I've got a small problem with enabling CnQ via the BIOS...when I enable it my system no longer boots to windows in a regular startup mode. I can boot into safe mode though. It goes through the loading windows screen and then switches to a solid blue screen (windows error screen) with nothing exept either the number 2 or the ^ character. When I disable CnQ I can boot into windows in either normal or in safe mode. Temperatures are not an issue and I have had CnQ enabled and running normally before.
Here's a look at my setup (other than whats in my sig):
- ambient: 22º C
- idle: CPU-28; System-28; RAM(on heatspreader surface)- 31
- load: CPU-40; System-35; RAM(on heatspreader surface)-41
- front & rear fans: Antec Pro 120mm BB fans @ ~1100 rpm via Thermaltake Hardcano13 Fan Controller
- PSU fan: stock Antec TPII fan @ ~900rpm
- Voltage readings via CoreCenter (motherboard utility): 12v -11.90v; 5v- 5.03v; 3.3v - 3.31v
With regards to my overclock...here are the relevant settings:
- CPU multiplier = 10x
- HTT speed = 250Mhz
- HTT multiplier = 4x
- RAM divider = 166MHz
- VCore = 1.400v (stock)
These settings are stable as far as I can tell....I haven't done the Prime95 24 hour test yet but it does not hang in the 4 hours I did do the test for. I also did it concurrently with 3DMark03 and it didn't hang. I know this doesn't mean its 100% rock stable but it should mean windows should boot at least.
I had it overclocked before to 2.4 GHz on an 11x multi with HTT multi of 4x and speed of 218Mhz and CnQ worked perfectly. I recently changed the settings to make it a little more stable and get a little more speed out of it (without making it hotter or louder) and now CnQ seems to hang up my boot. I'm not entirely sure what else could be causing it and I will be emailing MSI in the next few minutes to see if its hardware issue but I do have the latest bios from their website and the latest processor driver from AMDs website. Can anyone figure out why this doesn't seem to work?
Cool 'n Quiet problems
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Download and setup speedfan, and put togther a manual CnQ.
Your having problems because even though [email protected] is ok, your running [email protected] during CnQ which your CPU doesnt seem to be able to handle.
CnQ just drops the multipler and the voltage, it doesnt touch the HTT.
Your having problems because even though [email protected] is ok, your running [email protected] during CnQ which your CPU doesnt seem to be able to handle.
CnQ just drops the multipler and the voltage, it doesnt touch the HTT.
-
dukla2000
- *Lifetime Patron*
- Posts: 1465
- Joined: Sun Mar 09, 2003 12:27 pm
- Location: Reading.England.EU
CnQ can run fine on an OC'd CPU, BUT you have a second (or more, depending on which CPU) set of settings that you have to validate for stability. As per dragmor you have to know whether your CPU can run stable at 5*250 @ 1.1V. You also have to know whether your memory runs stable at the CnQ setting.
As example: my CPU OC's fine to 267HTT with DDR333 selected for the memory. Actually this means 9*267 @1.4 VCore at full speed CPU with memory at 2404/9 = 218. When CnQ kicks in my CPU is 5*267 @ 1.1VCore and the memory is CPU/6, or 1335/6 = 222.5. Note that not only are you dealing with CPU stability at a lower VCore, but also potentially HIGHER memory speed when CnQ kicks in depending on your settings!
You dont say which CPU you have, but you do say it ran at 11*multi before and you are running at 10* now? That is your problem: when CnQ executes it will reset the VCore AND multiplier to stock, so after an idle period it is setting your CPU to 11*250 @ stock VCore which your CPU can not handle.
As example: my CPU OC's fine to 267HTT with DDR333 selected for the memory. Actually this means 9*267 @1.4 VCore at full speed CPU with memory at 2404/9 = 218. When CnQ kicks in my CPU is 5*267 @ 1.1VCore and the memory is CPU/6, or 1335/6 = 222.5. Note that not only are you dealing with CPU stability at a lower VCore, but also potentially HIGHER memory speed when CnQ kicks in depending on your settings!
