AMD Phenom II X6 1090T OC Scaling With Voltage/Cooling

Overclocking/OC-Team.be by massman @ 2010-05-12

Does it, just like the Deneb, scale with voltage? Does it scale with temperature? Can we reach unseen heights in raw frequency? These are just a few of the questions we want to tackle in this article.

  • prev
  • next

AMD Thuban – Voltage Scaling

AMD Thuban – Voltage Scaling

The first graph of this article is the visual representation of the how the stable operating frequency of the Thuban scales with increased voltage. For this graph, we have filtered the data and chose only to use the most interesting results, this to keep the graph understandable and prevent it from becoming chaotic. So, you will find only the voltage scaling at 20°C, -40°C, -100°C and -180°C . Do note that we did not test certain temperature/voltage combinations, for instance 20°C and 1.8V.

This because of obvious reasons: for one to make sure the sample lives through the testing (it’s preferred to keep the silicon alive until you’ve done all the tests) and for two because of practical realism: it’s very unlikely that someone will be running this processor at combinations such as 40°C and 1.8V.

Madshrimps (c)


From 1.1V to 1.5V, we can see a very steep curve meaning that the added voltage has a very good effect on the overclockability of the processor. Beyond 1.5V we notice that the curve flattens out and see that even at -100°C, an increase in voltage from 1.7V to 1.8V offers no significant frequency improvement. Based on these findings, it looks like our initial thoughts when randomly testing this sample on an air-cooling-based system are making sense: The sweetspot for the CPU in terms of voltage scaling is probably somewhere in between 1.5V and 1.6V. For better scaling with higher voltage, subzero cooling is already required.

Comparison

As mentioned before, Finnish overclocker Sampsa has already done a similar test for his thesis. Luckily, I was allowed to have a peek at his results so I could compare the results to mine, if possible. Although he was using the Cinebench benchmark to test stability, we know that the results are +/_ comparable since both this benchmark and the 3DMark06 CPU test are multi-core stress benchmarks. His AMD Phenom II X4 965 C3 was running stable at 4.12GHz, using 1.4V and 20°C; looking at our results, the Thuban is hitting roughly 4.2GHz around the same temperature and voltage setting. Just a tad higher, possibly insignificant, possibly very interesting.
  • prev
  • next
Comment from jmke @ 2010/05/12
a follow-up article was posted @ HWBot: http://hwbot.org/article/news/update...rature_scaling

 

reply