Phenom II X4 945 OC limit

I cant get this damn thing stable over 3.5Ghz
Specs:
HSF Zalman 9900 yes properly seated with a thin layer of AS 5. Idles 27-30C
Case Antec 900 with so much airflow it could power a jet
MB: Gigabyte GA UD2H
Current stable settings:
FSB- 234= 3.5Ghz
HT 1920
NB 1920
Mem: 1600 @ 1599 auto timings (It's 1600 Patriot memory)

The comp will boot at 3.75 and can pass the Sandra Burn in for 2 hrs HOWEVER when I run Crysis, I eventually get a Crysis.EXE error. Sometimes in Explorer I get a Explorer.exe error

Obviosly the system is unstable over 3.5. I'm frustrated because I just dropped $70 on the Zalman 9900 and it made cooler temps but no increased OC.

Am I missing something or is the limit 3.5 and that's it?

Pls opine.

What kind of voltage are you pushing through the chip? Also, I'm sure you've done this already, but have you checked your RAM volts and timings? And Sandra's Burn in isn't the most stressful application and 2 hours isn't long enough to determine stability. I'd recommend you use either Prime95 or LinX. Also, what settings were you using to get to 3.75Ghz? What are you temps under load? The Phenom II's start hitting a heat wall around 55C. So if you're close to that, there isn't much you can really do, as they do not like anything really higher that 55-60C Max. If your temps are within range, I'd try starting with your 3.75Ghz settings and try tweaking your cpu, ram, and HT volts to see if you can get it stable. Your random crashes sound like it is succumbing to excessive heat.

Good questions

CPU and RAM are overvolted slightly. NB and RAM are running at spec speeds and timings. Sandra says the CPU load temp is 47C.

3 mins on Crysis @ 3.6 gives me a crash with all fans locked on full. 3 hrs on Crysis at 3.5 has no problem even using Cool and Quiet.

I'll try more volts and see what happens

NB is stock? I suggest bumping that then a touch.. The NB-CPU voltage could determine a stable system.. I don't know the specifics so I won't go into it (cos I can't), but I know the NB sometimes has to be upped itself to make the system more stable as the CPU communicates with the RAM.. Etc something along those lines.. Im not entirely sure to be perfectly honest, but I'd def say it your NB voltage.

fsb = 234 would suggest that your ram is not at specced? Cos to be at spec with its default multiplyer, the fsb would have to be 200 right? So it's running higher too, and thus you may need to drop an extra .01 or 0.02 volts into the ram.. Or perhaps loosen one of the main timings.. Play around ^^.

With the FSB at 234 the RAM runs at 1599 using a MULTIPLIER that makes it run at only 90% speed. It's 1600 ram.

As I type this the comp is at 3.75 with the increased volts Dave suggested. CPU is at 1.4v. NB and RAM multipliers are LOWERED to keep them below spec. I'll also try incr the NB volts even though the NB is not OCed.

Looks like 3.5 Ghz is the max. Overvolting and other things including a much better HS made no diff. I sugest to others that if your CPU can OC well you will see it with the stock sink. A better aid HSF will just increase stability and perhaps extend CPU life.

You have a good OCing CPU. AMD has always overclocked well. Athlon XP @ 2500Mhz lol, the good ole days,,, x800 etc lol

Samoyed said:

With the FSB at 234 the RAM runs at 1599 using a MULTIPLIER that makes it run at only 90% speed. It's 1600 ram.

As I type this the comp is at 3.75 with the increased volts Dave suggested. CPU is at 1.4v. NB and RAM multipliers are LOWERED to keep them below spec. I'll also try incr the NB volts even though the NB is not OCed.

Click to expand...

Your math for RAM speed is correct for 3510 (15x 234) but I believe if you check RAM speeds at 3750 MHz (15x 250) you'll find that your RAM is now running at a brisk 1666, well above it's rated speed and much faster then the 945 IMC can use - especially if your cpuNB is running at near stock speeds. For OC'ing at 3.6+ you need to have the cpuNB running faster than stock as well.


