OC BIOS options

Recheck your overclock by running LinX and / or Prime blend.

Here are two quick LinX tests. These usually pass. I am a little weary with my temps right now. I don't like going past 60C. The reason that they are high right now though is because I have the VCore pumped to provide stability.



I am going to run Prime overnight to see where it makes it to. If I remember correctly it was making it to about 15mins and then some of the cores were stopping.

EDIT: Here is a quick one that I did while I was out to eat. Seems to show proof of concept. I had read on the HOW TO sticky about how the OP was diagnosing the issues that arose. I am having a little difficulty figuring out how to diagnose the stopped workers. Any tips?



EDIT: The top two pictures will not post for some reason...

Bump for thoughts.

Try setting both CPU GTL REF Voltages to either 0.630 or 0.640, FSB Terminaton Voltage to 1.30V, AI Transaction Booster to Manual, Common Performance Level to 11, and NB Voltage to 1.34V. You could also try taking the NB GTL vREF off Auto or 0.680, and manually changing it to 0.670.

Edit: To give you a better understanding of setting up GTL vREFs:

How to set up GTL Ref Values for 45nm & 65nm
http://www.xtremesystems.org/forums/showthread.php?t=202292&highlight=gtl

... and for a more in-depth explanation:

Adjusting [Advanced] Gunning Transceiver Logic (A/GTL+) Voltage Levels for Increased Front Side Bus (FSB) Signaling Margins and Overclocking
http://www.thetechrepository.com/showthread.php?t=87

Yay reading! Something to do while I am in the server room at work. MUAUAHAHH!!! Thanks Red!

I have read the second link that you posted Red and was taken aback by it. I left it alone the first time that I re-read it.

However the second post was very useful. Per what I can tell from the GTL RefV spread sheet that the OP mentioned my GTL Ref should be .635 and .667

I am not sure if i mentioned it above but when I had changed the GTL Ref before to below .640 I had to reload the BIOS every time. Is there a particuler reason that that happens?

I have attached the excel spreadsheet that Seban had created with my value entered in the "Vtt for set 1 (in V)".

Testeing

Here are some test results that I got while trying out the settings that you suggested. For anyone else that has just tuned in I am trying to simplify my BIOS settings and test one thing at a time. When I had first started out there were so many people making suggestions that I lost track of what I had changed. Therefore I ended up with all the settings in my BIOS being changed. So I went back and cleaned it up and set a goal to reduce my Vcore so that I can lower my temps and possibly push a little farther. All of the tests that I did I used LinX on 10 passes for a quick test. I will run P95 at a later date to confirm. Thanks for the settings Red.

Notes:
There seems to be a 3-4C difference in cores 0 and 1 to 2 and 3.

Test: 1
Variable: CPU GTL Ref - .630
Vcore: 1.55000
Results: Completed 10 passes in 3m 3s. Lowered Vcore.

Test: 2
Variable: CPU GTL Ref - .630
Vcore: 1.53750
Results: Completed 10 passes in 3m 4s. Lowered Vcore.

Test: 3
Variable: CPU GTL Ref - .630
Vcore: 1.52500
Results: Completed 10 passes in 3m 4s. Lowered Vcore.
Temperature:
Ambient: 28.2
CPU Max: 66C on 2 and 3

Test: 4
Variable: CPU GTL Ref - .630
Vcore: 1.51250
Results: Completed 10 passes in 3m 4s. Lowered Vcore.
Temperature:
Ambient: 28.3
CPU Max: 65C on 2 and 3

Test: 5
Variable: CPU GTL Ref - .630
Vcore: 1.50000
Results: Froze after 2m 4s of testing. Raised Vcore.
Temperature:
Ambient: 28.3
CPU Max: 63C on 2 and 3

Test: 6
Variable: CPU GTL Ref - .630
Vcore: 1.50625
Results: Completed 10 passes after 3m 4s. Changed CPU GTL Ref to .640.
Temperature:
Ambient: 28.2
CPU Max: 63C on 2 and 3

Test: 7
Variable: CPU GTL Ref - .640
Vcore: 1.50625
Results: Froze after 1m 11s. Changed CPU GTL Ref to .630. Changed NB GTL Ref to .670.
Temperature:
Ambient: 28.1C
CPU Max: 63C on 2 and 3

Test: 8
Variable: NB GTL Ref - .670
Vcore: 1.50625
Temperature:
Ambient: 28.1
CPU Max:63C on 2 and 3
Results: Completed 10 passes after 3m 4s. Went to bed.

Ok, so I ran some more tests today; AKA played CoD4 and found that the settings that I was using before were the most stable. I don't know what it is but the Auto setting seems to be the best. I raised the Vcore back up one notch from the last test one night after my last BSOD. I have not tried out the AI Transaction Booster yet however.

