SAM tests on the RX 6800. AMD tunes more than it reveals

Ľubomír Samák

3 years ago

Methodology: performance tests

The Smart Access Memory feature is far from perfect. And although it is a useful technology overall, it also proves to be very unstable. What was true today may not tomorrow, and an in-depth analysis shows all those patches. On the one hand, those that eliminate the biggest problems that no one boasts about, and on the other hand, those that raise average performance in small steps.

Gaming tests

The largest sample of tests is from games. This is quite natural given that GeForce and Radeons, i.e. cards primarily intended for gaming use, will mostly be tested.

We chose the test games primarily to ensure the balance between the titles better optimized for the GPU of one manufacturer (AMD) or the other one (Nvidia). But we also took into account the popularity of the titles so that you could find your own results in the charts. Emphasis was also placed on genre diversity. Games such as RTS, FPS, TPS, car racing as well as a flight simulator, traditional RPG and sports games are represented by the most played football game. You can find a list of test games in the library of chapters (9–32), with each game having its own chapter, sometimes even two (chapters) for the best possible clarity, but this has its good reason, which we will share with you in the following text.

Before we start the gaming tests, each graphics card will pass the tests in 3D Mark to warm up to operating temperature. That’s good synthetics to start with.

We’re testing performance in games across three resolutions with an aspect ratio of 16:9 – FHD (1920 × 1080 px), QHD (2560 × 1440 px) and UHD (3840 × 2160 px) and always with the highest graphic settings, which can be set the same on all current GeForce and Radeon graphics cards. We turned off proprietary settings for the objectivity of the conclusions, and the settings with ray-tracing graphics are tested separately, as lower class GPUs do not support them. You will find their results in the complementary chapters. In addition to native ray-tracing, also after deploying Nvidia DLSS (2.0) and AMD FidelityFX CAS.

If the game has a built-in benchmark, we use that one (the only exception is Forza Horizon 4, where due to its instability – it used to crash here and there – we drive on our track), in other cases the measurements take place on the games’ own scenes. From those we capture the times of consecutive frames in tables (CSV) via OCAT, which FLAT interprets into intelligible fps speech. Both of these applications are from the workshop of colleagues from the gpureport.cz magazine. In addition to the average frame rate, we also write the minimum in the graphs. That contributes significantly to the overall gaming experience. For the highest possible accuracy, all measurements are repeated three times and the final results form their average value.

Tests with active AMD Smart Access Memory will not be part of the standard methodology yet. Of course, we will focus on SAM, but for better orientation, we will include these tests in a separate article. But we’re doing this just temporarily, until GeForce graphics supports it as well. Then we switch to the opposite model and all cards will be tested with SAM turned on. Until then, however, SAM will be turned off in standard tests, and we will publish the performance increase under its influence separately. No one will be cut short by anything (neither those who have pure AMDs in their cases, nor the owners of Intel platforms) and the clarity of the results will be nicely preserved. Still, putting multiple modes of one card into the same chart (or having 500 charts per article instead of 300) would no longer do any good.

We plan to do one more thing – once a quarter to measure the impact of various updates (drivers, OS, games, BIOS) on performance. This will result in percentage increases or drops in performance that you can work with when studying older tests. It’s a bit of a compromise, but it’s definitely a better option than releasing new tests with out-of-date software. Of course, it would be ideal to test all previous cards before doing every new test, but this is unrealistic. But we believe that you will also appreciate the continuous measurement with one GeForce graphics card and one Radeon and the inclusion of the appropriate coefficient in the criteria of interactive graphs.

Computational tests

Testing the graphics card comprehensively, even in terms of computing power, is more difficult than drawing conclusions from the gaming environment. Just because such tests are usually associated with expensive software that you don’t just buy for the editorial office. On the other hand, we’ve found ways to bring the available computing performance to you. On the one hand, thanks to well-built benchmarks, on the other hand, there are also some freely available and at the same time relevant applications, and thirdly, we have invested something in the paid ones

The tests begin with ComputeBench, which computes various simulations (including game graphics). Then we move on to the popular SPECviewperf benchmark (2020), which integrates partial operations from popular 2D and 3D applications, including 3Ds max and SolidWorks. Details on this test package can be found at spec.org. From the same team also comes SPECworkstation 3, where GPU acceleration is in the Caffe and Folding@Home tests. You can also find the results of the LuxMark 3.1 3D render in the graphs, and the remarkable GPGPU theoretical test also includes AIDA64 with FLOPS, IOPS and memory speed measurements.

