Test Methodology

Testing methodology has completely changed from our last GPU reviews, which were probably for the GTX 1070 Ti series cards. Most notably, we have overhauled the host test bench and had updated with new games. Our games selection is a careful one: Time is finite, and having analyzed our previous testing methodologies, we identified shortcomings where we were ultimately wasting time by testing too many games that didn’t provide meaningfully different data from our other tested titles. In order to better optimize our time available and test “smarter” (rather than “more,” which was one of our previous goals), we have selected games based upon the following criteria:

• Game Engine: Most games run on the same group of popular engines. By choosing one game from each major engine (e.g. Unreal Engine), we can ensure that we are representing a wide sweep of games that just use the built-in engine-level optimizations
• API: We have chosen a select group of DirectX 11 and DirectX 12 API integrations, as these are the most prevalent at this time. We will include more Vulkan API testing as more games ship with Vulkan
• Popularity: Is it something people actually play?
• Longevity: Regardless of popularity, how long can we reasonably expect that a game will go without updates? Updating games can hurt comparative data from past tests, which impacts our ability to cross-compare new data and old, as old data may no longer be comparable post-patch

Game graphics settings are defined in their respective charts.

We are also testing most games at all three popular resolutions – at least, we are for the high-end. This includes 4K, 1440p, and 1080p, which allows us to determine GPU scalability across multiple monitor types. More importantly, this allows us to start pinpointing the reason for performance uplift, rather than just saying there is performance uplift. If we know that performance boosts harder at 4K than 1080p, we might be able to call this indicative of a ROPs advantage, for instance. Understanding why performance behaves the way it does is critical for future expansion of our own knowledge, and thus prepares our content for smarter analysis in the future.

For the test bench proper, we are now using the following components:

GPU Test Bench (Sponsored by Corsair)

Sniper Elite 4 Benchmark – R9 290X 2019 vs. RTX 2070, 2060, GTX 970, RX 590

Sniper Elite 4 will start us out. Before displaying results, remember that one of the most interesting areas to look is going to be scaling performance between two fixed goalposts as we change resolutions. If the distance between those goalposts shrinks, that is indicative of an architectural advantage or deficit at the new resolution. We’ll set the RX 580 and R9 290X at stock settings for these posts. Sniper Elite 4 gives us a well-optimized DirectX 12 title to test with, which is valuable because we want higher framerates even at 4K to better illustrate some of those scaling gaps. Keep in mind that the 290X came out long before 4K was popularized. 1080p still had almost all of the marketshare.

At 4K first, we see the R9 290X stock card at 38FPS AVG, with lows surprisingly close by at 32FPS and 30FPS 1% and 0.1% low. We’ll look at frametimes in a moment. Overclocking headroom was limited and capped at about 1060MHz, getting us to 41FPS AVG and climbing 7.9% over the R9 290X stock card. These performance figures peg the R9 290X and its overclocked counterpart at rough equivalence with the RX 580 8GB card, not too distant from the new RX 590 Fatboy. This is without yet considering power consumption, mind you. The GTX 1060 is just surpassed by our R9 290X results, as is the GTX 970.

For our goal posts, the R9 290X stock GPU allows the RX 580 8GB stock GPU to hold a lead of about 2.9%.

Transitioning to the more limited 1080p results, we see now that the R9 290X has a stock framerate of about 98FPS AVG, allowing the RX 580 8GB card a lead of 9.4% with its 107FPS AVG. The fact that the R9 290X closed the gap at 4K suggests to us that the 580 becomes limited in its ROPs and TMUs, but primarily ROPs. The 290X is better equipped on this front leaving its biggest limitation as frequency, which is why the card has more trouble keeping up at the lower resolutions. Once the RX 580 gets pounded with higher demand on the pixel pipeline, where it becomes more limited, the R9 290X pulls ahead. This same pattern would emerge with anti-aliasing, as it’s effectively increasing the sample rate in the same way as increasing resolution, thus also becoming ROPs-bound rapidly.

