Important discovery in regards to the PCI lock question on nforce3
Ok here it goes.
The constants
MEM: 100mhz (auto timings)
CPU muliplier: 6x
LDT: 1x
I left the "cpu overclock" at the default 200mhz. And I went into clockgen to overclock I took it up to 235mhz
Notice the PCI and AGP clock in clockgen? I was able to increase it another 5mhz to 240mhz before if hard locked on me.
Now whats interesting is I had it up 335mhz yesterday, however I initially set my CPU Overclock setting to the max 300mhz. Now that got me thinking perhap u have to change the CPU Overlock to something else before the PCI lock engages. Ok so I set it to 201mhz. Lets try 235mhz again.
HELLO!!
Hmm perhaps clockgen is reading it wrong....lets take it up to 265mhz
Well this just about confirms the lock for me....I couldn't get past 240mhz when the CPU Overclock was left at 200mhz. Finally lets got for the whole smack 335mhz
Bingo!!
I discovered this the other day when I was playing with Clockgen...
Here was my writeup at 3AM:
The Mysterious nForce3 AGP/PCI Locks
After much debate on which chipset to choose for ordering on September 23, 2003 – the release day of the AMD Athlon 64, I decided upon the nForce3 chipset because of its claimed locked AGP/PCI frequencies. Soon after these motherboards hit the market, several articles popped up claiming that the nForce3 did NOT have a locked AGP/PCI bus. To my surprise, I tried overclocking my Athlon 64 3200+, and quickly ran into the wall at about 2230MHz (223MHz x 10). This coincided with an AGP frequency of 74.33MHz (37.17MHz PCI) if one were using a 1/6 divider…and this frequency causes my system to lockup and refuse to boot. I was frustrated by the fact that NVIDIA had indeed lied to the public and in the back of my head vowed to try to avoid such products in the future.
Well, that changed somewhat today. Today, I decided to try overclocking my Athlon 64 again, but instead of using the BIOS this time, I would try a program called Clockgen – ( http://www.cpuid.com/clockgen.php ). Because there are several different versions of the Clockgen program, I will specifically be speaking of the nForce3 version. This program theoretically allows you to adjust the voltage, multiplier, HTT clock, and the AGP/PCI clocks of any nForce3 150 based motherboards. At first I kept crashing my system, and I finally realized that I should probably lower my multiplier and AGP clock just to see if it would actually work. Using CPU-Z and Aida32 (http://www.cpuid.com/cpuz.php and http://www.aida32.hu/aida32.php ), I monitored the HTT, multiplier and overall CPU frequencies in real time. When I changed a setting within Clockgen, it was reflected by a change in both Aida32 and CPU-Z.
The best idea I had was to attempt to break my 223MHz HTT wall that had been reached while trying to overclock in the BIOS. To my amazement, while adjusting the HTT clock slowly upward, continually decreasing the AGP clock to around 60MHz and allowing it to slowly creep back to 66MHz, I passed right by my previous 223MHz HTT wall. At this time, I was forced to reboot and loosen my RAM timings (my BH-5 Corsair RAM is unable to do 2-4-3-6 at anything above 226MHz) to 2.5-4-4-8. I once again ran Clockgen and lowered the multiplier to 8, raised the core voltage to 1.55v, and began fiddling with the clocks until I got to 243MHz HTT. The computer locked up and I suspected that my RAM was to fault. In the BIOS, I reduced the “maximum” frequency of the memory to 166MHz so that the RAM would run at a 6:5 ratio to the HTT. I booted again, and dropped the multiplier – strangely, my memory frequency dropped from 166MHz to
At this point, I was pretty sure that I had a working AGP/PCI lock because I am fairly sure that very few AGP cards could survive 80.3MHz. I pushed the HTT to the max, still following the method, and finally arrived at 269MHz HTT. Anything higher simply would hard-lock and cause me to reboot.
