ASUS ROG STRIX RTX 2060 O6G Graphics Card Review

Asus hardly needs an introduction these days. They have been essentially the largest motherboard and graphics card manufacturer for decades, and they have branched out into other products that cover the spectrum of PC technology. They are generally known by their quality construction and high performing designs. Though the competition from Gigabyte and MSI has been fierce, Asus still has retained the impression that their products are the benchmark against which all others are measured.

The ROG (Republic of Gamers) moniker also covers a series of products that are a step above the standard/reference products. STRIX is the second part that essentially describes products that are designed for cool and quiet running. These have been pretty popular products with enthusiasts who are able to spend a little extra money to get the features they want out of a card. These products are over-engineered in terms of components and cooling to provide an experience that should extend well above what reference designs can deliver.

The Asus ROG STRIX RTX 2060 O6G that we are covering today is a card that currently is priced at $419, which is well above the MSRP of a generic RTX 2060 (and with the RTX SUPER launch this must assuredly come down – Ed.). It features an overclocked core, but the memory is left at standard speed. Further discussion of the benefits of the card will follow.

The ASUS ROG STRIX RTX 2060 O6G Card

This is the overclocked version of the STRIX RTX 2060 which pushes maximum boost clocks to 1860 MHz as compared to the standard 1680 MHz that NVIDIA specifies. The board has 6 GB of GDDR6 clocked at 14 Gbps (or 7000 MHz, depending on what measurement you apply). This memory is on a 192 bit bus rather than featuring 8 GB of memory on a 256 bit bus as seen on the RTX 2070. The 2060 and 2070 parts both share the same silicon, but the 2060 is partially disabled. The 2060 has a total of 1920 CUDA cores as compared to the 2304 of the 2070. The 1920 CUDA cores is identical to the specifications of the older GTX 1070. This should have been the biggest clue to me about how the card would perform in a variety of applications.

The Turing architecture increases overall performance per clock compared to the older Pascal GPUs, and it also adds the RTX features mentioned above. The die size of the RTX silicon is quite a bit larger than the previous GTX generation, even though it is made on an optimized 12 nm process from TSMC. The extra RTX features consume a respectable amount of die space with the included tensor cores and other hardware to accelerate the additional functionality. We also see a bit of a clockspeed boost going to the new Turing architecture due to both process and design decisions.

The biggest single feature of the card is the cooling. This is a 2+ slot affair that features 3 variable speed fans. It is a significant chunk of aluminum fins and nickel-plated copper heat pipes/GPU plate. The fans themselves are a Asus patented design which features finlets on the end of the blades to increase pressure/airflow as well as potentially cut down on noise. To help support this cooler there is a front metal frame that surrounds many of the components as well as a large, stiff back plate with strategically placed holes to improve airflow around hot components.

Asus has boosted the power delivery with their own custom design. It features more power phases and higher quality components as compared to the reference NVIDIA design. In theory this should allow the GPU to clock higher for longer with more adequate power and cooling. The reference RTX 2060 features a single 8 pin PCI-E power connector while the Asus ROG STRIX unit has both a 6 pin and 8 pin connection.

There is a switch that enables either a standard firmware setting which provides stronger fan response as well as a Quiet mode that can lower the fan profiles for moderate gaming. This does very little to reduce performance in Quiet mode, but once the GPU heats up then it falls back on the more aggressive fan profiles. If a user tends to play less graphics intensive games then this is a good option to keep fan noise down. Neither setting has any particularly detrimental effect on overclocking.

The back edge of the card features two external fan headers as well as the Aura RGB Header for external control. RGB is now pretty much standard on most enthusiast cards. It can be disabled by a button right next to the vBIOS switch. There are “light tubes” throughout the front side of the fan design which will display whatever color the individual desires in flashing/strobing/solid motifs.

