Trouble OCing QX6700

[zuffy]

Yes, I have read the Overclocking Guide but still need some help.

Most likely I am doing something wrong. I have a QX6700 in a Gigabyte GA-EP45-UD3R Ver. 1.1, using Crucial Ballistix 4 x 1GB DDR2-800 RAM. It's cooled with a Zigatek Dark Knight inside an Antec 300 case with 2 front, 1 rear and 1 top fans. Currently, I have the processor OCed to 3ghz (333 x 9), vcore @ 1.275, RAM voltage @ 2v because these are 2V DIMMs. RAM ratio are set to 2.4B, basically identical to what the offical RAM speed is rated for. Core Temp is showing 21c - 24c for the cores in idle. Real Temp shows a few C higher. Under load with Prime95, it's in the 40c-45c range under Core Temp. This tells me I have more room to OC but I don't.

I've tried OCing to 3.33ghz (333 x 10) with vcore 1.35. Bluescreen with Prime95. Also, I've tried 3.3ghz (300 x 11), 3.2ghz (350 x 9), 3.2ghz (266 x 12). All these bluescreen with Prime95.

What am I doing wrong? Any suggestion? I'm not trying to OC to the highest this processor can go but I know 3ghz is not the limit.

 

[Lazlo Panaflex]

What power supply do you have? Best I can do so far is 3.0 on my current setup (300x10, 1.28v, Earthwatts 430). These SOB's run hot & need a lot of juice.

Edit - you may want to run memtest on the ram, seems odd that you're getting BSOD'S.

 

[zuffy]

Power supply is a Corsair CMPSU-450VX. I have an 8800 GTS 640MB, VelociRaptor 300GB RAID 0, 750GB 7200rpm drive and a DVD-RW drive.

I can try the memtest but strange that it run super stable @ 3ghz (333 x 9) though.

 

[Beanie46]

I'd try the Memtest only because Crucial Ballistix hasn't got exactly the best reputation for durability or longevity......I quit using that stuff after having to RMA two sets. Went back to Corsair and Mushkin and never looked back nor regretted it at all.

 

[Wuzup101]

Have you tried setting your multiplier down to 8x and then just increase FSB (and core voltage / nb voltage) in small steps? I would be really suprised if you couldn't get a QX6700 to hit 3.2 (8x400) if there was no outside factors preventing it (power supply inadaquacy / bad ram or something like that). I really would assume that the psu you have would be good for the bit of extra power needed... but you might be on the boarder.

Honestly, you probably should be able to drop that chip in the mobo that you have and set it to 8x400 at default voltages (or damn close). My q6600 will do 3.2 at just a hair above default (and I have a chip with a crappy VID).

 

[Lazlo Panaflex]

That Corsair's a decent PS. Good idle/load temps, too...mine runs ~60c load on a Freezer 7 Pro; probably not the best cooler for Quads. Ah, if only these B3's were G0's (Wuzup, what stepping is your chip?). B3's ain't the greatest overclockers.

That seems like an awful lot of volts for DDR2-800, are you sure it's supposed to run at 2v? Mine is rated @ 1.8

I also tried 400x8, no dice. Board did a hard reset back to default settings. Which is annoying, considering it's an "extreme" chip...but it's an older board. oh well. Give the memtest a shot & let us know...

Edit - btw, are you running "small FFT'S" in Prime 95? This is the best setting to torture test the processor. You want to avoid "blend", since it involves the memory.

 

[Schmide]

Originally posted by: zuffy
RAM ratio are set to 2.4B, basically identical to what the offical RAM speed is rated for.

Set your ram to 1:1. (i.e. ddr2 533)

 

[Idontcare]

I owned a B3 QX6700.

What you need to do is eliminate some variables that might be limiting your CPU overclock, trying to overclock everything under the sun and then figure out why your cpu tops out is going to waste a lot of your time.

Drop your FSB to stock, drop your ram speed to 1:1 (synch). B3's QX6700s are a total PITA when it comes to overclocking thru the FSB, make that unlocked multiplier work for you.

Next thing is to make sure your core temp is using the right TJmax for your chip.

See slide 15 of this official Intel document regarding DTS and TJmax. TJmax is 90C but coretemp likes to default to 100C I've noticed.

Next thing to do is see where your chip can go in terms of clockspeed by upping the multiplier and then find the min voltage needed at that multiplier to be prime95 stable (or OCCT).

