Overclocking Q9450

[IanR]

My new build PC is currently overclocked to 3.61 MHz and it's stable. The following are the results from severals test utils:
BIOS Vcore 1.275v
FSB 450 MHz
Core Temp 45°C
CPU Clock 3600.0 MHz (8.0 x 450.0 MHz)
Memory Bus 750.0 MHz
Rated FSB (CPU-Z) 1800.0 MHz
Current sensor 0 (CPU) 19.54 Amps

Have I gone far enough or can anyone suggest any further enhancements?

PC Spec:
Case - Antec P182 Advanced Super Midi
Asustek P5E3 Premium/WiFi-AP @n X48
Intel Core2 Quad Q9450 2.66GHz
Zerotherm Nirvana NV-120 Heatsink/Fan
4Gb OCZ DDR3 PC3-12800 Platinum
Sapphire ATI Radeon HD4870 512Mb DDR5
Vista Home Premium x64

Thanks, Ian

 

[cmdrdredd]

Pretty much what I have with my setup. 3.61Ghz (425x8.5).

I need 1.3v for stable operation because my board has bad Vdroop and it's somewhere around 1.23v actual when at load.

Download realtemp and monitor temps when 4 cores are at full load.

 

[IanR]

Glad to see I'm on the right lines. Don't want to push things too far.
I've got RealTemp. Will run Prime95 tonight to stress test & check.

 

[cmdrdredd]

Also, keep cpu-z open and monitor your vcore. You'd be surprised how it drops when at load.

 

[BTRY B 529th FA BN]

Even though the volts are within specs, I just wouldn't push it. Its just good practice to keep it in the safe zone. Maybe when you are building a new machine and really want to push the cpu to see what it could do i'd try for further but you have a good OC out of that chip. Don't fry it.

 

[IanR]

I ran Prime95 and RealTemp simultaneously and the temp on 2 cores maxed out at 60°C. I gather this would be the absolute maximum you'd want to see. Trouble is the PC just locked up after about 5 hours, so it's difficult to tell if the problem was CPU or RAM. The RAM was set in the BIOS to 1350 MHz at a voltage well within spec., so I can only assume that I need to back off on the CPU a little.
This overclocking is a bit more complex that I'd thought. I've spent about 3 days at it now, but with so many possible BIOS combinations and testing taking so long, it's virtually impossible to check all of them.
Guess I'll give it another go at least.

 

[dust]

What values do you have for NB, SB, CPU PLL, FSB v TERM voltages?

I tried myself before 456 FSB and I had to bump the vcore to 1,4375 and that makes an oven from the comp and it is outside the safe limit. Right now I have 1,38v for the 440 FSB and it's nearly the same under load in Cpu-z, LLC being enabled. If I just set all to auto I can bump the FSB quite a bit and the vcore will be around 1,35 but the NB is set to 1,6v.

 

[dust]

Originally posted by: cmdrdredd
Pretty much what I have with my setup. 3.61Ghz (425x8.5).

I need 1.3v for stable operation because my board has bad Vdroop and it's somewhere around 1.23v actual when at load.

Download realtemp and monitor temps when 4 cores are at full load.

I noticed that Core Temp is usually indicating higher values than Real Temp, in some instances about 10c. I don't know if the tjmax is an issue but I think it is 100c in Core Temp, it may be different in the Real Temp, I haven't checked. Anyway I'm tempted to rely on the higher values/Core Temp since it keeps me from frying my chip. Also there was a topic over here somewhere about the updated reviews issued by Intel regarding the voltage safe limits and they have changed the 1,36v previously admitted as max safe value, into 1,45v as I recall,but I wouldn't recommend it even on water cooled systems.

 

[cmdrdredd]

Originally posted by: Drivenbyvoltage

Originally posted by: cmdrdredd
Pretty much what I have with my setup. 3.61Ghz (425x8.5).

I need 1.3v for stable operation because my board has bad Vdroop and it's somewhere around 1.23v actual when at load.

Download realtemp and monitor temps when 4 cores are at full load.

