Date: Tuesday , September 12, 2017
As of late we have been performing many follow-up evaluations with AMD’s new Radeon RX Vega line of GPUs. The AMD Radeon RX Vega 64 launched on August 14th, and we evaluated the video card against the competition in many games. Following it the AMD Radeon RX Vega 56 launched on August 28th and we also compared it against the competition in many games.
Since then we have also dived deeper into investigating AMD Radeon RX Vega 64 and 56 performances. We have compared both video cards to AMD’s previous high-end GPUs launched two years ago, the AMD Radeon R9 Fury and AMD Radeon R9 Fury X Fiji based GPUs. We looked at performance from the stance of an upgrade path from AMD’s previous high-end gaming experience, to today’s gaming experience with Vega. We found Vega to be a solid upgrade in performance for gamers running Fury and Fury X currently. One thing Vega does well is improve on Fury and allow a gaming experience at 1440p with much higher game IQ settings.
Continuing our in-depth look into AMD Radeon RX Vega performance we are going to do a unique test, at least for us, in today’s evaluation. We want to definitively find out if AMD Radeon RX Vega’s performance advantage over Fiji is because of architectural improvements, or just the sheer clock speed advantage Vega has over Fiji architecture. To do this we are going to do a clock-to-clock GPU comparison.
We are going to take the AMD Radeon RX Vega 64 and GPU clock and match it to the AMD Radeon R9 Fury X clock. In addition, we are also going to memory bandwidth-match both cards as well. In this way, both will have the same GPU clock speed and same memory bandwidth.
In this "clock-to-clock" test we can see if Vega 64 offers any performance advantages over Fury X when clock speeds are the same. If it turns out that Vega 64 is still faster than Fury X (and by how much of course) this could indicate architecture changes providing some level of performance improvements. If, however both video cards end up being identical performance, or very similar, then it would lean more toward the theory that clock speed increase alone is the reason for Vega’s performance lead over the previous Fiji GPUs. This is not a comparison born out of lack of things to do around here. In our previous reviews it stood out to us, and many of our readers, that our performance results between Vega and Fiji were almost identical in terms of performance profile. It just is not something we are used to seeing jumping from architecture to architecture.
And when you see the performance profiles come in almost mirror images of each other, we of course remember that both AMD Radeon RX Vega 64 and AMD Radeon R9 Fury X have the same number of compute units (64), the same number of stream processors (4096), the same number of ROPs (64) and the same number of texture units (256).
With GPU clock speed, and memory bandwidth equalized, we can find out if architecturally Vega is superior in performance to Fury X, or not. So this is much more of a "just because we wanted to know" review. Quite frankly, there is very little value in it when it comes to purchasing a video for gaming with today. Our test removes the clock speed as a variable and memory bandwidth variable to the best of our ability. While we struggled to get the numbers to match up exactly, we did not achieve that, but we came close enough to make the data valuable, if you care about it at all.
You can see below how our stock cards match up in terms of specifications, and what we are looking to make happen.
We will detail the process of "equalizing" both video cards on the next page.