Since I have everything on Auto is there any program that will allow me to see what the "Auto" setting is once I am inside the Windows environment? That way I will know what it is choosing and be able to change the settings in the BIOS accordingly.

Something else that I have noticed during POST. It seems that the CPU speed fluctuates ever so slightly from 3.94 to 3.93 every so often. I am not changing anything in the BIOS to make this happen. I have just always figured that the speed is being rounded and haven't worried about mentioning it. Would this cause an issue?

The good news is that I am using less Vcore now. It still doesn't leave any room for OCing farther but eh.

IRQL Error

I am going to crack this thread back open. I have been getting hangups and the occasional BSOD. The hangups occur at random times so I can't pin those down on anything. It happens rather I am doing something or the system is idle. All the temps seem to be fine at idle floating around 32C. It is really the IRQL BSOD that is bother me now. I only have the error when I am running OCed. It also seems like I have to have been using the system for most of the day. I read some posts online about it being hardware or software related. I want to lean towards software because of the intermittent nature of the error and the fact that it doesn't happen when I run stock. Now I am runnning an ATI card (main) and Nvidia card (secondary). I have tried reloading the drivers for both of those. The issue came back.

On a side note. Whenever I move from stock settings to OCed settings I have to hit the reset button for the comp to load into windows. Example: I am running stock, I switch settings to OC, comp turns off, turns on and nothing loads. I hit the rest button and the comp stays on, POSTS and shows stock settings, load settings from BIOS, comp stays on, loads into windows. Don't know if it related or not but I figure that it is worth mentioning.

The settings that I have chosen are still current with the ones from 5-24-10. I also ran some tests on 7-6-10 if those will help diagnose any issues. I am not currently at the hose and I will try to get the entire BSOD message when I get home. I always have to fight to get it. Any help is as always appreciated!

BAD_POOL_HEADER

I got that BSOD last night while running OCed. It happed just a little after another IRQL error. When I reset to stock BIOS options the comp didn't recognize my RAID setup! Anyway I got that fixed. I think that it might have been the options. This isn't the first time that I have gottent he BAD_POOL error. It all stops though once I run stock. Any suggestions?

It's hard (and sometimes impossible) to stabilize the FSB above ~475 w/ 45nm quads. I've found that the Large-FFT or Blend test will show errors the fastest when the FSB is unstable. FSB instability is the hardest 1 to nail down imo.

NB GTL has a big effect on it as you may have noticed.

Try turning your CPU multi down to 6, so you can focus on the FSB for now. Run Large-FFT, and take notes on how long it lasts before a failure. Then try playing w/ NB-GTL, NBv, FSB-Term, and possibly some other settings to see if you can get it to run longer.

The FSB is going to hold this chip back the same way it does on my Q9550. I'm only at 1.256v for 4GHz, but it has more potential if there was a higher multi. You're in a similar situation, and if you can get 465-475FSB stable then you're looking at 3.7-3.8GHz final OC and that might only need 1.2-1.25v to achieve.

Once the FSB is stable, switch focus back to the CPU by setting the multi to 8, and using small-fft to stress while adjusting vcore to stabilize.

Thanks for the info. I will have to give that a try. Yea this chip has treated me really well for everything that I have put it through.

I am glad to see that you didn't mention anything abuot the BSODs. If I am not mistaken I had read up on the POOL_HEADER being a HDD failure issue. These drives have been through thier paces as well.

Testing again

Sorry to revive this thread but it is an ongoing project for me. I am at it again testing and changing system settings. Here are three quick tests that I did just to test the multiplier theory. I feel like my temps are getting a little high but I know the reason for that.

Something that I noticed while I was testing. It seemed like the screen would freeze during testing for ~3 secs whenever my background wallpaper was set to change. This is curious because I only BSODs that relate to HDD issues when I OC. Could this have anything to do with the stability issue?

Here are the shots: I increased the multi on each one.




Jason, I will be taking your advice and running P95 overnight.

jason4207 said:

It's hard (and sometimes impossible) to stabilize the FSB above ~475 w/ 45nm quads. I've found that the Large-FFT or Blend test will show errors the fastest when the FSB is unstable. FSB instability is the hardest 1 to nail down imo.

NB GTL has a big effect on it as you may have noticed.

Try turning your CPU multi down to 6, so you can focus on the FSB for now. Run Large-FFT, and take notes on how long it lasts before a failure. Then try playing w/ NB-GTL, NBv, FSB-Term, and possibly some other settings to see if you can get it to run longer.