I noticed that Core Temp is usually indicating higher values than Real Temp, in some instances about 10c. I don't know if the tjmax is an issue but I think it is 100c in Core Temp, it may be different in the Real Temp, I haven't checked. Anyway I'm tempted to rely on the higher values/Core Temp since it keeps me from frying my chip. Also there was a topic over here somewhere about the updated reviews issued by Intel regarding the voltage safe limits and they have changed the 1,36v previously admitted as max safe value, into 1,45v as I recall,but I wouldn't recommend it even on water cooled systems.

It's not the accurate number though. Tjmax is 95c not 100c. I did a ton of readong about this so I could find an accurate temp monitor.

Reason is Intel specified that the tj max for a QX9650 is 95c and a QX977x was 85c. Ask yourself how that can be remotely correct. Then a Q9000 and Q8000 series was 100c max. How? Shouldn't the Extreme edition chips have a higher max by default? Then they changed the numbers yet again after they offered this presentation. It took 2 years for them to release that and they then changed it after they showed it publicly. Odd huh?

Now Intel is hiding the vid for the new i7 chips and people cannot find it.

Anyway, I normalize the tj max value to 95c on a .45nm quad and go. If I stay 30c or more away from the proposed 95c as the max limit, then I'm in the clear as far as temps go.

the temps reported as tj max by Intel, which aren't updated, are here http://www.xtremesystems.org/f...759&stc=1&d=1227810111

Now if you plug in that supposed 80c max on a B3 Q6600 you get idle temps BELOW ambient room temp which is just impossible on air. So they are incorrect and nobody really knows the correct values.

 

[IanR]

I've backed off to 1.3v at 435 MHz. Core temps under load seem to be about 57°C and voltage is stable. However, the PC still tends to just lockup running Prime95 after a couple of hours.

Out of interest BIOS settings are as follows which give 3.49 MHz:
Ai Clock Tuner: Manual (naturally)
FSB Strap to NB: 400 MHZ
FSB Frequency: 435 MHz
PCIE Frequency: 100
DRAM Frequency: DDR3-1740
DRAM Static Read Control: Disabled
DRAM Dynamic Write Control: Disabled
Ai Clock Twister: Stronger
CPUv: 1.3
LLC: Performance
CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled
Everything else is basically set to AUTO.

I don't know what effect setting other values might have, like NB, SB, PLL & FSB Term, and wouldn't really know where to start. I read the excellent overclocking article on this site, but I don't know enough to do a thorough job. My forte has always been software. Hardware at this level is a bit beyond me and I don't want to fry anything.

If anyone has the same chip (Q9450) with equivalent hardware, perhaps they can help.

 

[dust]

I use the Fsb strap at 333, the llc -enabled, the Ai clock-lighter( enhances stability/ could be the reason for your sys to lock up during prime, if you wish just put it normal and run prime again to see if there's a difference), The Ai clock tuner is @ level 13 for me( lower level might generate higher performance but you should double check that), all the spread spectrum crap is disabled as well.

To find the voltage used by various parts just go to the hardware monitor in Bios( under Power section) and scroll to the bottom of the page. You should see all the parts I've mentioned earlier and their respective voltage values. Since you probably have them set to auto you may notice how some values change as you look at them. It is generally admitted here that setting the voltage on auto is not a good practice, since most of the time the mb over estimates the voltage requirements.

You should ask more opinions of course( I'm not experienced at this, just started myself also), but probably the best way to oc is to start at stock clocks for everything and try to put all the voltages on fixed values, the lower the better. You have to do a lot of prime tests/ OCCT tests and once you get your rig stable at stock speeds you can build an oc. You'll probably manage to reach 20%oc with the stock voltage but it's not a rule as not all the chips/boards oc the same way.

Also you should probably do some research on the safe voltage limits for various parts and it's probably better to discuss with many people here since everyone wants different things from their rigs and are willing to accept different safety levels.

 

[dust]

Originally posted by: cmdrdredd

Originally posted by: Drivenbyvoltage

Originally posted by: cmdrdredd
Pretty much what I have with my setup. 3.61Ghz (425x8.5).

I need 1.3v for stable operation because my board has bad Vdroop and it's somewhere around 1.23v actual when at load.

Download realtemp and monitor temps when 4 cores are at full load.

