View Full Version : Intel Is Being Stupid!
OC-Master
04-28-03, 11:20 PM
Ok, there not stupid! I know that but..
Dont get me wrong, that new Canterwood chipset is awesome especially with the new 800MHz FSB but why the heck did Intel stick a lowesy amazingly slow bottleneck between the northbridge and the southbridge.
266MB/s between the northbridge and the southbridge.. Whats up with that??
Even VIA has implementions of V-link 8X which does 533MB/s between the two.
I think its time for Intel to invest in HyperTransport.
OC-Master
yes we know, it has already been posted thank u.
Rasputin'sLiver
04-29-03, 12:32 AM
How did i know this thread was started by an amd guy before i even looked at his siggy?
i dont know y but i see this alot. do intel guyz go in amd sections and point out dumb amd stuff too?
NookieN
04-29-03, 12:59 AM
A 266MB/s bus-link is more than enough for any communications between the north and south bridges. The PCI bus is still only 133MB/s whether you've got a 266MB/s link or 20GB/s link.
HyperTransport is good for SMP, but it's a waste of money for linking peripherals.
When PCI Express comes to the desktop next year it'll be a different story.
NookieN u smart guy ;) I have yet to grab that quarter out of your hand.
I'm not sure I get that...
If you have a link between SB and NB capable of 266MB/s and you are transfering 133MB/s over the PCI that would be one PCI-slot, would it not?
Or have I missed the basics? ..do all PCI-slots share 133MB/s? That doesn't make sense.
I mean, let's say you have 2 ATA-133 controllers (let's say they are RAID-controllers, because ATA-133-drives won't use 133MB/s anyway) then you would be using 266MB/s..?
Rasputin'sLiver
04-29-03, 02:36 AM
Originally posted by AZN
i dont know y but i see this alot. do intel guyz go in amd sections and point out dumb amd stuff too?
no, but we should. maybe we should go there and remind them about performance ratings instead of actual mhz's, the unending release setbacks for their products and amd's failure to ever develop a decent chipset of their own.;)
seamadan000
04-29-03, 02:42 AM
the PCI bus shares one 133mb/s link, no matter how many things need that bandwidth. Nookie is right b/c there isn't enough things that can even use the link in question to be able to saturate that channel. The only reason SiS has more is for a marketing advantage. This is why serial raid on board that uses PCI is pointless, the serial connection may be 150mb/s, but the channel it is attached to is only 133mb/s, which is why the native serial raid of the new southbridge in the canterwood is so important.
in short, its not a bottleneck at all, but thanks for trying AMDer :rolleyes:
________
Marijuana Vaporizers (http://weedvaporizer.info/)
jazztrumpet216
04-29-03, 07:55 AM
Alright guys, we've made our points, let's not let this thread get out of hand.
Warhawk
04-29-03, 09:09 AM
Originally posted by jazztrumpet216
Alright guys, we've made our points, let's not let this thread get out of hand.
Yeah what he said ;)
OC Noob
04-29-03, 11:10 AM
Originally posted by seamadan000
the PCI bus shares one 133mb/s link, no matter how many things need that bandwidth. Nookie is right b/c there isn't enough things that can even use the link in question to be able to saturate that channel. The only reason SiS has more is for a marketing advantage. This is why serial raid on board that uses PCI is pointless, the serial connection may be 150mb/s, but the channel it is attached to is only 133mb/s, which is why the native serial raid of the new southbridge in the canterwood is so important.
in short, its not a bottleneck at all, but thanks for trying AMDer :rolleyes:
Sorry to be off topic, but is SATA limited by PCI bus on all boards but the Canterwood?
ninthebin
04-29-03, 03:06 PM
Sorry to be off topic, but is SATA limited by PCI bus on all boards but the Canterwood?
as far as im aware once the data goes from the new SATA HDD, down the new SATA cable into the new SATA slot, it might as well be coming from a PATA or SATA drive as it hasnt been given any special bus that the older standard never.
imgod2u
04-29-03, 03:25 PM
There may be some merit to that. On all previous chipsets except Canterwood and Springdale, I'd say the hub-link isn't much of a limitation. However, with Canterwood/Springdale, Intel has opted to connect a lot of things directly to the South Bridge which could easily eat up that bandwidth. The dual SATA connectors are connected to the South Bridge each transfering 150MB/sec max, plus the ATA-100/RAID controller plus the USB 2.0 link plus the PCI bus. There are a few other things as well I think. Now, the likelihood of all of those channels being filled up to the brink is kind of unlikely. Very rarely I'd say but they still could max out the hub-link. PCI Express will help alleviate this significantly.
Krowa 02
04-29-03, 04:09 PM
This post is the main reason why there is an AMD forums and an Intel forum. If you dont like what Intel is doing and have a problem with it because yours is better, talk about it with people who wont get mad at you.
seamadan000
04-29-03, 04:20 PM
OC Noob: Yes, every MB with a SATA chip thus far has been attached to the PCI bus, the canterwood is the first without this limitation. Then again, the current ATA HDDs can't saturate the 133mb/s let alone the 150mb/sec, so this isn't a real problem for most setups :)
________
Sweety*****cat (http://www.girlcamfriend.com/cam/Sweety*****cat/)
If Intel's hub link speed was limiting, then SiS's chipsets would be noticeably faster. The SiS655 and 648 both have a 1 GB/s link with the SiS963 south bridge.
