• Home
  • Blog
  • Review of the AMD Radeon RX 7800 XT video accelerator based on the Gigabyte Radeon RX 7800 XT Gaming OC card (16 GB)

Review of the AMD Radeon RX 7800 XT video accelerator based on the Gigabyte Radeon RX 7800 XT Gaming OC card (16 GB)

24.11.2023 08:54

General information about Radeon RX 7800 XT

The new family of Radeon video cards entered the market in an unusual and rather gradual manner: it began with the release of top-end models, but instead of the expected mid-segment video cards, AMD introduced a budget model — the Radeon RX 7600. This decision passed by an interesting market segment in which gaming enthusiasts actively participate. The completion of the line of current-generation video cards occurred only in September, when AMD introduced the Radeon RX 7800 XT and RX 7700 XT, for which the wait was almost a year after the release of the flagship models.

These new graphics cards share the same GPU and are slotted between the RX 7600 ($250) and RX 7900 XT ($750). The price difference between the new mid-range graphics cards is small: the RX 7800 XT is priced at $500, while the RX 7700 XT has a suggested price of $50 less. This price range, although not the most widespread, is important for video card manufacturers.

The Radeon RX 7800 XT, like the previously discussed Radeon RX 7700 XT, is based on the Navi 32 graphics processor, part of the modern RDNA3 architecture. This architecture offers a number of benefits, including increased performance per clock, including improvements in ray tracing and support for artificial intelligence tasks. This became possible thanks to the use of a modern 5-nanometer technological process in the production of GPUs.

The Navi 32 GPU uses a chiplet design similar to the Navi 31, unlike the smaller Navi 33, which was monolithic. The core compute die (GCD) uses a 5nm process technology and the chip houses four dies with 16MB of Infinity Cache and a 64-bit memory controller. However, higher-end models like the RX 7800 XT have all four MCDs active, providing 64MB of Infinity Cache and a 256-bit memory interface, as opposed to the RX 7700 XT's 48MB and 192-bit bus.

Both video cards, released in early September, are oriented at 2560x1440 resolution. However, the RX 7800 XT offers significantly higher performance than the RX 7700 XT for a slight increase in price. AMD sees the GeForce RTX 4070 as the main competitor to the RX 7800 XT, despite the latter having a MSRP of just $500, which is $100 less than the RTX 4070 rival.

Yes, both of AMD's new graphics cards, including the Radeon RX 7800 XT, are excellent options for upgrading gaming systems, especially for those considering upgrading from previous generation GPUs such as the Radeon RX 5700 XT or GeForce RTX 2070 Super (according to AMD itself). The Radeon RX 7800 XT is an excellent graphics card for gaming at 2560x1440 resolution. It's capable of delivering over 60 frames per second at maximum settings in almost all games, with the possible exception of some of the most demanding ray tracing projects where performance needs may be higher. However, for most gaming scenarios and settings, the Radeon RX 7800 XT will offer an excellent gaming experience.

Graphics accelerator Radeon RX 7800 XT
Chip code nameNavi 32XT
Production technology5 nm and 6 nm (N5 and N6 TSMC)
Number of transistors28.1 billion
Core area350 mm²
Architectureunified, with an array of processors for stream processing of any type of data: vertices, pixels, etc.
DirectX hardware supportDirectX 12 Ultimate, supporting Feature Level 12_2
Memory bus256-bit: 4 independent 64-bit memory controllers with GDDR6 support
GPU frequencyup to 2430 MHz
Computing blocks60 CUs consisting of a total of 3840 ALUs for integer and floating point calculations (INT4, INT8, INT16, FP16, FP32 and FP64 formats supported)
Ray tracing blocks60 Ray Accelerator blocks for calculating the intersection of rays with triangles and BVH bounding volumes
Texturing blocks240 texture addressing and filtering units with support for FP16/FP32 components and support for trilinear and anisotropic filtering for all texture formats
Raster Operation Blocks (ROPs)12 wide ROP blocks of 96 pixels with support for various anti-aliasing modes, including programmable and for FP16/FP32 frame buffer formats
Monitor supportsupport for HDMI 2.1b and DisplayPort 2.1 interfaces
Radeon RX 7800 XT Reference Graphics Card Specifications
Core Clock (Gaming/Turbo)2124/2430 MHz
Number of universal processors3840
Number of texture blocks240
Number of blending blocks96
Effective memory frequency19.5 GHz
Memory typeGDDR6
Memory bus256 bit
Memory16 GB
Memory Bandwidth624 GB/s
Compute Performance (FP32)up to 37.3 teraflops
Theoretical maximum fill rate233 gigapixels/s
Theoretical texture sampling rate583 gigatexels/s
TirePCI Express 4.0 x16
Connectorsone HDMI 2.1b, two DisplayPort 2.1 and one USB Type C
Energy consumptionup to 263 W
Additional foodtwo 8-pin connectors
Number of slots occupied in the system case2.5/3
Recommended price$499