I noticed that Core Temp is usually indicating higher values than Real Temp, in some instances about 10c. I don't know if the tjmax is an issue but I think it is 100c in Core Temp, it may be different in the Real Temp, I haven't checked. Anyway I'm tempted to rely on the higher values/Core Temp since it keeps me from frying my chip. Also there was a topic over here somewhere about the updated reviews issued by Intel regarding the voltage safe limits and they have changed the 1,36v previously admitted as max safe value, into 1,45v as I recall,but I wouldn't recommend it even on water cooled systems.

It's not the accurate number though. Tjmax is 95c not 100c. I did a ton of readong about this so I could find an accurate temp monitor.

Reason is Intel specified that the tj max for a QX9650 is 95c and a QX977x was 85c. Ask yourself how that can be remotely correct. Then a Q9000 and Q8000 series was 100c max. How? Shouldn't the Extreme edition chips have a higher max by default? Then they changed the numbers yet again after they offered this presentation. It took 2 years for them to release that and they then changed it after they showed it publicly. Odd huh?

Now Intel is hiding the vid for the new i7 chips and people cannot find it.

Anyway, I normalize the tj max value to 95c on a .45nm quad and go. If I stay 30c or more away from the proposed 95c as the max limit, then I'm in the clear as far as temps go.

the temps reported as tj max by Intel, which aren't updated, are here http://www.xtremesystems.org/f...759&stc=1&d=1227810111

Now if you plug in that supposed 80c max on a B3 Q6600 you get idle temps BELOW ambient room temp which is just impossible on air. So they are incorrect and nobody really knows the correct values.

You're right of course, but ask yourself if it would be a good marketing practice to highlight the limits of everything is currently being sold( not only computers, but also cars, clothes, food and whatever else has got as lifetime). To give you an example, I bought two years back two TVs of the same brand and one quit on me just a few weeks back, the second one started having funny colors on one side. The point is I have to buy a newer one or maybe even two of them. I'm generally outside the house during the day since 8 am till 8pm and I maybe using one TV for a max of 4 hrs each day, more during the week-end. If you multiply this by two years it would probably give you a disappointing result, especially since older electronics equipment lasted three to five times longer. But hey they need to make sales to survive.

 

[soccerballtux]

Hm, what brand? We need to stay away

 

[dust]

Panasonic

 

[IanR]

Hi Drivenbyvoltage
I'm still having a bit of trouble getting a stable overclock. Your spec looks very similar to mine. Do you think you could let me know your BIOS settings, voltages etc?
Thanks

 

[Idontcare]

Originally posted by: cmdrdredd
Now if you plug in that supposed 80c max on a B3 Q6600 you get idle temps BELOW ambient room temp which is just impossible on air. So they are incorrect and nobody really knows the correct values.

That's an incorrect conclusion based on the data you are presented. Tjmax for B3 Q6600 is correct, Intel gets to define it as such. The problem is that DTS is not calibrated to have a linear response with decreasing temperatures.

So when the DTS is experiencing an 80C environment it correctly outputs that the temp is 80C, but when it is experiencing an environment <80C it can over-report or under-report the actual CPU temp (this is called calibration error). From the sounds of it the B3 Q6600 is under-reporting temperatures up until it gets to TJmax of 80C.

 

[dust]

You could try these, but I believe you'll probably need to alter them:

speed 3,52 mhz( 8x440)(fsb strap set to 333mhz)
cpu v 1.3875
fsb v term 1.3
nb 1.39
sb 1.125
dram 2.1
cpu pll 1.54
sb 1,5 1.6
dram contr ref 0.944( just put it on auto)
3.3 3.216
5 4.992
12 11.928
llc enabled
cpu gtl 0.67x
nb gtl 0.67x
ai tuner man/13
ai clock lighter
dram static disabled
whtever spectrum disabled

You have ddr3, I'm sure you'll have different values though. My ram doesn't clock very well, I'll get some 1066 and try again, if I'm getting different results I'll post again but it may take a while.

 

[IanR]

Hi Drivenbyvoltage
Thanks for the information and the trouble you've taken to provide it.
I'll do some more tests based on this and see what I can get.