> CPU voltage of 1.40v should be OK to 3.6 GHz but may need 1.45v for 3.7 GHz or more.
> Change the RAM speed to 1066 instead of 1333 but leave the RAM timings and voltage at the 1600 settings.
> Set the cpuNB multiplier to 9X. You may need to increase the cpuNB voltage (often labeled CPU NB VID or something similar - NOT the same as NB or NB chipset) to 1.225-1.25v. Sometimes even more than that is needed but start there. You can increase it up to 1.30v if needed.
> You're pushing the system clock quite a bit. You should increase the NB chipset voltage to at least 1.2v (or 1.25v if you have that option).
> Set the HT Link to 6-7X.

With those settings and starting out at 234 MHz slowly increase your clock 4-5 MHz at a time. Use Prime95 or OCCT for stressing and do this for 5 minutes or so before increasing the clock again. Keep an eye on load temps (while running Prime95 or OCCT) using CoreTemp. If you go above 55°C you'll start having problems.
http://www.mersenne.org/freesoft/
http://www.ocbase.com/perestroika_en/index.php
http://www.alcpu.com/CoreTemp/


Good luck and post back with results ...

Good suggestions. I have tried a few and tried yours with the volting and multipliers. Even with the ram running slower and the overvolting, the system won't play Crysis over 3.525Ghz.

I guess that's it with this MB and CPU. It's still a decent OC. I don't think heat is an issue. Tonight's testing was done with 62F room air and as I mentioned, a case with the 200mm fan on top and multiple fans.

Sandra reports idle temps of 27C and load temps of about 45C.

I guess I'll have to let the Artic Silver "set" ROFL (urban legend)

Samoyed said:

Sandra reports idle temps of 27C and load temps of about 45C.

Click to expand...

45°C is fine for load temps - how are your GPU temps?

Samoyed said:

I guess I'll have to let the Artic Silver "set" ROFL (urban legend)

Click to expand...

My rigs run 24/7, so there's no thermal cycling to speed up the process, and I water-cool, which reduces temps and further slows the curing time of AS5. Nonetheless my load temps drop ~3°C after a couple of weeks in a constant 70°F room. It's not a myth ...

I never noticed any difference after it set or cured. In fact all of the pastes "seemed" to work about the same but I used the AS5 just in case. I might add that the OC for this CPU was the same using the stock heatsink and that crap they put on it as it was with the 9900 Zalman and the AS5.

GPU temps are very good in the new case. As you know the 8800GTX is hot as HELL. What I do is lock the fan at 100% before I play. Doing that it gets up to about 65C max. Before I locked the fan and was in my old case it could get as high as 80C. It's only OCed to 625mhz. That's all it could do in either case but at least now I have cools temps.

In several OCing situations over the yrs it has always seemed that the silicon had a limit and better cooling only allowed a slightly better OC but with greater stability. Another recent machine I built has a 3.0Ghz Phen 2 dual core running at 3.6 with stock sink(small Aluminum with no pipes) and the gray crap that came with the HSF. Go figure.

The best OC I ever had was a Pent 2-350. It ran at 467mhz with a cheap alum sink and fan for 4 yrs before it died a noble death. Ah I feel nostalgic now. Here are my CPUs are the stable OCs I had with them:
XT- no OC
286 12 oc to 16 mhz!
386-33 no OC
486-33 no OC
486-66 used overdrive CPUto 100mhz
Pent 100 no OC
Pent 166 no OC
Pent 233 sock 7 @266 Soyo TX board
Pent 2 350 @ 467 Giga BX
Pent 866 @ 904 Soyo Via Apollo Pro
XP 1800 @ 180 fsb H2O
XP 2600@ 175 fsb H2O
XP 3000 @ 170fsb H2O
A64 3700 2.2 @ 2.5Ghz Asus
A64 4400 2.2 @ 2.5 Ghz DFI Lan party
Phen II X4 3.0 @ 3.5 Ghz Giga

There were probably other one night stands I had with a few P4s and various computers but I can't remember their faces.

I've never been too sure about the OC being better but temps go down and that's always good.

