ATA 100 & ATA133

[Timothy Daniels]

"Rod Speed" CANed:

Bullshit. And there is a CAN in Timmy's question.

Which is why I answered yep to the CAN in his question.

OK, under what conditions will one ATA133 hard drive alone
be limited in data transfer rate due to the bandwidth of a PCI
bus without any other data transfer on that PCI bus from other
devices?

*TimDaniels*

 

[Rod Speed]

Rod Speed wrote

OK, under what conditions will one ATA133 hard
drive alone be limited in data transfer rate due to
the bandwidth of a PCI bus without any other data
transfer on that PCI bus from other devices?

Basically the slower PCI implementations
combined with the fastest hard drives with
the worst type of data transfer to the drive.

Yes, thats not all that commonly seen.

 

[Mike Tomlinson]

32bit/33Mhz PCI gives a max of 132MB/s.

In theory, yes. In practice, the maximum appears to be around 110MB/s.

 

[Marc de Vries]

In theory, yes. In practice, the maximum appears to be around 110MB/s.

True.
But the same applies to harddisks too. The maximum there is also
around 110 MB/s:
SATA WD Raptor reaches 105 MB/s
SATA Maxtor Diamondmax plus9: 106 MB/s
SATA Hitachi 7K250: 112 MB/s
SATA WD 2500JD: 72MB/s

PATA Hitachi 7K250: 90 MB/s
PATA Maxtor Maxline II: 85 MB/s
PATA Seagate 7200.7: 85 MB/s
PATA IBM 180GXP: 88 MB/s
PATA WD 2500JB: 70 MB/s

Because of these figures I decided not to complicate the matter by
going into these details.

But to be fair, there are also some older VIA chipsets which are known
for their low PCI efficiency. (much lower than 110MB/s, more like
80MB/s IIRC)

Marc

 

[Marc de Vries]

But you forgot to mention that the possibility of that CAN is
EXTREMELY small, and that he should not worry about it.

So I decided to give a more detailed explanation to make sure that
Timothy wouldn't draw the wrong conclusion from your oneliner.
Too bad you don't understand that.

Basically the slower PCI implementations
combined with the fastest hard drives with
the worst type of data transfer to the drive.

Yes, thats not all that commonly seen.

Well, there is that old VIA chipset that had something like 80MB/s on
the PCI bus. If you couple that with a brandnew SATA disks than it's
burst rate would be limited by the PCI bus.

At least we can agree that that situation is quite rare.

Marc

 

[Rod Speed]

But you forgot to mention

Didnt 'forget' a thing.

that the possibility of that CAN is EXTREMELY small,

Not necessarily, particularly with the slower
PCI and other stuff going on over the PCI.

and that he should not worry about it.

I'm well aware of Timmy's background and that he
doesnt need that level of comment aimed at a newbie.

So I decided to give a more detailed explanation
to make sure that Timothy wouldn't draw the
wrong conclusion from your oneliner.

Pity you mangled the story considerably.

Too bad you don't understand that.

Too bad you dont understand what Timmy also knows.
I knew he wasnt some newbie who doesnt understand
the basics and wouldnt be assuming anything from that.

Well, there is that old VIA chipset that had
something like 80MB/s on the PCI bus.

There's more than just that. Plenty of systems dont have full
32bit PCI. Particularly the sort of older system that you're
likely to want to use one of those PCI cards in because the
motherboard bios doesnt have large drive support.

If you couple that with a brandnew SATA disks
than it's burst rate would be limited by the PCI bus.

You're overstating that too, just like you did with the 133MB
PCI thruput claim. That isnt in fact that commonly seen at all.

At least we can agree that that situation is quite rare.

I dont, particularly with the older systems where an
addon card is likely to be used because the motherboard
bios has a problem with large drives, and those where
the PCI is used for more than just high thruput to hard
drives, particularly with analog video capture etc.

Not that common is a better way of saying it than quite rare.