You dont say which CPU you have, but you do say it ran at 11*multi before and you are running at 10* now? That is your problem: when CnQ executes it will reset the VCore AND multiplier to stock, so after an idle period it is setting your CPU to 11*250 @ stock VCore which your CPU can not handle.
thanks for the replies people. I'm not sure you all understand my problem though. I've tested the OC at the speed posted above and in my sig and it's stable for a few hours (at least) on Prime95. I then restart my system and enable CnQ in the BIOS (previously not enabled). I then restart to apply settings and then the windows load screen comes on...the one with the blue/green loading progress bar. After that is done it changes to a solid blue sceen (same color as windows error screen) with the number 2 or the ^ character. It doesn't even get to starting windows so CnQ can't even be enabled...since the CPU would be under full load due to startup. The problem I have is with the BIOS setting. Any comments in light of this information?
-
dukla2000
- *Lifetime Patron*
- Posts: 1465
- Joined: Sun Mar 09, 2003 12:27 pm
- Location: Reading.England.EU
I thought I didsnowman59 wrote: I'm not sure you all understand my problem though.
You are trying to boot at 10 * 250 at stock VCore with CnQ, right? And your CPU stock multiplier is 11 or higher?
If so, as soon as your OS has a chance it will try set the CPU to full speed (as during boot things are somewhat CPU intensive, so the CnQ driver will go for max P state). For your CPU this full speed is actually multiplier 11 (or higher - still not sure exactly what CPU you have). So the driver sets you to 11 * 250 at stock VCore which is an overclock too far for the CPU, so it hangs. Your OS load is hanging as soon as the CnQ OS module loads because your overclock is not stable. And your overclock is not stable because although you thought you set a 10 multi, the CnQ driver knows your CPU can do 11* or more and uses that, not what you set in the BIOS.
QED.
If you want to use CnQ with an overclocked A64 it does not make sense to set a multiplier in the BIOS lower than the CPU spec.
i have a 3700+ San Diego (in my sig) with a stock multi of 11x and a stock speed of 2.2Ghz. I'm trying to boot at 10x because it makes the memory stable in both normal and CnQ enabled because they are multiples of each other and that doesn't make the memory run faster than its capable of with CnQ enabled. For example using the same HTT speed of 250 and multi of 4x:
10x CPU multi = 208 with and without CnQ
11x CPU multi = 196 normal and 208 with CnQ
I know my ram can run up to 214 without problems with 2.85v so I want to run it as close to that all the time as possible while not OC'ing my CPU too much. As a side note...when testing my overclock initially I accidentally left CnQ enabled and when my CPU idled it correctly returned to the 10x multi when I subsequently started Prime95. Dukla2000, is this the same effect as when Windows initally starts up since apparently it correctly returns to 10x? Is there anyway around this problem or is that just how it is...and I need to just stick with the 11x multi? Sorry for so many questions but this is all new info to me, i thought i had this CnQ thing down, but i guess I didn't know about this multiplier issue with startup.
10x CPU multi = 208 with and without CnQ
11x CPU multi = 196 normal and 208 with CnQ
I know my ram can run up to 214 without problems with 2.85v so I want to run it as close to that all the time as possible while not OC'ing my CPU too much. As a side note...when testing my overclock initially I accidentally left CnQ enabled and when my CPU idled it correctly returned to the 10x multi when I subsequently started Prime95. Dukla2000, is this the same effect as when Windows initally starts up since apparently it correctly returns to 10x? Is there anyway around this problem or is that just how it is...and I need to just stick with the 11x multi? Sorry for so many questions but this is all new info to me, i thought i had this CnQ thing down, but i guess I didn't know about this multiplier issue with startup.
-
dukla2000
- *Lifetime Patron*
- Posts: 1465
- Joined: Sun Mar 09, 2003 12:27 pm
- Location: Reading.England.EU
Well I am no expert on all the flavours of A64 or CnQ, but let me try explain why I think the way I do.