For obvious reasons, 3D rendering makes the largest portion of the tests. This is also the case, for example, in the Blender practical tests (2.91). In addition to Cycles, we will also test the cards in Eevee and radeon ProRender renderers (let AMD have a related test, as most are optimized for Nvidia cards with proprietary CUDA and OptiX frameworks). Of course, an add-on for V-ray would also be interesting, but at the moment the editorial office can’t afford it, we may manage to get a “press” license in time, though, we’ll see. We want to expand application tests in the future. Definitely with some advanced AI testing (we haven’t come up with a reasonable way yet), including noise reduction (there would be some ideas already, but we haven’t incorporated those due to time constraints).

Graphics cards can also be tested well in photo editing. To get an idea of the performance in the popular Photoshop, we’re using a script in PugetBench, which simulates real work with various filters. Among them are those that use GPU acceleration. A comprehensive benchmark suggesting the performance of raster and vector graphics is then also used in alternative Affinity Photo. In Lightroom, there are remarkable color corrections (Enhance Details) of raw uncompressed photos. We apply these in batches to a 1 GB archive. All of these tasks can be accelerated by both GeForce and Radeon.

From another perspective, there are decryption tests in Hashcat with a selection of AES, MD5, NTLMv2, SHA1, SHA2-256/512 and WPA-EAPOL-PBKDF2 ciphers. Finally, in the OBS and XSplit broadcast applications, we measure how much the game performance will be reduced while recording. It is no longer provided by shaders, but by coders (AMD VCE and Nvidia Nvenc). These tests show how much spare performance each card has for typical online streaming.

There are, of course, more hardware acceleration options, typically for video editing and conversion. However, this is purely in the hands of encoders, which are always the same within one generation of cards from one manufacturer, so there is no point in testing them on every graphics card. It is different across generations and tests of this type will sooner or later appear. Just fine-tuning the metric is left, where the output will always have the same bitrate and pixel match. This is important for objective comparisons, because the encoder of one company/card may be faster in a particular profile with the same settings, but at the expense of the lower quality that another encoder has (but may not have, it’s just an example).

Methodology: how we measure power draw

We have been tuning the method of measuring power draw for quite a long time and we will also be tuning it for some time. But we already have gimmicks that we can work with happily.

To get the exact value of the total power draw of the graphics card, it is necessary to map the internal power draw on the PCI Express slot and the external one on the additional power supply. For the analysis of the PCIe slot, it was necessary to construct an in-between card on which the power draw measurement takes place. Its basis is resistors calibrated to the exact value (0.1 Ω) and according to the amount of their voltage drop we can calculate the current. We then substitute it into the formula for the corresponding value of the output voltage ~ 12 V and ~ 3.3 V. The voltage drop is so low that it doesn’t make the VRM of the graphics card unstable and the output is still more than 12/3.3 V.

We measure power consumption on the card between the graphics card and the PCI Express slot. Rado Kopera took care of the design and implementation (thank you!)

We are also working on a similar device for external power supply. However, significantly higher currents are achieved there, longer cabling and more passages between connectors are necessary, which means that the voltage drop will have to be read on an even smaller resistance of 0.01 Ω, the current state (with 0.1 Ω) is unstable for now. Until we fine-tune it, we will use Prova 15 current clamp for cable measurements, which also measures with good accuracy, they just have a range of up to 30 A. But that is also enough for the OC version of the RTX 3090 Gaming X Trio. If a card is over the range, it is always possible to split the consumption measurement (first into one half and then into the other half of the 12 V conductors).

And why bother with such devices at all when Nvidia has a PCAT power draw analyzer? For complete control over the measurements. While our devices are transparent, the Nvidia’s tool uses the processor that can (but of course does not have to) affect the measurements. After testing the AMD graphics card on the Nvidia’s tool, we probably wouldn’t sleep well.

