Update - AMD Thuban Voltage and Temperature Scaling

  • News, Editorials
  • 8
  • HWBOT

Author: Pieter-Jan Plaisier

A few weeks ago we launched our first article of the AMD Thuban which dealt with the impact of voltage and temperature on the maximum stable operating frequency of the AMD Thuban. Of course, the findings presented in that article were based on the results of just one single Thuban sample, so possibly not representative for a larger group of samples. Thanks to Trouffman, the person behind Overclocking-TV, we can now present the results of a second sample and see if they match out initial findings.

The second sample, completely retail this time, has been put to the same kind of stability test procedure as the one we tested, be it using Cinebench rather than the 3DMark06 CPU tests. This benchmark still stresses all six cores at ones, so the results should be pretty comparable. Also, we’re not really comparing the overclockability of both samples directly, but we try to find out if there’s a common scaling pattern.

To provide you with all necessary information, here’s the stepping and batch information of both CPUs:


Voltage Scaling – Comparison


The graph on the left shows the overclocking results that we already know; on the right hand side is the new data. In general, we do not see a large difference between both samples: the general scaling findings of the first article still apply to the second sample. Looking more closely, we can see that the newly tested sample seems to be going just a tad higher in the high temperature and low voltage (e.g.: -40°C and 1.3V) test settings, but at -180°C, our first sample was again just a tad better.

Temperature Scaling – Comparison


Similar to our previous findings, both from the article as in the above section, we see that both samples behave pretty similar when it comes to the temperature scaling. The almost linear scaling at 1.5V is striking, almost mathematically beautiful, and seems to be once again the sweetspot for temperature scaling measurements. Looking at the absolute results it seems that the sample tested in the main article is able to hit tad higher speeds.

Conclusive thoughts

Having seen the results of a second AMD Phenom II X6, it looks like the findings in our original article still stand. The Thuban scales quite a bit with both voltage and temperature, but at the very end of both graphs we see less scaling as the Thuban loses the linear scaling around -160°C and 1.6V. Do note that the overclockability is not getting worse when you go lower in temperature and higher with the voltage, but the scaling isn’t as great. On the positive side, this means that AMD still has some room to tweak the current silicon process to make the Thuban scale better.

I’d like to thank Trouffman for running the tests and the local hardware shop Tones for setting me up with the necessary gear.

Thank you for reading.


8

Belgium richba5tard says:

There should be oc/voltage/temp scaling charts for each and every processor core. massman, have you got a few months spare time? :)

Romania Monstru says:

Dude!!!! Shouldn't you be partying, honeymooning, or smth like that right now? Get off teh forums :D

Belgium richba5tard says:

No honeymoon for another 5 months :( busy busy busy

United States steponz says:

How come no 1.8v on the second cpu?

Belgium Massman says:

I think Trouffman experienced instability issues.

Slovenia tiborrr says:

Nice reading again, mr. Massman. Like yourself I find maximal 6-core stability limited to high 5600's MHz, that's with ASUS C4F & X6 1090T BE 1010BPMW.

Slovenia tiborrr says:

And like Trouffman, I found ~ 1.7V to be the limit of some of the cores. Other cores worked great at ~ 1.8-1.85V.

FlanK3r says:

1015 looks good

Please log in or register to comment.