E8400 4GHz 9 hours orthos stable, but not in L4D

keyser

keyser

keyser

Associate
Joined
23 Aug 2005
Posts
1,274
New problem, new thread! Following this thread: http://forums.overclockers.co.uk/showthread.php?t=17973083

My PC can do 4005MHz for 9 hours in orthos, but keeps crashing in L4D. To be fair, the first time was my fault, I forgot to attach the gpu fan molex :o. But after I fixed that, after 5-10 mins of L4D, I get sound loop for 2 secs then it either quits to desktop or resets PC (even though I have auto reboot disabled). Thats with speedstep disabled, so its not that.

No problems if I clock down to 3.6Ghz. Should I run Orthos and FurMark overnight? :) Or just up clock until it crashes l4d again? In bios, what does CPU Margin Enhancement mean? options are optimized (set to that, default), compatible and performance.

L4D stable is 3.6ghz, cpu at 1.1625v and DRAM 2.04v (I read that P5Q board add +0.08v to DRAM), but for 9 hours of orthos it was 4ghz, cpu 1.15625v, DRAM 2.02. L4D is more taxing than orthos?

d.png
 
gurusan has a lot of experiences with these chips on P5Q so I'm sure he will be here later but I think you need to start setting other voltages manually instead of leaving them at auto.

First I'd enable load line calibration to eliminate vdrop. Also NB voltage seems a bit too low.

Just for reference I have my GTL/PLL/FSB/NB at 0.63x/1.56/1.26/1.28.

Source engine based games including L4D are very sensitive to overall sys stability I find not just the CPU.
 
keyser said:
My PC can do 4005MHz for 9 hours in orthos, but keeps crashing in [Left 4 Dead]
Click to expand...

Hey keyser,

Prime is a very basic diagnostic used to test CPU stability (Small FFTs) and has no bearing on overall system stability! . . . If you have successfully passed a 9 hour run of Prime but games crash then it sounds like your Northbridge or Memory are falling over, hopefully this is something you can correct with the right BIOS tweak!

Have you tried running this diagnostic?

IntelBurnTest v1.6

That literally floods the system with data and finds errors quickly, it also heats up a processor a lot more than Prime Small FFTs did. From what you have described I expect IBT to fail when you do a 5-10 loop test. It's easy to use, just say YES to error checking, select maximum stress #1 and choose 5-10 loops to start (5 loops takes about 15 mins).

Also post up some CPU-z screenies (inc memory tabs) showing us where your at! :cool:
 
Thanks for help. OK, I found the limit to how low my cpu volts can be at 4ghz, now at 1.15v and 8 hours orthos stable. I increased the NB to 1.28, DRAM now 2.04 and enabled load line calibration. I haven't yet touched GTL/PLL/FSB because I have no idea what I'm doing with those :) It passes 10 runs of IntelBurnTest, so will now test some L4D! In Orthos the cpu got to 60c, in a cold room overnight, running IntelBurnTest cpu gets to 74/75c!!

ss_4ghz_orthos_intelburntest_stable_1.png


Edit: L4D reset pc, increased DRAM to 2.06, NB to 1.3, tried again but BSOD this time. I read on here somewhere that the P5Q mobo adds +0.08v to ram, is this true? My ram is rated for 2.1v. Also is there a temp sensor on the northbridge? Maybe Left 4 Dead needs more volts on the cpu?
 
Last edited:
Yeah I'll run that now, how long should that run for? My PC just reset while it was doing nothing! Lowered to 3.8Ghz.

Update: Ran memtest twice and passed. Probably need to test it overnight.
 
Last edited:
Its not the common L4D fault where you need to put mat_queue_mode 0 into console to stop it looping sound and freezing up? :)

It did that on mine until i added that command line and now its smooth as hell
 
Nah, well mines all default settings, mat_queue_mode is -1 which is default and I assume off. But, good news, L4D doesn't crash when at 3.8GHz, I can probably live with that :) I'll run memtest overnight.