To read and record measurements, we use a properly calibrated multimeter UNI-T UT71E, which exports samples to XLS. From it we obtain the average value and by substituting into the formula with the exact value of the subcircuit output voltages we obtain the data for the graphs.

We will analyze the line graphs with the waveforms for each part of the power supply separately. Although the 3.3 V value is usually negligible, it needs to be monitored. It is difficult to say what exactly this subcircuit powers, but usually the consumption on it is constant and when it changes only with regard to whether a static or dynamic image is rendered. We measure consumption in two sort of demanding games (F1 2020 and Shadow of the Tomb Raider) and one less demanding one (CS:GO) with the highest graphic details preset and UHD resolution (3840 × 2560 px). Then in 3D rendering in Blender using the Cycles renderer on the famous Classroom scene. However, in addition to high-load tests, it’s important to know your web browser consumption (which, in our case, is accelerated Google Chrome), where we also spend a lot of time watching videos or browsing the web. The usual average load of this type is represented by the FishIE Tank (HTML5) website with 20 fish and the web video in our power draw tests is represented by a sample with the VP9 codec, data rate of 17.4 mb/s and 60 fps. In contrast, we also test offline video consumption, in VLC player on a 45 HEVC sample (45.7 mb/s, 50 fps). Finally, we also record the power consumption of the graphics card on the desktop of idle Windows 10 with one or two active UHD@60 Hz monitors.

Noise measurement…

Noise, as well as other operating characteristics, which we will focus on, we’re measuring in the same modes as consumption, so that the individual values overlap nicely. In addition to the level of noise produced, we also record the frequency response of the sound, the course of the GPU frequencies and its heating.

In this part of the methodology description, we will present something about the method of noise measurement. We use a Reed R8080 sound level meter, which we continuously calibrate with a calibrated Voltcraft SLC-100 digital sound level meter. A small addition to the sound level meter is a parabola-shaped collar, which has two functions. Increases the sensitivity to distinguish the sound produced even at very low speeds. It is thus possible to better compare even very quiet cards with the largest possible ratio difference. Otherwise (without this adjustment) it could simply happen that we measured the same noise level across several graphics cards, even though it would actually be a little different. This parabolic shield also makes sense because, from the outer convex side (from the back), it reflects all the parasitic sounds that everyone who really aims for accuracy of the measurements struggles with during the test. These are various cracks of the body or objects in the room during normal human activity..

To ensure the same conditions when measuring the noise level (and later also the sound), we use acoustic panels with a foam surface around the bench-wall. This is so that the sound is always reflected to the sound level meter sensor in the same way, regardless of the current situation of the objects in the test room. These panels are from three sides (top, right and left) and their purpose is to soundproof the space in which we measure the noise of graphics cards. Soundproofing means preventing different reflections of sound and oscillations of waves between flat walls. Don’t confuse it with sound-absorbing, we’ve had that solved well in the test lab for a long time.

During the measurements, the sound level meter sensor is always placed on a tripod at the same angle and at the same distance (35 cm) from the PCI Express slot in which the graphics card is installed. Of course, it’s always closer to the card itself, depending on its depth. The indicated reference point and the sensor angles are fixed. In addition to the “aerodynamic noise” of the coolers, we also measure the noise level of whining coils. Then we stop the fans for a moment. And for the sake of completeness, it should be added that during sound measurements, we also switch off the power supply fan as well as the CPU cooler fan. Thus, purely the graphics card is always measured without any distortion by other components.

… and the frequency response of the sound

From the same place, we also measure the frequency of the sound produced. One thing is the noise level (or sound pressure level in decibels) and the other thing is its frequency response, coloration.