Most of my rigs are s939 and Phenom Is with a couple of others thrown in for flavor. Best so far on s939 was 2.94 GHz on an Opty 165 - hard to beat a 60+% OC ...

Phenom II 945 95watt c2

For some reason, im having the same problem getting this chip stable at 3.6ghz. Memtest was done for 2.5 hours without errors prior to Prime95. I then run Prime95 for 2 hours, and passes no probs. So I go a step further and try to run for 12 hours, but only after 3 hours into Prime95 blend test, it BSODs and reboots. No errors logged though. So im thinking voltage. Yes? No?

Here's the settings and screenshots for the 3.6ghz oc.

CPU: 945 Phenom II X4 Deneb 95 watts C2
FREQUENCY: 3600 MHZ
REF*MULTI: 240*15
CPU VOLTAGE: 1.46250
CPU-NB: 2160 (9x multiplier)
CPU-NB VOLTAGE: 1.25
CHIPSET VOLTAGE: 1.22
HT LINK FREQUENCY: 1800 MHZ
HT LINK VOLTAGE: 1.215
RAM: OCZ Reapers 4GB DDR2 PC-6400 downclocked to 333mhz / 5-5-5-15-24 @ 800mhz
RAM VOLTAGE: 1.905
TEMPERATURE: 32C IDLE (with cool and quiet disabled) - 54C (100% load 3 hours Prime95)
MOTHERBOARD: ASUS M4A785-M
COOLING: Zalman 9500A 92mm
OS: Windows 7 Professional x86




So ive backed it down to 3.525ghz, which is rock solid stable through 12 hours Prime95 blend test. Ill post the settings for this stable OC later.

Peace....

Prime 95 can be fickle...I've had a machine that wouldn't make it 5 minutes on Prime 95, but would do 12 hours of OCCT...and run BFBC2 for days on end.

That said, blue screens are usually Ram or CPU/NB related. I would bump the CPU/NB voltage and try again. Well, actually I'd call it good, but if your after a 12 hour Prime run then I would try that.

to OCF!

piranhamoose said:

For some reason, im having the same problem getting this chip stable at 3.6ghz. Memtest was done for 2.5 hours without errors prior to Prime95. I then run Prime95 for 2 hours, and passes no probs. So I go a step further and try to run for 12 hours, but only after 3 hours into Prime95 blend test, it BSODs and reboots. No errors logged though. So im thinking voltage. Yes? No?

Click to expand...

Or heat. Core temps above 50°C or so can cause problems. The rough range for problems is 52-58°C, so we usually say 55°C but it can run higher or lower depending on the CPU. It looks like you're running a 54°C load core temp and that's most likely the cause of the instability. Your Zalman 9500 was never meant for cooling highly overclocked quad-core CPUs ...

Phenom II 945 95watts C2

I dont think its heat as you can see from my screenshot of 3.525ghz OC after 12 hours Prime95, temps hit 55C also. Plus, according to specs this 95watt c2 chip has a max temp of 71C (10C higher than it's 125 watt brother).

I tried your suggestions combined with many other's suggestions to no avail to try and get this chip stable at 3.6ghz. No go. 3 hours Prime95 longest run til bsod. I think 3.525ghz is the chip's limit with my setup as well. 12 hours Prime95 makes me feel comfortable running my rig 24/7 @ 3.525ghz.

Any other suggestions from the OC guru's out there?

Like I posted (recently about your set-up), until you get your load core temps down you will likely not be able to go any farther ...

Those screenshot pics are tiny! Do you expect anyone to be able to read the info they present?

trents said:

Those screenshot pics are tiny! Do you expect anyone to be able to read the info they present?

Click to expand...

you must be retarded.... let me explain so you understand. those are thumbnails which you click to make them larger.

piranhamoose said:

you must be retarded.... let me explain so you understand. those are thumbnails which you click to make them larger.

Click to expand...

^^^ This (insults) is not tolerated at OCF. Better check yourself before the Green Monsters send you on a little vacation.

Not being rude, but that's not how we roll around here.

trents said:

Those screenshot pics are tiny! Do you expect anyone to be able to read the info they present?

Click to expand...

He is right, just click on them, lol.