P5Q-E + E8400 : stuck at 4770 Mhz : need help

[overclock]

Hello,

I would really need help since I don’t manage to maximize the CPU frequency.

My configuration is the following :
ASUS P5Q-E bios 0803, E8400, 4x1 GB DDR2-800 4-4-4-12 Super Talent, Corsair 750 W, phase change cooler, 80 mm fans on memory and northbridge.

I obtain a stable overclock at 9x530 Mhz (30 min Prime 95 CPU blend test OK, Memtest 2.01 OK)

The parameters are the following :

AI Overclock tuner: MANUAL
CPU Ratio Setting: 9
FSB Frequency: 530
FSB Strap to North Bridge: 400
PCI-E Frequency: 100
DRAM Frequency: 1060
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: AUTO

1st Information :
CAS# Latency: 5
DRAM RAS# to CAS# Delay: 5
DRAM RAS# Precharge: 5
DRAM RAS# Activate to Precharge: 15
RAS# to RAS# Delay : 3
Row Refresh Cycle Time: 42
Write Recovery Time: 6
Read to Precharge Time: 3

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

3rd Information :
WRITE to PRE Delay: 14
READ to PRE Delay: 5
PRE to PRE Delay: 1
ALL PRE to ACT Delay: 5
ALL PRE to REF Delay: 5
DRAM Static Read Control: ENABLE
DRAM Read Training: ENABLE
MEM. OC Charger: ENABLE
AI Clock Twister: STRONG
AI Transaction Booster: MANUAL
COMMON PERFORMANCE LEVEL: 10
PULL-INS: ENABLED (ALL OF THEM)

CPU Voltage: 1.4
CPU GTL Voltage Reference (0/2): 0.63
CPU GTL Voltage Reference (1/3): 0.67
CPU PLL Voltage: 1.58
FSB Termination Voltage: 1.26
DRAM Voltage: 2.1
NB Voltage: 1.3
NB GTL Reference: 0.585
SBridge Voltage: 1.20
PCIE SATA Voltage: 1.50

Load Line Calibration: ENABLED
CPU Spread Spectrum: DISABLED
PCIE Spread Spectrum: DISABLED
CPU CLOCK SKEW : NORMAL
NB CLOCK SKEW : NORMAL

Whatever the FSB from 534 to 580 Mhz, I always obtain a few errors (1 -6) in the Memtest test 7, even if I change the following parameters :

FSB Strap to North Bridge: 333 or 400
DRAM Frequency: always synchronized with FSB
DRAM CLK Skew on Channel A1: every possible option
DRAM Timing Control: MANUAL

1st Information :
CAS# Latency: 5 or 6
DRAM RAS# to CAS# Delay: 5 or 6
DRAM RAS# Precharge: 5 or 6
DRAM RAS# Activate to Precharge: 15 or 18
Row Refresh Cycle Time: 42 or 60

DRAM Static Read Control: ENABLE or DISABLE
DRAM Read Training: ENABLE or DISABLE
MEM. OC Charger: ENABLE
AI Clock Twister: STRONG or MODERATE
AI Transaction Booster: MANUAL or AUTO
COMMON PERFORMANCE LEVEL: from 10 to 15
PULL-INS: ENABLED (ALL OF THEM)

CPU Voltage: 1.4 - 1.45
CPU GTL Voltage Reference (0/2): 0.6 - 0.65
CPU GTL Voltage Reference (1/3): 0.66- 0.68
CPU PLL Voltage: 1.52 -1.62
FSB Termination Voltage: 1.22 – 1.38
DRAM Voltage: 1.9 – 2.14
NB Voltage: 1.2 – 1.5
NB GTL Reference: 0.53 – 0.63
SBridge Voltage: 1.20 – 1.3
PCIE SATA Voltage: 1.50 – 1.6

CPU CLOCK SKEW : every possible option
NB CLOCK SKEW : every possible option

I checked that the RAM is Memtest stable at 1173 Mhz 6-6-6-18. I tested with these validated parameters in order not to be RAM-limited.

I did not check the maximum FSB but I was able to launch Memtest at FSB=580 Mhz.

Since I always have a few errors in Memtest test 7, whatever the bios parameters, I assume that only 1 or 2 parameters are the cause. Or, the bios is faulty and I must wait another one.

Really, I don’t know what to do now. Your help would be much appreciated.

Thanks.

[overclock]

I am finalizing my tests : the cause of my failure was the Vcore too low. I did not want to increase it more than 1.45 - 1.475 V. I did not think that only a few errors in test 7, whatever the FSB and the core frequency, may be caused by the Vcore too low, I rather thought that I would have had more and more errors in test 7 by increasing the CPU frequency.

So finally, my E8400 reaches 9x542 Mhz 1 h Prime 95 stable, at 1.475 V (bios). I am trying to decrease the voltages now while keeping stability.

The board is great with plenty of parameters but is quite sensitive.

[B.E.E.F.]

You running this 24/7?

[overclock]

Yes, I finally managed to solve my problem. Now, my E8400 runs at 4803 Mhz, fully stable, 24/7, under phase.

[B.E.E.F.]

I'm new to overclocking. I've got the E8400 with the P5Q-E. C0.

So far I have my E8400 at 3.96Ghz. 9*440. CPU voltage set to 1.4v in BIOS. Stable. CPUz tells me its at 1.36v. Everything else is on the default setting. Even speedstep, and all the other power savers.