 

[Rod Speed]

True.
But the same applies to harddisks too. The maximum there is also
around 110 MB/s:
SATA WD Raptor reaches 105 MB/s
SATA Maxtor Diamondmax plus9: 106 MB/s
SATA Hitachi 7K250: 112 MB/s
SATA WD 2500JD: 72MB/s

PATA Hitachi 7K250: 90 MB/s
PATA Maxtor Maxline II: 85 MB/s
PATA Seagate 7200.7: 85 MB/s
PATA IBM 180GXP: 88 MB/s
PATA WD 2500JB: 70 MB/s

Because of these figures I decided not to complicate the matter by
going into these details.

But to be fair, there are also some older VIA chipsets which are known
for their low PCI efficiency. (much lower than 110MB/s, more like
80MB/s IIRC)

And there aint just 32bit PCI too.

 

[Marc de Vries]

And there aint just 32bit PCI too.

Ever seen anybody with a system that has 64/33 or 64/66 pci who is
using that for videoediting AND using a single IDE disk for that?

Of course it "CAN" happen ;-)

 

[Rod Speed]

Ever seen anybody with a system that has 64/33 or 64/66 pci who
is using that for videoediting AND using a single IDE disk for that?

Completely and utterly irrelevant to what was actually being discussed.

Of course it "CAN" happen ;-)

Careful you dont end up blind, now.

 

[J. Clarke]

Ever seen anybody with a system that has 64/33 or 64/66 pci who is
using that for videoediting AND using a single IDE disk for that?

Of course it "CAN" happen ;-)

Most if not all Athlon 64 and Opteron boards have 64/66 PCI, but they
are often used with single IDE drives.

 

[Marc de Vries]

I'm well aware of Timmy's background and that he
doesnt need that level of comment aimed at a newbie.

So you thought the oneliner without any explanation was enough?
It is clear from his next reply that it wasn't.

But we'll probably never agree on that.
(Sometimes I really wonder if your aim is to help people, or if your
goal is to confuse them)

Pity you mangled the story considerably.

Pity you cannot point out where it is mangled, so we can have a normal
discussion about it.

There's more than just that. Plenty of systems dont have full
32bit PCI.

What do you consider "not full 32bit pci", if it is not connected to
the slow pci implementations like the one I mentioned from via?

Particularly the sort of older system that you're
likely to want to use one of those PCI cards in because the
motherboard bios doesnt have large drive support.


You're overstating that too, just like you did with the 133MB
PCI thruput claim. That isnt in fact that commonly seen at all.

I am not overstating anything. I have already made it perfectly clear
that it isn't commonly seen at all. I don't understand how you got the
impression that I would claim anything else.

I have just given an example of how to create one of those rare
situations.

I dont, particularly with the older systems where an
addon card is likely to be used because the motherboard
bios has a problem with large drives, and those where
the PCI is used for more than just high thruput to hard
drives, particularly with analog video capture etc.

How much pci bandwith do you guess is used during analog videocapture?
My experience is with DV which is only 6MB/s, but lets take a high
load for analog: 720*480 at 24bit color at 30 fps. That is 31MB/s
Let's be generous and add 1MB/s for sound.

Then we also need a modern IDE harddisk that can write that fast.
the harddisk is waiting for data from the capture card, so it's
limited to that and can't make use fo the burst speed to the cache.

so suppose we have 32MB/s going over the PCI bus for the capture card.
(actually I don't think that it will do that for capturing)
Then we have again 32MB/s for writing it to the harddisk.

So that makes 64MB/s Even counting the overhead and poor
implementation of the pci bus we are well clear of the maximum for
32/33Mhz pci.

Now we have sound, video and harddisk using pci. I can't think of much
else that will be used at the same time for videoediting that we can
add on top of this that will overload the pci bus.

But I'm open to suggestions.

Marc

 

[Marc de Vries]

Most if not all Athlon 64 and Opteron boards have 64/66 PCI, but they
are often used with single IDE drives.