The new graphics card, designated Radeon RX 7800 XT, follows the previously introduced RX 7700 XT model, which is based on a stripped-down version of the Navi 32 GPU. The RX 7800 XT is the older of the two mid-priced graphics cards in this line and is codenamed RX 7800 and suffixed XT.

The recommended price for the Radeon RX 7800 XT is $500. This is a significantly better deal than the RX 7700 XT, which is priced at $450, despite the similarities in using different versions of the same GPU. The difference between them, especially in memory (capacity and speed), is noticeable. Therefore, the RX 7700 XT, with its lower price/performance ratio, seems unreasonably expensive.

The $500 price for the Radeon RX 7800 XT seems like an attractive proposition compared to the GeForce RTX 4070's MSRP, which is $100 higher. Even though Nvidia's graphics card is faster at ray tracing and more energy efficient, it has 4GB less VRAM and is even slightly slower on average. The RTX 4060 Ti with 8 GB of video memory is also viable, as it costs more than $100 less.

The 256-bit video memory bus width led to a choice between 8 and 16 GB for the RX 7800 XT. Currently, 8 GB is the minimum acceptable size for a mid-priced solution, but 16 GB may be more profitable in the future. AMD itself recently showed that less than 16 GB may not be enough for modern games, so 16 GB is considered the optimal size for mid-priced graphics cards today.

AMD has released the Radeon RX 7800 XT in its reference design. This card is similar in style to the RX 7900 series, but has a compact size of 27 cm in length. It requires three slots of installation space and offers efficient cooling for the Navi 32 GPU. In terms of power consumption, the RX 7800 XT consumes 50 W less than the RX 6800 XT, but the RTX 4070 uses even less power at 50 W. The cooling system of the RX 7800 XT includes two 85 mm fans and two 8-pin power connectors, delivering up to 300 W of power, as well as an HDMI 2.1 connector and three DisplayPort outputs.

However, there are already many models on the market with their own designs and cooling systems from AMD partners, including overclocked options.

Architecture Features

The characteristics of the RDNA3 graphics architecture were thoroughly discussed in the review of the Radeon RX 7900 XT video card (link to review), so in this review we will only focus on the key features and changes that this architecture makes compared to the previous version.

The RDNA3 architecture is an evolution of the previous version with a number of significant improvements:

  1. Dual Instruction Parallel Compute Units: RDNA3 introduces units that can process two instructions simultaneously, improving overall computation efficiency.
  2. Optimizations for more efficient use of resources: The new architecture offers improved use of available resources, which improves productivity and efficiency.
  3. Support for new math formats: RDNA3 supports FP16, BF16 and INT8 formats, providing a significant performance increase over conventional FP32/FP16 operations.
  4. Improved ray tracing engines: To improve performance in this area, the hardware blocks have been improved to make ray tracing more efficient.
  5. Artificial Intelligence Operations Accelerators: RDNA3 now has units specialized for performing matrix calculations, which significantly improves performance in artificial intelligence tasks.

These changes make the RDNA3 architecture more efficient and powerful than previous versions. Improvements in ray tracing, new math formats, and AI accelerators have made the architecture promising for solving AI problems, which could be an important factor in FSR 3's scaling technology.