Connectivity is provided by two HDMI 2.0 as well as two DisplayPort 1.4 ports. The card can drive four monitors at once, but users will obviously have to mix and match connections if they choose to use 3 or 4 monitor configurations. Surround gaming is seemingly being pushed aside on the lower end cards as users typically like to utilize 3 identical monitors with the same connection due to potential G-SYNC use. This is not a killer in terms of functionality, as the availability of ultra-widescreen displays is very common and affordable. I myself moved from a triple screen display to a single ultra-wide with a resolution of 3440×1440 and have been quite happy with that transition.

Included in the software bundle is the GPU Tweak II overclocking and monitoring utility. This is a handy way to monitor clockspeeds, fan speed, and temperature to optimize your gaming experience. The software works very well to easily adjust clockspeeds and fan profiles. The only error I had is that it reports GPU Power Target settings going up to 125%. The cards I used with this application would go nowhere near that high in reality. Less expensive cards might go to around 106%, but the program would report that it could be set to 125%.

The packaging with the card is a highly protective one. Stiff cardboard surrounds a large amount of packing. The card is nestled in the middle with an anti-static bag. The box contains the quick manual, a driver CD that also includes GPU Tweak II and a full version of XSplit Gamecast, as well as several velcro cable ties. These are not exactly the pack-ins of yesterday that could contain multiple cable solutions and adapters as well as bundles of free software and games. Still, Asus does provide some extra value with the XSplit software and the cable ties.

A Personal History of RTX

Last year the release of RTX products was met with nearly a collective “meh”. I was part of the group that had read some of the reviews and came away thinking, “That’s it?” Performance did not seem to match well with the prices that the cards were being offered at. The highest performing part was the RTX 2080 Ti which was introduced at a pretty eye watering $1200 US. That is Titan territory for what is still considered a desktop/gaming card. During the past 9 months we have not seen the prices drop at all, but we have seen the introduction of the lower end RTX 2060 this past January and the non-RTX Turing based cards a few months later.

Benchmarks from last year did not seem all that impressive to me. My impression was that the RTX 2070 was only slightly faster than the GTX 1070 OC cards. When the RTX 2060 was released it was again only faster than the GTX 1060. Sure, it added RTX features to the mix, but there were still relatively few games that supported those. Also, the RTX performance penalty was pretty hefty, though it could be offset by utilizing DLSS (if the game supported it).

I never purchased any of these cards for myself as I simply thought they were not worth it. This old MSI GTX 1070 Gaming Z card that was overclocked to nearly 2 GHz was going to be perfectly fine, as the only reasonable upgrade for me would be the RTX 2080. At nearly $700 US that was just not an option for my penny pinching heart. I was then given the opportunity to review an RTX 2060. Originally it was supposed to be a GTX 1660, but mistakes were made and I was sent the 2060. I’m actually quite glad it happened that way. I was not really prepared for what I discovered when actually testing these cards by myself. The numbers are far better than what I remembered from reading previous reviews.

Has this changed my mind about RTX? Yes, it has. Do I think the price is justified? Well, mostly. As I am writing this review we are closing in on two major GPU launches from AMD and NVIDIA with their respective RX 5700 series and the Super cards. The pricing will have to change, but I do not have any new information I can share here that will change what my impressions are here at the end of June, 2019. Once we receive more information I will revisit the price/performance ratio that currently exists for this card.

Did I expect to have as much fun with this card as what I did? Frankly, no. I was expecting less than GTX 1070 performance for a price that was more than what 1070 cards could be had for. I’m glad that mistakes were made and I could experience this product for myself. I really was not expecting it to perform as well as it did.

Test Setup

I have decided to shake a few things up with this review. Instead of testing at multiple resolutions I thought I would see how these midrange cards can handle ultra quality settings while trying to push a 144 Hz 1080P monitor. Hitting 60 fps on a modern title at this resolution is something that the RX 570 and GTX 1050 and above cards can do, but pushing upwards to 144 fps is a lot more of a struggle than one would think. The availability of quality 24” monitors featuring 144 Hz operation with IPS and VA panels along with FreeSync is high with prices often found below $250. I enabled FreeSync/G-SYNC on the tested cards with the Acer ED242QR which currently retails for $180, but is often for sale around $160.