The FSB is going to hold this chip back the same way it does on my Q9550. I'm only at 1.256v for 4GHz, but it has more potential if there was a higher multi. You're in a similar situation, and if you can get 465-475FSB stable then you're looking at 3.7-3.8GHz final OC and that might only need 1.2-1.25v to achieve.

Once the FSB is stable, switch focus back to the CPU by setting the multi to 8, and using small-fft to stress while adjusting vcore to stabilize.

Click to expand...

Ok, I ran the large FFT test overnight, or what was supposed to be overnight. The FSB is 490 and the multi was 6. Any thoughts on what the test means? I am not as familiar with how to diagnose P95.

Coreyoliseffect said:

Ok, I ran the large FFT test overnight, or what was supposed to be overnight. The FSB is 490 and the multi was 6. Any thoughts on what the test means? I am not as familiar with how to diagnose P95.

Click to expand...

FSB or RAM is failing...most likely the FSB since you have it so high.

See my previous comments for some BIOS settings to play with, but keep in mind that you may not be able to find complete stability w/ the FSB that high.



Your vcore looks awfully high for the speeds you are running...

Is there any software that can tell me what those settings are once I am in Win7 or do I just have to guess? I have read some articles on the these settings but can't ever find a setting that seems like it is more stable.

Maybe I'm confused, but I just write down the settings while I'm in the BIOS, and take notes on what helps and what hurts during my OCing. You shouldn't need to monitor them from w/n Windows 7 if you see what they are via the BIOS.

If you have settings on Auto it's time to take off the training wheels. Plus, w/ the FSB that high there is no telling how badly your mobo is overvolting everything. PLL & VTT might be set dangerously high. Your vcore looks way higher than it should be for those CPU speeds.

Just try 1 setting at a time, and record when P95 1st fails and what exactly happens (does a core fail, and if so at what fft size; BSOD, freeze, etc.). Hopefully, w/ time and patience you will start to get a feel for how each setting interacts w/ your OC.

I still think you might not be able to run the FSB that fast and maintain 100% stability. I tried for 4 weeks on my rig, and finally settled for 471FSB. I tried every setting possible including clock skews. I could have gotten 475...maybe 480 stable, but I was shooting for 500. Best I could do is run some benchmarks at 495, but it wasn't P95 stable.

Most of the high FSB results I've seen w/ 45nm quads have been on GB boards.

jason4207 said:

Maybe I'm confused, but I just write down the settings while I'm in the BIOS, and take notes on what helps and what hurts during my OCing. You shouldn't need to monitor them from w/n Windows 7 if you see what they are via the BIOS.

If you have settings on Auto it's time to take off the training wheels. Plus, w/ the FSB that high there is no telling how badly your mobo is overvolting everything. PLL & VTT might be set dangerously high. Your vcore looks way higher than it should be for those CPU speeds.

Just try 1 setting at a time, and record when P95 1st fails and what exactly happens (does a core fail, and if so at what fft size; BSOD, freeze, etc.). Hopefully, w/ time and patience you will start to get a feel for how each setting interacts w/ your OC.

I still think you might not be able to run the FSB that fast and maintain 100% stability. I tried for 4 weeks on my rig, and finally settled for 471FSB. I tried every setting possible including clock skews. I could have gotten 475...maybe 480 stable, but I was shooting for 500. Best I could do is run some benchmarks at 495, but it wasn't P95 stable.

Most of the high FSB results I've seen w/ 45nm quads have been on GB boards.

Click to expand...

I remember why I was asking about the software now. I was wanting to go ahead and manually set the items that I do have on Auto. I had a bunch of settings changed the first time around and found that I got a much more stable OC if I just left everything on Auto. I was wanting to get a baseline where I could start from. Hence me wanting to know if there was a piece of software that could monitor the settings from inside Win7. I don't mean to sound like I am heckling you; for lack of a better term. I just don't have a base line to start off with. I figured that if I had that I could tweak the settings from there.

... I think I have to type what I am saying before I realize what other people are trying to tell me. I had a revelation right as I typed the last sentence. Jason, you are suggesting that I set the options using the values that the mobo suggests then tweak them from there. Got'ya now.

I will have some extra time on my hand Friday. I hope that I can get some screens up. On another note, is LinX a decent stress testing software? It seems like it dosen't show the same issues that turn up when I run P95. I know that OCing is a long process but darn P95 can take a long time.

Linx is a bit tricky sometimes. I use IBT (Intel Burn Test) which is just an interface for Linx, but through reading and trial & error I've found that if you don't set it up right it can give a false sense of stability.

