| Author |
Message |
Thomas D
Guest
|
 Posted:
Mon Nov 07, 2005 6:27 pm Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Sat, 05 Nov 2005 09:16:48 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | The Outsider wrote:
On Sat, 05 Nov 2005 05:32:12 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
What part of "probably the clocking of the parallel bus going into the
bridge" did you have trouble with?
"Probably" isn't conclusive. That's the problem I have with it.
Provide some hard data.
The "hard data" is that PCI Express moves 2 billion bits per second on a
single pair with a 2.5 GHz clock and 8/10B encoding. Every technical
discussion of PCI express by any company that actually makes the chips that
support it explains this.
Whatever is clocked at 100 MHz on the nforce4 chipset is not the PCI Express
lanes unless _you_ can explain how they get 2 billion bits per second on a
single pair with a 100 MHz clock and 8/10B encoding and why Intel and
National Instruments and the rest are wrong and some random overclocker is
right.
|
You really believe that some chinese workers can come up with some 2,5
Ghz Maiboard, which is gigantic in comparison to the worlds famous cpu
makers, struggling around 3 Ghz on highly integrated chips ?
Come on... |
|
| Back to top |
|
 |
J. Clarke
Guest
|
| Posted:
Mon Nov 07, 2005 7:44 pm Post subject:
Re: PCI vs. AGP 8x? |
|
|
Thomas D wrote:
| Quote: | On Sat, 05 Nov 2005 09:16:48 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
The Outsider wrote:
On Sat, 05 Nov 2005 05:32:12 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
What part of "probably the clocking of the parallel bus going into the
bridge" did you have trouble with?
"Probably" isn't conclusive. That's the problem I have with it.
Provide some hard data.
The "hard data" is that PCI Express moves 2 billion bits per second on a
single pair with a 2.5 GHz clock and 8/10B encoding. Every technical
discussion of PCI express by any company that actually makes the chips
that support it explains this.
Whatever is clocked at 100 MHz on the nforce4 chipset is not the PCI
Express lanes unless _you_ can explain how they get 2 billion bits per
second on a single pair with a 100 MHz clock and 8/10B encoding and why
Intel and National Instruments and the rest are wrong and some random
overclocker is right.
You really believe that some chinese workers can come up with some 2,5
Ghz Maiboard, which is gigantic in comparison to the worlds famous cpu
makers, struggling around 3 Ghz on highly integrated chips ?
|
???? PCI Express was not created by "some chinese workers". The first
chips that supported it came from Intel.
The whole point of it is that a single pair can be clocked much higher than
a parallel bus. Those 3 GHz microprocessors have to keep 64 signal lines
synchronized at that speed, where PCI Express only has to worry about one
line at a time. 10 gig Ethernet gets 10 billion bits per second to go 40
kilometers on a single optical fiber and they've demonstrated 40 meters on
CAT5E cable in the laboratory--2.5 billion for a few inches on circuit
board traces isn't all that much of a trick.
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
The Outsider
Guest
|
| Posted:
Mon Nov 07, 2005 11:46 pm Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Sat, 05 Nov 2005 09:16:48 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | The "hard data" is that PCI Express moves 2 billion bits per second on a
single pair with a 2.5 GHz clock and 8/10B encoding. Every technical
discussion of PCI express by any company that actually makes the chips that
support it explains this.
Whatever is clocked at 100 MHz on the nforce4 chipset is not the PCI Express
lanes unless _you_ can explain how they get 2 billion bits per second on a
single pair with a 100 MHz clock and 8/10B encoding and why Intel and
National Instruments and the rest are wrong and some random overclocker is
right.
|
My mb uses a ULI chipset and it says right in the bios - PCI-E clocked
at 100mhz. |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Tue Nov 08, 2005 1:06 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
The Outsider wrote:
| Quote: | On Sat, 05 Nov 2005 09:16:48 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
The "hard data" is that PCI Express moves 2 billion bits per second on a
single pair with a 2.5 GHz clock and 8/10B encoding. Every technical
discussion of PCI express by any company that actually makes the chips
that support it explains this.
Whatever is clocked at 100 MHz on the nforce4 chipset is not the PCI
Express lanes unless _you_ can explain how they get 2 billion bits per
second on a single pair with a 100 MHz clock and 8/10B encoding and why
Intel and National Instruments and the rest are wrong and some random
overclocker is right.