The test machine is powered by the AMD Ryzen 5 2600X. This has more than enough horsepower to push these cards to high framerates at even 1080P settings. The motherboard is the MSI X470 Gaming M7 AC along with 2 x 8GB XPG 3200 memory from ADATA being run at 3200 MHz with timings of 16-18-18. Storage is comprised of a 250 GB OCZ Agility 3 drive for OS, 2 TB Hitachi 7200 RPM drive for extended storage, and the Intel 660P 2TB NVMe drive for gaming/benchmarking installation. Power supply is a Mushkin Enhanced Joule 1000 watt. Windows 10 Pro was used as the OS and updated to version 1903. NVIDIA cards were powered by the 430.86 drivers while the AMD cards utilize 19.6.2.

The RTX 2060 was tested against the MSI GTX 1070 Gaming Z, MSI RX Vega 56 OC, and the MSI RX 580 Armor. All three are overclocked versions of those chips. These make for an interesting comparison as the RX 580 is often found well below $200 and is considered a good deal for 1080P gaming. The Vega 56 is the last generation midrange card from AMD that can now be found for around $279 when on sale. The 1070 Gaming Z was the highest overclocked offering of that model from MSI and originally retailed for around $460.

Benchmark Results

Unigine Heaven 4.0

This is a much older benchmark, but I was curious how geometry/tessellation performance is with the latest GPUs. This is the one benchmark that at 1080P with tessellation disabled actually went above 144 fps with two of the cards. Ultra preset enabled with 4X AA.

Unigine Superposition

The Unity Engine is highlighted by this much newer benchmark. It delivers quite stunning visuals and at the Extreme setting can hammer even the highest end cards. I tested the Medium and Extreme presets at 1080P.

This really is a good looking benchmark with some excellent material and lighting effects.

All of the cards except the RX 580 are performing very well, with the 2060 taking a commanding lead. At extreme settings we see framerates drop pretty dramatically. The 2060 is still pretty smooth with minimum rates at 28, but when the screws are applied we see the RX Vega 56 take 2nd place from the 1070. The RX 580 is getting close to being a slideshow.

DiRT Rally 2.0

The true successor to the surprise racing hit DiRT Rally features upgraded graphics, more advanced physics and handling, and greatly improved lighting effects. The game has a built-in benchmark that can be enabled by adding the variable -benchmark to the executable/shortcut. Ultra settings were enabled for the test.

None of the cards are sitting anywhere near 144 fps. The 2060 again takes a commanding lead, while the Vega 56 again grabs second. The Vega and 1070 both exhibit some pretty low minimum framerates while the 2060 never gets below 58. This test was done several times for each card and they exhibited these characteristics on each run. The poor RX 580 is well behind, but at least is still playable for the most part, but minimum framerates drop pretty low.

Far Cry 5

While not the absolute latest iteration of the Far Cry series, it still uses the same Dunia engine as Far Cry: New Dawn. The built-in benchmark was used with Ultra settings enabled.

The cards are still mostly bunched together, but the 2060 again takes first place with a nearly 10% lead. Even the RX 580 is doing well by staying around 70 average. Minimum framerates are all pretty acceptable here with the 2060 hanging on with the lead. In future titles the lack of a full 8 GB of memory might come back to haunt it. We are already creeping up towards 6 GB of used video memory in many of these latest titles.

The Division 2

The sequel to The Division takes place during Summer in Washington, DC rather than Winter in NYC. This makes the color palette far more interesting than its predecessor. DC is meticulously recreated in this very ambitious title. The built-in benchmark was used at the Ultra preset.

This is a huge win for the 2060 over what should be a pretty competitive 1070. The Vega 56 holds its own, but the 2060 is just miles ahead in this particular benchmark. The RX 580 again brings up the rear.

Battlefield V

The latest Battlefield installment on the latest version of Frostbite. Though Johan Andersson and some of the principle software engineers have gone to start up a new company, the technology still pushes hardware. It is also highly tuned so that it plays smoothly on a wide variety of configurations. The Ultra preset was enabled and a manual runthrough of the first single player episode was recorded using FRAPS.

