Performance Validation and Conclusions

Performance Considerations

As I stated before, there should be no performance changes at all to the Radeon RX 480 because of the power division change instituted with the 16.7.1 driver. However, AMD did say that some other driver performance improvements were made and implemented to help offset any performance hit from enabling the Compatibility Mode.

Finally, we’ve implemented a collection of performance improvements for the Polaris architecture that yield performance uplifts in popular game titles of up to 3%. These optimizations are designed to improve the performance of the Radeon RX 480, and should substantially offset the performance impact for users who choose to activate the “compatibility” toggle.

To run a quick spot check with my short time with the driver, I tested the RX 480 8GB reference card with the 16.6.2 driver and with the 16.7.1 driver with the compatibility mode set to on.

You’d be hard pressed to find any noticeable difference in performance in our Metro: Last Light performance results at 4K. The average frame rate is actually highest with the new 16.7.1 driver WITHOUT compatibility mode enabled, giving credence to the notion that the driver offers some very slight performance tweaks as AMD stated. With the compatibility mode enabled, performance matches nearly perfectly with the performance we saw with the 16.6.2 launch day driver. There isn’t any frame time variance change to speak of either. Considering the extreme power delta we witness from the base 16.6.2 driver to the 16.7.1 with compatibility mode on, measuring no performance delta is a fantastic result for AMD.

The results for The Witcher 3 running at 1920×1080 look equally as good as the Metro: Last Light results above. The 16.7.1 driver without compatibility mode shows a slightly, barely measureable, improvement in performance. Meanwhile, the original 16.6.2 driver and the 16.7.1 with compatibility ON are indistinguishable.

Polishing it all up

So what can we take away from this whole experience and this new driver that AMD has released into the wild today? First, it’s important to recognize these types of problems as early as possible, and again, props to Tom’s Hardware for pointing it out first with its launch review. Despite initial reaction that we were making this up, or that it wasn’t a big deal, AMD did the right thing by recognizing and addressing the issue head on. First with acknowledgement, then with a statement and finally with worthwhile and easy to implement solution.

It’s clear now that the RX 480 should not have shipped with the power configuration it had. It was out of spec, and sometimes by a surprisingly high amount. I have talked several times about how this could have been avoided (8-pin power, more validation) so I don’t feel the need to harp on it again. I’m told that internally there are more changes being made to the design and product team to make sure this doesn’t happen again. But for AMD, there can’t be any more issues; the Radeon Technologies Group needs to focus on the straight and narrow going forward.

With the two new fixes, AMD has brought the reference boards back to where we expected them to be. The power phase weighting adjustment is in my mind, the most crucial change. By allowing the 6-pin power connection to provide more of the power to the card than the +12V on the PCI Express graphics slot, there is significantly reduced worry about power draw through the motherboard. Even though we still found that the current readings were over the 5.5A level that the specification says is a maximum, the change gets the RX 480 much closer, and with zero impact on card performance.

The compatibility mode, despite the awful name that would leave consumers believing that the card is broken without enabling it, allows AMD to drop power draw on both +12V power sources slightly, by the same relative amount. In my testing, this actually brings the card right down to the 5.5A power draw recommendation for the PEG slot in our 1080p-based testing. I still measured slightly higher draw in Metro: Last Light at 4K, but 5.75A is at least in the right ballpark.

Should these issues have been discovered and fixed before releasing and selling the Radeon RX 480 to the consumer? Absolutely. Is the power phase weighting adjustment and Compatibility Mode combination enough to make the card at or under spec for ALL workloads? Based on my testing, no. But I do believe that AMD has done it’s best to address the power consumption concerns without a hit to performance, getting the RX 480 to a much more reasonable power situation. I no longer believe that consumers should be worried about the stability of their PCs running the RX 480 with the 16.7.1 driver installed.

What a week.

« PreviousNext »