ED NOTE: Cem sent me this email:
I have been lost for some time for building a new overclockers site,
with my close friend Ozgur and some other fellow Turkish overclockers. The site is almost finished but it is only in Turkish for now. In fact people have been using it a lot, especially
the downloads section, but we are a bit slow with the news section at the moment.
The site is called
It has “local Pricewatch” and “configuration suggestions” sections
besides common stuff like news, articles, reviews and downloads. Most of our downloads have a Turkish “how to use” added in – I think that’s why turkish people prefer using TurkOC*.
I don’t know if you have met any, but we Turks are strange. A lot of
people, some even without basic knowledge of hardware, are interested in overclocking. So I believe we would be of some help.
After spending some real time for TurkOC* (small detail – we always
write the site’s name that way) we eventually got back to overclocking. As anyone with common sense, we contiunued the quest through the AMD path.
We have a nice sponsor so we were able to try almost all the Socket A
boards out there. Their reviews are coming out slowly though (Turkish
“lazy” Overclockers). The best turned out to be Abit KT7 (RAID) with the AOpen AK73Pro closely following up. We also had some fun with a T-Bird 1 GHz running at 1.3 GHz, and a Duron 700 @ 1045 MHz.
The site is progressing slowly at the moment – due to huge amount of
initial work to be done – and even if we published these stuff, it will be in Turkish, so I wanted to share a few things with you that we have done lately.
Well, we believe we have made one of the best mounting clips out there. It is designed for mounting water jackets with or without peltiers on Socket A CPU’s. What’s so special about it is this is strictly another “pencil” case as it is completely home made, out of Lego Technics parts (I believe most people have some laying around or at least you can steal from a younger brother). The “How-To” tutorial will be published (with photos of building steps) at TurkOC* soon.
As we experimented with different boards and CPU’s, we discovered that the maximum bus speed Socket A systems can handle depends on the type of the CPU used as well. Sometimes T-Birds can handle up to 5-6 MHz higher than Durons (or vice versa, Durons can handle less). We were able to replicate results on many combinations.
We tested different combos both with T-Birds and Durons. In cases where RAM is tuned for max performance (and there are no other limiting factors) T-Birds tend to be stable at higher FSBs than Durons can handle. We believe this is because of the lower cache (hence more need for CPU-RAM transfers) of the Durons.
So we think a board has to be tested with both CPU’s as stability
Well, we all know this is the biggest problem AMD overclockers have at the moment. This has been bothering us since the beginning but we were not able to compile our findings into a proved statement up to now. Things are beginning to clear up, though. A lot of people have good things to say about this issue (like your article or the article at Burning Issues) but they all need to be summed up.
So here are our additions (derived through many experiments) to the
We are bothered by another aspect of the situation: Temp readings under extreme water cooling (with cold water or with water+peltier setups):
- If the CPU is cooler than the board (or below ambient temp) things tend to get worse
- In the range of “Sub board” to “extreme – near 0” cooling, the effect of CPU heat in backside readings seems to decrease a lot, and what you read is merely a combination of side factors with a bit of CPU heat mixed in. In these cases, local temperature of the board at the probe’s connection point and the heat from the capacitors seem to be dominant. So you can get a significant decrease in readings by cooling the contact spot (this can even be done by cooling the backside of the board at the exact location) or the capacitors, which leads to even more errors in measurement.
- On most of the boards, so called “System Temperature” readings do not seem to be very related with the CPU readings. Temperature of the
capacitors seem to be more effective. Local capacitor cooling can yield low CPU temp readings even if the system temp is high.
- We believe the behaviour of the measured CPU temperature (the path of increase) at the boot time is definitely a “Capacitor Behaviour”. The readings perfectly fit a capacitor heat-up curve, and it is very different from normal ambient or CPU core behaviour. But we don’t have enough recordings to prove that yet.
- We suspect a reversal in backside heat transfer under extreme CPU
core cooling, as we recorded a few cases where CPU plate (and pins) were hotter than the core. This may be caused by some measurement error, so we are still working on replicating the results. But if we are right, after some point the heat is transferred from the board to the cooler CPU, and as we try measure CPU temperature on that pathway, we measure something completely different.
- With sub zero cooling, things seem gto et a bit “back to normal”. We believe that the extreme temp difference causes the real CPU temp to get dominant in the readings again. But because of the matter stated above, we think all we measure is a pale reflection of what we would like to measure.
- AMD CPUs seem to be waterproof :-)))) We were not able to burn even a single CPU under “extreme cooling without insulation”, even though half of the board was flooded with condensing water. We only had a few lockups. Something like that could have easily killed 10+ Intel CPUs. That does not mean everybody will be so lucky of course. We tried so hard that we now consider Socket As “Immortal” :-)))))
Hope these help,