My mb uses a ULI chipset and it says right in the bios - PCI-E clocked
at 100mhz.
|
Which means that either (a) that clock speed is not the lane clock or (b)
your board is only able to run PCI Express at 5% of the rated speed.
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
The Outsider
Guest
|
| Posted:
Wed Nov 09, 2005 3:46 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Mon, 07 Nov 2005 14:06:08 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | Which means that either (a) that clock speed is not the lane clock or (b)
your board is only able to run PCI Express at 5% of the rated speed.
|
It's the frequency it runs at, not the throughput. WTF would it say
100mhz if it doesn't run at a frequency of 100mhz? Is the mb
manufacturer on drugs or are you? |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Wed Nov 09, 2005 5:12 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
The Outsider wrote:
| Quote: | On Mon, 07 Nov 2005 14:06:08 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
Which means that either (a) that clock speed is not the lane clock or (b)
your board is only able to run PCI Express at 5% of the rated speed.
It's the frequency it runs at, not the throughput. WTF would it say
100mhz if it doesn't run at a frequency of 100mhz? Is the mb
manufacturer on drugs or are you?
|
I have no idea why the motherboard manufacturer would claim that.
Presumably they're running _something_ at that speed, but that something is
_not_ the PCI Express lanes themselves.
Why is it that you are having so much trouble grasping that it is not
possible to move 2 billion bits per second down a single pair using 8/10B
encoding at 100 MHz? You keep saying that "it's the throughput". Well,
with a single pair, using 8/10B encoding at 100 MHz, you get 80 Mb/sec
throughput, not 2 billion.
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
The Outsider
Guest
|
| Posted:
Thu Nov 10, 2005 1:31 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Tue, 08 Nov 2005 18:12:39 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | Why is it that you are having so much trouble grasping that it is not
possible to move 2 billion bits per second down a single pair using 8/10B
encoding at 100 MHz? You keep saying that "it's the throughput". Well,
with a single pair, using 8/10B encoding at 100 MHz, you get 80 Mb/sec
throughput, not 2 billion.
|
http://www.directron.com/a8nslid.html
PCIe Frequency: allowing PCIe frequency from 100MHz up to 200MHz at
1MHz increment.
http://www.digit-life.com/articles2/mainboard/abit-fatal1ty-aa8xe-i925xe.html
PCI/PCI-E frequency divider setup
PCI-E = Auto, 99 - 255 MHz at 1 MHz steps
PCI = 33.33 MHz, 36.36 MHz, 40.00 MHz, Auto (modify in accord with the
PCI-E frequency), CPU (modify in accord with the FSB frequency)
http://www.msi.com.tw/html/service/techexpress/mainboard/7160/page2.htm
PCIE Frequency (MHz):
Its adjustable range from standard maximum is 100MHz to 148MHz. Please
aware that any adjustment over 110MHz may cause your VGA card damaged. |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Thu Nov 10, 2005 3:48 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
The Outsider wrote:
| Quote: | On Tue, 08 Nov 2005 18:12:39 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
Why is it that you are having so much trouble grasping that it is not
possible to move 2 billion bits per second down a single pair using 8/10B
encoding at 100 MHz? You keep saying that "it's the throughput". Well,
with a single pair, using 8/10B encoding at 100 MHz, you get 80 Mb/sec
throughput, not 2 billion.
http://www.directron.com/a8nslid.html
PCIe Frequency: allowing PCIe frequency from 100MHz up to 200MHz at
1MHz increment.
http://www.digit-life.com/articles2/mainboard/abit-fatal1ty-aa8xe-i925xe.html
PCI/PCI-E frequency divider setup
PCI-E = Auto, 99 - 255 MHz at 1 MHz steps
PCI = 33.33 MHz, 36.36 MHz, 40.00 MHz, Auto (modify in accord with the
PCI-E frequency), CPU (modify in accord with the FSB frequency)
http://www.msi.com.tw/html/service/techexpress/mainboard/7160/page2.htm
PCIE Frequency (MHz):
Its adjustable range from standard maximum is 100MHz to 148MHz. Please
aware that any adjustment over 110MHz may cause your VGA card damaged.
|
I've already posted numerous links to sources far more authoritative than
any of those that say otherwise. And you have still not explained how they
accomplish what you claim that they accomplish. Further, it is
increasingly clear that you have not the slightest clue how PCI Express
works or what it is, other than a name.