As for 1080p performance on the whole, the 290X performs behind a GTX 1060 6GB card and ahead of an RX 570 4GB card.

Frametimes are what we’re most curious about. As a reminder, frametime plots demonstrate the frame-to-frame variance in time to present a new frame. This is a measure of frame-to-frame intervals in milliseconds, so lower is better, versus benchmark progression. The more consistent each point on the line is to the previous, the better the experience. Deviation from the mean in excess of 8-12ms becomes noticeable to most gamers.

The R9 290X does well in this department with Sniper Elite 4. To Sniper’s credit, the game is remarkably well-built, but the 290X still needs the right hardware to keep frame pacing consistent. In this title, we don’t see too much deviation from the mean frametime, with the biggest variance in the form of 3-4ms swings. This is completely acceptable and, as you can see, isn’t too distant from the modern RX 590’s performance. The 590 has fewer peaks on average, but the difference in consistency is unnoticeable overall for most players. The RTX 2060 is also plotted as an example of the most modern architecture, where we’re nearing an ideal frametime plot. The takeaway is that the 290X does well in frametime consistency in this particular title, and that’s despite some early life issues with frametime consistency. Many of these were patched-up with later driver launches, but the rest would likely be more game- or API-dependent.

F1 2018 – R9 290X Benchmark in 2019

F1 2018 is next, giving us a DirectX 11 game that uses the same API as most of the market back when the 290X released. Just for scaling reasons, we’ll look at 4K results, despite this card not really being meant for it in 2019.

At 4K, the R9 290X 4GB card ends up at about 33FPS AVG, ranking it as similar to the GTX 1060 6GB and GTX 970. The RX 580 8GB outperforms the 290X 4GB by 4.6%, landing at 34FPS AVG. We can also learn from the 390X result, which shows a 34FPS AVG. This card is a refresh of the 290X, with a higher frequency and double the memory capacity. In this title, it rapidly becomes clear that memory is not the primary limitation, as performance only increases by a few percentage points. The 290X and 390X are more limited by the GPU than by the memory.

Moving on to 1440p, we see similar resolution scaling as the previous game: The RX 580 stock GPU’s 56.6FPS AVG is 9.1% ahead of the R9 290X’s 51.9FPS AVG, posting a relative gain in performance for the RX 580. Again, we think this is because the R9 290X can leverage its increased ROPs and texture units at higher resolutions or higher anti-aliasing values, closing the gap as resolution increases. That doesn’t mean it’s playable at those higher resolutions, but does illustrate how the GPUs scale.

For 1440p resolution, the 290X is still reasonably playable in this title. Dropping settings from ultra-high to just ‘high’ or similar would make for a consistent 60FPS and beyond. Comparatively, the 290X does about as well as the GTX 970, although the 290X’s lows manage higher results, with the 390X not too distant. The 390X’s extra memory doesn’t get leveraged in a meaningful way for this benchmark. Versus some modern cards, the 290X is outperformed by the GTX 1060 and RX 580 alike.

We don’t see too much improvement for the 290X at 1080p, moving up to 65FPS AVG and with still minimal gains from overclocking. The RX 580 8GB runs at 72FPS AVG, for a lead of 11.6%. To recap this title, we see 11.6% improvement in the RX 580 at 1080p, 9.1% at 1440p, and 4.6% at 4K, showing very clear performance improvements in the higher-frequency, newer cards at lower resolutions.

Shadow of the Tomb Raider - R9 290X Benchmark vs. GTX 980, 970, RTX 2070

Shadow of the Tomb Raider is up next, giving us a DirectX 12 title for another modern look at performance. Dx12 didn’t officially launch until 2015, so the 290X was made well before the new API saw any adoption.

At 4K, the 290X obviously struggles at 24FPS AVG, making it largely unplayable with these settings. The RX 580 doesn’t do much better at 25FPS AVG, with the 590 at 28FPS AVG. Let’s move on to something more reasonable.