The Navi 32 GPU is an evolution of the chiplet design similar to Navi 31 in high-end graphics cards, but scaled up for more affordable devices. Unlike the monolithic Navi 33 die used in the Radeon RX 7600, most solutions of this generation are based on chiplets.

The structure of Navi 31 and Navi 32 includes a central chiplet (GCD) and several chiplets with memory controllers and cache (MCD). For example, Navi 31 contains six MCDs, while Navi 32 has only four for the Radeon RX 7800 XT and three for the Radeon RX 7700 XT.

The MCD chiplets feature 16MB of Infinity Cache, giving the Navi 32 a total cache of 64MB and a 256-bit memory bus. The main GCD compute die uses the EUV 5nm process technology, while the MCDs are manufactured using DUV 6nm, which accounts for the difference in overall area between the Navi 31 and Navi 32.

Navi 32, used in the Radeon RX 7800 XT, has 60 RDNA3 compute units, 3840 stream processors, 240 TMUs, 96 ROP units, 60 RT units and 120 AI accelerators. This GPU also supports AV1 video encoding and decoding at 8K/60 FPS and provides high-bandwidth GDDR6 memory.

The improved subsystem for outputting information to displays and the capabilities for encoding and decoding video formats in Navi 32 are at the level of older solutions, including support for AV1 and SmartAccess Video.

The technological capabilities of the new Radiance Display Engine, including support for DisplayPort 2.1 with increased bandwidth, allow high resolutions and refresh rates without loss of video stream quality.

FidelityFX Super Resolution 3 (FSR 3) technology

FidelityFX Super Resolution 3 (FSR 3) is AMD's answer to Nvidia's DLSS 3 feature. Both technologies work on the principle of generating additional frames based on information about the movement of objects in successive frames, creating intermediate images. This method has limitations when carefully analyzing each frame or pixel, but in real-time games the effects of these artifacts are almost negligible, especially at higher frame rates.

FSR 3 allows you to double frames with virtually no additional performance impact, making it possible to enable higher graphics settings or increase resolution, especially in ray-traced games.

However, for now, the number of games that support FSR 3 is limited, as is Nvidia's DLSS 3. Despite this, the release of FSR 3 overall improves the value of Radeon RX 7800 XT graphics cards and reduces the gap between AMD and its competitors. It should be noted that Nvidia is now ahead of AMD by introducing DLSS 3.5 with improved ray tracing noise reduction, but this feature is only supported in a small number of games for now.

One of the downsides of FSR 3 and DLSS 3 technology is the added latency, but to combat this, both companies offer technologies such as AMD's Radeon AntiLag+ and Nvidia Reflex.

FSR 3 runs on all modern Radeon GPUs starting with the RX 5700, while DLSS 3 requires the latest GeForce RTX 40 series graphics cards with matching hardware.

AMD is actively pushing FSR 3 support into games and is working on integrating this feature at the driver level, which will improve frame rates in games even if they don't natively support FSR. Hypr-RX is an AMD driver feature that automatically applies various technologies (Radeon Boost, Radeon AntiLag+ and Radeon Super Resolution) to improve performance in DirectX 11 and DirectX 12-compatible games.

Theoretical performance evaluation

Let's look at the characteristics of a pair of recently released current generation AMD video cards based on two modifications of the Navi 32 GPU, as well as the corresponding pair of video cards from the previous family: RX 6800 XT and RX 6700 XT. Using this table, you can estimate how much more productive AMD's new mid-price models have become than their predecessors — however, so far only in terms of their theoretical characteristics.

