If my bios warns me that max voltage for the processor is 1.55v...is that "no matter what"? or is that assuming I use stock cooling? I have a ZALMAN CNPS9900ALED cpu cooler. I am currently running 3.7 @ 1.55v stable and cool at about 40C in Bad Company 2. I just want to be able to try 3.8 or 3.9 but in order to do that, I'll need to up the voltage some. Am I capped out? Any thoughts?
1.55v is, iirc, the max recomended voltage for Phenom II chips. Admitting to running to anything above will void your warranty should the CPU fail.
Regardless of the temps, I would not run 1.55v for daily usage. Something seems out of wack, I wouldn't think a 940 needed that much volts for 3.7ghz.
Agreed. I decided to drop to 3.68 ghz @ 1.55v (was actually using 1.56 and 1.57 with poor results) but increase NB to x12 and 1.35v. It passed the medium stress test in Prime95. So, I was pretty thrilled with that.
I will try lowering to 1.54v on the core voltage and run the test again... I would prefer less voltage. Another interesting note: At higher voltages, I began to see core failures in OCCT and Sandra. So, I thought I was screwed and fried my chip. But luckily, that wasn't the case.
How much should I push the NB? I have experienced the problem of it not booting and having to reset the bios manually with a jumper on the motherboard (is there an easier way?)...
Playing around with the NB is one of the only ways I have forced the use of the clear cmos jumpers, and no there is no easier way, unless you have a motherboard with a button replacing the jumpers.
I have only been able to run 2400mhz NB, bumping the volts to 1.3v, but it wouldn't pass stability. All attempts to run 2600mhz, regardless of the voltage, required a cmos clear. My daily NB is at 2200mhz stock volts, and that passes all stability tests.
AMD chips will see synthetic benchmark improvements from NB oc, but real world advantages are debatable. on the flipside, NB oc can stabilize high cpu overclocks, as I have only been able to stabilize 4.0ghz with a 2200 NB, and any attempts to run stability tests at 2000 NB failed.
Results will vary.
I respectfully suggest you start your OC from scratch. Reset your Bios to default, then change all auto settings to manual (especially memory timings, but also including NB + HT multipliers, and voltage controls) but replicating stock settings.
Some settings, when on Auto, will vary during the operation of your PC, eliminating a stable platform from which to deduce and eliminate issues. These variations have to be eliminated for a stable OC procedure.
From this point with all stock values on manual control, I would test the memory to be sure that it isn't the weak-link. Memtest is an accepted memory testing program. If it passes, you can move on. If it fails, then you may need to lower the speed of the memory, or raise timings, or both to achieve stability. If nothing works to stabilize, your memory may be a hardware issue that needs to be resolved.
Assuming the memory passes stability testing, I would then move on to seeing if there is any OC headroom at stock voltages (not auto!) Since the 940 is a BE chip, then this can be done by bumping just the CPU multiplier, probably best in 0.5 increments. Stability test each bump. If you are going to use Prime95, use small FFTs test, as it is known to produce the most heat and find CPU errors the quickest among the P95 presets. It barely stresses the memory. Blend is the opposite, as it stresses the memory the most out of the 3 presets, and is also a decent test for memory stability, but doesn't produce as much heat on the CPU as small FFTs.
I would leave the NB alone until you hit a stability wall that even massive volts can overcome, and even then, only increase the NB by the smallest increments when in this situation.
Always work on keeping the voltages as low as possible. The speed of the CPU doesn't matter for longevity as much as the volts and heats. In fact, I would venture that a chip OC'd on stock voltages would live longer than a stock speed chip being overvolted.
If none of this helps, then you may just have a leaky chip that likes the volts and makes a lot of heat. While for daily use, OCing chips like these are limited, for extreme overclocking these types of chips are bees knees, hah.
Thanks for the thorough reply! I read 4ryan6s AMD overclocking "how-to" and it pretty much lays out everything I've learned on my own except the very first step which is what you also suggested where I increment in .5 steps and monitor the voltage being applied until I hit a wall.
The way I started isn't anything I'd recommend to anyone which was basically googling what others with the same proc were oc'ing at and starting there to see what happened. I realize every proc is unique in its tolerances, but I also feel a little lucky that I have been able to hit 3.8 (briefly) and 3.7 fairly successfully as opposed to others I have read to be limited to 3.5 or less. Bear in mind, this pc is not a 24/7 machine. It's my enthusiast, gaming rig. I am the type of person that gets more excited in the performance of a game than the game itself.
This weekend will be a perfect time for me to start over.
Thanks again...wish me luck, and I'll post the results if anyone is interested.
Good luck, it is always a joy to help out someone who has actually done a fair chunk of research as opposed to the HALP NOOB OVERCLOCKER NEEDS 1337 HELP! topics we see quite often.
