How Much Longer does 32 Bit, 33 MHz PCI Have Left?

[BDawg]

32 Bit/33 MHz == 256 MB of bandwidth, correct? With drives getting faster and faster (hopefully, we'll break the 70 MB/sec top end with Serial ATA), sound getting more and more complex, and network cards nearing consumer level gigabit, how much longer does PCI have?

I suppose that's not even mentioning Firewire 2 and USB using the same bus (do they?). Are there any alternatives or advances to PCI in the works? I know that most servers are including 64 Bit slots now, but that seems evolutionary rather than revolutionary.

 

[paralazarguer]

70MB/sec? What are you talking about? We now have 133mb/sec transfer between the controller and the cache in the hard drive. Yes, usb and firewire travel across the PCI bus as does pretty much everything else except AGP and memory. There is a new standard in the works called PCI express or 3GIO *third generation IO* Read this

 

[BDawg]

Originally posted by: 7757524
70MB/sec? What are you talking about? We now have 133mb/sec transfer between the controller and the cache in the hard drive. Yes, usb and firewire travel across the PCI bus as does pretty much everything else except AGP and memory. There is a new standard in the works called PCI express or 3GIO *third generation IO* Read this

Sure, ATA-133 can handle 133MB/s, but the non-burst rate for the actual hds is under 70MB/s.

 

[GoodRevrnd]

Is the 256 MB/s per slot or for the whole controller?

 

[BDawg]

I believe it's for the entire bus; I might be mistaken though.

 

[GoodRevrnd]

I was thinking entire bus but that kind of doesn't make sense. A firewire alone would kill that at 400MB/s... I guess? So do our firewire cards not accept data at peak?

EDIT: Ooops I'm retarded firewire is megabit, not megabyte. Still, I think we are definitely nearing a limit, esp. w/ Serial ATA debuting at 150, going to 300 quickly thereafter, and eventually 600.

 

[BDawg]

Originally posted by: GoodRevrnd
I was thinking entire bus but that kind of doesn't make sense. A firewire alone would kill that at 400MB/s... I guess? So do our firewire cards not accept data at peak?

Firewire is 400Mb/s, that's bits, not bytes. Divide by 8 for about 50MB/s.

 

[Bullhonkie]

32-bit 33MHz PCI only has ~132MB/sec of bandwidth if I remember right.

33,000,000 * 32 = 1,056,000,000 bits/sec
1,056,000,000 / 8 = 132,000,000 bytes/sec

Hopefully my math is right and I didn't butcher something there.

Things like Ultra160 SCSI and gigabit ethernet can pretty much saturate the whole bus by themselves, especially once you take into account any overhead on the PCI bus. Which is why you'll usually see those kinds of peripherals as 64-bit PCI cards.

 

[GoodRevrnd]

Originally posted by: Bullhonkie
32-bit 33MHz PCI only has ~132MB/sec of bandwidth if I remember right.

33,000,000 * 32 = 1,056,000,000 bits/sec
1,056,000,000 / 8 = 132,000,000 bytes/sec

Hopefully my math is right and I didn't butcher something there.

Things like Ultra160 SCSI and gigabit ethernet can pretty much saturate the whole bus by themselves, especially once you take into account any overhead on the PCI bus. Which is why you'll usually see those kinds of peripherals as 64-bit PCI cards.

Hehehe, that's what I calculated for PCI bandwidth but decided not to post it at the risk of sounding stupid again.
So it sounds like each slot has it's own amount of bandwidth, otherwise the 64bit slots wouldn't help that much?

 

[boi]

nm

 

[paralazarguer]

well, you could get the answer to all of your questions in the link that I provided. PCI shares its bandwidth across all slots. It's not a per slot thing.

 

[paralazarguer]

While having 133MBs of bandwidth can easily handle a few hard drives, a network card and an optical drive, things become a problem when you have many devices sharing the same I/O bus. Take for example gigabit network cards. Each card is capable of sending and receiving 100MB/sec - still within the realm of PCI, but put in two gigabit cards however, and you have a situation where the PCI bus is being heavily saturated.

Mainstream technologies like IDE RAID are also contributing to PCI's death. A single hard drive is pushover for PCI, but when people stripe together 4 or more drives, transfer rates can climb up to 100MB/sec easily. Add a network card, a sound card and maybe a SCSI card and you can begin to see the strain.

 

[paralazarguer]

once again here's the link.

So it sounds like each slot has it's own amount of bandwidth, otherwise the 64bit slots wouldn't help that much?

64bit PCI is faster because it doubles in parallelism. Your logic doesn't make sense. PCI's 133mb/sec of bandwidth is SHARED!!!!!
Seriously, read the article and don't forget to click on the little blue arrow at the bottom. As you'll see, PCI is horribly out of date and some of the main problems with it are with its parallel nature which takes up too much space on motherboards and can't use devices outside of the computer without the use of something like USB or IEEE1394. There's a lot more to it than that. read the article. Guessing, especially wrong guessing is pointless. Just read up on it.

 

[TNTrulez]

I say a couple more years.