Video card modelRX 7800XTRX 7700XTRX 6800XTRX 6700XT
GPU modelNavi 32XTNavi 32 XLNavi 21XTNavi 22XT
Transistors, billion28.128.126.817.2
Number of ALUs3840345646082560
Number of TMUs240216288160
Number of ROP blocks969612864
Game frequency, MHz2124217120152424
Turbo frequency, MHz2430254422502581
Computing power FP32, tflops37.335.220.713.2
Texturing speed, Gtex/s583550648413
Fill rate, Gpix/s233244288265
Video memory capacity, GB16121612
Infinity Cache volume, MB644812896
Video memory frequency, GHz19.5181616
Video memory bus, bit256192256192
Bandwidth, GB/s624432512384
Energy consumption, W263245300230
Recommended price, $499449649479

We have already talked about the comparison of the RX 7800 XT with the RX 7700 XT above — both video cards are far from the flagship solutions of the RX 7900 subfamily, the younger solution differs from the older one in fewer functional units and a quarter smaller amount of video memory and its bus, which is even more important. At the same time, the number of ROP blocks in the younger model is not reduced, and the frequency is even higher than that of the RX 7800 XT, so according to some theoretical parameters, the RX 7700 XT is even faster. But in practice, it is strongly limited by bandwidth, which makes the RX 7800 XT much more productive and attractive, especially considering the price of $500 compared to $450 for the younger model.

We have already said that the Radeon RX 7700 XT and RX 7800 XT are very similar to the RX 7900 XT and XTX pair in terms of pricing strategy — the low-end model in both pairs was located too close to the higher-level solution — the same RX 7900 XT was 10% cheaper XTX, but was more than 15% behind and offered less video memory, so it seemed less attractive. AMD later lowered the price, making the RX 7900 XT more affordable, the RX 7700 XT needs the same, but the RX 7800 XT is okay with the price, as it competes well with the more expensive GeForce RTX 4070.

Compared to the RX 6800 XT, the video card model we are considering today has fewer computing units, but they operate at an increased frequency, and even under certain conditions they can work twice as productive. As for the memory subsystem, the Infinity Cache has become half as large, but its throughput has increased. Given that the new GPU uses the same 256-bit memory bus, the chips operate at a significantly higher effective frequency of 19.5 GHz, and the increase in memory bandwidth relative to the RX 6800 XT corresponds to 22%. So if we compare the video card we are considering today with the Radeon RX 6800 XT from the previous generation, then at close prices the old model simply has no advantages left. The same can be said about the RX 6900 XT modification, which costs more and gives almost the same level of performance.

According to benchmark results from AMD, the RX 7800 XT, based on the full Navi 32 GPU, showed better performance than the previous generation RX 6800 XT and came close to the RX 6900 XT. This performance boost allows you to play at maximum settings at 2560x1440 resolution and even use ray tracing in many games. At the same time, the new product is available for $150 cheaper than the recommended price of the RX 6800 XT. Compared to previous models of Nvidia's competing graphics cards, the RX 7800 XT is more than 15% ahead of the RTX 3070 Ti and remains slightly behind the RTX 3080. However, when using ray tracing in games such as Cyberpunk 2077 or Minecraft RTX, even the RTX 4060 Ti can beat the RX 7800 XT, but in less demanding games like Control, Metro Exodus Enhanced and Spider-Man: Miles Morales, the new AMD graphics card beats the RTX 4060 Ti by about 25%.

Based on AMD's data, the RX 7800 XT often outperforms the GeForce RTX 4070, especially at 2560x1440 resolution, although the RTX 4070 has some advantages, such as more efficient ray tracing, better power efficiency, and support for DLSS 3.5 and FSR 3. However, overall The RX 7800 XT remains competitive with its performance and more affordable price. While the difference between the amount of video memory (16 GB for the RX 7800 XT and 12 GB for the RTX 4070) is not always a decisive factor, it can be important when using maximum settings and high resolutions.

The key competitor to the RX 7800 XT is the $600 RTX 4070, which is 20% more expensive, although it has its advantages. Overall, the RX 7800 XT is a good choice in its price range, and potential buyers should consider it along with other alternatives in this price segment.

Video card Gigabyte Radeon RX 7800 XT Gaming OC 16G 16 GB

Gigabyte Technology (Gigabyte brand) was founded in 1986 in the Republic of China (Taiwan). Headquarters in Taipei/Taiwan. Initially created as a group of developers and researchers. In 2004, the Gigabyte holding was formed on the basis of the company, which included Gigabyte Technology (development and production of video cards and motherboards for PCs); Gigabyte Communications (production of communicators and smartphones under the GSmart brand (since 2006).