For the first time we see the 2060 take second place to the Vega 56. The two are pretty close together, but Vega edges it out. The 1070 is barely edging out the RX 580. Results were consistent across several tests. I did notice some unpleasant stuttering in certain cutscenes with the RX 580, but in actual gameplay it smoothed right out. All of the cards performed well and provided a good experience.

Power Consumption

One of the areas where NVIDIA has had an exceptional lead over AMD is that of power efficiency. The performance of their architectures from Kepler on proved to be very high, all the while doing it without extreme power consumption. AMD’s GCN architecture was pretty good for the time, but it was seemingly unable to scale up while improving power characteristics. So while AMD was running towards 300 watt TDP chips, NVIDIA was staying under 200 watts for most of their desktop chips. Power was measured from the back of the system power supply so it includes all subsystems and not just the GPU.

The results here are not entirely unexpected. At idle all of the cards go into a similar low power state. There is no real winner as the differences are not significant enough. At load the GTX 1070 has the lowest power consumption of the four tested cards at 279. The RTX 2060 chip is pretty large and features more potential functional units, but it comes in at a still pretty reasonable 310 watts. RX Vega is pretty power hungry. While it benefits from lower power HBM2 memory, the chip itself is consuming a lot of power and pumping out some serious heat. The RX 580 sits at 324 watts, but is consistently about half the performance of the RTX 2060 in the above games and benchmarks.

The RTX 2060 achieves a good boost in performance and a significant increase in capabilities without breaking the bank when it comes to power consumption as compared to the previous generation of NVIDIA chips.

Overclocking, Temperatures, and Sound

Expectations for overclocking were very high for this card. The reworked VRMs, the dual power connections, the extra cooling, and a Power Target that approached 120% as well as voltage modifications promised that this card should exist quite happily above 2000 MHz, if not approaching stable 2100 MHz speeds. Sadly, these promises were mostly hollow.

Without overclocking the card at all I was seeing speeds bursting to around 1920 MHz, which is 60 MHz higher than the stated maximum boost clock. The card ran around that speed regardless of the application or how long the application was played. Raising the Power Target to its maximum saw the burst go up to 1960 MHz while extended play resulted in speeds dropping to around 1940 MHz and staying there.

Raising the voltage allowed the burst to achieve 2000 MHz without touching the clock offset and speeds normalizing to 1960 MHz while under load. Pushing the offset as high as it could go without instantly freezing the system did allow speeds to hit 2100 MHz under boost, but very quickly those speeds would drop below 2000 MHz. The highest speed I was able to achieve consistently and under load was between 1980 MHz and 1990 MHz. It would simply not stay above 2 GHz.

Temperatures would go from 42C at idle and max out around 68C at full load and overclocking. At stock speeds/target/voltage the card would hit around 63C with the fans being audible, but certainly not overpowering. Under overclocking and that 68C temperature the fans barely made much more noise. I did not see temps go over 70C at any time during testing and monitoring. The cooling solution on this card is highly effective and the fans are not overly loud or annoying.

I was able to get a few more MHz out of the already overclocked design, and heat and noise were never problems during that experience. Memory overclocking proved to be ineffective. Raising the memory speeds did not show any appreciable performance differences even by going to 7200 MHz. The monitor reported that clocks did indeed go up, but it simply had no positive performance impact. Users can certainly experiment, but my suggestion here is to leave memory clocks alone and focus on GPU core speed.

Conclusion

I came away from testing the RTX 2060 with a much greater respect for the card. I did not honestly expect it to be consistently faster than the previous 1070 which shares the same number of CUDA cores and relative clockspeed. The new RTX features are slowly being integrated into many titles with perhaps one of the more interesting twists is that NVIDIA released a RTX version of Quake II.