Should I begin messing with the NB voltage next?

[overclock]

What are your bios settings in extreme tweaker part ?

Did you already test stability with Prime 95 or Orthos ?

What is your motherboard bios release ?

Did you check your overcloking ability batch ?

What are your cooling and temperatures on load ?

[B.E.E.F.]

I tested with wPrime. I'm using the m1402 BIOS. Cooling and temperatures on load are about 50C. I have to baseline realtemp to figure it out.

I used the same settings, except disabled load line calibration, and DRAM is at defaults. I achieved a OC of 4050mhz. 450x9.0. Core voltage at 1.368v

[overclock]

I tested with wPrime. I'm using the m1402 BIOS. Cooling and temperatures on load are about 50C. I have to baseline realtemp to figure it out.

I used the same settings, except disabled load line calibration, and DRAM is at defaults. I achieved a OC of 4050mhz. 450x9.0. Core voltage at 1.368v

Do you watercool your CPU ? Indeed, 50°C is obtained with water.

Did you check the overclockability of your batch ?

[B.E.E.F.]

I'm running a Thermalright HR-01 Plus with a Noctua 120mm.




I'm looking for the batch listings. I've got batch Q814A330. Nothing on Google about it.

[overclock]

So 4050 Mhz on air is not bad. Did you search for the frequencies attained with your batch ? Indeed, that would enable to know if your are likely to overclock further or not.

[B.E.E.F.]

For the Q814A batch its anywhere from 4ghz to 4.3ghz.

I want to push this to 4.5ghz stable without exceeding 1.4v vCore. Doable?

[overclock]

No, I do not think that 4.5 Ghz is posible with 1.4 V max.

If 1.368 V is given by CPUz, it means that the bios voltage is higher. Consequently, your current CPU voltage is near 1.4 V. You should reach 4.1 Ghz maximum, fully stable. I don't recommend to increase the Vcore further because of the damage to the CPU : if you apply 1.5 V or higher, your CPU will still work but its performance will deteriorate (tested with AMD 3700+ at 1.75 V at -50°C : around 7 % performance loss).

First, enable load line calibration. Be careful, that increases every voltage but the Vcore : VRAM is increased by about 0.1 V, other voltages by 0.05 V.

In order to find the maximum stable frequency, let's assume 4.3 Ghz, you have to validate :
- the maximum stable frequency of your RAM (477 Mhzx2 = 954 Mhz). You have to use dividers in order to obtain 954 Mhz for the RAM and a frequency inferior for the CPU (to test only the RAM, not the CPU). For the stability test, you can use Memtest first (tests 5 then 7 then the whole test). If you have no errors, then you RAM will be 99 % stable. If errors appear, you can increase the memory voltage first. If you always have errors, you can search for users who have both your motherboard and RAM, in order to copy their RAM settings.

- test the motherboard stability at FSB=477 Mhz. You choose the CPU multiplier = 8 and the divider RAM:FSB = 1:1. You test with Prime 95 (CPU and RAM, the third test, the most stressful) during 1 hour. It should be OK since the P5Q-E can overclock far higher than 477 Mhz. If it is not stable, you can increase 3 voltages : VNB, VCPU PLL and VFSB termination. For 477 Mhz, VNB should be around 1.3 V, VCPU PLL about 1.6 V and VFSB termination roughly 1.25 V.

When you validated the RAM (954 Mhz) and motherboard stability (477 Mhz), you can focus on the CPU : now, if Prime95 (CPU and RAM test) fail, you are sure that Vcore will be too low. So, in order to improve the CPU frequency, increase the Vcore. The CPU temperature must not exceed 65°C on load (Prime95) and its voltage must be inferior to 1.45 V (max Vcore for 45 nm CPUs).

I hope that should help.

[B.E.E.F.]

So when I increase the FSB the ram becomes unstable too?

I got DDR2-1000 specifically because I didn't want to OC the ram.

Yeah. I will be keeping this CPU under 1.4v. My PSU fan is already running harder than it should be... it adjusts to load.


I'll shoot for 533 FSB to match the ram, and lower the multi. Sounds good? Or is it better to increase the multi and go for less FSB?

[overclock]

No definite answer : by increasing the FSB, the bandwidth between the CPU and the northbrdige incrases which is good. But the gain is not substantial, only 1 - 2 % performance increase. And, you have to lessen the internal NB timings in order to attain 533 Mhz FSB. Furthermore, the higher the RAM frequency, the higher its timings which damages performance. And the level performance will have to be increased too in order to attain high FSBs like 533 Mhz.

The best is to keep the FSB lower and enhance performance by 3 ways :
- improve NB timings by decreasing "FSB strap to NB" parameter
- decrase RAM timings (especially the first 4 ones).
- decrease performance level (Ai transaction booster : manual then performance level appears). This parameter has a very strong impact on performance.

A few another RAM parameters have a minor performance impact : DRAm static read control, DRAM read training and mem OC charger on enabled and Ai clock twister on strongest.

Consequently, the RAM bandwidth will be strongly increased. In a few applications like Winrar, the gain is huge, in dozens of percent.

I answer to your 1st question : even if your RAM is rated at 1000 Mhz, you cannot be sure that it is stable on your motherboard. The RAM has to be certified on your motherboard at 1000 Mhz. You can check your RAM compatibility with the P5Q-E by reading the QVL on ASUS website. If it is not in the list, then you must check its stability with memtest and change its timings if necessary.