Update: Memtest passed overnight

Current settings:
FSB 428
FSB Strap to Northbridge: 333
RAM 847MHz (It wasn't stable when at 1066MHz on 3GHz cpu)
5-5-5-15 and Row Refresh Cycle: 55
CPU 1.15v
GTL 0.63x (GTL/PLL/FSB based on info from google/other forums)
PLL 1.5v
FSB 1.2v
RAM 2.02v (P5Q overvolts by 0.08v, apparently)
NB 1.3v
SB 1.2v (guess)
PCIE SATA volts 1.6v
Load Line Calibration Enabled
CPU & PCIE spread spectrum Disabled
CPU Margin Enhancement Optimized (no idea what this is)
C1E disabled
 
Last edited:
3.9GHz is my limit, on 4GHz games crash, even when NB is up to 1.4v. I got ram working at 869 / 4-4-4-12, as it doesn't work at its rated 1066 / 5-5-5-15. Orthos and memtest 8 hours stable. 1mil superpi in 12secs yay :)

VID is 1.15, so I'm still on stock volts? :eek: Temps: idle/orthos (small) load 41/64

CPU Ratio Setting : Auto (9)
FSB Frequency : 434 (cpu 3906)
PCIE Frequency: 100
FSB Strap to North Bridge : 333
DRAM Frequency: 869Mhz
DRAM Timing Control: 4-4-4-12
Row Refresh Cycle Time: 55

DRAM Static Read Control: Auto
Dram Read Training : Auto
MEM OC Charger : Auto
Ai Clock Twister : Auto
Ai Transaction Booster : Auto

CPU Voltage : 1.15
CPU GTL Reference: Auto
CPU PLL Voltage : 1.52
FSB Termination Voltage: 1.22
DRAM Voltage : 2.02
NB Voltage : 1.3
SB Voltage : 1.2
PCIE SATA Voltage: 1.5
Load Line Calabration : Enabled
CPU Spread Spectrum : Disabled
PCIE Spread Spectrum : Disabled
CPU Clock Skew: Auto
NB Clock Skew: Auto
CPU Margin Enhancement : Optizimed

C1E : Enabled
Intel Speedstep : Enabled
 
Last edited:
I've found DDR2 needs at least 2.1 to be stable at higher speeds... and dont be scared to try some more cpu voltage.. I had 1.45 on my 8500 under a TRUE and its temps didn't get above 65.
 
Try these...I found LLC doesn't play nicely when using vcores below 1.2ish

DRAM Static Read Control: DISABLED
Dram Read Training : DISABLED
MEM OC Charger : ENABLED
Ai Clock Twister : Auto
Ai Transaction Booster : Auto

CPU Voltage : 1.15
CPU GTL Reference: Auto
CPU PLL Voltage : 1.5
FSB Termination Voltage: 1.2
DRAM Voltage : 2.04
NB Voltage : 1.26
SB Voltage : 1.2
PCIE SATA Voltage: 1.5
Load Line Calabration : DISABLED
CPU Spread Spectrum : Disabled
PCIE Spread Spectrum : Disabled
CPU Clock Skew: Auto
NB Clock Skew: Auto
CPU Margin Enhancement : Optizimed
 
gurusan said:
DRAM Static Read Control: DISABLED
Dram Read Training : DISABLED
MEM OC Charger : ENABLED
Click to expand...

Hey gurusan,

sorry to disturb keyser's thread but I'm interested to know more about those three options above, especially the two you have set as DISABLED :confused:

Thanks in advance and hopefully keyser can benefit from the insight too! :)
 
I found something out about that here

Thanks for the advice gurusan, I got up to 4.1Ghz with 1.225, but temp was high at 67c, then orthos failed after 43mins. At 4Ghz / 1.2v / 61c now, 1 hour orthos, will go for 8 hours orthos and 4-4-4-12 next.
 
Last edited:
Was at 4Ghz and 1.2v cpu, 5-5-5-15 then this happened in L4D

l4d_vs_hospital02_subway0001.jpg


Upped cpu to 1.20625v, played L4D again, but within 10 mins it quit to desktop. I think I'm gonna give up as testing just takes too long. Back at sig speeds. Seeing as 4GHz seems to be as fast as my cpu could go for 24/7 its only 100Mhz difference :)
 
keyser said:
3.9GHz is my limit, on 4GHz games crash, even when NB is up to 1.4v. I got ram working at 869 / 4-4-4-12, as it doesn't work at its rated 1066 / 5-5-5-15. Orthos and memtest 8 hours stable. 1mil superpi in 12secs yay :)

*snip*

C1E : Enabled
Intel Speedstep : Enabled

You sure about these last two? Thought you were supposed to turn these off when oc'ing?
Click to expand...
 
Sorry, yeah I turn them off when testing, but everything works fine at 3.9GHz with speedstep on. But at 4GHz, all tests pass (IntelBurnTest 10-15 runs and Orthos small + blend, memtest for 9 hours) it's just L4D, even tried PCIE at 101MHz. Very odd!
 
You must log in or register to reply here.
Back
Top Bottom