UD3P and Q9550 Overclock Problems

[Ken429]

Re: UD3P and Q9550 Overclock Problems

I am embarrassed to admit it but every time I have tried to back the memory multiplier off to 200 B or D the system won't even post! Is there something in the above stable settings that I need to change short of setting everything back to Auto to get the Bios to accept the new multiplier?

[profJim]

Re: UD3P and Q9550 Overclock Problems

See if you can use either of the 2.00 memory multipliers at 3.60GHz.
With 450MHz and 2.00D, tRD might need to be set 1 click higher than when using 2.00B.
With upper 400MHz settings, the B-strap usually requires tRD values that are one step higher compared with the D-strap at the same cpu host frequency.

Try the following, but don't apply all the settings at the same time:
CPU Clock Ratio ------------------------ : [ 8 X]
Fine CPU Clock Ratio ------------------- : [+0.0]
CPU Frequency -----------------------: 3.60GHz( 450X8.0)

System Memory Multiplier --------------- : [2.40B]
Memory Frequency (Mhz) -----------------: 1066 1080

Advanced Timing Control <--- optional, set all to AUTO temporarily
CPU
Voltage Types ----------- Normal ---------- Current
Load-Line Calibration ------------------ : [Disabled] <--- disable for now
CPU Vcore --------------- 1.28750V ----- : [1.3250V] <--- with LLC disabled, you might need somewhat higher volts
CPU Termination --------- 1.200V* ------ : [1.220V] <--- try up to 1.28 volts (starting with 1.24)
CPU PLL ----------------- 1.500V* ------ : [1.500V] <--- set to 1.55 or 1.57

See if setting the reference volts to AUTO helps at the lower overclock.
Keep in mind that changing CPU Termination (VTT) also changes the manually set values for the following Reference Voltages:
MCH/DRAM Ref
DRAM Termination
Channel A Ref
Channel B Ref
- - - -
Init Display First --------------------- : [PCI] <--- set to PEG with PCI-e graphics

[profJim]

Re: UD3P and Q9550 Overclock Problems

Additional info to further muddy the 2.40B waters:
With 2x1GB of Crucial Ballistix DDR-800 (2.2 volt) memory, my system passed stability testing using:

CPU Host Frequency: 492 MHz
CPU Clock Ratio ------------------------ : [ 7 X]
Fine CPU Clock Ratio ------------------- : [+0.0]
CPU Frequency ---------------- : 3.44GHz ( 492X7.0 )

System Memory Multiplier --------------- : [2.40B]
Memory Frequency (Mhz) ------ 800 -----: 1181 MHz

I didn't push my luck by testing with 4x1GB of memory.

[oc nub]

Re: UD3P and Q9550 Overclock Problems

I am embarrassed to admit it but every time I have tried to back the memory multiplier off to 200 B or D the system won't even post! Is there something in the above stable settings that I need to change short of setting everything back to Auto to get the Bios to accept the new multiplier?

Using 200D-1:1 ratio you will need to use PL9, only way you can use PL8 200D is at cas4. PL=Static t ReadValue

[profJim]

Re: UD3P and Q9550 Overclock Problems

Using 200D-1:1 ratio you will need to use PL9, only way you can use PL8 200D is at cas4. PL=Static t ReadValue

Maybe, maybe not.
Even if we had identical hardware setups, some of our bios settings would not be identical.
My system has passed extended stability testing with tRD = 8 with 5-5-5-15 (1066 - 1100MHz) using 2.40B. With cpu settings greater than about 470MHz, tRD = 9 or 10 was required when using 2.00B. My best memory benchmarks were with tRD = 9, using 2.00D, from 480 - 494MHz.

Some have had more trouble getting their systems stable when using 2.00B versus using 2.00D. If tRD (aka Performance Level) is too tight, then your system will have stability and/or boot problems.

My preference is to initially get my system stable with memory timings that are somewhat loose and voltages that are probably a bit higher than might be needed.

[oc nub]

Re: UD3P and Q9550 Overclock Problems

Maybe, maybe not.
Even if we had identical hardware setups, some of our bios settings would not be identical.
My system has passed extended stability testing with tRD = 8 with 5-5-5-15 (1066 - 1100MHz) using 2.40B. With cpu settings greater than about 470MHz, tRD = 9 or 10 was required when using 2.00B. My best memory benchmarks were with tRD = 9, using 2.00D, from 480 - 494MHz.

Some have had more trouble getting their systems stable when using 2.00B versus using 2.00D. If tRD (aka Performance Level) is too tight, then your system will have stability and/or boot problems.

My preference is to initially get my system stable with memory timings that are somewhat loose and voltages that are probably a bit higher than might be needed.

Yes 240B allows you to use PL8, but 200D does not normally unless your ram will run at cas4, for cas5 it requires PL9. I have an old 2x1gb kit of dominators that will do cas4 1000mhz, that was the only way possible for me to boot using PL8 200D. Ken429 asked "Is there something in the above stable settings that I need to change short of setting everything back to Auto to get the Bios to accept the new multiplier?: His performance level in his template is set to 8, using 9 should allow him to get into windows for testing.

Some results on the ep45 dq6, this also worked on my ud3p.

XtremeSystems Forums - View Single Post - The E0 Q9550 SLB9V thread (finally!)

XtremeSystems Forums - View Single Post - The E0 Q9550 SLB9V thread (finally!)
XtremeSystems Forums - View Single Post - The E0 Q9550 SLB9V thread (finally!)

XtremeSystems Forums - View Single Post - The E0 Q9550 SLB9V thread (finally!)

