Sorry for such an arrogant title I'm from a small forum in Korea and recently wrote a review on AMD Fury X - not only the article about Fury X itself alone but also about its crossfire configuration as well as the competitor's. I carefully guess this is the first 4-way crossfire/sli benchmark for either Fury X, TITAN X or 980 Ti. Hope you enjoy this thread
Note that the images below all contain Korean characters, that you may not understand and may bother you. I'm quite regretful of that, but I'm sure that the images themselves can tell most of the story.
Okay. This is my test rig configuration table.
...and this is what I covered during the test. 12 game titles plus two version of 3DMark (11 and Fire Strike) are used.
I'm going to skip the chapter showing detailed result for each game title due to # of images limit in this forum. Let's go directly to the performance analysis chapter instead. (The original publisher (http://iyd.kr/753) contains all detailed data for each game.)
I analyzed the whole data via two different ways - one is just summing up the whole framerates measured in each game title while the another one is take average of the performance numbers normalized by that of single TITAN X in each game title. I named the former "Accumulated Framerates" and the latter "Average Relative Performance (ARP)".
In addition to that, in taking ARP for all 12 games + 2 variants of 3DMark, there surely are some games not supporting (or at least the scale is very poor) multi GPU configuration. To avoid bias, I decided to offer both ARP analysis : ARP of -literally- all games and 3DMark, as well as the ARP without three multi-GPU-incapable titles.
Well, let's first put an eye on FHD (1920 x 1080) result. Here they are : AF, ARP, and ARP w/o multi-GPU incapable titles.
Seconly the QHD result.
Lastly the UHD result.
Below is multi-GPU scaling chart.
Okay. You've overseen the whole numbers at a glance. The point is that, whether Fury X shows poor performance in low resolution or not, the efficiency of crossfire is pretty good. To tell the truth, what's really good is not the "efficiency" but pure performance itself : Crossfired Fury X actualy beat out TITAN X SLI in most of test scenarios. To make it simple to understand, let's summarize the whole result in one table. Here it is.
See what I mean?
Well, my thread is finally over. Thanks for reading such long (and written in poor English) article.