Don't want to get this thread side-track from its intended purpose but since we are on the subject of downdraft coolers, the Reeven Brontes is one that I used recently in an HTPC build.
It's good to about 110 watt TDP. Best thing going for it is high quality construction, low profile design and quiet operation.
https://www.newegg.com/Product/Product.aspx?Item=1DR-003S-00006
That is an attractive price compared to the noctua or bequiet stuff. Plus the noctua does not seem to support 775
Back to my C2D-OC progression:
I switched for one of the E0 E8500s. Not lapped, still standard. Had to reinstall windows a 4th time because during stability testing with this one it got messed up again. I´m burning through clean windows installs so fast that I checked my SDD just to make sure it is not the cause.
This new-old one performs very interesting. It produces much less heat and clocks much worse. Best I could do for a cinebench run was 4.34 GHz:
After this it demands tons of voltage and can´t be convinced to run a single cinebench pass without throwing an error. It does not bluescreen or lock-up, just after about 35% CB crashes. No slight bump in Vcore nor VTT could help me. Relaxing memory timings was not helping either and GTL-offsets did nothing at all. Neither positive nor negative ones. NB voltage one more step did not help either and even trying a bump PLL after all this as a last resort did nothing. It crashed when reaching ~ 60°C.
EDIT: A comparison for the heat output on the same voltage
So it seems that ~4.5 GHz is a hard wall for my E8500s so far. This one had a very high VID (1.3V) too, don´t know how much that is telling about overclocking potential.
Lets discuss voltages for a moment, these are my settings to get 4.3GHz stable on this chip:
1.5V CPU (reading 1.56 in software)
1.5V PLL (reading 1.52 in software)
1.34V VTT (any lower and I can´t get it to boot beyond 400FSB, while below 400 even 1.18 is stable! Trying something as low as 1.2 with 400 FSB results in the mainboard bricking completly, with only CMOS reset as the last resort) (1.36 in software)
LLC is enabled and working, none or very little droop.
1.66V RAM (1.71 in software)
1.6V NB (1.63 in software)
1.5V SB (1.5 in software)
Now I don´t know how much my boards sensors are off during load, have to measure that now which is a bit tricky. All points I need to access are on the back of the board. What makes me a bit uneasy is that I found a german forum thread for the Striker II Extreme with a chart that shows the actual voltage should be pretty close to what the software readings are. Could this be right? Did I really need 1.56V for my 4.3 GHz run and how on earth does that only produce 60°C on a mid-range air-cooler?
I do notice that other people seem to be able to run their 45nm C2Ds at lower voltages. For 4.4GHz I see values typically ranging between 1.35V and 1.47V. Same goes for VTT, many seem to stay below 1.3V. I can´t even post at that level.
On the mentioned thread I found one guy testing multiple E8500s back in 2009 and he only had to use 1.43V for 4.4GHz as an average with disabled LLC. When I disable LLC I have trouble to maintain 4GHz. Only big difference I see is that he uses 1.9V on the RAM, which I´m a bit hesitant to try on my 1.6V rated sticks.
Could the voltage regulation and power delivery of a mainboard degrade over time? Maybe I should check the caps on the board.
Speaking of this damn thing, I have to ask if this is common behavior for the platform or if my board is just haunted:
If you change bios settings it has a 50% chance to get stuck during post, you have to reset it and apply the settings again. If you re-apply them the second time it always works. In 20% of the cases it can not do a clean restart from OS. Instead it gets stuck somewhere during post. This happens on rock solid stock/auto settings too and does not seem related to memory OC issues.
Changing settings is like walking a minefield. One single step from 1600 rated FSB to 1605, memory on the same divider results in the board becoming a brick needing a full CMOS reset. Oh but 1610 is fine again on the same voltages.
When finding an unstable setting and trying to get it more stable by reverting the step, e.g. memory timing change causing crashes and setting it back to the old value, trips it up even more resulting in a system that suddenly no longer boots with values that ran benchmarks not even a minute ago just fine.
Maybe it is just me and my inexperience with overclocking, but some of this is made me question my sanity in between the 4th windows install.
It might be time to let go of this board and look into other options.