My First Overclocking Adventure A Success!!!

[KingstonU]

Finally upgraded after 4 years on an Athlon 3200+ and 1Gb of DDR1 ram (which has been good to me and is still working perfectly even after a bent pin).

Got the bug and ended up upgrading some of the parts that I planned. Roughly a 15% budget increase for ~10% performance increase :laugh: Whatever I deserve it. I ended up spending about $1400 CA for everything including 22" screen from DirectCanada.

Specs:

Q9550 (C1 Stepping)
P5Q Pro
2x2Gb G.Skill DDR2-1000 5-5-5-15
OCZ Vendetta 2 + OCZ Freeze TIM
Enermax modu82+ 625W
Antec 300 case
HD 4870


Finally have the time to tinker with it. I am using AT's own Overclocking guide :beer: thanks for everyone's advice thus far on the build.

Decided to put all of my notes that I've been keeping for anyone who can benefit from it.

Any tips appreciated. Comments welcome.


<-------------------->

*UPDATE*

Reached my 3.6GHz goal.

Thanks to AT OC guide and to everyone for their tips! :beer::beer::beer:

Before getting to the numbers I would just like to say that overall the system feels much snappier and responsive, XP seems to but up quicker and applications seem to load faster, which is what ultimately counts. Full BIOS settings at bottom.


PCMark05 Score: 9468 (+1107 over stock.
3DMArk06 Score: 15960 (+1620) over stock
Prime 95 benchmark: 1024k FFT = 15.926ms (-4.0ms over stock)
I know should have let the test run longer.

RealTemp is supposed to be the most accurate for 45nm quads and it reports my temperatures as 5'C lower than CoreTemp and HWMonitor. This has been the best way to lower my temperatures yet! :laugh:

My load temps are now 55/54/50/55.



My complete step by step over-clocking process below.

<-------------------->

Here goes:

23'C ambient temperature
Using HWMonitor v1.11
Prime95 v25.7 b3

Stock PCMark05 Score = 8361
Stock Prime 95 benchmark: 1024k FFT = 19.924ms

I will likely buy shortly a 120mm fan for the front and and one for the side.
Currently one 120mm fan on top and one on back + CPU cooler fan.

<-------------------->

Stock Settings, 4th push pin not in properly

idle: 38/39/41/41
load (10 minutes): 43/49/50/50

After reseating the heat sink:

idle: 38/39/40/40
load (10 minutes): 43/49/50/50 = -1'C improvement

After reducing amount of thermal compound:

idle: 37/37/38/38
load (10 minutes): 41/48/49/50 = -2'C improvement


<-------------------->


Overclocking log, Iteration Set 1
comments: Window open, no idea what VTT is. Left NB and SB on AUTO
FSB = 353 (3.0 GHz)


FSB = 353
CPU Voltage = 1.18750
NB Voltage = AUTO
SB Voltage = AUTO
DRAM Voltage = 2.00
VTT = ?

No Errors after 15 minutes

Temps:
idle: 34/34/35/38
load (15 minutes): 44/50/51/51 (window open, will close from now on)

Max fan speed = 1278 RPM


<-------------------->


Overclocking log, Iteration Set 2
comments: Window closed, VTT is FSB Termination Voltage, left on AUTO.
Only change is FSB = 389 (3.3 GHz)


FSB = 383
CPU Voltage = 1.18750
NB Voltage = AUTO
SB Voltage = AUTO
DRAM Voltage = 2.00
FSB Term (VTT) = AUTO

Errors = Worker #2 after 30 seconds

Temps:
idle: 38/39/40/40
load (1 minute): 44/49/51/51 = window now closed

Max fan speed = 1240 RPM


<-------------------->


Overclocking log, Iteration Set 3
comments: Increase CPU Voltage to 1.20v

FSB = 383
CPU Voltage = 1.20 (CPUZ and HWMonitor still only see 1.18)
NB Voltage = AUTO
SB Voltage = AUTO
DRAM Voltage = 2.00
FSB Term (VTT) = AUTO

Errors = Worker #2 after 3 minutes

Temps:
idle: 40/39/38/40 = +2'C over stock
load (4 minutes): 56/55/50/56 = +7'C over stock

Max fan speed = 1430 RPM


<-------------------->


Overclocking log, Iteration Set 4
comments: Lowered FSB to 377 (3.2 GHz)


FSB = 377
CPU Voltage = 1.20 (CPUZ and HWMonitor still only see 1.18)
NB Voltage = AUTO
SB Voltage = AUTO
DRAM Voltage = 2.00
FSB Term (VTT) = AUTO

No Errors after 1 hour

Temps:
idle: 38/38/38/37 = +1'C over stock (1'C improvement)
load : 55/54/48/54 = +6'C over stock (1'C improvement)


Max fan speed = 1406 RPM


Suggestions for NB, SB, and FSB Term (VTT) voltages?
Will try and buy 2 fans tomorrow because that is already close to my target 60'C.






