Q9450 Performance Difference at Various Speeds

[themattman]

I have a Q9450 paired up with a Gigabyte P35-DS3L v2.0 board running the F8f BIOS. For one reason or another (I've spend over a month trying to find the problem), I can't go over 380FSB. With 380, I can run the board on all normal settings @ 1.275vcore. This gives me a 3.040Ghz speed, close to a 400Mhz overclock. Not bad, but I was hoping for better. To get to even 385 I need to up the voltages a ton on almost every part of the board and I fail Prime 95 5 minutes later.

My question goes out to those who can overclock a Q9450 higher. Is there a noticeable difference in performance at 3.2/3.4/3.6? In terms of 3dmarks (I know it isn't everything, but I like high numbers ), do you score much higher?

I'm hoping the performance difference is little to none so I don't make myself go out and get something like an Abit IP35 Pro which is still around $170.

 

[GuitarDaddy]

3dmark scales very well with clockspeed, so yes there is a big difference between 3.0 and 3.6. It's definately not the chip or vcore thats stopping you at 3.0, without your full settings it's hard to tell exactly what the problem is but here are a couple of guesses

1. Is your memory on the lowest divider? If so CPU-z should be showing 380mhz on the memory tab, if your showing higher than that you need to drop the mem ratio

2. Internal timings too tight. Most mobo's automatically switch from the 1066 strap to the 1333 strap for internal timings when you hit 400 or 401mhz. On most boards FSB of 380-399 is very difficult to get stable and requires a lot of NB chipset volts, once you cross 400 and hit the 1333 strap things get much easier

Here is what I would try

FSB 401
Mem ddr2-802
vcore 1.35 (If it boots and passes prime you can start working it down)
nb volts/MCH +.1
everything else on Auto

I would be shocked it it doesn't pass prime at those settings, my x3350 overclocks like a beast even on an old P965 board I've gotten it prime stable up to 3.8ghz. I run it 24/7 at 3.2ghz with the settings show above except I was able to work the vcore down to 1.28

 

[themattman]

My RAM is on a 2.0 divider. It passed 13 rounds of memtest at 5 5 5 15 at DDR2 1000, so I assume it should be good until I hit those speeds.

I will try your suggestions.

EDIT: Tried 401 FSB
+.2 FSB
+.2 MCH
1.350vcore
+.3 DDR2

Fails Prime 95

 

[GuitarDaddy]

Which prime test are you running? small,large or blend? How long before it failed?
What cpu volts does CPU-Z show while prime is running? What are your temps under load? Which HSF are you using?

And generally just uping volts across the board like that doesn't work, too much volts on a particular component can cause instability just as too little. If your ram does ddr1000 you definately don't need +.3ddr, and you shouldn't need to increase FSBv either, and probably even MCH could stay on auto at 401.

At this point vcore is all you should have to raise, and how much depends on the vdroop on the board (diff between bios vcore and cpu-z vcore under full load)

 

[themattman]

I am using a lapped Thermalright Ultra 120 Extreme with a Scythe S-Flex Fan in an Antec 900 case.
I was running blend in prime 95.
Temps under load when I was looking at core temp showed around 58-62C.

I'll re-check my vdroop, but I remember it being something like 0.04v

EDIT: Currently testing 401 @ 1.375v and +1 to FSB. I'll run it through the night to see if it is stable.

Load Temps: 66 66 61 57 (should go down overnight)
^^Core Temp's measurements. Realtemp ate 10C lower.

 

[GuitarDaddy]

Looking good, temps are nice. And the concensus is that realtemp is correct and coretemp is +10

I also have to set 1.375v in bios for 3.2ghz but my droop is horrible, that gives me about 1.28-1.31 idle and about 1.21v prime loaded per CPU-z

Thats quite a diff between core temps 66-57, not quite sure what that means? sensor off?

I'm running an unlapped TRUE 120 w/Sflex in a TT Tsunami case(good flow but not like your 900) in TX where it almost hit 100 today and prime loaded I get

62 62 60 60 per realtemp, so your temps are better than mine but you have the advantage of lapping, the excellent airflow of the 900 and I hope for your sake lower ambient temps

 

[themattman]

I decided to lap my heatsink a week ago, and I believe I did a decent job. It isn't perfect, but it should be better than the slight bulge that was in the middle of the TRUE.

I'm assuming the last two sensors are off, so I am just going really by the first two temps.

I forgot to mention that I was using the smallFFT test this time to try something new. So far I am at around 8 hours stable. If I make it to 24 I'll try the other two tests, just to be sure. I really wanted at least a 3.2Ghz OC on this chip.

Thanks for your help, I will keep updating.