AMD Navi and Nvidia RTX with OC, UV and PCIe 4.0

RX 5700 (XT) und RTX 2070 Super: Overclocking, Undervolting und PCIe 4.0 im Test

How do AMD Radeon RX 5700 (XT) and Nvidia GeForce RTX 2070 Super do in terms of overclocking and undervolting? The editorial team has so far failed to provide an answer due to time constraints. Now BitcoinMinersHashrate answers the open questions and also takes a look at the advantage of PCIe 4.0 with Navi 10.

This summer the graphics card market was buzzing. Nvidia updated the Turing portfolio and started the warm days with the faster models GeForce RTX 2060 Super and GeForce RTX 2070 Super (test), followed by the GeForce RTX 2080 Super (test). Nvidia provided the new variants, because shortly afterwards AMD sent the navigation generation in the form of Radeon RX 5700 and Radeon RX 5700 XT (test) as direct opponents into the race, which could beat GeForce RTX 2060 and 2070 without super. RDNA, the first truly new GPU architecture since the launch of GCN in late 2011, made a successful debut.

RDNA does a lot differently than GCN and brings great improvements in performance per shader unit. The new graphics cards also react differently to overclocking and undervolting. In view of the Navi-Custom-Design-Test-Marathon, the editors have so far not had the time for corresponding tests, but this is now being made up for.

Many tests with Radeon RX 5700 (XT) and GeForce RTX 2070 Super

In this article, the Radeon RX 5700, Radeon RX 5700 XT and GeForce RTX 2070 Super clarify which architecture has the most potential for higher clock rates and lower voltages. The latter two models also investigate whether there have been performance improvements with the latest drivers in the past three months. Finally, the Radeon RX 5700 XT still has to show whether PCIe 4.0 offers advantages in games compared to PCIe 3.0.

The well-known test system

The well-known graphics card test system, which is equipped with a Core i9-9900K, was used for all tests. With the exception of the driver section, the respective launch driver was used throughout. In the case of the GeForce RTX 2070 Super, this is the GeForce 431.16, with the Radeon RX 5700 (XT) it is a beta version of the adrenaline 19.7.1. The exact test procedure can be found in the Benchmark & ​​Methodology article.

Overclocking brings more FPS with sometimes big disadvantages

The performance of modern graphics cards can be improved by two changes: a higher power limit, so that the GPUs can absorb more energy in order to achieve a higher clock released from the factory, and classic overclocking. In the following test series, both were done: the power limit was maximized and then the last stable clock was set.

The reference design of the Radeon RX 5700 and the GeForce RTX 2070 Super were used for the graphics cards. In contrast, the Sapphire Radeon RX 5700 XT Nitro + was selected for AMD's fastest GPS offshoot. The reason: The cooler design of the XT variant has its problems in the reference design when overclocking. The founding edition of the competing GeForce RTX 2070 Super was chosen because it is easier to overclock than many custom designs. Unlike the AMD, the cooling system is strong enough for this, and the power limit can be screwed higher than that of many other custom models.

The following table shows the exact changes in power limit and clock rates. The GPU voltage is not adjusted.

GeForce RTX 2070 Super and Radeon RX 5700 XT in OC flight

The GeForce RTX 2070 Super FE and the Sapphire Radeon RX 5700 XT Nitro + are comparatively easy to overclock. The Nvidia graphics card enables 9 percent more FPS and 6 percent better frame times via overclocking. This makes the graphics card as fast as a GeForce RTX 2080. The Sapphire accelerator enables 10 percent more images per second and 8 percent better frame times than the reference design. This pulls the graphics card in line with the GeForce RTX 2070 Super and leaves the Radeon VII just behind. The Radeon RX 5700, on the other hand, is slowed down more by driver limitation and only allows a 5 percent higher performance. This means that the graphics card works slightly faster than a GeForce RTX 2070 without OC.

The individual games show that the advantage fluctuates due to higher clock rates. The more the power limit limits in the standard setting, the greater the advantage of overclocking. The GeForce RTX 2070 Super gained the most with 12 percent in F1 2019 and Rage 2. The Radeon RX 5700 XT, on the other hand, makes its biggest leap in F1 2019 with a plus of 13 percent. In other games, the performance increase is sometimes only around 5 percent.

AMD gives the Radeon RX 5700 XT "free rein"

For the additional power, the GeForce RTX 2070 Super FE with 256 watts or 15 percent more electrical power than with the standard settings. The Radeon RX 5700 approved itself for the performance plus with 209 watts an additional 19 percent. The two graphics cards are therefore designed to be comparable with regard to the scope of the power limit. With the Radeon RX 5700 XT, on the other hand, AMD leaves much more leeway, which makes the power consumption explode unsurprisingly.

Read This Now:   Is the end of power cables coming? USB Type-C Cables Announced Supporting 240W

The Sapphire Radeon RX 5700 XT Nitro + works with a full 329 watts at maximum overclocking, which corresponds to an additional 119 watts compared to the reference settings. So the power consumption increases by 57 percent and that naturally destroys any energy efficiency, while the Radeon RX 5700 and also the GeForce RTX 2070 Super still work quite efficiently with the higher clock rates.