According to the data on the noise level, you can quickly find out whether the graphics card is quieter or noisier, or where it is on the scale, but it is still a mix of different frequencies. Thus, it does not say whether the sound produced is more booming (with a lower frequency) or squeaking (with a high frequency). The same 35 dBA can be pleasant but also unpleasant for you under certain circumstances – it depends on each individual how they perceive different frequencies. For this reason, we will also measure the frequency response of the sound graphics card in addition to the noise level, via the TrueRTA application. The results will be interpreted in the form of a spectrograph with a resolution of 1/24 octave and for better comparison with other graphics cards we will include the dominant frequency of lower (20–200 Hz), medium (201–2,000 Hz) and higher (2,001–20,000 Hz) sound spectrum into standard bar graphs. For measurements, we’re using a calibrated miniDSP UMIK-1 microphone, which accurately copies the position of the sound level meter, but also has a collar, even with the same focal length.

At the end of this chapter, it should be noted that measurements of noise and frequency response of sound will be performed on most cards only in load tests, as out of load and at lower load (including video decoding) operation is usually passive with fans turned off. On the other hand, we must also be prepared for exceptions with active operation in idle or graphics cards with dual BIOS setup, from which the more powerful one never turns off the fans and they run at least at minimum speed. Finally, as with measuring the noise level in one of the tests, we also record the frequency response of whining coils. But don’t expect any dramatic differences here. It will usually be one frequency, and the goal is rather to detect any potential anomalies. The sound of the whining coils is of course variable, depending on the scene, but we always measure in the same scene (in CS:GO@1080p).

Methodology: heat tests

We’re also bringing you heat tests. You are at HWCooling after all. However, in order to make it sensible at all to monitor temperatures on critical components not only of the graphics card, but anything in the computer, it is important to simulate a real computer case environment with healthy air circulation. The overall behavior of the graphics card as such then follows from this. In many cases, an open bench-table is inappropriate and results can be distorted. Therefore, during all, not only heat tests, but also measurement of consumption or course of graphics core frequencies, we use a wind tunnel with equilibrium flow.

Two Noctua NF-S12A fans are at the inlet and the same number is on the exhaust. When testing various system cooling configurations , this proved to be the most effective solution. The fans are always set to 5 V and the speed corresponds to approx. 550 rpm. The stability of the inlet air is properly controlled during the tests, the temperature being between 21 and 21.3 °C at a humidity of ±40%.

We read the heat from the internal sensors via GPU-Z. This small, single-purpose application also allows you to record samples from sensors in a table. From the table, it is then easy to create line graphs with waveforms or the average value into bar graphs. We will not use the thermal camera very much here, as most graphics cards have a backplate, which makes it impossible to measure the PCB heating. The key for the heating graphs will be the temperature reading by internal sensors, according to which, after all, the GPU frequency control also takes place. It will always be the heating of the graphics core, and if the sensors are also on VRAM and VRM, we will extract these values into the article as well.

Test rig

RAM Patriot Blackout (4× 8 GB, 3600 MHz/CL18)

2× SSD Patriot Viper VPN100 (512 GB + 2 TB)

Power supply BeQuiet! Dark Power Pro 12 1200 W


Test configuration
Processor	AMD Ryzen 9 5900X
CPU Cooler	Noctua NH-U14S@12 V s NT-H2
Motherboard	MSI MEG X570 Ace
Memory (RAM)	Patriot Blackout, 4× 8 GB, 3600 MHz/CL18
SSD	2× Patriot Viper VPN100 (512 GB + 2 TB)
PSU	BeQuiet! Dark Power Pro 12 (1200 W)

/* Here you can add custom CSS for the current table */ /* Lean more about CSS: https://en.wikipedia.org/wiki/Cascading_Style_Sheets */ /* To prevent the use of styles to other tables use "#supsystic-table-979" as a base selector for example: #supsystic-table-979 { ... } #supsystic-table-979 tbody { ... } #supsystic-table-979 tbody tr { ... } */

Note.: At the time of testing, graphics drivers AMD Adrenalin 20.12.2 are used, and the OS Windows 10 Enterprise build is 19042.

3DMark

For the tests we use 3DMark Professional and the Night Raid (DirectX12) is suitable for comparing weaker GPUs, for more powerful ones there is Fire Strike (DirectX11) and Time Spy (DirectX12).

Age of Empires II: DE

Test platform benchmark, API DirectX 11; graphics settings preset Ultra; no extra settings.

Assassin’s Creed: Valhalla

Test platform benchmark; API DirectX 12; graphics settings preset Ultra High; no extra settings.