<-------------------->


Overclocking log, Iteration Set 5
comments: Added 2 S-Flex 1600RPM case fans and lowered GPU to lower PCI-E slot


FSB = 377
CPU Voltage = 1.20 (CPUZ and HWMonitor still only see 1.18)
NB Voltage = AUTO
SB Voltage = AUTO
DRAM Voltage = 2.00
FSB Term (VTT) = AUTO

No Errors after 1 hour

Temps:
idle: 38/37/38/39 = +1'C over stock (no change)
load : 51/51/45/51 = +2.5'C over stock (3.5'C improvement)

Max fan speed = 1430 RPM


Added 2 intake fans for $36, 1 front, 1 side. Load temps dropped by 3.5'C. Though these tests were taken later in the evening. I wish I had a digital thermometer to note the ambient temperature each time I benchmark, (though the idle temps are the same) to know weather the money was worth it or not.


<-------------------->


Overclocking log, Iteration Set 6
comments: Increase FSB to 400 (3.4GHz) would not pass Prime95 until CPU vcore, VTT and NB were increased.


FSB = 400
CPU Voltage = 1.24
FSB Term (VTT) = 1.30
NB Voltage = 1.26
SB Voltage = AUTO
DRAM Voltage = 2.00


No Errors after 3 hours 16 minutes

Temps:
idle: 40/38/38/40 = +1'C over stock (no change)
load : 57/56/50/56 = +8'C over stock (+5.5'C increase)

Max fan speed = 1534 RPM


PCMark05 overall: 9160 (+799 over score at stock)
CPU: 10954
Memory: 7232
Graphics: 13047
HDD: 4949

couldn't complete the CPU portion of 3DMark06.


<-------------------->


Overclocking log, Iteration Set 7
comments: Increase FSB to 424 (3.6GHz) increase vcore, VTT, CPU PLL, NB, SB


FSB Frequency: 424
CPU Voltage: 1.30v
CPU GTL Voltage Reference: AUTO
CPU PLL Voltage: 1.50v
FSB Termination Voltage (VTT): 1.36v
DRAM Voltage: 2.10v
NB Voltage: 1.36v
SB Voltage: Auto


No Errors so far after 1 hour

Temps:
idle: 40/38/38/40 = +1'C over stock (no change)
load : 62/60/58/62 = +13.5'C over stock (another +5.5'C increase)

With RealTemp (supposed to be more accurate)
idle: 37/37/33/37
load : 55/54/50/55

Max fan speed = 1534 RPM

PCMark Score: 9468 (+1107 over stock)
CPU: 11596 (+642 over 3.4)
Memory: 7396 (+164 over 3.4)
Graphics: 13151 (+104 over 3.4)
HDD: 4800 (-149 over 3.4) ???

Prime 95 benchmark: 1024k FFT = 15.926ms (-4.0ms over stock)
I know should have let the test run longer.

3DMArk Score: 15960
SM2.0 Score: 5820
SM3.0 Score: 7212
CPU Score: 5727


Reached my 3.6GHz goal (still need long term testing). Now I want to investigate as to whether these temps are in fact abnormally high and figure out why that is. Perhaps I'll order a bolt-thru kit.

And then I may try to up my ram to 1100MHz. I figure 100MHz over rated shouldn't be too bad unless research or someone tells me otherwise.

 

[BTRY B 529th FA BN]

Raise the northbridge volts

 

[KingstonU]

Yes but to what? 1.2v? 1.25v? I have no idea what is min and what is max safe with regards to NB, SB and VTT.

 

[BTRY B 529th FA BN]

try 1.3 then 1.4

 

[KingstonU]

Ok up to 3.4GHz, but it would not pass Prime95 until I upped the CPU vcore, VTT and NB were increased to as follows.

FSB = 400
CPU Voltage = 1.24
FSB Term (VTT) = 1.30
NB Voltage = 1.26
SB Voltage = AUTO
DRAM Voltage = 2.00


I lowered my temperatures by reseating the CPU cooler properly, lowering my GPU to the bottom PCI-E slot (better air flow) and adding 2 S-Flex 1600RPM case fans. All of these gave me a combined ~7'C improvement. Though the fan I installed in the front has an annoying whine that is bothersome.

At 3.4GHz I am now reaching the high 50's. I hope I can reach my goal of 3.6 without needing more vcore. Will attempt after the weekend.