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
The Outsider
Guest
|
| Posted:
Thu Nov 10, 2005 8:26 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Wed, 09 Nov 2005 16:48:01 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | I've already posted numerous links to sources far more authoritative than
any of those that say otherwise. And you have still not explained how they
accomplish what you claim that they accomplish. Further, it is
increasingly clear that you have not the slightest clue how PCI Express
works or what it is, other than a name.
|
Those are mb manufacturer links, explain why they give a frequency for
PCI-E if you say one doesn't exist.
And you are wrong about what I know.
http://www.computerforum.com/showthread.php?t=13239
PCI Express. Formerly known as 3GIO, also denoted as PCIe, this is an
extremely highspeed serialized interface. The individual serial lanes
can be grouped together and when done as such, they are denoted as
PCIxN where N is the number of lanes that are grouped together. Each
lane is capable of 250MB/s |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Thu Nov 10, 2005 8:26 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
The Outsider wrote:
| Quote: | On Wed, 09 Nov 2005 16:48:01 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
I've already posted numerous links to sources far more authoritative than
any of those that say otherwise. And you have still not explained how
they
accomplish what you claim that they accomplish. Further, it is
increasingly clear that you have not the slightest clue how PCI Express
works or what it is, other than a name.
Those are mb manufacturer links, explain why they give a frequency for
PCI-E if you say one doesn't exist.
|
Liar.
I did not say that "one does not exist". I said that the frequency that
they were specifying was not for the PCI Express lanes, which run at 2.5
GHz.
| Quote: | And you are wrong about what I know.
http://www.computerforum.com/showthread.php?t=13239
PCI Express. Formerly known as 3GIO, also denoted as PCIe, this is an
extremely highspeed serialized interface. The individual serial lanes
can be grouped together and when done as such, they are denoted as
PCIxN where N is the number of lanes that are grouped together. Each
lane is capable of 250MB/s
|
Which proves that you can cut and paste, not that you have the tiniest iota
of understanding of what you have read. You still have not explained how
one can get 2 Gb/sec out of a 100 MHz signal using 8/10B encoding on a
single pair.
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
Smack'n Rat
Guest
|
| Posted:
Fri Nov 11, 2005 8:26 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
J. Clarke wrote:
| Quote: | The Outsider wrote:
On Wed, 09 Nov 2005 16:48:01 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
I've already posted numerous links to sources far more authoritative than
any of those that say otherwise. And you have still not explained how
they
accomplish what you claim that they accomplish. Further, it is
increasingly clear that you have not the slightest clue how PCI Express
works or what it is, other than a name.
Those are mb manufacturer links, explain why they give a frequency for
PCI-E if you say one doesn't exist.
Liar.
I did not say that "one does not exist". I said that the frequency that
they were specifying was not for the PCI Express lanes, which run at 2.5
GHz.
And you are wrong about what I know.
http://www.computerforum.com/showthread.php?t=13239
PCI Express. Formerly known as 3GIO, also denoted as PCIe, this is an
extremely highspeed serialized interface. The individual serial lanes
can be grouped together and when done as such, they are denoted as
PCIxN where N is the number of lanes that are grouped together. Each
lane is capable of 250MB/s
Which proves that you can cut and paste, not that you have the tiniest iota
of understanding of what you have read. You still have not explained how
one can get 2 Gb/sec out of a 100 MHz signal using 8/10B encoding on a
single pair.
|
The lanes are in fact LVDS signaling schemes (lanes are limited to 20
inches by spec) that yell at each other at 2.5 Gbps (that's Gigabits per
second or 250 MegaBytes per second). The 100 MHz clock is usually a
reference clock (on pins 13 and 14 on side A of the connector).
Technically the clock for RX and TX are embedded in the signaling scheme
in which they communicate with each other. However, when the
transmitter runs in spread-spectrum for EMI purposes the receiver needs
the common clock (which is usually the 100MHz ref...)
The interesting inclusion of spread-spectrum clock generation for PCs
can be seen here (under "Spread-spectrum clock generation"):
http://en.wikipedia.org/wiki/Spread_spectrum
Another good link that shows the way that spread spectrum works and why
mobo folks implement it:
http://zone.ni.com/devzone/conceptd.nsf/webmain/EE6BC161F0C17F0986256FEE005D3797
Unfortunately, I don't know the details of the spread-spectrum clock and
its effect on the transmission of PCIe data, but I assume it controls
how fast the transmission 'sweeps'.
Hope it helps.