Undervoltage brings more FPS with less watts

The GCN offshoots have traditionally always brought a lot of scope for undervolting, i.e. the manual lowering of the GPU voltage. This pleases the enthusiast, but the less experienced user has to accept the disadvantages of the relatively high voltage ex works. But what about the successor RDNA? Has AMD managed to operate the voltages used in the graphics cards closer to the optimum? The editors check this using the reference models of both navigation variants. At Nvidia, the GeForce RTX 2070 Super Founders Edition is again used for the tests. The aim of the test series is to set the voltage as low as possible without changing the preset clock rates.

Read This Now:   Samsung churns out 24 gigabit LPDDR4X memory chip

Undervolting is easy to operate on AMD GPUs, because the included WattMan offers the possibility to reduce the voltage of the graphics card. The reference design of the Radeon RX 5700 is configured so that 1,054 volts is applied at 1,750 MHz. At lower clock rates, correspondingly lower voltages are preset, which can be seen from a clock-voltage curve in the WattMan.

The test pattern allows a voltage of 0.972 volts at 1.750 MHz before it crashes. The voltages are also correspondingly lower at lower frequencies, the reduction being less due to the non-linear clock-voltage curve than at the highest frequency.

The variant of the Radeon RX 5700 XT from Sapphire, on the other hand, provides for significantly higher 1,192 volts at an equally higher 2,014 MHz. The scope for tension is very similar to the slower model. With 1.125 volts, the graphics card can be operated error-free, with 1.111 volts there is a crash after a short time.

UV at Nvidia is more precise, but also more complex

With an Nvidia graphics card, the clock-voltage curve can be configured much more precisely than with AMD, but an external tool such as the MSI Afterburner is required. The GeForce RTX 2070 Super Founders Edition is configured so that the 3D accelerator works with around 1.037 to 1.050 volts when playing. However, the same clock rates in the test can still be achieved with 0.925 volts, which represents a significant reduction.

All three graphics cards can be properly "undervolt", whereby RDNA already works much closer to the optimum than GCN, which reduces the tuning potential.

Undervolting can bring minimal performance

Depending on how much the power limit limits, the graphics cards react differently to a lower GPU voltage. The explanation: If the power limit is higher than the maximum allowed consumption, lowering the voltage can drive higher clock rates until the limit is reached again.

The GeForce RTX 2070 Super FE is braked by the power limit ex works, so that the lower voltage increases the performance by an average of 2 percent. Now other factors limit the power limit is out of the game. As a result, the graphics card with the reduced voltage also works audibly quieter than in the original state.

The reference design of the Radeon RX 5700 XT reacts almost identically to the Nvidia graphics card and thus also differently than the Vega predecessor. With a lower voltage, Vega automatically slowed down – this is no longer the case with navigation systems. The Radeon RX 5700 XT thus increased by an average of 2 percent. The volume, on the other hand, remains at the same level, only the temperature drops slightly.

The Radeon RX 5700 does not get slower or faster on average with undervoltage. In some games there is a slightly higher frame rate, in others a slightly lower one. What becomes strange at first glance becomes clear at second, because unlike the larger XT brother, the variant internally called "XL" only gets slightly and not always within the preconfigured power limit.

The result: If the GPU was limited by the power consumption in a game, it gets faster with a lower GPU voltage. However, if this is not the case, the performance drops minimally. This is due to the voltage curve shown in WattMan, which can only be shifted to a limited extent. And so it happens inevitably that in the event of a reduced voltage and a GPU not braked by the power limit, the clock is automatically reduced slightly. This has hardly any practical effects, but is reflected in the benchmarks. Unlike the Radeon RX 5700 XT, the Radeon RX 5700 is quieter during undervoltage because the fan operates at 350 revs less.

The GeForce RTX 2070 Super is the UV king

Nvidia uses a fairly high GPU voltage in the GeForce RTX 2070 Super FE, which results in a relatively high power consumption. For this reason, the graphics card also achieves the best results in the event of undervoltage, since the scope is great. With UV, the GeForce RTX 2070 Super FE only works with 182 watts instead of 222 watts – that's 18 percent less.

Read This Now:   GeForce GTX 1630 arrives on May 31: here are the specs stripped down to the bone

In contrast to the previous generations, AMD lets the navigation reference cards work much more on the “sweet spot” in terms of clock rates and voltages. For this reason, the graphics cards already work efficiently by themselves and the UV potential is not too great. The power consumption of the Radeon RX 5700 can still be reduced by 7 percent to 165 watts, that of the Radeon RX 5700 XT by 5 watts to 199 watts. As a result, the performance per watt increases by 6 percent with the Radeon RX 5700 and 8 percent with the Radeon RX 5700 XT. In contrast, the GeForce RTX 2070 Super FE has a significantly higher 24 percent, so that the Nvidia graphics card is clearly ahead in this discipline, while in the factory state it is only enough for a tie.

Auf der nächsten Seite: PCIe 4.0, neue Treiber und das Fazit

Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/ on line 5373

Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/ on line 5373