 

[BTRY B 529th FA BN]

Enable Load Line Calibration (LLC), set NB to 1.3, set memory voltage to specified voltage... that is a C1 stepping, NOT E0???

 

[KingstonU]

Yes E0s were sadly not around yet when I bought it. Will try to go further after the weekend. Thanks for the advice!

 

[KingstonU]

At 3.4 here are my scores in PCMark05, they seem to be low and I couldn't complete the CPU portion of 3DMark06.

PCMark overall: 9160 (+799 over score at stock)
CPU: 10954
Memory: 7232
Graphics: 13047
HDD: 4949

My temp seems quite high considering I have the Vendetta 2 and Freeze TIM on there. (58'C load) Where as a friend just bought the Vendetta 2 for his Q9400 @ 2.8 and his temps don't go above 45'C load in CoreTemp and 40'C in HWMonitor.


Going to try for 3.6 shortly I hope and then maybe try to get my ram higher.


Thanks

 

[BTRY B 529th FA BN]

I was just reading about VTT and i've seen reports from respectable answers that 1.15 is the max VTT for 45nm from what Intel says and the other thing I read is that increasing the VTT correspondingly increases the temp, that may may not be a reason for the temps being that high. Then again, are you testing with the side cover off? For me on two of my cases lowers temps worth leaving the covers off.

 

[KingstonU]

My case is fully assembled and closed now.

If I Enable Load Line Calibration (LLC) what should I set it as? Also what about SB, CPU PLL, CPU GTL, NB GTL?

I read everwhere that 1.4 VTT is max for 45nm quads. I will try 1.3 as you suggested.

 

[KingstonU]

Wow found a great site from a guy who got his C1 to 4GHz. Q9550 C1 at 4GHz

Though I have no idea why my temps are so much higher than his, higher than most people in fact.

 

[BTRY B 529th FA BN]

Wait, I did not suggest 1.3 I said 1.15v is the max VTT -

LLC can either be enabled, auto, or disabled as far as i know. When I was overclocking my 8400 and made it to 4.050Ghz using an 8gb 2*(2x2gb) config I had to raise my NB to 1.5 to stabilize it on an Asus P5Q Premium P45 - the funny thing is i failed at the same vcore until i only raised my NB volt, but at that volt on the NB I suggest using some type of active cooling like a fan blowing directly on it. EDIT: temps ~ 42-46c on the NB will cause failure

 

[KingstonU]

Currently at 3.6 with no errors in Prime95 for 1 hour. Load Temps = 62/60/58/62

FSB Frequency: 424
CPU Voltage: 1.30v
CPU GTL Voltage Reference: AUTO
CPU PLL Voltage: 1.50v
FSB Termination Voltage: 1.36v
DRAM Voltage: 2.10v
NB Voltage: 1.36v
SB Voltage: Auto


I think you had previously said 1.4v for max VTT, and every other source I've read says 1.4v. Could you give me the link to where you get 1.15v from?

How do I know what my NB temperature is? Does HWMonitor show it somewhere?

Wish I could figure out why my temps seem to be higher than most people. I guess I will try ordering a bolt through kit to see if it is my push pins that are still not in properly.

 

[BTRY B 529th FA BN]

I suggested to a response about the NB v that I was referring to in the first place..

Link XS 1.15v was specified by Intel, read most of those responses on what VTT is and you'll come across it. I for one would not go over specified voltage range. The SYStin in HWmonitor is your NB temps from what I've heard.

 

[yh125d]

Originally posted by: KingstonU
Currently at 3.6 with no errors in Prime95 for 1 hour. Load Temps = 62/60/58/62

FSB Frequency: 424
CPU Voltage: 1.30v
CPU GTL Voltage Reference: AUTO
CPU PLL Voltage: 1.50v
FSB Termination Voltage: 1.36v
DRAM Voltage: 2.10v
NB Voltage: 1.36v
SB Voltage: Auto


I think you had previously said 1.4v for max VTT, and every other source I've read says 1.4v. Could you give me the link to where you get 1.15v from?

How do I know what my NB temperature is? Does HWMonitor show it somewhere?

Wish I could figure out why my temps seem to be higher than most people. I guess I will try ordering a bolt through kit to see if it is my push pins that are still not in properly.

Those temps aren't bad at all for a quad at 3.6. Also, your motherboard doesn't have a NB temp sensor

 

[vj8usa]

Originally posted by: KingstonU
And then I may try to up my ram to 1100MHz. I figure 100MHz over rated shouldn't be too bad unless research or someone tells me otherwise.

I wouldn't recommend that; there's no noticeable performance benefit to running the RAM faster than the FSB with Core2s. You might as well run the RAM at stock and not risk instability from OCing it.