I'm going to try to finish up my article and see if it will be published - O_P - I think you've arrived at the same results as I have which confirms to me that the nF3 does have AGP/PCI locks.
Originally posted by Namek on 02-28-2004 at 02:36 PM I discovered this the other day when I was playing with Clockgen...
Here was my writeup at 3AM:
The Mysterious nForce3 AGP/PCI Locks
After much debate on which chipset to choose for ordering on September 23, 2003 – the release day of the AMD Athlon 64, I decided upon the nForce3 chipset because of its claimed locked AGP/PCI frequencies. Soon after these motherboards hit the market, several articles popped up claiming that the nForce3 did NOT have a locked AGP/PCI bus. To my surprise, I tried overclocking my Athlon 64 3200+, and quickly ran into the wall at about 2230MHz (223MHz x 10). This coincided with an AGP frequency of 74.33MHz (37.17MHz PCI) if one were using a 1/6 divider…and this frequency causes my system to lockup and refuse to boot. I was frustrated by the fact that NVIDIA had indeed lied to the public and in the back of my head vowed to try to avoid such products in the future.
Well, that changed somewhat today. Today, I decided to try overclocking my Athlon 64 again, but instead of using the BIOS this time, I would try a program called Clockgen – ( http://www.cpuid.com/clockgen.php ). Because there are several different versions of the Clockgen program, I will specifically be speaking of the nForce3 version. This program theoretically allows you to adjust the voltage, multiplier, HTT clock, and the AGP/PCI clocks of any nForce3 150 based motherboards. At first I kept crashing my system, and I finally realized that I should probably lower my multiplier and AGP clock just to see if it would actually work. Using CPU-Z and Aida32 (http://www.cpuid.com/cpuz.php and http://www.aida32.hu/aida32.php ), I monitored the HTT, multiplier and overall CPU frequencies in real time. When I changed a setting within Clockgen, it was reflected by a change in both Aida32 and CPU-Z.
The best idea I had was to attempt to break my 223MHz HTT wall that had been reached while trying to overclock in the BIOS. To my amazement, while adjusting the HTT clock slowly upward, continually decreasing the AGP clock to around 60MHz and allowing it to slowly creep back to 66MHz, I passed right by my previous 223MHz HTT wall. At this time, I was forced to reboot and loosen my RAM timings (my BH-5 Corsair RAM is unable to do 2-4-3-6 at anything above 226MHz) to 2.5-4-4-8. I once again ran Clockgen and lowered the multiplier to 8, raised the core voltage to 1.55v, and began fiddling with the clocks until I got to 243MHz HTT. The computer locked up and I suspected that my RAM was to fault. In the BIOS, I reduced the “maximum” frequency of the memory to 166MHz so that the RAM would run at a 6:5 ratio to the HTT. I booted again, and dropped the multiplier – strangely, my memory frequency dropped from 166MHz to
At this point, I was pretty sure that I had a working AGP/PCI lock because I am fairly sure that very few AGP cards could survive 80.3MHz. I pushed the HTT to the max, still following the method, and finally arrived at 269MHz HTT. Anything higher simply would hard-lock and cause me to reboot.
hmm but I didn't have to decrease my agp and then bring back up. Also when I OC via the bios i can get all the way to 300mhz HTT. Its only when I leave the CPU OC in the bios at default 200mhz and use clockgen the the PCI lock doesn't work for me. What did u have your CPU Overclock settings too?
Originally posted by Optimus_Prime on 02-28-2004 at 02:51 PM hmm but I didn't have to decrease my agp and then bring back up. Also when I OC via the bios i can get all the way to 300mhz HTT. Its only when I leave the CPU OC in the bios at default 200mhz and use clockgen the the PCI lock doesn't work for me. What did u have your CPU Overclock settings too?
Whether I leave everything alone or not in the BIOS, I have to lower my AGP and let it slide back up....but it works just fine if you do it that way.
Originally posted by Namek on 02-28-2004 at 03:09 PM Whether I leave everything alone or not in the BIOS, I have to lower my AGP and let it slide back up....but it works just fine if you do it that way.
