Testing For Thermal Throttling
System Throttling
Running Prime95 and FurMark at the same time places a very high load on any system, with both the CPU and GPU drawing maximum power while also generating maximum heat. The Panther 5D's CPU is rated for up to 150 W and the each of the GPUs are at least 100 W. Add in 25-50 W for the desktop platform components and you're looking at 400 W or more. Obviously, that's significantly more power than most notebooks draw, and the reason why Eurocom needs to offer dual power adapters. If the system can't get more than 400 W from its supplies, or isn't able to dissipate that amount of heat efficiently, it will throttle the CPU, GPU, or both in order to honor its thermal and electrical specifications.
Synthetic Heat Run
In order to really push the Panther 5D to its limits, we fully load its Core i7-3970X CPU and both GeForce GTX 680M GPUs. The goal is to tax the machine until its internal and external temperatures plateau. The only reason to end a run early is a hardware failure.
In the chart above, you can see the load and temperatures for the CPU and both GPUs.
The CPU took 12 minutes to heat up both of its big copper sinks enough that the fans were spinning at their highest rotational speeds. You can see the slight dip between the 12- and 14-minute mark where the amount of air being drawn in momentarily pulled temps down a bit.
At the 60-minute mark, I shut down FurMark to alleviate the GPU load, dropping them to idle. The CPU temps were unaffected. It was not until I also shut down Prime95 at 65 minutes that the thermals started to relax.
The two CPU fans are on the right half of the computer. When they both kick in, the amount of air they move is startling the first time you feel it.
After close to 20 minutes, the CPU hit 90 °C and throttled down to 3.4 GHz. It stayed at that clock rate for most of the remaining run.
You can see the CPU's clock rate through our run in the chart above.
The first GPU is located next to the GPU, and it sits in the middle of the machine. Its air intake is third from the left. Cooling that graphics processor is naturally a little more difficult, so its temperatures were a few degrees higher than the second GPU, situated all the way to the right.
With that said, neither GPU ran particularly hot during our test, even though they were both fully-loaded for an hour. In the 70-minute graph, you can see a slight dip in their temperatures, which was caused by lifting the machine to measure heat on the desktop under it.
Both GPUs maintained their minimum clock rate during testing.
We saw power drawn from the wall peak at 464 W. Allowing for 10% losses at the power adapters, this means that the Panther 5D was dissipating more than 400 W. Even the power adapters themselves heat up quite a bit. It's naturally important to keep everything around Eurocom's machine well-ventilated if it's under a demanding load.
Then again, we didn't see temperatures spike so high in any real-world workload we through at the Panther 5D.