I've read or tried to read many blogs like that. I feel sorry for the one who has to filter through those, but I very much appreciate that you stopped by to read my inquiry. I'll play with my bios some more tonight and let you know how it goes.
Well, here is my OC info:
3.69 ghz mult: x18 bus:205 vcore:1.55 NB: 2.05 NBv: 1.17
Mem flex enabled with mem timings of 5,5,5,15 Mushkin DDR2
CPU spread spectrum disabled. cool & quiet disabled (obviously). It's still quiet though!
idle temp: 27C. under stress: 51C.
It passed the Prime95 small FFT test and I stopped it manually shortly after it began the next phaze. I got no errors. If I bump the mult to 18.5, the machine is very unstable and crashes easily unless I increase the voltage to 1.56 or 1.57. But, I don't make it through the test before the machine shuts down. I noticed CPU-Z is showing my memory as PC2-6400 (400 mhz) at max bandwidth, but the memory is supposed to be rated at 1066 mhz, so I'm a little confused.
I still have a lot to learn about all this overclocking stuff. There are many, many options for my memory that I don't know anything about in my bios. Should I mess with the link or bridge stuff? It's all set to auto except the CAS, tRCD, tRP, and tRC.
Also, I keep reading about people running there Phenoms into the 70C and 80C temps...??? I thought that was way past critical melt-down temps. Am I still safe running past 54C? How hot should I go and is there a way to disable the auto-shutdown (or do I dare?)? My fan seems maxed out at 1900 rpm. Auto fan control is off, which, to my understanding auto-maxes the speed.
Also, I'm leaving NB relatively at spec speeds and voltage. In order for me to get to 18.5 or higher, I have to increase those, but it's extremely sensitive and if not right, it either won't boot or is highly unstable.
I'm still drooling over the 4Ghz club members...they are taunting me.
I appreciate any tips, thoughts, suggestions you offer. Thanks!
Phenom IIs memory controller can only do 1066ghz memory on 2 sticks. Otherwise you are limited to 800mhz (400 x2). If there are things you don't understand about your memory settings, do some research. Having stable memory is a must, and far more important than running it faster or with tighter (lower) timings.
Anyone who runs a Phenom II past 62c is risking meltdown. Anyone claiming to run it 70-80c has no problem replacing fried hardware, or lieing. I personally set a limit of 55c as a safety barrior, as I don't want to pay out for new hardware, or attempt a warranty claim.
From what you have described, it appears as though you have hit the limit of your processor. There may be a few mhz to be found here or there with some tweaking, but generally, I think you might be SOL for getting into the 4ghz club without extreme cooling solutions.
Ok, thanks for the info. I think I've hit my ceiling as well. It's running perfectly at the moment and whizzing through everything I throw at it and that was my goal. I will start my wish list for a future upgrade with high-end water cooling system...if the wife allows that is. Do you recommend one?
About the memory, I'm interpreting that you are saying that with more than 2 sticks, it won't go to 1066, correct? And since I do have 4 sticks of memory, 800 is the cap?
I thought I'd turn off mem flex and force it to 2.1v as the mem specs suggest since I haven't done that yet.
I also keep a record of most of my clocks.
About watercooling: There is no easy all in one awesome solution. Read the stickies at the top of the forum and do some HEAVY research on the topic. It is my ignorance (and lack of funds) that has prevented me from going the LC route as yet.
4 sticks = x4.00 memoy max. 2 sticks = x5.33 max. You cannot work around this, I am sorry to say.
I tried wc once years ago at the bottom of the barrel store...as in the cheapest I could find. Air was better. It depended on all the components being top performers, such as: a good radiator, good pump able to move lots of water through large hoses and of course, a drip-free circuit. It's a pain to setup and unless you spend hundreds...forget about it.
I have heard of a peltier setup, but since it is a refridgeration device, you must go through a painstaking process to prevent condensation in your case.
Basically, if your not a science student at a prestigious college funded by your local government, it's not worth it.
Your description about the memory didn't help me much. I have discovered that as long as I have memflex enabled, I can run the DRAMv at 2.0 and force the mhz at 533 without it griping. No problems so far. I'm running 18.5, bus 202, vcore 1.53, nb x11 @ 1.2v...which gives me 3.75 Ghz. It's running WoW with no issues, but I'm worried to run Prime95. But seriously, who uses 100% of their proc for long times anyways?
Update: 3.75 bombed in BC2 in under 10 minutes. The best performance I've seen out of my machine is at 3.66 at the moment. Think I'll keep it at that for a while.
BTW, I posted about an issue I'm having with the bios on my ASrock board and was wondering if someone could stop by chipsets and bios threads and take a look at it. I'd like to not have to hit F1 every time I turn on my machine.
P.S. When I look at my profile it's only showing me 2 of the 3 threads I've participated in...makes it difficult to track. I assume there needs to be a reply for it to show up???