Alex
yes but if you have to lower the agp clock and then let it increase...wouldn't that mean your agp isn't really locked?
Originally posted by Namek on 02-28-2004 at 03:25 PM Well, it's not "locked" per say - but it is managable. Which is good. I think that I would actually prefer that in some cases...
Alex
that still isn't sitting right with mewhy would your nforce3 be different then mine, unless it doesn't have to be implemented by each mobo manufacter like some people suggested??
Try these settings and see if your agp increases
In the bios:
CPU multi- 6x
AGP- 66mhz
Mem: 100mhz
CPU OC: 201mhz
LDT: 1x
Now use clockgen to increase the fsb to see if your agp clock increases.
Everyone that has an nforce3 try that. And post your results. Perhaps we can put this PCI/agp question to rest.
Every board isn't the same. ClockGen reads my PCI locked at 33.40 MHz no matter where I put the HTT, BIOS or not. I have tried putting the HTT in the BIOS @ 201 and then using ClockGen and nothing is different. 33.40MHz is what ClockGen reports.
I can however lower my multiplier and have tested going higher with my HTT and my HTT cap is identical with no change. So even with a confirmed lower multiplier and my HTT in the BIOS set to 201, ClockGen reports 33.40MHz on the PCI bus and the HTT will not go any higher than 228.
My RAM has been tested in my ChainTech ZNF3-150 and hit 234 MHz @ cas 2-2-2-11 1:1. So I feel confident my board isn't locking the PCI bus.
Originally posted by J-Bo on 02-28-2004 at 04:21 PM Every board isn't the same. ClockGen reads my PCI locked at 33.40 MHz no matter where I put the HTT, BIOS or not. I have tried putting the HTT in the BIOS @ 201 and then using ClockGen and nothing is different. 33.40MHz is what ClockGen reports.
I can however lower my multiplier and have tested going higher with my HTT and my HTT cap is identical with no change. So even with a confirmed lower multiplier and my HTT in the BIOS set to 201, ClockGen reports 33.40MHz on the PCI bus and the HTT will not go any higher than 228.
My RAM has been tested in my ChainTech ZNF3-150 and hit 234 MHz @ cas 2-2-2-11 1:1. So I feel confident my board isn't locking the PCI bus.
OK one thing I found out, when I had my CPU OC setting to 200mhz and set 235mhz in clockgen it did show the PCI at 33.4 and AGP at 66.8. It was when I clicked "get values" that the speed changed. For some reason when u first change the FSB setting in Clockgen the PCI/agp changes aren't reflected until u hit get values. What mem divider are you using? Try using 100mhz so u know its not the mem holding u back even though u could get to 234mhz on your chaintech doesn't mean u can with the shuttle. If it still isn't locking then throw it up too 234mhz to see if you at least have another divider that kicks in .
Wow, you guys been hard at work! Great job. O_P, what do you mean by the "cpu overclock" max setting? Is that your FSB/HTT setting in the BIOS? Or something different. And I saw you mention before about a "PCI Changeable" setting? It don't sound like a good thing, but did you fiddle with it at all?
Nicely explained Namek. If I'm understanding you correctly. When you raise the FSB/HTT using clockgen, your AGP/PCI frequencies also rise with it? And you manuallhave to set them back with the AGP slidder in Clockgen?
One more thing for anyone stuck at 230-260.
Quote:
Originally posted by Optimus_Prime on 02-28-2004 at 03:39 PM
Try these settings and see if your agp increases
In the bios:
CPU multi- 6x
AGP- 66mhz
Mem: 100mhz
CPU OC: 201mhz LDT: 1x
That's the setting that's most likely holding you back. You have to lower it to 1x or 2x to get over 300.
Processors
Motherboards
Chipsets
Memory
Graphics Cards
Storage
Cases and Cooling
Mobile
Systems
Displays
Shows and Expos
Linear Mode