Battlefield V

Test platform custom scene (War stories/Under no flag); API DirectX 12, graphics settings preset Ultra; TAA high; no extra settings.

Battlefield V with DXR

Test platform custom scene (War stories/Under no flag); API DirectX 12, graphics settings preset Ultra; TAA high; extra settings DXR.

Note: This game also supports DLSS, but considering it is an older title and there are already more than enough tests, we will no longer test this setting. But we can test it if you ask for it.

The Radeon RX 6800 is the cheapest graphics card equipped with the Navi 21 graphics chip. This chip is naturally weaker in this card than in the RX 6800 XT and RX 6900, but the RX 6800 clearly offers the most attractive price/performance ratio of the three. And the efficiency is remarkably very decent here, despite the large, partially deactivated core. However, some flaws can be found on the AMD reference card.

Borderlands 3

Test platform benchmark; API DirectX 12, graphics settings preset Ultra; TAA; no extra settings.

Control

Test platform custom scene (chapter Polaris); API DirectX 11, graphics settings preset High; no extra settings.

Control with DXR

Test platform custom scene (chapter Polaris); API DirectX 12, graphics settings preset High; extra settings DXR and DXR.

DXR (native)

Eighteen selection games in various resolutions and settings, performance measurement in computing applications, detailed analyses of power draw and generated sound, in which we go up to the level of frequency response. This captures the new robust methodology for graphics card testing. And right at the beginning, popular high-end models from both camps will face each other.

Counter Strike: GO

Test platform benchmark (Dust 2 map tour); API DirectX 9, graphics settings preset High; 4× MSAA; no extra settings.

Cyberpunk 2077

Test platform custom scene (Little China); API DirectX 12, graphics settings preset Ultra; no extra settings.

Cyberpunk 2077 with FidelityFX CAS

Test platform custom scene (Little China); API DirectX 12, graphics settings preset Ultra; extra settings FidelityFX CAS.

FidelityFX CAS (50 %)

DOOM Eternal

Test platform custom scene; API Vulkan, graphics settings preset Ultra Nightmare; no extra settings.

F1 2020

Test platform benchmark (Australia, Clear/Dry, Cycle); API DirectX 12, graphics settings preset Ultra High; TAA; extra settings Skidmarks blending off*.

*on GeForce graphics cards, the Skidmarks blending option is disabled. This option is missing on AMD graphics cards. However, the overall quality of Skidmarks is otherwise set to High on both GeForce and AMD.
Note: The game also supports DLSS 2.0 and FidelityFX for upscaling and sharpening, but due to the relatively low hardware requirements in the native settings, we will not address them in standard tests. However, some measurement on request is possible if you ask for it.

FIFA 21

Test platform custom scene (Autumn/Fall, Overcast, 9pm, Old Trafford); API DirectX 12, graphics settings preset Ultra; no extra settings.

Forza Horizon 4

Test platform custom scene; API DirectX 12, graphics settings preset Ultra; 2× MSAA; no extra settings.

Mafia: DE

Test platform custom scene (from the Salieri’s Bar parking lot to the elevated railway gate); API DirectX 11, graphics settings preset High; no extra settings.

Metro Exodus

Test platform benchmark; API DirectX 12, graphics settings preset Extreme; no extra settings.

Metro Exodus with DXR

Test platform benchmark; API DirectX 12, graphics settings preset Ultra; extra settings DXR and DXR.

DXR (native)

Microsoft Flight Simulator

Test platform custom scene (Paris-Charles de Gaulle, Air Traffic: AI, February 14, 9:00 am) autopilot: from 1000 until hitting the terrain; API DirectX 11, graphics settings preset Ultra; TAA; extra settings Motion Blur off.

Red Dead Redemption 2 (Vulkan)

Test platform custom scene; API Vulkan, graphics settings preset Favor Quality; no extra settings.

Red Dead Redemption 2 (Dx12)

Test platform custom scene; API DirectX 12, graphics settings preset Favor Quality; no extra settings.

Shadow of the Tomb Raider

Test platform custom scene; API DirectX 12, graphics settings preset Highest; TAA; no extra settings.