Opeteron is developed for use in servers. And in servers boards 64/66
pci is normal. (and usually combined with multiple processor sockets)

But Athlon 64 and Athlon 64 FX are meant for desktop and workstation.
And all of those boards, that I have seen so far just had 32/66 pci.

But we are getting to the point were it can be usefull to have faster
pci slots in a desktop machine, so I wouldn't be suprised if a
manufacturer will design or has already designed such a board.

I'm thinking about buying a Athlon 64 in about two months. So could
you point me to one of those Athlon 64 boards with 64/66 pci? I would
like to take a look at it.

Marc

 

[Marc de Vries]

Completely and utterly irrelevant to what was actually being discussed.

The OP specifically asked about a single IDE disk.
So it is very relevant.

 

[J. Clarke]

Opeteron is developed for use in servers. And in servers boards 64/66
pci is normal. (and usually combined with multiple processor sockets)

But Athlon 64 and Athlon 64 FX are meant for desktop and workstation.
And all of those boards, that I have seen so far just had 32/66 pci.

What I get for not checking the current market.

But we are getting to the point were it can be usefull to have faster
pci slots in a desktop machine, so I wouldn't be suprised if a
manufacturer will design or has already designed such a board.

I'm thinking about buying a Athlon 64 in about two months. So could
you point me to one of those Athlon 64 boards with 64/66 pci? I would
like to take a look at it.

I must be getting Alzheimers or something--I could have sworn that every
board I looked at had 64 bit 66 MHz PCI but it turns out that all of them
that also have AGP seem to have regular 32/33 PCI as that appears to be
what the consumer chipsets support. Three Bronx cheers for Via and nvidia.

 

[Rod Speed]

So you thought the oneliner without any explanation was enough?

It wasnt very clear exactly what he was asking
given that I know he understands the basics.

It is clear from his next reply that it wasn't.

Because it wasnt very clear exactly what he was
asking given that I know he understands the basics.

But we'll probably never agree on that.

Yep, you're far too arguementative and obnoxious for that.

(Sometimes I really wonder if your aim is to
help people, or if your goal is to confuse them)

See just above.

Pity you cannot point out where it is mangled,

Pity I already did.

so we can have a normal discussion about it.

Unlikely to be possible when you are always into point scoring.

What do you consider "not full 32bit pci", if it is not connected to
the slow pci implementations like the one I mentioned from via?

I am not overstating anything.

You already admitted you did on that thruput number.

You're clearly now trying to claim that black is white.

And since you're now into just plain lying, here
goes the chain on the rest of your puerile shit.

 

[Timothy Daniels]

It wasnt very clear exactly what he was asking
given that I know he understands the basics.


Because it wasnt very clear exactly what he was
asking given that I know he understands the basics.

Actually, I do know the basics, but I didn't know the specifics
of real-world PCI bandwidth and whether there was a wasted
capability in using an ATA133 device on a PCI bus. I think that
since there *are* newbies who read these NGs to learn more
about their PC systems, answers are of value to a wider audience
if we underestimate the background of any inquirer.

*TimDaniels*

 

[Rod Speed]

Actually, I do know the basics, but I didn't know the specifics
of real-world PCI bandwidth and whether there was a wasted
capability in using an ATA133 device on a PCI bus. I think that
since there *are* newbies who read these NGs to learn more
about their PC systems, answers are of value to a wider audience
if we underestimate the background of any inquirer.

Trouble is that you have to be a lot more long winded if
you do it like that. Sometimes I bother, sometimes I dont.

 

[Marc de Vries]

Pity I already did.

Maybe in your fantasies, but not in this ng.

But since you are only interesting in flaming, instead of having a
normal discussion I wont continue this any further.

 

[Marc de Vries]

Trouble is that you have to be a lot more long winded if
you do it like that. Sometimes I bother, sometimes I dont.

Yep. Instead of a oneliner you might have to type three lines to make
the post usefull (or at least not confusing) for everyone.
Those 30 seconds extra are certainly not something you should bother
about.

 

[Rod Speed]

message just the usual puerile sniping that it always ends up with.