Q9400 + P5B Deluxe overclocking in progress

[pogiman]

Like the many who jumped on the P965 boat, I wanted to upgrade my old E6600 to the newer quad cores. It is pointless for me to spend extra moola on a mobo when my current mobo supports the newer yorkfields. The only advantage to a Pxx is overclocking potential. (i dont need the extra features) Well, I want to let others see that a p965 still has ample amount of juice left and there is no need to upgrade to a Pxx if your mobo BIOS supports the new 45nm quads. I just got in a Q9400 and Im still on the benches to see what this p965 can do. (BTW, I know for sure the mobo can do 500FSB since I ran that on my E6600, with quad core though, thats another story...)

My current settings

Asus p5B Deluxe Rev 1.05 BIOS ver. 1238
Q9400 @3.4GHz Vcore=1.3 (BIOS) VID=1.28 (core temp) Multi = 8
8GB DDR800 5-5-5-15 @ 2.1V
FSB @ 425
FSB Volts @ 1.4V
NB @ 1.45V\
Win 7 RTM x64

From here on, I am only messing with FSB and vcore. I upped the FSB and NB more than I need to make sure that this is not the problem if it fails any tests. So far, no tests failed......on to 3.6Ghz

I will have SS when everything is final

UPDATE:

With the above settings, I was able to do 3.6 Linpack stable. FSB at 450. Temps are 30C idle and 55C load (looks a bit high, I will reseat later) C1E and SpeedStep disabled.

UPDATE2:

So after two days of tweaking and testing with games, various apps, etc, I found that 3.6 @ 1.3V is the sweet spot for this combination. I can get 3.7 everything stable, but it requires 1.33V Thats a huge jump in vcore for just 100 mhz. I was never able to even boot into windows at 3.8, but i can get into BIOS, even at 3.9! Maybe its is the mobo, but 3.6@ 1.3V, I cant complain. I did a reseat of the CPU with an extra thin layer, almost transparent coat of AS5. Temps seem to gone down a bit, but the same problem plagues. Core 1 and 2 are cooler than 3 and 4. Under load core 2 has a lower temp. Also, during tests above 3.7, core 2 usually goes first. Any ideas?





idle temps, looks like a bad IHS solder to the cores?

Another update:
I was trying to see how far I can undervolt and underclock this and I think i found a bug. Right now, I have it at 1.8GHz at 1.12V and on idle, I get the same temps: 32, 31,35,35. Now this cant be true at all since I am running much lower volts and speed. Under load, i get temps 35,34,37,37. This looks very weird and it seems that either the ASUS board is reporting the wrong temps or its a bug in the chip? any one have similar experiences?

[unclewebb]

The P5B Deluxe screams with Dual Cores but doesn't have a great reputation with Quads. I'd suggest pulling the northbridge heatsink and make sure it sits square and makes good contact with the chip. After that, add lots of chipset voltage and crank up the MHz.

I've been thinking about trying a Q9550 on my P5B Deluxe so post some good numbers to encourage me. It would be great to see 500 MHz FSB so then I wouldn't have to buy a UD3P.

to Asus for updating the bios for the 45nm chips.

[pogiman]

My entire system is WCed so temps are not a problem. Looks pretty strong at 3.7...still going

[unclewebb]

When it hits 8x500 I might be reaching for the Visa. A 500 FSB with a Q9550 (8.5 multi) is good enough for me.

[pogiman]

So far, it looks like 463 looks like the limit, ironically, its also the FSB wall most people get with quad cores. I was able to get 498 FSB on a E6600 with GSkill 2GBHZ ram. I dont have that ram anymore, so that may be the limit. Changing vcore doesn't affect anything. Changing NB and FSB to max volts doesnt have affect. I guess I can live with 3.7 as my daily, or I might lower to 3.4. temps are the same and vcore is at 1.3V. Sadly, I know this chip has alot more, maybe even 4GHz, but that means I need to spend another 150 for a p45.

[Leeghoofd]

any screens ?

[pogiman]

Not yet, still benching. I think I am going to settle for 3.6 @ 1.3V This chip looks to be a keeper but my mobo is the weakest link. I will have some IBT and p95 screenies up soon.

