Performance summary charts & graphs

Here are the summary charts of 26 games and 3 synthetic tests. The highest settings are always chosen and it is DX11 when there is a choice, and the settings are ultra or maxed. Specific settings are listed on the Main Performance chart.  We have added the “Kite Demo” on Unreal Engine 4 using the default settings at 1920×1080 as measured by Fraps.

The benches are run at 1920×1080, 2560×1440, and 3840×2160.  All results, except for Firestrike, show average framerates and higher is always better. In-game settings are generally maxed out and they are identically high or ultra across all platforms.  “NA” means that the benchmark was not tested at all.

The main summary chart: The Big Picture

This main summary chart is what we call, “The Big Picture” since it places our test card into a much larger test bed.

The first column represents the GALAX GTX 970 EXOC and the second column is its competitor, the reference non-throttling 290X.  The third column represents the 290X results when overclocked to its maximum 1080MHz/1425MHz, compared with the R9 390X at the same clocks in the fourth column; it is a -OC (minus overclock) on the 390X memory.  The fifth bright red column represents the MSI R9 390X results at its out of the box clocks next to the sixth column (purple) which is overclocked as far as we could go on stock voltage and fan profile.  In the seventh column (green: third to the last) is the reference GTX 980 results followed by the stock Fury X results in the eighth, and finally by the reference GTX 980 Ti in the ninth and last column.  Please bear in mind that the Fury X results are on slightly older drivers (15.15B on Windows 8.1) for a rougher comparison, so there will be some slight variability until we receive our replacement.

Big-PicThis chart tends to be information overload, so we also break it down into two smaller charts for ease of viewing and for analysis.  Our chart provides a lot of information across 26 games and 3 synthetics.  What we can take away from the results generally is that the GTX 980 Ti is the fastest single GPU video card followed by the Fury X, the GTX 980, and then the MSI R9 390X. The R9 390X is faster than the overclocked 290X or aftermarket 290Xes which trade blows with the factory overclocked GTX 970.

When we compare the $429 MSI 390X Gaming 8G  against the $479 GTX 980, it does OK at 3840×2160 but the gap widens further in favor of the GeForce below 4K resolutions.  Neither card is particularly suited for 4K except in multi-GPU configurations in which case the 8GB of video memory will benefit the Radeon.  We can see the benefits of the 390X with 8GB vRAM in Wolfenstein at 4K where the 290X with 4GB of vRAM chokes completely, yet a 4GB card like the GTX 980 is still faster.  However, at these extreme settings, no single video card can run this benchmark well at 4K, not even the 12GB vRAM-equipped TITAN X.

The 390X versus the GTX 980.

This chart places the reference GTX 980 between the stock and the overclocked results of the MSI 390X Gaming 8G.

390xv980

The MSI card almost catches up to the GTX 980 at 4K which is impressive until we realize that the GTX 980 can be overclocked much further that it can.

The Overclocked 290X 4GB versus the 390X at the same clocks.

This chart will give us a comparison directly between overclocked results for the 290X at 1080/1425MHz versus the 390X at identical clocks; the memory of the 390X is underclocked by 75Mhz.  Look at the two middle columns for the direct clock for clock comparison.

390xv980-fix

What we see is that at the same clocks, the 390X is slightly faster than the 290X at the same clocks.  We also checked the 390X against the 290X at reference 290X speeds to make sure that the 290X was not throttling at all – at 1000MHz/1250MHz. At 290X reference Uber speeds, the performance improvements of the 390X were still there but now much smaller.  And these performance improvements of the 390X are most apparent at 1920×1080 where they are less useful than at 2560×1440 for which these card appear to be designed.  And it also appears that these small performance improvements may be mostly explained by the faster timings of the upgraded and faster 8GB of Hynix GDDR5.  It is doubtful that there are core enhancements to Grenada over Hawaii.

Let’s head for our conclusion.

8 COMMENTS

  1. “Why would AMD repackage their R9 290 4GB series into 390 8GB series to tout 8GB for a relatively weak card”

    Because in DX12 titles (i.e. Windows 10), an R9 290x will dog a 980 and almost match a Ti (at over $650). The 390x will test fine. Just wait.

    • I have been told to wait … over and over. When I buy a video card I look for performance in today’s games and for what I am playing now; not next year. When the real built-from-the-ground-up DX12 games arrive, they will have far more demanding visuals that will require next generation GPUs from both AMD and Nvidia.

  2. This review would be far more relevent if the 390X and GTX980 were both retested using the latest drivers as AMD have done considerable work to their drivers over the last 9 months or so. Also, overclocking the 390X and showing those results but not the OC performance of the GTX980 is some of the most lop-sided reviewing I’ve ever seen.

  3. I have an fx6300 at stock speeds with 8 gb of 1800mhz ram hyperxram.
    I recently got a really good deal on an xfx r9 390x DD black addition.I can run it at around 1160 core stable. for some reason
    overclocking the memory decreases performance so i left it at stock

    Anyhoo.
    I ran the tomb raider benchmark at 1080p( same settings as in this review) on my card at the same oc they achieved in this review 1125 core and 1500 mem. i got a result of avg fps of 72.9. Just goes to show how much drivers have improved
    considering my cpu and hardisk(which is 5200 rpm btw) is probably bottlnecking my card and i still got those results.

Comments are closed.