF slideshow about PCI-X, fairly detailed

And the overall PCI-X page that Compaq has

The PCI Sig page had lots of "Member Only" Info, so I just dug the above info off one of the members.

The whole one device on a bus is nice, but I dislike the way it's done with a propritary slot compaired to the others. Why not a network slot, sound slot, and storage slot next? They could all benefit frol their own bus, but noone does it this way. Even the CMR and such slots tie into the PCI bus. And on servers, PCI-ES slots are mainly for SCSI or array controllers, but they are still normal PCI components. (PCI-ES has 3 SCSI connectors routed to the PCI slot to allow for redundancy in controllers without the need to move cables or do duplexing tricks at the back of a hard drive cage).

Thanks for the links on PCI-X. I'll take a look at them.

As far as the separate bus thing, I tend to agree with you. But the consumers tend to push the market in this area, and really, really fast graphics cards are the things to have these days, and they do push a lot of bandwidth, especially when you consider all the texture maps and whatnot that they're pushing around. And separating that off from the rest of the data being pushed around is actually an advantage for both the graphics cards and the rest of the devices. Of course, this has been attempted before in the name of VESA Local Bus, and that fell off the planet in short order. AGP may go the same route, but change is less likely to happen these days, IMO, because of the much larger, less knowlegeable group of people that own computers these days. Of course, since most of those people just swap computers out on the whole, that may not be an issue.

I'd also like to point out that there is a specification for a separate networking bus that can accomodate modems and NICs. CNR, maybe? Of course, no one uses it, because the manufacturers see no advantage in it, and they're right, so the consumers don't push it. (Edit: In rereading your post, I notice that you talk about CMR. Maybe you typo'd CNR? If so, I didn't know that it was just attached to the PCI bus. What possible reason is there in having a different connector?)

Oh this is sweet. Not only an answer, but ammunition!

Thanks heaps

edit: Drak, those links are blocked for some reason.

I notice that you talk about CMR. Maybe you typo'd CNR? If so, I didn't know that it was just attached to the PCI bus. What possible reason is there in having a different connector?)

The first version of the CNR slot was actually called AMR (Audio/Modem Riser). Maybe that's where the typo originated ? Anyway, here's a little info on the subject. AMR/CNR cards relie heavily on the CPU and do little more than provide the connector for the output. That, in combination with the fact that an AMR/CNR card costs practically the same as a real, self-sufficiant modem/audio/NIC card has prevented this technology from becoming a success.

AMR was originally designed so you can put all of the country specific bits and anything else you wanted onto a small card which plugged into a special slot. This would reduce the cost of the device because some of the parts would already be integrated with your motherboard.

In the UK anything that attaches to the phone line directly must be approved otherwise it is illegal to use it. So you can get your little AMR card design approved and not have to go through the hassle of approving your entire motherboard design as well.

CNR is the same type of thing but you've got network support as well.

- Trevor

I'd never noticed the technical aspects of the CNR. When I saw the prices for the cards weren't any better and I'd have to go out of my way to get a motherboard that supported it, why bother?

I love that at the same time that there's this big push to move graphics processing off of the CPU, there's a big push to move the networking onto it. I suppose that the majority of people that are okay with their WinModem have never seen how much better a real modem works. (Not that I've really used a modem much in the last ten years.)

No, but I see your point. The principle is the same. In fact it's the same with everything they try to combine into one. e.g. That's also why the most expensive stereo systems consist out of separate modules.

I guess when everything's combined, Instead of several separate little units that do a great job at what they're made for, you get one unit that does a half-assed job of everything it's supposed to do. The only pro I see is the way more cheaper price and (for developers, so what so I care ) way less chance of compatibility issues.
Also, If one function of your mobo is shot, you'll have to get an entirely new one.
There's no way in hell I'll ever consider anything that's built with the el cheapo idea in mind.

edit: Drak, those links are blocked for some reason.

Strip the HTTP off of them, they are FTP links that the BBS UBBCode toasted. Or hit the third link, they are also linked off that page.

And yes, I was intending CNR. There are a few standard for this type of slot, but basicially yes they are just connectors for the actual devices integrated somewhere on the motherboard.

I'll admit, AGP is nice, but thats only after a few revisions. If only that effort has been towards the standard... PCI-X for example is just as fast as current AGP implementations, and no extra support from a PCI capable OS is needed. I'd rather have a "Fast PCI" slot on it's own bus on my motherboard instead of a slot on it's own bus restricted to one device. Well, I'd also be happy if PCs even saw 64 bit PCI slots someday soon.