Started by disabling C1E, EIST, DOT, Spread Spectrum and all those features that could cause troubles,
Set memory timings by hand (5-5-5-18). Set RAM voltage to 1.8V (normal voltage and timings, according to manufacturer,)
Set the FSB to a value over 333MHz,
Set the ram divider so that the memories don't go over 800MHz,
Setting the MCH, ICH and VTT FSB voltages to "normal" values (just to get them off the automatic option),
and finally, set the CPU voltage (PCI speed comes auto-locked to 100MHz, in theory.)
I can manage to get stable at 375 FSB with a multiplier of 9 (3375MHz), even at 1.1475V, which is less than stock voltage. So far so good.
When I try 385 FSB, everything goes to hell. Sometimes it reboots before booting (rare but possible), and every time I run Prime95, it always fails withing 3 minutes.
I played with voltages a lot, giving everything a little boost, first at different times, and then all at the same time, and raising the the CPU voltage to 1.2475 (and also values in between.) I don't wanna go higher than that on the processor voltage because I belive its temprature readings are not reliable (I think the digital thermometer is broken) and also because I raised a whole milivolt (edit: there should say 100mV there ) just for 90MHz sake (10FSB x9), which I think is too much and still doesn't work.
I also tried 385 FSB x 8.5 multiplier (lowering it from 9), giving as a result 3272MHz, which is less than the 3375MHz I manage to get stable, but the instability persisted.
The memories at 375FSB are at 750MHz, and at 385 FSB, 770MHz, which is always lower than 800MHz.
I don't know what else to do. Does anybody have any other idea?
Also, there is one thing I could use some explaning of, which I didn't find anywhere else:
On the ram frequency dividers option, I have several dividers to choose from, however there are a few extra numbers apart from the divider value. For example:
1:1,00 1600:800 (same divider, diferent weird number, and the BIOS projects the same memory frequency)
and so on.
What is the meaning of those numbers?
What CPU cooler do you have? My old E8400 was the easiest overclock ever. I also had Corsair DDR2 800 ( same memory as you the DHX, but mine was factory overclocked at 2.1v cas 4-4-4-12 ) but I had a Gigabyte P45 board and a ZeroTherm Nirvana NV 120 CPU cooler.
Anyway all I needed to do was leave everything at stock as far as C1E, EIST and Spread Spectrum goes. Left CPU voltage at auto. My first overclock was to 3.6Ghz so I just set the FSB to 400. That allows for a 1:1 RAM divider so your RAM runs at the rated 800Mhz. Worked great for almost a year. Downclocked it's self at idle and everything.
My second overclock was to 4.050Ghz. Once again I left all the throttling settings alone and just raised the FSB to 450. I left the RAM at 1:1 and this overclocked the RAM to 900Mhz so I loosened the timings to 5-5-5-15. I sold this system to a friend who still uses it everyday. I used it at this speed for a year as well.
4.050Ghz with the RAM at 900Mhz cas 5-5-5-15. Still downclocks it's self at idle and with the ( old ) ZeroTherm Nirvana NV120 idles at 28-30C and hits the mid 60C stressing with Prime 95. Still going strong after 3 years of being overclocked.
I'm on stock cooler. If I rase FSB to 400, it won't even POST (in fact, what you describe is pretty much what I tried to do after trying FSB 375).
I also tried downclocking the mems a bit and nothing. Memories "shouldn't" be a problem anyway since they are working at most at 385MHz (at FSB 385). And then again the processor shoulnd't be a problem either with FSB 385 and multiplier x8.5 since it's less frequency than 375x9. I don't know what else to try.
Right now I have a 120mm cooler sitting on top of the stock heatsink with the stock cooler. I reseated the heatsink _three_ times in a row, always using new TIM (and the last time, AS5), and always getting the same crappy readings. It just can't be a temperature problem.
I've read people getting to 4GHz on stock cooling (on 90°C on full load, but that's another story). There has to be something else. I can run the processor at a "high" voltage for a given stable frequency so it generates heat and it won't crash.
You know you can go to 1.4v on these chips right? That will make your temps go sky high though so that's why you need aftermarket cooling. I really don't remember what my vcore was at 4Ghz but it was pretty low. I lucked out and got a "good" chip.
I think that the sensor is broken because before I started OCing (that is, with everything at stock settings) I ran Prime95 and it heated to over 80°C. That's when I started fiddling with the cooler.
I reseated it three times in a row with different grease each time, and each time I got almost the same readings.