I guess your OPN is something like ADA3700DAA5BN. Unfortunately AMD have not updated their A64 Power and Thermal Data sheet for over a year, so I am guessing that your processor, like the 2200MHz ADA3400AEP5AR in that guide, has 4 P-states. These are the things CnQ toggles between. Although the Data sheet talks about MHz at each P-state, in fact the crucial point is the multplier needed for that frequency assuming 200HTT. So your P-states are (I think):
CnQ will toggle up and down those multipliers depending on load, and despite any multiplier you have selected in the BIOS.
CnQ will also toggle up and down some VCore options, unfortunately because your CPU is not in the Data sheet I am not 100% sure what they are. At Max P-state it will be stock VCore (1.35 or 1.4 methinks for your CPU) and probably 1.1V at Min P-state.
CnQ will also cause the memory divider to change to as closely match your BIOS selection depending on the theoretical CPU speed (theoretical because CnQ assumes HTT is 200)
CnQ will not adjust or consider actual HTT. So you are free to adjust this (overclock) and CnQ will leave it be.
When you overclock with HTT 218, CnQ will toggle between the 11, 10, 9 & 5 multipliers (or 2398, 2180, 1962 and 1090MHz actual). The point of this long winded ramble is that it is possible, with CnQ enabled, for you to see 10 multi sometimes but for a different reason to what you thought. If my logic is correct it is when CnQ is in Int 1 P-state, not as (I think) you think when it is at Max P-state. Certainly when I ran Prime in Win with CnQ enabled, my system only ran at Min P-state - i.e. Prime95 did not have sufficient priority to kick CnQ to Max P-state.
My assertion is CnQ ignores the CPU multi you set in the BIOS: it should retrieve the CPU capabilities direct from the Processor status registers. I am not very good at reading code, but that is what I think I decipher from the AMD Linux CnQ driver. There are exceptions that seem to be for less able mobo/BIOS that do not play ACPI properly though.
The other hiccup is the RAM divider. First the important point is the RAM speed is always derived from the CPU speed via a divider (because the memory controller is on the CPU). HTT is academic (although obviously it plays a big part in generating CPU speed). For your CPU the RAM dividers for 200 (DDR400), 166 (DDR333) and 133 (DDR266) are 11, 14 & 17 respectively. And these remain constant (for Max P-state) whatever you do to HTT. I am confused by your
At HTT 218, multi 11 CPU is 2398 and with RAM defined as 166 (DDR333) the A64 should pick the 14 divider giving 2398/14 = 171. At Min P-state the memory controller assumes CPU speed is 1000 and thus selects a divider of 6, which actually results in a RAM speed of 181. How are you reading your RAM speeds - I know CPU-Z reported my dividers correctly.
I guess this is all theory though. And there may well be huge deficiencies in my version of the theory! What I would suggest is set HTT to 218, multi 11 and RAM to 166 (DDR333) and boot. Then watch what happens in CPU-Z to your CPU speed, VCore and RAM divider as CnQ does its thing. With Prime in Win2k I had to go to control panel & disable CnQ to see my full overclock in CPU-Z. I am not sure what to recommend to get a stable view of the Int 1 & 2 P-states.
One last thought - I presume you have APM disabled in BIOS and ACPI enabled? As per a couple of paras back, in the Linux CnQ code this may make a difference but I never tested it. And that you have the latest version (from AMD) of the CnQ driver for your OS?
ps - yet another last thought! As you are (I think) constrained by a high CPU multi, low CPU overclock (around 10% at stock VCore), lower mem overclock and the huge gap between DDR400 and DDR333 (16%), it may be possible to improve your options by using Clockgen (or similar) to set your RAM speed to 183. Not sure how well this plays with CnQ though.
pps - an even more meaningful test would be BIOS multi 10, HTT 218. I suspect at full, high priority load CnQ will kick the multi to 11.