While the RTX features do not turn Quake II into a modern looking title, the changes in rendering with regards to light and effects is subtle, yet powerful to experience. It makes it an entirely new game in many ways. Watching light stream through windows as the sun starts to set is beautiful to behold. Reflections, water effects, and materials upgrades also contribute to a fully revitalized game that is a fun trip down memory lane. While not nearly worth the price of admission into the RTX world, it shows the promise of this type of hybrid rendering which can have dramatic effects on not only the visuals, but potentially the mechanics of the game.

The 2060 is billed as a 1440P solution. On a 60 Hz monitor this is perfectly reasonable. It might be the best option for those looking at these 1080P, high refresh rate gaming monitors. Not only can it push framerates in a lot of games, but it supports G-SYNC compatible (eg. FreeSync) monitors. It also is able to apply RTX effects at 1080P resolutions without causing rates to go below 30 fps.

Asus has made a very good card the keeps extremely cool. It is not able to overclock as well as I had hoped, but it never broke 70C in an office that was at a pretty constant 20C. Users in hotter climates can expect higher temperatures, but it did very well under the circumstances I tested it in. Not everyone likes RGB, so Asus has given us the option of turning it off to go into “Stealth Mode”. It was extremely stable in performance and when tweaking power it was able to sustain near 2 GHz clockspeeds without excessive heating or fan noise.

The bundle with the card is not exactly inspiring, but it does include a few pieces of software that can be useful. The packaging keeps the card very safe and the box art is certainly pleasant to look at, but ultimately will make its way to a shelf or the recycle bin.

The price is a sticking point. The card is listed as of today for $419.99 on most sites. This is a very large increase from the $349 MSRP of the standard RTX 2060. Then again, it is clocked significantly higher than the standard RTX 2060 and approaches base RTX 2070 type performance. When looking from that angle the user is getting a pretty reasonable deal considering the 2070’s $499 MSRP. It does not quite reach that performance and users get the memory bus hit as well as 2 GB less memory. We also do not yet know the pricing of the “Super” products that will launch soon. For now $419 is reasonable, but it is going to have to adjust to stay reasonable in the face of the RX 5700 and Super releases.

Asus makes a very solid and appealing card, but the user pays a price for that luxury. It is very flexible and it certainly is fast for the price range. It is an attractive card that packs in the cooling power, all the while keeping noise to a minimum. I can find very few faults with the card other than the price. I will be revisiting it when we see the release of these new cards in the next week. If Asus is able to lower the price, then it will continue to be a compelling buy for the enthusiast not wanting to spend much more than $350. In a vacuum it is really an outstanding card in terms of quality and performance. Let’s just hope that as the RTX 2060 continues, it will be adjusted down in price to be more competitive with the latest offerings.

One last point to make is that the 2060 and 2070 parts do not support the NVLINK functionality. Users cannot double up these cards in SLI as that is only reserved for the RTX 2080 and above parts. This is not necessarily a bad thing as multi-GPU rendering is slowly being phased out in most applications. It is rare to see new titles coming out with multi-GPU support and it sometimes takes a while for drivers to be released that will optimize that type of usage. Once SLI was a value proposition, now it seems far too expensive to effectively implement in any reasonable number of games. Perhaps we will see scalability come to the forefront again with improvements in software and hardware implementations. I would not necessarily bet on it though.

In closing the Asus ROG STRIX RTX 2060 O6G is a very good card with a justifiable price point. The hardware implementation is extremely solid and high performing for this class of card. It can easily replace a GTX 1070 and give the user a noticeable boost in performance all the while delivering on the new RTX features.

Update: This article was written before the Super release last week, so some factors have changed in that time. As good as this card is, the performance it provides at nearly $420 does not justify the price in the face of a stock 2060 Super. Yes, it is a superior build and has outstanding cooling, but we are not that far away from non-reference designs of the 2060 Super which provide these same features with a big uptick in performance that will debut around this price point. Asus has made a great card, but we need to see some serious price reductions in the $370 range before I would recommend buying it. Even at that point, we must ask if spending an extra $30 is a better idea? Regardless, the Strix card is an impressive product, but adjustments must be made.