Abit IP35 + Q9550 - how to get above 400 FSB?

kasakka

2[H]4U
Joined
Aug 25, 2008
Messages
3,385
I just got a Q9550 and have had trouble overclocking it above 400 FSB. I had the exact same issue with my previous CPU, the E6400. It went to 3.2 GHz (8x400) but not above that. On the Q9550 I can still run at ~415 FSB but it doesn't seem to be stable. 420 FSB and above won't work at all.

How high should I raise the voltages? At the moment they are:

CPU 1.35V (CPU-Z reports 1.32)
VTT 1.2375V or something like that
MCH 1.30V
ICH 1.55V
Memory 2.0V
GTL Ref 63%

The machine:
Abit IP35 with beta bios for 45nm processor support
Q9550 E0
4x1GB G.Skill DDR2-800 5-5-5-15
8800GT (GTX260 soon after I get my case modded so it'll fit... :rolleyes:)
2xHDD

I don't know if my Antec NeoHe 430W power supply has something to do with it. It doesn't have the EPS12V connector so per my motherboard user manual I use the ATX12V 4-pin connector instead. BIOS reports the following voltages: +12V = 11.89, +3.3 = 3.2, +5 = 4.01.

It's definitely NOT heat related. My system (CPU + GPU) is watercooled and the temps are around 35-45 degrees for the CPU (idle-load).

Not sure if it's memory related. 2x1GB G.Skill or 2x1GB A-Data DDR2-800 didn't improve overclocking at all over 4x1GB G.Skill.
 
By looking at those voltages readings, you're definitely pushing that poor PSU. I wouldnt try overclocking until you get a QUALITY 500+ watt.
 
By looking at those voltages readings, you're definitely pushing that poor PSU. I wouldnt try overclocking until you get a QUALITY 500+ watt.

Whaa!? At least 1200W...

Never trust BIOS readings of voltages. If you really want to know how your PSU is doing get a multimeter...
As for the OC, try upping MCH volts a bit.
 
Got it up to 430 FSB now. Seems stable. 440 will get to Windows and you can even start a game but it will crash eventually. For 430 FSB the memory voltage needed to be upped to 2.20V, raised MCH volts a few notches too and raised GTL Ref to 67% (although I still don't quite understand what it does but it seemed more stable on 67% than 63 or 65).

It seems I may have to improve the cooling of the NB and PWM MOSFETs because they seem to get really hot. PWM is reported as 80°C on load. Possibly watercooling the NB and then some sort of big heatsink for the PWM would be in order?
 
Assuming IP35's got the same PCB layout as IP35-e you'll have hard time finding a heatsink for PWM... I've checked ThermalRight, Zalman and pretty much any company that's advertising on the net and none seems to be making it :/ The best [and only] solution I can think of is finding someone who'd make you a custom waterblock for PWMs... I'm not saying it'd be easy but you don't have much choice if you want to keep this OC running or keep pushing it.

[this is all assuming you've already reseated the stock heatsing replacing whatever goo Abit's used instead if real TIM and installed a fan blowing directly to it]

In case you actually find any method of cooling PWMs that's better than stock - let me know.
 
You should at least have a 1600W PSU if your going to be pushing that chip.

Way overkill. My X3360 (xeon version of the q9550) @ 8.5x400 only uses 212 W full load. I have an 8800GTS fully loaded as well that contributed to that figure, as well as 2x HDDs, 5x120mm fans. That 212 W figure was measured, not estimated. He definitely does NOT need that much in a PS.

It's a scam perpetrated by the companies marketing these things. Even w/ two vid cards I doubt you'd need over 700 W PS.
 
Way overkill. My X3360 (xeon version of the q9550) @ 8.5x400 only uses 212 W full load. I have an 8800GTS fully loaded as well that contributed to that figure, as well as 2x HDDs, 5x120mm fans. That 212 W figure was measured, not estimated. He definitely does NOT need that much in a PS.

It's a scam perpetrated by the companies marketing these things. Even w/ two vid cards I doubt you'd need over 700 W PS.

There you go:

Wiki's definition of SARCASM
 
I would turn the Southbridge back to stock, you should never have to overvolt it.
 
I would turn the Southbridge back to stock, you should never have to overvolt it.

I'll do that. I just added some voltage to everything just in case.

Now I'm back to 400 FSB though, it seems that even 440 will run but will eventually crash and 430 wasn't stable in the long run.

I did the "bolt mod" for the northbridge and PWM heatsinks. With this you replace the stock push pins with screws so the heatsinks are held better on the parts. With the stock pins some of the PWM MOSFETs weren't even touching the heatsink. Also replaced the goop on the NB with Arctic Silver2. Kept the grease pad on the PWM because I was worried that any goop I have would be slightly electricity conductive and since the PWM is multiple components I just stuck the heatsink on better. Improved PWM temps about 5-10 degrees.

I don't think I'll bother spending the money on watercooling the north bridge and PWM because even at 3.4 GHz everything that requires lots of processing power runs without a hitch.
 
Bolt mod thread @ XS. Well, at Google cache, I think XS is upgrading their servers or something, I can't get it to load.

But to be honest, I've got the board, and long ago got the parts at Lowe's, but still haven't done the mod. 80c at load is nothing to worry about - I've seen 90c+ on a regular basis, and saw 100+ in the summer months at times, and no stability problems at all.

 
and saw 100+ in the summer months at times

You're system would have been rebooting when it saw those kind of temps... 105c or 100c is when the proc will just force the system to shut-down.

Honestly I wouldn't trust anything over 80c unless it was like during an intel burn test in a room with high ambient temps...

Personally mid 70's is my limit for under 100% load, 40's in idle roughly...
 
zzz he's talking about pwm temps. 100'c is completely fine.


and on both my dfi p965-s and abit ip35-e, i've needed to add onto vsb to get stability at higher fsb's.
 
Back
Top