 

[Pariah]

It's got a long way to go. No mainstream boards have anything but 32bit now, with nothing faster imminent either. Once something faster does come around, it will be 2 or 3 years of backwards compatibility before we see 32bit really start to wind down. Look how long it took to get rid of ISA. There's no reason to get rid of 32bit completely as it will be useful for quite some time to come. There are a number of faster PCI variations currently available on server boards besides just 33MHz 64 bit, including every combination of 33MHz/66MHz, and 32bit/64bit, as well as PCI-X running at 66MHz, 100MHz, and 133MHz. At 133MHz 64bit, PCI-X is capable of 1GB/s which is well beyond what a typical user will need for some time to come, especially if the video card remains seperate on AGP.

 

[paralazarguer]

No mainstream boards have anything but 32bit now, with nothing faster imminent either.

OH MY GOD!!!!!! Why do people insist on just guessing and making stuff up rather then learning and knowing the facts!!??

I guess you can lead a horse to water but you can't make him drink.

Click on the word drink and learn something dammit!

 

[Pariah]

2 years off is not imminent. To my knowledge there have not been any single CPU consumer level products (actual products or chipsets) announced with either 64bit or 64MHz PCI let alone anything faster. Clawhammer may change that next year, though it doesn't appear it will with the first gen boards. There isn't need for anything more in the consumer market currently. If by some fluke you do, there are server board options out there for you.

 

[paralazarguer]

The article is 6 months old and it says 1 year to 2 years. that's 6 month to a year and a half. That's pretty freakin close.

 

[Deleted member 4644]

Any idea when this stuff is coming out? If its in the next year or so, I will hold off on any more hardware purchases. ....hehe thats gonna be hard.

 

[Pariah]

I can pretty much guarantee you 6 months is not going to happen. There is no way there is going to be an Athlon chipset that uses this, probably ever. Intel's roadmap is pretty much set for the rest of the year, and if Itanium II and Xeon don't have near term implementations, there is no way the P4 will. There is no reason even to release it with so many other options already available on production products. There is absolutely no demand for it whatsoever either at this point in time. If you think SATA has taken a long time to market since its announcement, just keep waiting for this. Early 2004 on a consumer product at earliest.

 

[Deleted member 4644]

Hm... Why wont it show up in Athlon products? Is it Intel?

 

[Pariah]

The Athlon platform is rapidly losing its position as a technology leader and innovator. The P4 platform has already surpassed it, and by the beginning of next year AMD will have a new platform with Clawhammer that will be far more advanced. Putting 3GIO on an athlon platform would be like putting a Ferrari engine on a gokart. By the time it is finally released, no one with a need for it will be using the Athlon platform.

 

[bluemax]

I must admit, that article on Intel's 3GIO looked awfully good. I'll be rooting for them.....
Until then, PCI gets the job done.

 

[Athlon4all]

I doubt its going anywhere anytime soon. Remember, the old 16-bit ISA bus? Its first "successors" the Micro Channel Architechure (MCA) and the EISA, were "launched" in what 1986? Yet, neither gained ground, and we really didn't see anything other than 16-bit ISA until I think 1992 or 1993 (the PCI bus). And it took until 1998-1999 for ISA to totally die, and really, the normal "Mainstream" PCI devices hardly saturate it. Its not going anywherre anytime soon IMHO.

 

[McCarthy]

Reading that article 7757524 mentioned. Are the pure 3GIO slots going to be just that little connector? Looks like an itty bitty ah heck, wonder if our expansion cards will get any smaller and if braces will have to be made for some cards if they stay current sized. Yeah, boards are tough to break, but they can break/warp. My motherboard's warped from my alpha 8045, not enough to hurt anything (yet anyway), but just shows it does happen. Then again the cards are held pretty firm by the L brackets so even a SCSI-160 card with lots of connectors probably wouldn't take too much stress. Ok, disregard everything up to this point, just thinking outloud again.

LOOKS like it could be introduced on boards without taking away any PCI functionality if I'm reading this right. Add the extra connector at the end of the PCI slot for a "low-band 3GIO slot" and someone with only PCI cards isn't hurt in the least, right? Wonder why it said only PCI 2.2 cards though. Oh well, anyway, not like fighting for board space with VLB/PCI/ISA. Actually kinda like VLB in it's day. Could have 6 "low band 3GIO" slots and have them all populated with PCI devices without using the GIO at all, right? Not clear on if you plug a pure 3GIO card into one of those extra connectors in a combo slot if it'll work or not. Would go back and reread, but there will be plenty more articles on it before it comes out.

I think I like it. Unlike going from parallel to serial ata, or isa to pci or serial to usb or or or...having 3GIO in place won't be preventing you from having the legacy PCI or require adapters or new equipment right away. That little difference will allow it to take off so much easier, with the biggest roadblock being the additional board costs of introducing it when few if any know about it, muchless are planning to use it. Seems like when it's ready to launch it should take off quickly and fairly painlessly.

Looks nifty overall. Needs a better name though. 3GIO, sounds like a virtual pet or something.

--Mc