I guess your OPN is something like ADA3700DAA5BN. Unfortunately AMD have not updated their A64 Power and Thermal Data sheet for over a year, so I am guessing that your processor, like the 2200MHz ADA3400AEP5AR in that guide, has 4 P-states. These are the things CnQ toggles between. Although the Data sheet talks about MHz at each P-state, in fact the crucial point is the multplier needed for that frequency assuming 200HTT. So your P-states are (I think):
Code: Select all
State Freq Multi
Max 2200 11
Int 1 2000 10
Int 2 1800 9
Min 1000 5CnQ will also toggle up and down some VCore options, unfortunately because your CPU is not in the Data sheet I am not 100% sure what they are. At Max P-state it will be stock VCore (1.35 or 1.4 methinks for your CPU) and probably 1.1V at Min P-state.
CnQ will also cause the memory divider to change to as closely match your BIOS selection depending on the theoretical CPU speed (theoretical because CnQ assumes HTT is 200)
CnQ will not adjust or consider actual HTT. So you are free to adjust this (overclock) and CnQ will leave it be.
When you overclock with HTT 218, CnQ will toggle between the 11, 10, 9 & 5 multipliers (or 2398, 2180, 1962 and 1090MHz actual). The point of this long winded ramble is that it is possible, with CnQ enabled, for you to see 10 multi sometimes but for a different reason to what you thought. If my logic is correct it is when CnQ is in Int 1 P-state, not as (I think) you think when it is at Max P-state. Certainly when I ran Prime in Win with CnQ enabled, my system only ran at Min P-state - i.e. Prime95 did not have sufficient priority to kick CnQ to Max P-state.
My assertion is CnQ ignores the CPU multi you set in the BIOS: it should retrieve the CPU capabilities direct from the Processor status registers. I am not very good at reading code, but that is what I think I decipher from the AMD Linux CnQ driver. There are exceptions that seem to be for less able mobo/BIOS that do not play ACPI properly though.
The other hiccup is the RAM divider. First the important point is the RAM speed is always derived from the CPU speed via a divider (because the memory controller is on the CPU). HTT is academic (although obviously it plays a big part in generating CPU speed). For your CPU the RAM dividers for 200 (DDR400), 166 (DDR333) and 133 (DDR266) are 11, 14 & 17 respectively. And these remain constant (for Max P-state) whatever you do to HTT. I am confused by your
I thought your problem was getting CnQ to work if you set HTT 250, multi 10? But in any event your CPU memory controller has decided to use 200*10 as its baseline to calculate the RAM divider and thus selects 12 (2000/12 = 166) but with HTT 250 you get 2500/12 = 208.10x CPU multi = 208 with and without CnQ
11x CPU multi = 196 normal and 208 with CnQ
At HTT 218, multi 11 CPU is 2398 and with RAM defined as 166 (DDR333) the A64 should pick the 14 divider giving 2398/14 = 171. At Min P-state the memory controller assumes CPU speed is 1000 and thus selects a divider of 6, which actually results in a RAM speed of 181. How are you reading your RAM speeds - I know CPU-Z reported my dividers correctly.
I guess this is all theory though. And there may well be huge deficiencies in my version of the theory! What I would suggest is set HTT to 218, multi 11 and RAM to 166 (DDR333) and boot. Then watch what happens in CPU-Z to your CPU speed, VCore and RAM divider as CnQ does its thing. With Prime in Win2k I had to go to control panel & disable CnQ to see my full overclock in CPU-Z. I am not sure what to recommend to get a stable view of the Int 1 & 2 P-states.
One last thought - I presume you have APM disabled in BIOS and ACPI enabled? As per a couple of paras back, in the Linux CnQ code this may make a difference but I never tested it. And that you have the latest version (from AMD) of the CnQ driver for your OS?
ps - yet another last thought! As you are (I think) constrained by a high CPU multi, low CPU overclock (around 10% at stock VCore), lower mem overclock and the huge gap between DDR400 and DDR333 (16%), it may be possible to improve your options by using Clockgen (or similar) to set your RAM speed to 183. Not sure how well this plays with CnQ though.
pps - an even more meaningful test would be BIOS multi 10, HTT 218. I suspect at full, high priority load CnQ will kick the multi to 11.