Something like this will help you see where your CPU is capable of playing at without concerns of ram or FSB limitations. (that data there is for my QX6700, albeit at -40C on vaporphase...I'm surprised you can sustain 3GHz with a mere 1.275V at 40-50C...very surprised)

Once you find out the limits of your chip then you can play the game of maximizing your FSB and ramspeed while keeping your CPU below its clockspeed limits and thus have confidence in knowing if your rig is unstable as you change FSB and ram clocks that it isn't an unstable CPU to blame (or vice versa).

 

[Lazlo Panaflex]

Well said, Mr. Idontcare :beer: btw, I got "page not found" after clicking your photobucket link.

FYI, I'm using Real Temp 3.00, and TJmax is correct (90c).

 

[Idontcare]

whoops, typo in the photobucket link, fixed now, thanks!

 

[Lazlo Panaflex]

Welcome

 

[zuffy]

Thanks for help folks. Please excuse my overclocking newbieness.

Lazlo Panaflex, my original Ballistix RAM was rated 2.2v. I had it RMA back to Crucial because of memtest failure in all 4 sticks. In the email, Crucial specifically said the new pairs are rated @ 2v and that I should set the motherboard to 2v for the RAM.

OK, I tried 3.2ghz (400 x 8) with vcore 1.35. My Core Temp is 48c - 52c with Prime95, running the In-place large FFTs test. No crashes so far after 15 minutes.

Question though. Is it "safe" to run the bus speed @ 400mhz and FSB @ 1600mhz? Wouldn't it be safer with a lower bus speed and a higher multipler? Before, I tried 266 x 12 with vcore @ auto or 1.35v, it would crash after only running Prime95 for less than 30 seconds.

Regarding my memory multipler, this is the description for the memory multipler option:
Set memory freq by DRAM SPD data
(G)MCH Strapping
x.xxA --> FSB 266
x.xxB --> FSB 333
x.xxC --> FSB 200
x.xxD --> FSB 400

These are the option I have:
Auto
2.50A
3.00A
4.00A
2.00D
2.66D
3.33D
2.00B
2.40B
3.20B
4.00B
2.66C
3.33C
4.00C

For the current 400 x 8, I selected 2.00B, that set the SPD to 800. In CPU-Z, it says the DRAM freq is 400mhz, FSBRAM is 1:1. So my question is, since my RAM are DDR2-800, shouldn't I keep the multipler SPD to 800 as close as possible?

 

[zuffy]

Just curious, why is Core Temp and Real Temp different by a few degree? Aren't both apps getting the temp reading from the core sensor?

http://mysite.verizon.net/zuffy/realtemp.jpg
http://mysite.verizon.net/zuffy/coretemp.jpg

 

[Schmide]

Ok I don't totally understand this board, but from what I can understand poking around here:

http://www.clunk.org.uk/forums...r-living-review-2.html

The 1:1 mode would be 2.0b or 2.0d.
The 2.4b option would be a 6:5 ratio.


You really won't receive any benefit for running the FSB over 1:1 so weather you run at 333 or 400 pick the 2.0x multiplier. Especially since you have 4 sticks.

 

[zuffy]

So regardless of what FSB I set it to, I should always keep my RAM to 2.0B or 2.0D?

 

[Schmide]

Yes.

 

[zuffy]

Play around a bit more. I kept the vcore @ 1.35. Tried the 333 x 10 again but this time with the memory multipler @ 1:1. It still BSOD. Next, tried 370 x 9, same result. Now, I have it at 420 x 8 @ 1.375v and seem to be pretty stable. Core Temp has it @ 46 - 53c and Real Temp has it @ 52 - 58c.

 

[Ratman6161]

Originally posted by: zuffy

Question though. Is it "safe" to run the bus speed @ 400mhz and FSB @ 1600mhz? Wouldn't it be safer with a lower bus speed and a higher multipler?

According to gigabyte your motherboard "Supports 45nm Intel® Core 2? multi-core processors with FSB 1600 MHz " so the 400 MHz setting is not really even overclocking your board.

 

[Schmide]

It could be that your memory is running at too tight a timings for 4 sticks at 333+fsb. When you put the FSB to 400, it loosens these timings and it runs fine. If you wanted to check this, you could record the timings for 400 and plug them into 333 and see if it makes a difference.