Research object: commercially available 3D graphics accelerator (video card) Gigabyte Radeon RX 7800 XT Gaming OC 16G 16 GB 256-bit GDDR6

Gigabyte Radeon RX 7800 XT Gaming OC 16G 16 GB 256-bit GDDR6
ParameterMeaningNominal value (reference)
GPURadeon RX 7800 XT (Navi32)
InterfacePCI Express x16 4.0
GPU operating frequency (ROPs), MHz (BIOS OC/BIOS Silent)2565(Boost)—2874(Max)/
2565(Boost)—2910(Max)
2430(Boost)—2770(Max)
Memory operating frequency (physical (effective)), MHz2425 (19400)2425 (19400)
Memory bus width, bits256
Number of computational units in the GPU60
Number of operations (ALU/CUDA) in block64
Total number of ALU/CUDA blocks3840
Number of texturing units (BLF/TLF/ANIS)240
Number of rasterization units (ROP)96
Number of Ray Tracing blocks60
Number of tensor blocks
Dimensions, mm300×130×56300×130×50
Number of slots in the system unit occupied by a video card33
PCB colorblackblack
Peak power consumption in 3D, W (BIOS OC/BIOS Silent)254/260255
Power consumption in 2D mode, W2222
Energy consumption in sleep mode, W44
Noise level in 3D (maximum load), dBA (BIOS OC/BIOS Silent)36.8/30.837.8
Noise level in 2D (video viewing), dBA18.018.0
Noise level in 2D (idle), dBA18.018.0
Video outputs2×HDMI 2.1, 2×DisplayPort 2.11×HDMI 2.1, 3×DisplayPort 2.1
Multiprocessing supportNo
Maximum number of receivers/monitors for simultaneous image output44
Power: 8-pin connectors22
Power: 6-pin connectors00
Power: 16-pin connectors00
Weight of the card with delivery set (gross), kg1.771.5
Card weight (net), kg1.331.0
Maximum resolution/frequency, DisplayPort3840×2160@144 Hz, 7680×4320@60 Hz
Maximum resolution/frequency, HDMI3840×2160@144 Hz, 7680×4320@60 Hz

Memory

The card has 16 GB of GDDR6 SDRAM memory, located in 8 16 Gbit chips on the front side of the PCB. SKhynix memory chips (H56G42AS8DX014) are designed for a nominal operating frequency of 2500 (20000) MHz.

The card has 16 GB of GDDR6 SDRAM memory, located in 8 16 Gbit chips on the front side of the PCB. SKhynix memory chips (H56G42AS8DX014) are designed for a nominal operating frequency of 2500 (20000) MHz.

Card features and comparison with Gigabyte Radeon RX 7700 XT Gaming OC 12G (12 GB)

Gigabyte Radeon RX 7800 XT Gaming OC 16G (16 GB) front view
Gigabyte Radeon RX 7700 XT Gaming OC 12G (12 GB) front view
Gigabyte Radeon RX 7800 XT Gaming OC 16G (16 GB) back view
Gigabyte Radeon RX 7700 XT Gaming OC 12G (12 GB) back view

To confirm the hypothesis that the Radeon RX 7700 XT/7800 XT cards have identical printed circuit boards (PCBs) with differences only in the GPU and memory chipset, we conducted a comparison. The results showed that the PCBs of both cards are completely identical and are based on the AMD reference design. Even the power systems on them are exactly the same. The original PCB provides a 256-bit memory bus, which corresponds to the Radeon RX 7800 XT with 16 GB of memory. The only difference is that the card based on the RX 7700 XT lacks a pair of memory chips, which reduces the bus width to 192 bits and reduces the amount of video memory from 16 to 12 GB.

The core information indicates an encrypted digital marking, and the release date is the 28th week of 2023.

The total number of power phases on the Gigabyte card is 14 (10+3+1).

The core power circuit is marked in green, memory in red, and SoC power in blue.

The 10 phases of core power are controlled by an MP2856 PWM controller (Monolith Power Systems), designed for a maximum of 12 phases. It is located on the front of the card.

