Power Consumption
Measurement Methodology and Graphical Representation
In addition to the high-resolution measurement curves we normally present, we're adding a second, more reader-friendly look at the same data using significantly longer intervals supplemented by a hardware-based low-pass filter and a variable filter in the evaluation software. This helps identify short power spikes (and troughs) without bombarding you with all of the raw data.
If you want to know more about how we test and the equipment we use, check out The Math Behind GPU Power Consumption And PSUs.
Compared to previous features, our evaluation hardware hasn't changed:
Power Consumption | |
---|---|
Test Method | Contact-free DC Measurement at PCIe Slot (Using a Riser Card) Contact-free DC Measurement at External Auxiliary Power Supply Cable Direct Voltage Measurement at Power Supply |
Test Equipment | 2 x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function 4 x Rohde & Schwarz HZO50 Current Probe (1mA-30A, 100kHz, DC) 4 x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) 1 x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function |
Power Consumption Idle Mode
We had to re-run the first test a second time to confirm our measurements: a reading of almost 34W at idle is ludicrous. In our launch coverage, AMD's Radeon RX 480, based on the same GPU, used just 16W. A GPU clock rate of 300 MHz eliminated that as the source of this inflation. Ultimately, the cause was determined to be memory operating at full speed.
Update: Troubleshooting the Driver
After working with AMD post-review to reproduce this issue, it no longer occurs under the latest driver. We therefore updated the old idle measurements with new values. Power consumption at idle dropped to 20 watts, which is still higher than the Radeon RX 480, but much better than before.
A different picture emerges after dividing total consumption into individual rails. Under the old driver, draw from the motherboard slot is particularly high, since that's where the memory gets its power.
Old Driver:
New Driver:
Considering where each power phase is connected, it becomes pretty clear that the additional consumption is attributable to the memory. In fact, we detected wild swings in frequency, with the old driver spending most of its time at 1750 MHz and only sporadically throttling down. This is fixed with AMD's new driver.
Power Consumption Under Different Load Scenarios
At the start of this review we noted the Radeon Pro's restrictive power limit that tried to keep the card below its 130W ceiling. Demanding workloads appear to be pegged at 137 watts without any room to push higher, which is good.
Power Supply Connection Loads
This high-resolution view of a one-minute measurement includes data for each rail. It includes all peaks and valleys, illustrating how quickly Power Tune can respond. The new driver has no influence on these results; the values from both builds are approximately the same:
But the spikes may cause more confusion than they need to. For this reason, we're only using the lower-resolution charts with a variable low-pass filter applied for our analysis.
The balance between the motherboard's and six-pin power connector's 12V rails is important to us. Given what we previously observed about how each power phase is fed, the results match well. These graphs now reflect measurements from the new driver, replacing our findings from the original launch.
The 137-watt limit we mentioned earlier is slightly higher than the 130W typical board power cited on AMD's webpage. This ceiling comes partly from the motherboard's 66W power delivery, which is close to what the PCI-SIG considers to be this connection's maximum. The oft-cited 75W limit is incorrect, since it's the product of a 5.5A peak and the ATX's standard's supply voltage (this can be slightly higher than 12V).
Adding current measurements to our test suite allows us to confidently assert that the Radeon Pro WX 7100 remains within AMD's specifications, even when it's pushed hard.
The Radeon RX 480's maximum power consumption seems a bit high for a mainstream product. But the workstation-class Radeon Pro WX 7100 is more reasonable. Solid performance through our real-world benchmark suite and closer proximity to Ellesmere's sweet spot make WX 7100 a more efficient product.