Since you're looking for FSB stability try setting the the Memory to High in Linx instead of 772 where you have it now. Don't use "All", though as sometimes it will over allocate and then you end up stressing the HDD I/O instead of the FSB-RAM interface.

To speed up P95 a bit I use custom settings. 1st select Blend, and then select Custom (this will populate the Custom settings just like Blend). You can leave the spread where it is or change it to 1024-4096k like Large-FFT (8-4096k will obviously take longer, but it will test the CPU more when the lower fft sizes come up). Change the RAM usage to something much higher, but don't over allocate. I see you have 8GB, so try something like 6144MB. Then set the time per fft size to 5mins instead of 15. Like this you can do a full spread run (8k-4096k) in about 4hrs, and a full large-fft run (1024k-4096k) in about 2hrs iirc. Plus, I've found this gets things a bit hotter and finds errors faster.

Hopefully, you can identify problematic FFT sizes more quickly this way. Once a problematic FFT size is found you can do a custom run on just that FFT size (set the same size for the start and finish fft sizes in the Custom options) for even faster feedback if your settings are helping or hurting.

As for settings, I would start below as a baseline, and then start a systematic approach to finding out what helps/hurts. This may involve backing off on the FSB a bit, and then trying to work your way back up. B/c if you fail P95 right away it's hard to tell if your changes have any effect. If you can get it to run for at least a few mins of P95 before failing then it's much easier to see if your BIOS change had any effect. If it goes from lasting 3mins to lasting 10mins then you know you've made a positive change. But if it fails immediately both before and after a change you really don't know if that change helped.

ASUS P5Q-E Template :

JumperFree Configuration Settings
AI Overclock tuner: MANUAL
CPU Ratio Setting: 8.0
FSB Frequency: 490 (may need to drop down to ~470 and work your way back up)
PCI-E Frequency: 100
FSB Strap to North Bridge: 400 MHz (Use 333 if you can)
DRAM Frequency: DDR2-980MHz
DRAM CLK Skew on Channel A1: AUTO
DRAM CLK Skew on Channel A2: AUTO
DRAM CLK Skew on Channel B1: AUTO
DRAM CLK Skew on Channel B2: AUTO
DRAM Timing Control: Manual

1st Information : 5-5-5-15-3-54-6-3
CAS# Latency: 5
DRAM RAS# to CAS# Delay: 5
DRAM RAS# Precharge: 5
DRAM RAS# Activate to Precharge: 15
RAS# to RAS# Delay : AUTO
Row Refresh Cycle Time: AUTO
Write Recovery Time: AUTO
Read to Precharge Time: AUTO

2nd Information : 8-3-5-4-6-4-7
READ to WRITE Delay (S/D): AUTO
Write to Read Delay (S): AUTO
WRITE to READ Delay (D): AUTO
READ to READ Delay (S): AUTO
READ to READ Delay (D): AUTO
WRITE to WRITE Delay (S): AUTO
WRITE to WRITE Delay (D): AUTO

3rd Information : 14-5-1-6-6
WRITE to PRE Delay: AUTO
READ to PRE Delay: AUTO
PRE to PRE Delay: AUTO
ALL PRE to ACT Delay: AUTO
ALL PRE to REF Delay: AUTO
DRAM Static Read Control: Disabled
DRAM Read Training: Disabled
MEM. OC Charger: Enabled
AI Clock Twister: AUTO Moderate
AI Transaction Booster: Auto (10 for now, but you want this as low as possible for the final OC; 9 should be easily stable, 8 is possible, and 7 probably won't POST)

CPU Voltage: 1.53750 1.35v
CPU GTL Voltage Reference (0/2): AUTO
CPU GTL Voltage Reference (1/3): AUTO
CPU PLL Voltage: AUTO 1.55-1.6v
FSB Termination Voltage: Auto 1.35v
DRAM Voltage: 2.20
NB Voltage: AUTO 1.35v
NB GTL Reference: 0.680 (I found this to have a huge effect of high FSB OCing...you'll have to experiment.)
SBridge Voltage: Auto
PCIE SATA Voltage: Auto

Load Line Calibration: Auto Enabled
CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled
CPU Clock Skew : Auto (Experiment between Auto and Normal)
NB Clock Skew : Auto (Experiment between Auto and Normal)
CPU Margin Enhancement: Optimized

Advance CPU Settings
CPU Ratio Setting: 8.0
C1E Suppport: Disabled


Keep in mind that higher is not always better. I've found that getting stable high FSB speeds is an art, and just adding a little extra voltage for "cushion" can actually make things worse. There is a definite sweet spot.

There is a reason that some people have blue stars. Thanks for the extra mile on your post. Now I feel obligated! Maybe I can get off work early tomorrow.