At 1440p, the 290X runs at about 42FPS AVG, with the GTX 1060 functionally tied with the 290X. The differences are inside of error margins, so we can’t state if one is better than the other. The RX 580 8GB leads at 45FPS AVG, with the RX 590 at 50FPS AVG and Vega 56 at 62FPS AVG.

1080p positions the R9 290X in playable territory even with these higher settings, at 60FPS AVG for the overclocked version – about where most partner cards would fall – and 58FPS AVG for the stock model. That puts the 290X as comparable to the GTX 1060 6GB and behind the RX 580 8GB.

Far Cry 5 Benchmark – R9 290X Revisit

Far Cry 5 uses geometrically complex meshes and longer view distances, making it one of the more draw call-intensive games we benchmark.

At 4K, Far Cry 5 positions the 290X at 26FPS AVG, right between the GTX 970 and GTX 980 cards, and affording the RX 580 a lead of 8.3% at 28FPS AVG. This isn’t particularly playable under these settings so, once again, we’re mostly using them for perspective.

At 1440p, performance climbs significantly to 43FPS AVG, which is about where the RX 570 and 390X perform. NVIDIA’s GTX 1060 6GB outperforms the 290X by a few percent here, with the most modern cards posting significant leads. The RX 580 holds a 16.7% lead, showing one of the largest gaps we’ve seen between the two yet, but still following the trends we saw previously.

1080p really carries this trend, now allowing the RX 580 8GB card a lead of 22%, which is the biggest gain we’ve seen thus far. As a reminder, that’s against 8.3% at 4K and 16.7% at 1440p, so our earlier theory remains consistent. As for raw framerate, the 290X is still adequate for 60FPS in Far Cry 5 at 1080p and with these settings, but it is getting long in the tooth. Vega 56 and GTX 1070s or even RTX 2060s would offer considerable performance improvements, as you can see in the chart.

GTA V Benchmark – R9 290X Revisit

GTA V is a 2015 game and is the oldest on our benchmark, but is also the most-played game out of everything in this test suite.

At 4K, GTA V lands the R9 290X at 24FPS AVG, which is within error of the RX 580 or slightly leading it, for the first time all bench, and not distant from the GTX 970. This is more of a synthetic look, of course, since it’s not particularly playable.

1440p again posts the R9 290X and RX 580 as roughly equal performance. We run GTA V with 2x MSAA, so it is likely that we’re seeing a potential ROPs limitation on the 580. The 390X does actually post meaningful improvement over the R9 290X here, landing at 53FPS AVG, but this is clearly more of a change in frequency than memory capacity, as the overclocked 290X is not too distant from the 390X.

Conclusion: R9 290X in 2019

It’s really not all that bad, although is considerably aged in its ability to run some of these titles at higher resolutions.

If willing to occasionally drop settings to medium or high, shying away from ultra, and if willing to stick to 1080p, the R9 290X still does reasonably well in these games. Going for better graphics settings might suggest that it’s about time to replace the GPU, though, and in such instances, you’d want to shoot higher than an RX 580, RX 590, or GTX 1060. The R9 290X maintains performance roughly equivalent to these devices – or close enough that the differences aren’t worth a purchase – in modern gaming scenarios, even including DirectX 12. Keeping in mind that Dx12 didn’t even publicly exist when the R9 290X was released, the fact that the 290X maintains overall acceptable performance in Dx12 titles is impressive.

Jumping to Vega-class cards or RTX 2070 would sort of be the entry point for this one, minimally, as anything short of that is a pointless endeavor and more of a “side-grade.” The most interesting take-away, we think, is how the R9 290X managed to close the gap between itself and the RX 580 when playing at increasing resolutions. The natural downside is that the R9 290X isn’t particularly well-equipped for these games at 4K, anyway; while closing the gap is interesting, it doesn’t change the story that these games are functionally unplayable at such high settings and resolutions. The biggest motivator for an upgrade will likely be supporting those higher resolution displays, especially that they’re properly affordable now, unlike when the 290X launched.

The card is getting long in the tooth, but it’s still good a little while longer for anyone truly pinching pennies.

Editorial, Testing: Steve Burke
Video: Andrew Coleman