The 3 phases of power supply to the memory chips are managed by exactly the same PWM controller, located on the reverse side of the PCB.

And the third controller of the same type controls the power supply of the SoC (which is in demand in processors with chiplet architecture).

The core power converter uses DrMOS transistor assemblies — in this case MP87997 (Monolith Power Systems), rated for a maximum of 70 A.

There is no separate controller for monitoring (monitoring voltages and temperatures): all these functions are assigned to the GPU itself.

Backlight control, traditionally from Gigabyte, is assigned to the Holtek controller.

The card has two operating modes, they are embedded in two BIOS options, which are switched using a switch on the top end of the card: OC and Silent. The difference between the modes lies mainly in the fan speeds; the maximum operating frequencies of the GPU are also slightly different (and for some reason in favor of the Silent option).

Standard memory frequencies are equal to reference values. The boost core frequency in both BIOS versions is 5.5% higher than that of the reference card, the maximum core frequency in BIOS OC mode is 3.7% higher than the reference one (in BIOS Silent mode — by 4%). The real performance increase in games in BIOS OC mode was 3.5% relative to the reference accelerator.

The power consumption of the Gigabyte card in tests reached 254 W in BIOS OC mode and up to 260 W in BIOS Silent mode.

I tried manual overclocking with the consumption limit increased by 15% and received maximum frequencies of 3158/20768 MHz, however, even with such an increased consumption limit, the increase in games at 4K resolution averaged only 6.4% relative to the reference values. The card's power consumption increased to 331 W.

Power is supplied to the Gigabyte card via two conventional 8-pin PCIe 2.0 connectors.

The connector has an LED indicator that the power supply is connected correctly (if the connection is incorrect and there is no power, it lights up white).

The dimensions of this card are generally standard, the thickness is 5.6 cm, so it occupies 3 slots in the system unit.

It is also worth noting that the board has the usual 4 video outputs: two HDMI 2.1 and two DP 2.1.

Control of the card's operation is provided using the proprietary Gigabyte Control Center utility, we have written about it many times: this program provides control of fans, card operating frequencies and core voltage, and monitoring of the card's state (monitoring).

Heating and cooling

The main component of the cooling system is a powerful multi-section nickel-plated heatsink with heat pipes that effectively transfer heat from the GPU to the heatsink fins.

These heat pipes are directly attached to a large copper pad. The memory modules are also provided with cooling through the same large thermal conductive structure using thermal pads. In addition, additional pads are provided on the same radiator to cool the VRM power supplies.

The back plate, in addition to protecting the printed circuit board (PCB), enhances structural rigidity and improves GPU cooling (since it contacts the back of the PCB through a thermal pad).

A casing with three ∅100 mm fans with grooved blades is installed on top of the radiator

It is also worth noting the popular technology of rotating the central fan in the opposite direction, when a “gear effect” is achieved, reducing the turbulence of the air flow.

Fans stop when the video card is under low load if the GPU temperature drops below 50 degrees, and the memory chips heat up below 80 degrees. When the PC starts, the fans work, but after loading the video driver, the operating temperature is polled and they turn off. Below is a video on this topic.

Temperature monitoring

BIOS OS mode:

After continuous operation for two hours under maximum load, the maximum core temperature was 56 degrees Celsius, the highest temperature (hot spot) reached 78 degrees Celsius, and the memory chips reached 70 degrees Celsius. These results are an excellent indicator for video cards of this level. The card's power consumption reached 254 W. It is important to note that AMD drivers use a maximum hot spot temperature of up to 110 degrees Celsius as the limit to determine whether to reduce the GPU frequency.

The maximum heating is near the GPU and at the PCIe connector.

We filmed 9 minutes of heating and sped it up 50 times.

During manual overclocking (BIOS OC mode), when the consumption limit was raised by 15%, the heating parameters turned out to be slightly higher (62/93/72 °C), and the CO noise became more noticeable (fan speeds increased significantly).