Shadow of the Tomb Raider with DXR

Test platform benchmark; API DirectX 12, graphics settings preset Highest; extra settings DXR.

Note: This game also supports DLSS and FidelityFX CAS, but since this is an older title and there are more than enough tests, we will not address this setting in standard tests. However, some testing on request is possible if you ask for it.

Total War Saga: Troy

Test platform benchmark; API DirectX 11, graphics settings preset Ultra; 4× AA, no extra settings.

Wasteland 3

Test platform custom scene; API DirectX 11, graphics settings preset Ultra; no extra settings.

Overall game performance

Výkon za euro

CompuBench 2.0 (OpenCL)

Test platform benchmark; API OpenCL; no extra settings.

Game Effects

Advanced Compute

High Quality Computer Generated Imagery and Rendering

Computer Vision

SPECviewperf 2020

Test platform benchmark; API OpenGL and DirectX; no extra settings.

SPECworkstation 3

FLOPS, IOPS and memory speed tests

Test platform benchmark; app version 6.32.5600; no extra settings.

LuxMark

Test platform benchmark; API OpenCL; no extra settings.

Blender@Cycles

Test platform render BMW and Classroom; renderer Cycles, 12 tiles; extra settings OpenCL.

Blender@Radeon ProRender

Test platform render BMW and Classroom; renderer Radeon ProRender, 1024 samples; extra settings OpenCL.

Blender@Eevee

Test platform animation renderEmber Forest; renderer Eevee, 350 images; extra settings OpenCL.

Photo editing

Adobe Photoshop: Test platform PugetBench; no extra settings.

Affinity Photo: Test platform built-in benchmark; no extra settings.

Adobe Lightroom: Test platform custom 1-gigabyte archive of 42 raw photos (CR2) from DSLR; no extra settings.

Broadcasting

OBS Studio and XSplit: Test platform F1 2020 benchmark; extra settings – enabled encoders AMD VCE/Nvidia Nvenc (AVC/H.264), output resolution 2560 × 1440 px (60 fps), target bitrate 19,700 kbps.

Password cracking

Test platform Hashcat; no extra settings. You can easily try the tests yourself. Just download the binary and enter the cipher you are interested in using the numeric code on the command line.

GPU clock speed

GPU heating

Net graphics power draw

Performance per watt

Analysis of 12 V subcircuit power supply (higher load)

Analysis of 12 V subcircuit power supply (lower load)

Analysis of 3.3 V subcircuit power supply”

A look at the in-between card for power draw measurement from the PCI Express slot. 3.3 V subcircuit

Noise level

Frequency response of sound

Measurements are performed in the TrueRTA application, which records sound in a range of 240 frequencies in the recorded range of 20–20,000 Hz. For the possibility of comparison across articles, we export the dominant frequency from the low (20–200 Hz), medium (201–2,000 Hz) and high (2,001–20,000 Hz) range to standard bar graphs. However, for an even more detailed analysis of the sound expression, it is important to perceive the overall shape of the graph and the intensity of all frequencies/tones. If you don’t understand something in the graphs and tables below, you will find the answers to all the questions in this article. That explains how to read the measured data below correctly.

The microphone we use for coolers and coils noise analysis


Graphics card	Dominant sound freq. and noise level in F1 2020@2160p	NF-F12 PWM		NF-A15 PWM
	Low range		Mid range		High range
	Frequency [Hz]	Noise level [dBu]	Frequency [Hz]	Noise level [-dBu]	Frequency [Hz]	Noise level [-dBu]
AMD Radeon RX 6800 (SAM on)	100,794	-71,019	1076,347	-66,494	9665,273	-81,252
AMD Radeon RX 6800 (SAM off)	100,794	-71,759	1107,887	-67,416	2091,412	-75,288
Asus TUF RTX 3080 O10G Gaming	100,794	-76,045	1107,887	-77,850	7034,643	-74,423
AMD Radeon RX 6800 XT (SAM on)	100,794	-71,589	1107,887	-74,742	10848,902	-76,306
AMD Radeon RX 6800 XT (SAM off)	100,794	-72,991	1107,887	-74,724	10848,902	-76,519