SmacknRat |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Fri Nov 11, 2005 5:30 pm Post subject:
Re: PCI vs. AGP 8x? |
|
|
Smack'n Rat wrote:
| Quote: | J. Clarke wrote:
The Outsider wrote:
On Wed, 09 Nov 2005 16:48:01 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
I've already posted numerous links to sources far more authoritative
than
any of those that say otherwise. And you have still not explained how
they
accomplish what you claim that they accomplish. Further, it is
increasingly clear that you have not the slightest clue how PCI Express
works or what it is, other than a name.
Those are mb manufacturer links, explain why they give a frequency for
PCI-E if you say one doesn't exist.
Liar.
I did not say that "one does not exist". I said that the frequency that
they were specifying was not for the PCI Express lanes, which run at 2.5
GHz.
And you are wrong about what I know.
http://www.computerforum.com/showthread.php?t=13239
PCI Express. Formerly known as 3GIO, also denoted as PCIe, this is an
extremely highspeed serialized interface. The individual serial lanes
can be grouped together and when done as such, they are denoted as
PCIxN where N is the number of lanes that are grouped together. Each
lane is capable of 250MB/s
Which proves that you can cut and paste, not that you have the tiniest
iota
of understanding of what you have read. You still have not explained how
one can get 2 Gb/sec out of a 100 MHz signal using 8/10B encoding on a
single pair.
The lanes are in fact LVDS signaling schemes (lanes are limited to 20
inches by spec) that yell at each other at 2.5 Gbps (that's Gigabits per
second or 250 MegaBytes per second). The 100 MHz clock is usually a
reference clock (on pins 13 and 14 on side A of the connector).
Technically the clock for RX and TX are embedded in the signaling scheme
in which they communicate with each other. However, when the
transmitter runs in spread-spectrum for EMI purposes the receiver needs
the common clock (which is usually the 100MHz ref...)
The interesting inclusion of spread-spectrum clock generation for PCs
can be seen here (under "Spread-spectrum clock generation"):
http://en.wikipedia.org/wiki/Spread_spectrum
Another good link that shows the way that spread spectrum works and why
mobo folks implement it:
http://zone.ni.com/devzone/conceptd.nsf/webmain/EE6BC161F0C17F0986256FEE005D3797
Unfortunately, I don't know the details of the spread-spectrum clock and
its effect on the transmission of PCIe data, but I assume it controls
how fast the transmission 'sweeps'.
|
Finally someone who actually does have some clue what's going on enters the
discussion. Thank you.
| Quote: |
Hope it helps.
SmacknRat
|
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
The Outsider
Guest
|
| Posted:
Sat Nov 12, 2005 3:55 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Thu, 10 Nov 2005 00:27:12 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | Liar.
I did not say that "one does not exist". I said that the frequency that
they were specifying was not for the PCI Express lanes, which run at 2.5
GHz.
|
From the links I provided one can indeed see they are specifying the
PCI-E graphics slot. Contact Asus and ask them about it. |
|
| Back to top |
|
|
The Outsider
Guest
|
| Posted:
Sat Nov 12, 2005 3:59 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
On Fri, 11 Nov 2005 06:30:07 -0500, "J. Clarke"
<jclarke.usenet@snet.net.invalid> wrote:
| Quote: | Finally someone who actually does have some clue what's going on enters the
discussion. Thank you.
|
Well, you couldn't explain the 100mhz clock on PCI-E so you know fuck
all too. |
|
| Back to top |
|
|
J. Clarke
Guest
|
| Posted:
Sat Nov 12, 2005 8:26 am Post subject:
Re: PCI vs. AGP 8x? |
|
|
The Outsider wrote:
| Quote: | On Thu, 10 Nov 2005 00:27:12 -0500, "J. Clarke"
jclarke.usenet@snet.net.invalid> wrote:
Liar.
I did not say that "one does not exist". I said that the frequency that
they were specifying was not for the PCI Express lanes, which run at 2.5
GHz.
From the links I provided one can indeed see they are specifying the
PCI-E graphics slot. Contact Asus and ask them about it.
|
I'm sorry, but there is no statement there to the effect that the 100 MHz
clock is for any part of the signalling that is part of the PCI Express
standard, just that something related to PCI Express is clocked at that
speed.
Look, if it's so important to you that you believe something that is not
true, I'm not going to try to stop you, but when Intel and ASUS disagree on
a standard that was pretty much created by Intel and is implemented using
an Intel chip, I'll believe Intel before I'll believe ASUS.
--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net) |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in