Q9550 Overclock to 400 FSB

Feb 23, 2018
5
0
10
Hi all,im new to overclocking world,but i would like to learn.I read so many tutorials,but im guessing im doing something wrong.
I own Q9550 with default clock and GA-P35T-DS4 motherboard with 4x2Gb DDR2 Ram,but when i try to change FSB,it wont let me start my pc..Do i need to change more settings?
I know every motherboard has to have different settings,but i really appreciate some help..Sorry for stupid questions,but like i said,im lost.
Thanks for your time and reply.
 
Solution
That is a remarkably low max temperature, so I would just leave things alone if you are happy with the speed. If not, then further tinkering with voltages and perhaps ~417 x 8.5 = 3.54GHz would be reasonable too.

After you are pretty certain things are dialed in where you want, run the Prime95 overnight to make sure all the RAM is OK. After all, the IntelBurnTest runs have so far only been testing the 1st 1GB because we've really only been using it to check the CPU. While it could be used to test the rest of the RAM in a hurry too, I'm a bit leery of hammering a 45nm CPU for that long because they sort of have a reputation for electromigration and degrading overclocks over time so it just seems like needless wear & tear. 65nm...
In general you will be most likely limited by your RAM--you'll need at least DDR2-800/PC2-6400 stable RAM because except for laptop chipsets, Intel didn't let you run DDR2 memory slower than FSB. Quads also generally do not work at as high FSB as duals, but 400 should be easily doable without having to tinker with VTT voltages.

Set the FSB at 400, Memory divider at the lowest 2.00 (for DDR2-800), vCore at +0.05v from whatever your VID is, MCH at +0.1v and DDR2 at +0.2v with everything else on Auto. That should post and work fine. If you go into CPU-z and find your RAM's SPD is rated for example 400MHz 5-5-5-18 at 1.8v then Turbo tRd 8 is about the best you can expect. Or if it's rated 400MHz 4-4-4-12 at 2.1v then you could set that and Turbo tRd 7 for lower latency, if you raise DDR2 voltage to +0.3v. If it's only DDR2-667 rated for 333MHz then you could attempt 400MHz at slow timings and +0.4v but no guarantees.

You shouldn't need much extra CPU voltage for just 3400MHz. Extra MCH voltage is useful when all RAM slots are filled but usually +0.2v is only needed for very high FSB. DDR2 often doesn't live long at very high voltages but locally 2GB 1066 sticks are only $7 here so if they die, I'll replace them with better ones.
 
Feb 23, 2018
5
0
10
Very thanks for your fast reply,i really aprecciate your time,but i must assume that im newbie in this and im surely missing or doing something wrong..
I post here some pictures of settings in BIOS and my CPU-Z details for more info,because i dont really know,if im doing this correctly,the system still wont boot,so i assume,you or other clever people here will see,where i made my mistakes..
Very thanks to anyone for their or yours help.

CPU-Z
2zevzv7.png

20fxr4i.png


BIOS
219v1hx.jpg

1rtd21.jpg
 
Hmm, looks good except you have neglected to supply the CPU any extra voltage. That's a fairly low VID chip so you have plenty of safe headroom. Try 1.275 to 1.300v to start, or if you are tired of clearing CMOS and manually resetting things, just go for the maximum common VID sold (for 45nm) of 1.325v and go down later if it's stable.
 
Feb 23, 2018
5
0
10
Hey,very thanks for your reply,i really appreciate it.
I set the voltage to the max like you said thats 1.325 and system boots fine and then i slowly dropped down to 1.2 and my temperatures is around 40C in safe state.
Should i run some kind of benchmark program or ensure its safe to use it on this voltage?
Thanks again for all your help.
 
I like to run at least 5 passes of IntelBurnTest standard to verify the CPU is stable and its thermal solution is adequate under any circumstance. It's quick so you could use this to find a lowest stable voltage, then bump it up 1 or 2 notches. If you are happy with the temperatures though, 3.4GHz is faster than any stock C2Q ever made so 1.325v isn't really high. The published VID table did go to 1.3625V so Intel would've undoubtedly gone there if AMD had gave them any competition. Max safe voltage spec was listed at 1.45v (as in no damage but normal operation not guaranteed).

And finally I like to run at least 12 hours of Prime95 blend to verify the RAM is stable. It's not so quick but after all that you can be sure it is as stable as any unoverclocked OEM machine.
 
Feb 23, 2018
5
0
10
Okay,so i runned Intel Burn Test,and it was finished in 0,11 seconds,and says thats success,am i doing something wrong or is it so quick?So you suggesting to increase voltage a bit if its stable?
I could run some other tests,if you guide me some,because i think in BurnTest,CPU dont get any kind of stress,or like i said im doing something wrong.
Thanks for reply mate.
 
Sounds like you tested 1 pass at 5MB.

IntelBurnTest at 1024MB (which is "standard") should take about 25 seconds per pass on your machine. And the object is to lower the voltage until it is unstable and fails the test--that's how you know where the limit is. If you just wanted to increase the voltage past where it's stable you would not need to run any test at all because the test couldn't tell you anything anyway.
 
Feb 23, 2018
5
0
10
Okay,now i tested it correctly with 5 passes,here are results..
2m5nhp0.png

In 2-5 test i have max 65C,is it okay to use or should i run something else?
That Prime 95 for blend to be sure?
Thanks so much i really appreciate your help,you surely know what your talking about.
 
That is a remarkably low max temperature, so I would just leave things alone if you are happy with the speed. If not, then further tinkering with voltages and perhaps ~417 x 8.5 = 3.54GHz would be reasonable too.

After you are pretty certain things are dialed in where you want, run the Prime95 overnight to make sure all the RAM is OK. After all, the IntelBurnTest runs have so far only been testing the 1st 1GB because we've really only been using it to check the CPU. While it could be used to test the rest of the RAM in a hurry too, I'm a bit leery of hammering a 45nm CPU for that long because they sort of have a reputation for electromigration and degrading overclocks over time so it just seems like needless wear & tear. 65nm quads are nearly indestructible but you'd instead have a difficult time getting rid of their heat!

Perhaps this is too cautious as it's only a little hotter than Prime95, and people run Prime95 for years at a time looking for new prime numbers with no apparent ill effects. So it really is a real-world workload of the scientific/cryptographic kind, while Linpack is instead a worst-case kind of synthetic test which no real-world workload can approach.
 
Solution