 

[KingstonU]

Before getting to the numbers I would just like to say that overall the system feels much snappier and responsive, XP seems to but up quicker and applications seem to load faster, which is what ultimately counts.

Thanks to AT OC guide and to everyone for their tips! :beer::beer::beer:

Final settings and results at 3.6 GHz (until I decide I want to try and push it further).

FSB Frequency: 424
CPU Voltage: 1.30v
CPU GTL Voltage Reference: AUTO
CPU PLL Voltage: 1.50v
FSB Termination Voltage (VTT): 1.36v
DRAM Voltage: 2.10v
NB Voltage: 1.36v
SB Voltage: Auto


Temps:
idle: 40/38/38/40 = +1'C over stock
load : 62/60/58/62 = +13.5'C over stock


PCMark Score: 9468 (+1107 over stock)
CPU Score: 11596 (+642 over 3.4)
Memory Score: 7396 (+164 over 3.4)
Graphics Score: 13151 (+104 over 3.4)
HDD Score: 4800 (-149 over 3.4) ???

Prime 95 benchmark: 1024k FFT = 15.926ms (-4.0ms over stock)
I know should have let the test run longer.

3DMArk Score: 15960 (+1620) all over stock
SM2.0 Score: 5820 (+295)
SM3.0 Score: 7212 (+784)
CPU Score: 5727 (+1057)

 

[n7]

Nice job

What are you using to measure temps?
RealTemp is the closest to accurate for 45nm; CoreTemp for 65nm (5C too high for 45nm).

Your temps aren't bad IMO.

You can likely reach at least 8.5x450 (3.825 GHz) with a bit more voltage, though you will likely have to work with GTLs to get past the 440-450 range.

Obviously, results will vary for different people, but i'll post the voltages that i ran for 3.95 GHz on my P5Q-D
I tried many other peoples' without success, likely partially because i run four DIMMs not two.

These may be helpful if you decide to go higher, or they may not be.

8.5x465 (3.95 GHz w/ 8 GB P95 Blend/IBT max RAM stable) settings

CPU Voltage: 1.4
CPU GTL Voltage Reference(0/2): 0.68x
CPU GTL Voltage Reference(1/3): 0.68x
CPU PLL Voltage: 1.58
FSB Termination Voltage: 1.34
DRAM Voltage: 1.96 (P5Qs overvolt by 0.08v, so apply lower than desired vDIMM accordingly)
NB Voltage: 1.36
NB GTL Reference: 0.67x
SB Voltage: Auto
PCIE SATA Voltage: Auto

Load-Line Calibration: Enabled
CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled
CPU Clock Skew: 100
NB Clock Skew: 200
CPU Margin Enhancement: Optimized

 

[KingstonU]

Thanks n7.

My P5Q Pro doesn't have the options of CPU GTL Voltage Reference(0/2) and (1/3): just one for both I guess. And I think I only have the 4 options: AUTO, 0.63x, 0.65x and 0.67x.

And I don't think I have a NB GTL Reference at all.

I had forgotten that the dram voltage will be higher than what is actually input so I have to change that ASAP.


I am trying RealTemp right now and look at that, exactly 5'C lower on all cores! My max load temps are now: 55/54/50/55. This has been the best way to lower my temperatures yet! :laugh:

 

[KingstonU]

Having a hard time getting 3.7GHz stable. Upped my voltages on everything: 1.335 vcore, 0.67x CPU GTL, 1.52 CPU PLL, 1.36 VTT and 1.36 NB.

3.65 GHZ seems t be fine but going the extra little bit has been challenging and I am hesitant to push the voltages further. Perhaps I'll just be happy with 3.6 GHz.

With regards to my ram, which are rated at 1000MHz, my choices are either 860MHz or 1032MHz. Is it fine to put them at 1032 and have no damage or should I stick to 860?

I also reduced my ram voltage from 2.1 to 2.02 after you said that P5Qs overvolt them by 0.08v.

 

[Tempered81]

471 FSB should be your goal! Good luck!

 

[Duvie]

That 1.15v max VTT seems odd since I believe my board at stock is higher then that....

I want to see some proof about that being max....every oc I have seen that is decent and that thing is 1.3-1.4v....I needed 1.35v to get my E8400 to 4ghz

 

[Duvie]

I was going to only 600mhz until I read the stepping was only c1...seems about right then....you could get 3.7ghz stable with a bit more tweaks but that extra performance versus the stress on the system is not worth it...

 

[n7]

It's not 1.15 max, it's ~1.45v max IIRC.
1.2 VTT is default 65nm, 1.1 45nm.
Obviously the lower you can keep it, the better.

 

[soccerballtux]

Temps are fine you can go to 3.825Ghz. Do it!