Well, it has been a long time since I posted about my OC on this chip, and have run into a new ceiling.
I recently installed a new HSF (Scythe Samurai Zz) that has dramatically lowered my operating temps on the CPU. Prior to putting this HSF in, I was limited in my OC efforts to the restrictive stock HSF, and thus, could only OC to a temp threshold I was comfortable with, 55c. This limit in place, I ended up settling on a daily 3.25ghz clock, as it was P95 stable with a max temp of 54c. I was able to run up to 3.4 fairly well, but wasn't comfortable with the 58-60c P95 peak temps. I was able to bench @ 3.6ghz, but without even a hint of stability beyond a CPUID validation and a single 3Dmark06 run.
Well, as I already said above (and on numerous other threads since I got it, heh) the new HSF has dropped my temps significantly and has thus allowed me to OC the chip further without worry of hitting thermal thresholds. but I have run into another wall, and am looking for any insight or hints that you all might have to offer.
I have run this setup p95 stable (8hrs blend). I am not dissappointed with this as my games and programs are running awesome, but was hoping to go higher. My current BIOS settings:
x7 HT Link - (1.4ghz)
4.00 memory @ 1092mhz, 2.1v
CPU Volts + .175 for 1.5v
24/49c idle/load temps
Since installing the new HSF, I have been unable to replicate 3.6ghz (@277fsb), even if only to post. This baffles me, as I had played around with those settings as short as a month before with enough success to run 3dmark06.
My system BSoD memory dumps at 274mhz FSB about 10seconds into doing anything at all. At 273, everything is tight. I figured that this was pointing to either my memory not getting enough volts, or just being clocked too high. Since I am already running the max warrantied volts for my ram (2.1v) I tried instead to drop the multiplier to 3.33 keeping the rest of my stable settings intact. It woouldn't post. I tried loosening up the timings and lowering the voltages, but to no avail, my computer refuses to post @ 3.33 memory multiplier. If this is as high as I am going to see out of this non-BE chip, I am fine with that, but if there is anything I can do to squeeze a bit more out of it, I might be willing to give it a shot.
Might upgrading to 1200mhz rated memory open up some more headroom?
Also, I have read that 64bit OS often run lower stable OC than 32bit OS across the same hardware. If this is true, then I have to assume that upgrading to Windows 7 Ultimate 64 won't change my situation, correct me if I am wrong, please. I understand that XP64 is rather dated and lacking current support, so I would understand if that is contributing to my limits, though I haven't come across anything to confirm or deny that theory.
Thanks everyone for helping expand my understanding of overclocking, and re-igniting my interest in this dark art. I wouldn't have gotten this far without all your support (and I try to repay in kind) and hope someone has an answer to my questions, good or bad, heh.
you're not going to see any differences in 3.55 vs 3.6Ghz. If you want something faster, you might consider one of the new 965 125W cpus. the 720 IS a budget cpu. I bought one and am happy with 3.4Ghz. If I want something faster, I'll need to move to a "real" quad core instead of the disabled "three core" defect budget cpu. As long as your pc plays the games you have well, I wouldn't worry about the extra 50mhz
I am not unhappy with 3.55ghz. I also understand that there won't be a real difference between that and 3.6ghz. I have greatly enjoyed the experience I have had with overclocking this PC, and as with most hobbies I take up, tend to push it as far as I can go. Also, as a learning experience and practice in overcoming certain obstacles I am ignorant about, this sort of approach is pretty essential.
I am planning a new build for January, and want to have some fun with this one, see what I can extract from it. You can be damn sure, that even though my next build will in stock form blow this one away, I will still pursue overclocking on it as well, heh.
By saying that it would be more cost effective to buy a 720be, I was highlighting the pointlessness of purchasing DDR2 1200, not an intention of buying a new, slightly superior CPU.
Also, the triple core "defect" budget CPU may indeed be budget, but they are not all defects. Would it make sense to sit on perfectly usable unsold quads, when they can repackaged into the higher volume market where they wll actually sell? You can be sure that there are more good quads produced than there is demand for quadcores, as there hasn't been a shortage on the quad PhIIs since launch, yet many times I have seen the triples on back order in that same time frame. While i still would never purchase one of these chips depending on it unlocking, it is equally foolish to write them all off as defective, as has been proven by many successful unlock and overclock testimonials.
I purchased my 710 before I even knew unlocking was an option (and they were out of 720s on that day) so I was lucky it is from the high-unlock-rate 904 batch, and have even had it unlocked in another Mobo, with some short term, but completely error free P95 testing. I built a rig for a friend at about the same time, and used the same mobo/cpu as I have now. He upgraded to a 965, gave me his 710, and low and behold, it is also batch 904. Even if only one of them is a stable unlock, I will be using the other in my HTPC build. So as you can see, I have absolutely no need for a new CPU in the near future, heh. And if it performs well enough unlocked, I won't need a CPU for the next build, either.
don't get me wrong, like I said, I own a 720BE b/c it is more of a budget cpu. They are "defect cpus" not because they are bad, but as you said one core was bad. This comes from statistical testing. If you get a batch of say 1000 and you test 20 of them and you get a certain percentage of defective 4th cores (maybe 1 or 2 out of those 20... whatever their requirements are). They will deem this entire batch "defective 4th core" and disable the defective core on all cpus. I'm sure there is a cost effective equation saying it's more expensive to test ALL cpus than it is to test some and disable if some CPUs from the batch are defective. So, not all trip core cpus are bad, just enough to be considered statistically all bad.