The only thing left to discard was a faulty fan, which I did by placing the 120mm fan blowing on the stock heatshink and fan arrangement. The temperature drop was 1-2°C which is almost negligible for a processor running almost 90°C. Right now, at full load and OCed it gets to 80°C (which is less than it would get at stock settings since it's under-voltaged.)
In any case, if it dies (or finishes dying) then it would not be a big loss, as I'm getting a new computer soon.
EDIT: saw the new post:
Yes, I know that they can go to 1.4V, however it seems very silly getting almost 3.4GHz with less than stock voltages, and needing more than 100mV (which still won't do it) to get 90MHz more.
EDIT 2: Furthermore, as I said, if I lower the multiplier and maintain the "high" 385 FSB, therefore not demanding as much of the processor, it still won't work. It would appear to point to something else the FSB might affect.
So you have a jury rigged stock cooler? A 120mm fan blowing from somewhere above the stock heatsink onto it? That could be your temp problem right there. Even with the small stock fan it is positioned so that it is pushing air through the heatsink at a high pressure. Just backing a fan off a few inches will majorly reduce the effectivness.
Stock coolers are just that....for stock speeds. They are not designed for any heat but what your processor puts out at stock speed. Artic Silver 5 can help but it also has a 200 hour cure time. That's why I don't use it. I use IC Diamond because it is just as good as AC5 but needs 10min to 2 hours to reach maximum effectivness.
If in fact I read it wrong and you do have a fan pushing high pressure air through your heatsink and not just blowing on it, and a 120mm fan even making direct contact with the stock heatsink will probably not be very effective due to the large center area of the fan moving no air....the fan would literally be too big for the heatsink...it would only be blowing on the outside edges of the heatsink then correct me.
The E8400 should be a very cool running chip even with stock cooling if that cooling is functioning correctly. Case airflow matters as well. A well ventilated case like my Antec 1200 with 7 x 120mm and 1 x 200mm case fans ( I also have a 120mm fan on my Scythe Mugen 2 CPU cooler ) will have temps 5 to 10 C cooler than say a case with a single 120mm exhaust fan and thats it.
The 120mm cooler is a consequence, not a cause. When the processor started overheating with P95 on stock settings, then I unmounted the stock heatsink and reseated it with new grease.
I did that a couple of times just to be sure that I didn't screw up the seating of the heatsink or the spread of the termal compound, and when that didn't work, then I put the 120mm fan on top.
The stock heatsink and fan are still in place working normally. The 120mm fan just blows a little more air into the heatsink and moves a lot of air to keep the surroundings cool. The idea was to discard a stock fan malfunction.
And, I didn't just put the 120mm fan on top of the stock one, I moved it a little to the side to maximize the air flow into the heatsink, and taking advantage of that, I can also keep cool the northbridge (I think) and the memories (completely unnecesary ). I just tried the same trick with a 80mm fan and the results were roughly the same.
The stock fan works fine, and the heatsink is seated properly. I can't think of any other test I could run to ensure that the problem isn't the processor's thermometer, so I'm blaming it.
In any case, I understand the risks about the temperature. How do I make it go to 3.6GHz?
I am sure you have done this but make sure the heatsink is not full of dust. If so blow it out with a can of compressed air.
My advice is to either clear CMOS or at least load failsafe defaults in BIOS.
Clearing CMOS would probably be better. This will reset any voltage changes you have made.
Next make sure you have the latest BIOS version for your motherboard.
Next with everything at default set the FSB to 400 and the RAM ratio to 1:1
Make sure the RAM is set to 1.8v and the timings are 5-5-5-18. This should be default.
You should be able to run 3.6Ghz with the stock cooler. If this does not work I would say probably your motherboard is not up to the task. MSI only recently became a brand to recommend. Back in the LGA 775 days it was more of a budget brand. It could also be your power supply. I really can't find anything on that model in English but I have never heard of it.
I would leave the voltage on auto just to see if it will boot. Just check it with CPU-Z, Speccy or HW Monitor to make sure it is under 1.4. Use Coretemp and Realtemp for accurate temperature monitoring.
Well, I've made some more tests and now there is no way I can get into windows even with 385 FSB (always reboots while booting, or BSOD's, and sometimes it doesn't even POST. LOL, sometimes it even hangs while on the BIOS). Even though I've yet to try clearing the CMOS (I put everything back manually)... I doubt it'll work, but I'll keep it in mind for the future.
Anyway, I'm probably gonna wait for the new computer before trying to push this one over the limit again. Thanks for everything.
Oh well, either the motherboard or the power supply was not up to the task. Probably the motherboard. MSI is starting to gain a reputation as a solid manufacturer but back in the LGA 775 days you pretty much wanted a Gigabyte or Asus board for overclocking.