[profJim]

Re: UD3P and Q9550 Overclock Problems

Sorry for the confusion, when I read post #124:
"Using 200D-1:1 ratio you will need to use PL9, only way you can use PL8 200D is at cas4. PL=Static t ReadValue"
I mis-read the the first part of your statement and thought that you were stating that PL = 8 would only work with cas4 settings when using the 2.00 memory multiplier.

Nice info in the XtremeSystems thread. I don't recall any TweakTown threads where tRD = 8 was stable on P45 mobos with cpu host frequency above 500MHz. I wish I had the time to read all 34 pages of the 833 posts!

Suggestion: for screen-shots, include the Memset SPD panel that includes the memory manufacturer's name and memory model/part number.
I recommend that TT posters post screen-shots that include:

1. Memset using the SPD and ">>" tabs
2. cpu-z CPU and Mainboard screens (several instances can be run at the same time)
3. EasyTune6 --> TUNER tab --> ADVANCE tab --> VOLTAGE tab (only the Voltage tab is needed)
4. CoreTemp screen [for your cpu's VID (Voltage ID)]

[oc nub]

Re: UD3P and Q9550 Overclock Problems

profJim: Have you been able to get stability using PL8 with the 2.00D memory multiplier? I was not able to, it required PL9.
Using the G.Skill F2-9600CL5D-4GBPI's a lot of guys were able to use PL8 and the 2.40B memory multiplier at well over 500 FSB.

[profJim]

Re: UD3P and Q9550 Overclock Problems

oc nub:
The quick answer to your question is: yes, my system was stable with 2.00D and tRD = 8 with some mid 4xxMHz cpu settings, but 2.00B was stable with tRD = 7 at the same speeds.

My E6550 could boot and run benchmarks up to 529MHz, but failed extended stability testing with cpu host frequency higher than 494MHz.
The 7X cpu multiplier severely limited my o/c options.

I found that with lower to mid 4xxMHz settings, 2.00B and tRD = 7 or 8 gave better memory performance, about 200 - 250MHz higher bandwidth compared with 2.00D and tRD = 8 or 9.

With my cpu host frequency about 475MHz or higher, 2.00D gave better memory bandwidth because 2.00B now required tRD to be one step higher than the D-strap at the same speed setting.

As I recall, I tried several setups where I loosened the primary memory timings with the memory strap that required the higher tRD setting, but didn't perform stability testing because the memory benchmarks were worse when I tried lower tRD with somewhat looser primary memory timings.

With my 3.444GHz, 492x7, 984MHz, 5-5-4-12 (2.00D) settings, the Everest read/write/copy scores were 9261/8846/9251 MB/sec.

FWIW, my 2x2GB Mushkin 996599 memory could run with 4-4-4-12 primary memory timings up to 875MHz. This was somewhat disappointing because with with 4x1GB of Crucial Ballistix 2.2 volt DDR-800 memory, the 4-4-4-12 settings were stable up to 920 - 925MHz speeds.

Thanks for mentioning that the G.Skill F2-9600CL5D-4GBPI modules allowed you and others to use 2.40B with 1200+ MHz memory speeds.

[oc nub]

Re: UD3P and Q9550 Overclock Problems

Very nice info, Thank you. I had little success with my Mushkin kit as well.

[Ken429]

Re: UD3P and Q9550 Overclock Problems

Back to my issues! Just kidding. I am not proud of how I got there but we did get this thing to run at 4GHz with the 2.40B (471*2.4=1130) memory multiplier. I had to increase the Vcore to 1.35v (LLC Enabled), the CPU Termination to 1.24v and the MCH Core to 1.32v with the DRAM Voltage at 2.1v. The system ran Prime95 for 6+ hours and reached a maximum CPU temperature of 52C a few times but spent most of the time at 48-50C. I probably need a temperature probe to see just how hot the Intel Controller gets at 1.32v? Probably don't want to run at those settings every day? I got pictures if you want to see what it looked like. I did not get the Mem Set display since I had not downloaded the latest version which works under W7. For some reason the one I used a couple of months ago would not run under W7. Will I do damage if I try to get to 480*2.4=1152 which is what Mushkin says is the Maximum or... move on to the 2.00 settings?

[oc nub]

Re: UD3P and Q9550 Overclock Problems

Those voltages should be fine for 24/7, just put a fan blowing on your NB if you are concerned.

[Ken429]

Re: UD3P and Q9550 Overclock Problems

Prime95 has been running for 4 hours with the FSB at 475*2.4=1140 and the CPU temperature has not gone above 52C. Looks like 480*2.4=1152 maybe in the cards. Any suggestions on what kind of fan I need to comfort and cool the P45 chip?

[oc nub]

Re: UD3P and Q9550 Overclock Problems

It doesn't have to be anything special, I use a 120mm scythe, it is sitting on my gfx card. Any fan will help.

[kick]

Re: UD3P and Q9550 Overclock Problems

Nb temp runs fairly close to the system temp with a nb/memory fan installed.
There is a 2-3 degree variation

I find it doesnt like much above 42C,thats the temp it reaches for me with 1.42 v applied and running 2020 fsb on a warm day,like 25c indoors.

Of course I cant guarantee thats what may give errors it could be the memory ,it could be the cpu/fsb limit.
It may be coincidental that nb temp hits 43C and errors occur,but I havent found a way to make it hotter below 2000 fsb,I suppose i could if i turned off the fan and applied more voltage,but why would I want to :0

There is an intel article about installing a temperature sensor on the nb,but id have to trawl back thru my posts.
Its probably easy enought to find on google.