[In-Fluence]

Have you tried to go further with minimum NB volts? My old P5B could manage 3.69GHz with my Q6600 with 1.25v nb (more than my curent Maximus II formula POS could do, even after a month of tweaking)!

Great clock nonetheless, shows that this board is still a great contender in terms of 24/7 Quad o/c.

[pogiman]

Yea i set the NB to 1.25V and its fine, I tried 1.45, 1.55, 1.65 and theres no change. 1.25 is pretty stable. But the FSB voltage I had to crank to 1.45. It looks like 3.6 is my final 24/7. All games stable, encoding apps stable, p95 stable, running OCCT now for an hr and no problems. Nonetheless, I am happy to be able to reuse my old board and just upgrade to a new CPU. Saved me a ton and now I have more to spend on the next gen CPU in 3-5 years ahead.

[FunkyRider]

Good on you man, I am having the same combination here: P5B Deluxe, Q9400

I think my 4 sticks of D9GMH puts too much pressure on the P965 MCH. My system is only stable at 425x8. If I set FSB to 450, it results instant freeze when the OS start to load (Win7 / Ubuntu x64)

I have heard from many places that putting too much FSB Termination Voltage would kill the 45nm Quads so I am being conservative and setting FSB Term. Volt to 1.3v

Multi: 8x
FSB: 425
CPU Volt: 1.325v
FSB Term: 1.3v
NB Volt: 1.45v
RAM: 2.1v, 4-4-4-12
Others all manually lowest

I was previously having P5K-E (P35) + Q9550 but sold it at appropriate time and recovered most of my money. The P35 is clearly having an edge when you put 4 sticks of RAM and crank up the FSB, but I am happy with my system for now. The saved money can well be used towards next gen build.

Also, when testing stability, try to run UT2004, Race Driver GRID, Left 4 Dead, preferably under x64 system. Those thing in 64-bit OS are extremely picky about stability. I could run SP2004 for whole afternoon and have instant crash in GRID with 1-2 games

[pogiman]

Glad to see others in the same boat as me. I will have to read up more about FSB voltage if it kill the quads. I have played quite a few games, including L4D and seems to run fine. Once i drop the FSB to 1.35, it will freeze right when it loads into windows. Have you tried raising your mem timings to get higher speeds? I pushed that ram beyond 1050mhz in my old system. They were crazy!

[FunkyRider]

This is the post which states on Anandtech that they managed to kill the QX9650 on a test using 1.5v VTT
:
http://www.anandtech.com/weblog/showpost.aspx?i=428

There is also a post of guide of ocing Core 2 Duo / Quad which mentioned above article:

http://forums.anandtech.com/messagev...VIEWTMP=Linear

Also, my 4 sticks of RAM used to work at 2.1v 1000 5-5-5-15 on P5K-E (P35) rock solid for almost a year, so this is definitely not the RAM's problem

[nugzo]

So far, it looks like 463 looks like the limit, ironically, its also the FSB wall most people get with quad cores.

With ASUS maybe. Grab you a Gigabyte UD3 and see 500fsb in your near future.... Great job though. I just picked up a new UD3P for myself, Was hoping for that extra 100mhz for my 9650, no go though. I used the old UD3P for a Q9400 i had laying around at the shop. I'm at 3.6 but havent pushed it. Its for resale so i dont want to be on the edge with it, must be 100% stable everywhere. So far so good. What batch is your Q9400? I think i'll see what the max is on mine either monday or tuesday. Good luck!

[scharnhorst]

I don't know If it might help you, but I tried my Q9650 with my P5B Deluxe Wifi-AP (1238) + 2*1024MB Corsair XMS2 + 2*1024MB Geil Ultra and I obtained, more or less, 450MHz for FSB while on my P5Q deluxe I reached 470MHZ for around 1.3v for both (because I don't want to increase more).
On the P5B deluxe Wifi AP I was using the following settings

Vcore : 1.3
vnb : 1.25v
And all the others to the lowest value
+Pencil Mod