But guess what? The 845PE, with its "inadequate" 266 MB/s link, is consistently faster.
Rasputin'sLiver
04-29-03, 06:09 PM
Originally posted by Krowa 02
This post is the main reason why there is an AMD forums and an Intel forum. If you dont like what Intel is doing and have a problem with it because yours is better, talk about it with people who wont get mad at you.
roflol
OC-Master
04-29-03, 06:46 PM
This is not a rat people, and was never intended to be. The only thing I wanted to talk about was the fact that Intel stuck with a 266MB/s link between the north bridge and south bridge.
Lets not let this get out of hand... and I still need information such as..
Since the Live! and Audigy/2 can use the full potential of PCI bandwidth (133MB/s) what happens to the performance of the remaing periphiels?
I mean, say your two Serial ATA hard drives in RAID0 can do 80MB/s sustained..
Thats 80MB/s + 133MB/s
Now last, say you were doing something freaky like playing music over the internet (using the potential of sound card to decode compressed audio) and at the same was sending some information over your new gigabit lan interface, would the lan potential actually be bottlenecked?
Or is that 266MB/s just an estimate? I heard PCI could do over 160MB/s with some chipsets, which is well over the rated 133MB/s (33.3MHz) spec.
OC-Master
ninthebin
04-29-03, 08:02 PM
the gigabit ethernet has been given the CSA, so thats fine
Originally posted by OC-Master
This is not a rat people, and was never intended to be.
...
Or is that 266MB/s just an estimate? I heard PCI could do over 160MB/s with some chipsets, which is well over the rated 133MB/s (33.3MHz) spec.
It is a fair question, but it *is* a bit silly to come to the conclusion that Intel's engineers are stupid and never considered this issue.
The only benchmark I've seen that reveals N-S bus saturation is the Intel numbers they used to justify using CSA for gigabit ethernet. Apparently that makes a significant difference.
OC-Master
04-30-03, 09:05 AM
Originally posted by Soapy
It is a fair question, but it *is* a bit silly to come to the conclusion that Intel's engineers are stupid and never considered this issue.
The only benchmark I've seen that reveals N-S bus saturation is the Intel numbers they used to justify using CSA for gigabit ethernet. Apparently that makes a significant difference.
You never know :D
they could be???
hehe
OC-Master
All right, time to say how VIA C3 owns both pentium and athlon.
jk of course
OC-Master
04-30-03, 11:17 PM
Originally posted by efini
All right, time to say how VIA C3 owns both pentium and athlon.
jk of course
In some ways the VIA C3 is the best because it uses passive cooling and its cores are smaller than the P4 and AthlonXPs. Thats pretty sweet if your looking for a rig in a recording studio and cant afford a water cooling setup :)
OC-Master
Originally posted by OC-Master
In some ways the VIA C3 is the best because it uses passive cooling and its cores are smaller than the P4 and AthlonXPs. Thats pretty sweet if your looking for a rig in a recording studio and cant afford a water cooling setup :)
OC-Master
only if you want to wait 50 hours for the chip to finish processing. :rolleyes:
Originally posted by OC-Master
In some ways the VIA C3 is the best because it uses passive cooling and its cores are smaller than the P4 and AthlonXPs. Thats pretty sweet if your looking for a rig in a recording studio and cant afford a water cooling setup :)
OC-Master
it is nice, cuzz u know how much people in the recording business like to OC. :rolleyes:
imgod2u
04-30-03, 11:57 PM
Originally posted by OC-Master This is not a rat people, and was never intended to be. The only thing I wanted to talk about was the fact that Intel stuck with a 266MB/s link between the north bridge and south bridge.
Lets not let this get out of hand... and I still need information such as..
Since the Live! and Audigy/2 can use the full potential of PCI bandwidth (133MB/s) what happens to the performance of the remaing periphiels?
This is very unlikely. Sound data does not take that much PCI bandwidth. Many of the SB Live! and Audigy problems from PCI timing and latency, not bandwidth.
I mean, say your two Serial ATA hard drives in RAID0 can do 80MB/s sustained..
Thats 80MB/s + 133MB/s
And still well within the limits of the Hub-Link. Of course, maxing out the PCI bus is a little stretched.
Now last, say you were doing something freaky like playing music over the internet (using the potential of sound card to decode compressed audio) and at the same was sending some information over your new gigabit lan interface, would the lan potential actually be bottlenecked?
Not in Canterwood/Springdale, the only chipsets in which the SATA channels are actually separated off the PCI bus. The gigabit ethernet is directly connected to the North Bridge.
Or is that 266MB/s just an estimate? I heard PCI could do over 160MB/s with some chipsets, which is well over the rated 133MB/s (33.3MHz) spec.
Nope, it's 266MB/sec maximum just like PCI is 133MB/sec maximum.
vBulletin® v3.8.7, Copyright ©2000-2013, vBulletin Solutions, Inc.