/* Here you can add custom CSS for the current table */ /* Lean more about CSS: https://en.wikipedia.org/wiki/Cascading_Style_Sheets */ /* To prevent the use of styles to other tables use "#supsystic-table-770" as a base selector for example: #supsystic-table-770 { ... } #supsystic-table-770 tbody { ... } #supsystic-table-770 tbody tr { ... } */


Graphics card	Dominant sound freq. and noise level in SOTTR@2160p	NF-F12 PWM		NF-A15 PWM
	Low range		Mid range		High range
	Frequency [Hz]	Noise level [dBu]	Frequency [Hz]	Noise level [-dBu]	Frequency [Hz]	Noise level [-dBu]
AMD Radeon RX 6800 (SAM on)	100,794	-71,769	1140,350	-66,111	9948,487	-81,293
AMD Radeon RX 6800 (SAM off)	100,794	-71,603	1140,350	-67,765	9665,273	-80,642
Asus TUF RTX 3080 O10G Gaming	100,794	-75,410	1076,347	-72,321	7240,773	-74,199
AMD Radeon RX 6800 XT (SAM on)	100,794	-73,222	1107,887	-73,892	10848,902	-76,328
AMD Radeon RX 6800 XT (SAM off)	100,794	-73,170	1107,887	-75,262	10848,902	-75,397

/* Here you can add custom CSS for the current table */ /* Lean more about CSS: https://en.wikipedia.org/wiki/Cascading_Style_Sheets */ /* To prevent the use of styles to other tables use "#supsystic-table-771" as a base selector for example: #supsystic-table-771 { ... } #supsystic-table-771 tbody { ... } #supsystic-table-771 tbody tr { ... } */

/* Here you can add custom CSS for the current table */ /* Lean more about CSS: https://en.wikipedia.org/wiki/Cascading_Style_Sheets */ /* To prevent the use of styles to other tables use "#supsystic-table-772" as a base selector for example: #supsystic-table-772 { ... } #supsystic-table-772 tbody { ... } #supsystic-table-772 tbody tr { ... } */


Graphics card	Dominant sound freq. and noise level in Blender (Cycles), Classroom	NF-F12 PWM		NF-A15 PWM
	Low range		Mid range		High range
	Frequency [Hz]	Noise level [dBu]	Frequency [Hz]	Noise level [-dBu]	Frequency [Hz]	Noise level [-dBu]
AMD Radeon RX 6800 (SAM on)	100,794	-71,913	987,015	-89,190	7452,944	-88,332
AMD Radeon RX 6800 (SAM off)	100,794	-71,136	987,015	-89,041	7452,944	-88,237
Asus TUF RTX 3080 O10G Gaming	106,787	-81,541	1659,995	-80,568	6834,380	-77,967
AMD Radeon RX 6800 XT (SAM on)	97,924	-79,763	1208,159	-89,625	7671,332	-85,188
AMD Radeon RX 6800 XT (SAM off)	100,794	-72,980	1243,561	-95,235	7671,332	-84,980

/* Here you can add custom CSS for the current table */ /* Lean more about CSS: https://en.wikipedia.org/wiki/Cascading_Style_Sheets */ /* To prevent the use of styles to other tables use "#supsystic-table-773" as a base selector for example: #supsystic-table-773 { ... } #supsystic-table-773 tbody { ... } #supsystic-table-773 tbody tr { ... } */


Graphics card	Dominant sound freq. and noise level in CS:GO@1080p (coils only)	NF-F12 PWM		NF-A15 PWM
	Low range		Mid range		High range
	Frequency [Hz]	Noise level [dBu]	Frequency [Hz]	Noise level [-dBu]	Frequency [Hz]	Noise level [-dBu]
AMD Radeon RX 6800 (SAM on)	100,794	-71,813	987,015	-87,658	7452,944	-80,420
AMD Radeon RX 6800 (SAM off)	100,794	-72,013	1659,955	-90,354	8863,094	-84,530
Asus TUF RTX 3080 O10G Gaming	100,794	-75,576	1140,350	-81,739	9948,487	-78,734
AMD Radeon RX 6800 XT (SAM on)	100,794	-73,593	1659,955	-79,766	7452,944	-73,997
AMD Radeon RX 6800 XT (SAM off)	100,794	-73,272	1659,955	-83,327	7452,944	-76,372

/* Here you can add custom CSS for the current table */ /* Lean more about CSS: https://en.wikipedia.org/wiki/Cascading_Style_Sheets */ /* To prevent the use of styles to other tables use "#supsystic-table-774" as a base selector for example: #supsystic-table-774 { ... } #supsystic-table-774 tbody { ... } #supsystic-table-774 tbody tr { ... } */

Conclusion

Behavior model of the SAM varies a lot, multiple tests confirm this. 3DMark results don’t show it yet, but games and practical apps do. If we look at it chapter after chapter, then in a very short time due to some (de)optimization, AOE II: DE does not benefit from SAM turned on (and the performance increase from a month ago was not low). The difference (SAM on/off) does no longer include this game. But Assassin’s Creed: Valhalla keeps its 10–15% difference, that the Smart Access Memory adds on the RX 6800. The percentage increase in Battlefield V is already bigger (in QHD up to 15%), that we measured by the end of January on the RX 6800 XT (+5%). Control does not seem to like SAM, where it consumes a bit of performance.

Strange things going on in Counter-Strike: GO. With SAM on, the performance increase depends on the resolution 43–80% (in terms of minimal FPS in FHD even up to 94%). Naturally, the difference in power draw is also adequate. From these results, however, it is clear that there is something wrong. The performance with SAM is finally amazing, but without this technology, CS:GO only draws 90 W out of the card, which is way too little. Results in Cyberpunk 2077 are also remarkable. In the native setting, SAM also increased the performance before in this game, but the situation was reversed with FidelityFX CAS enabled, where active Smart Access Memory decreased the performance (at least on the RX 6800 XT, but there is no reason why it should be different on the RX 6800). That’s clearly not true today and the 10–17% increase is present here as well, like without SAM.

Positive impact of SAM can also be seen in DOOM Eternal. It is not a big deal, but instead of a negative percentage it is now positive (1–5 %). SAM also nicely increases the performance in Forza Horizon 4 or Red Dead Redemption 2, but compared to the RX 6800 XT, it’s also slightly higher. The question is how much of it would happen “under the same conditions” and how much of it is due to new drivers and how much is caused by game updates. One such update has just been released recently for RDR2.

But there are also games where the ratios (and percentage increases across the RX 6800 XT and RX 6800 cards) have more or less not changed, at least not in such a way that it immediately strikes the eye. This typically includes F1 2020, Shadow of the Tomb Raider, which are titles that constantly benefit from SAM. Unlike devil-may-care MS Flight Simulator, Total War Saga: Troy or Wasteland 3, that do not respond to SAM.

From the point of view of application performance, nothing fundamental changes and it is still true that SAM can decrease it in some cases. Especially with screen capture in Xsplit, where FPS drops while recording are more significant than in OBS. But even here, there has been a significant improvement since the last SAM test. And we’ve also seen small percentage increases in other applications, including various cipher-cracking tests.

And conclusion? After 20 hours of pure SAM on testing, I will not avoid the frustration of how its behavior fluctuates. The question is therefore whether to bother with this technology at all within the standard methodology. The original plan was to test SAM additionally until Nvidia began to support this technology. But again … it will also give us numbers that can be misleading in one chart. Of course, performance and thus operational characteristics are always formed over time, but we still remain at the level of relevant comparisons apart from the exclusive and proprietary technologies that companies pride on and tune them. For better orientation, you will find information about the test date and, of course, the version of the graphics driver and OS build in each bar of the graphs.

Anyway, we’ll figure something out. Quarterly testing of one card from each manufacturer and creating some sort of progress coefficient could be at least a little helpful. But for you, the users, of course, the constant SAM testing and improving the methodology is good news. It’s just a hassle for us to somehow test it reasonably and get some of the desired valuable result databases of graphics cards. Well, it’s going to be a struggle and it seems we’re going to fight windmills. 🙂

Thank you to Spacebar for providing us with games for our tests

Continue: Methodology: how we measure power draw