BIOS Silent Mode:

In this case, the maximum core temperature reached 66 degrees, and the memory chips reached 78°C, which is also a very good result for video cards of this level. The card's power consumption was slightly higher (260 W) because the fans were running at a reduced speed. The maximum GPU hotspot temperature is 88 °C.

Noise

The noise measurement technique assumes that the room in which the measurements are taken is soundproofed and muffled to reduce the reverberation effect. The system unit in which the video card noise is tested does not have fans and is not a source of mechanical noise. The background noise level is 18 dBA, which is the room noise level and the meter level. Measurements are taken at a distance of 50 cm from the video card at the level of its cooling system.

Noise level assessment occurs in various modes:

  • Less than 20 dBA: relatively silent
  • 20 to 25 dBA: very quiet
  • 25 to 30 dBA: quiet
  • 30 to 35 dBA: Clearly audible
  • 35 to 40 dBA: loud but tolerable
  • Above 40 dBA: very loud

In idle mode in 2D and when watching movies, the noise level did not differ from the background — it was 18 dBA. The temperature in this mode did not exceed 29 °C, and the fans did not turn on.

At maximum load in 3D mode, temperatures reached 56/78/70 °C (core/hot spot/memory). At the same time, the fans operated at a speed of 1650 rpm, the noise level increased to 36.8 dBA, which can be described as loud, but still tolerable. Some ringing noise from the fans was also noted.

The noise spectrogram is below.

BIOS Silent Mode:

At maximum load in 3D, temperatures reached 66/88/78 °C (core/hot spot/memory). At the same time, the fans spun up to 1200 rpm, the noise increased only to 30.8 dBA: this is on the verge of a clearly audible level.

Backlight

The card only has the company logo on the top edge highlighted. Management is traditionally carried out through the Gigabyte Control Center utility.

Delivery and packaging

Apart from the traditional quick start guide, there is nothing else included in the package.

Testing: synthetic tests

Of course, here is the revised text, preserving its essence:

«We have tested the new AMD graphics card model using standard frequencies in our synthetic test suite. The test suite is constantly updated with new tests added and outdated ones removed. While increasing the number of calculation examples presents certain challenges, we are always striving to expand and improve our test suite. If If you have suggestions for improving the test methodology, do not hesitate to write them in the comments to the article or send them to us.

We've introduced several new benchmarks to evaluate the performance of ray tracing and resolution scaling and performance enhancement technologies such as DLSS, FSR, and XeSS. As semi-synthetic tests, we also used a set of subtests from the popular 3DMark package: Time Spy, Port Royal, DX Raytracing, Speed Way, etc. However, we removed sample DirectX 11 and 12 applications from various SDKs, as they recently gave incorrect results.

We conducted synthetic tests on the following video cards:

  • Radeon RX 7800 XT with standard parameters (RX 7800 XT)
  • Radeon RX 7700 XT with standard parameters (RX 7700 XT)
  • Radeon RX 6800 XT with standard parameters (RX 6800 XT)
  • GeForce RTX 4070 with standard parameters (RTX 4070)
  • GeForce RTX 4060 Ti with standard parameters (RTX 4060 Ti)

To analyze the performance of the Radeon RX 7800 XT video card, we used the Radeon model from the previous generation based on the RDNA2 architecture — RX 6800 XT, since it is a complete analogue of the new product in terms of market positioning. We also compared the new product with the RX 7700 XT, a junior model based on the same Navi 32 GPU, to assess how much it is inferior to today's hero.

In addition, we chose the GeForce RTX 4070 as a direct competitor for the new product, although it costs 20% more. For a more complete picture, we also included the RTX 4060 Ti, which is already noticeably cheaper and does not directly compete with the model in question. This comparison with two Nvidia rivals at once will help us understand how good the RX 7800 XT is compared to similarly priced competing models.»

3D Mark Vantage tests

Feature Test 1: Texture Fill

The first test measures the performance of texture fetch blocks. This involves filling a rectangle with values read from a small texture using multiple texture coordinates that change every frame.

Benchmark tests, especially texture tests provided by Futuremark, generally show that AMD and Nvidia graphics cards perform well. These tests usually show results close to theoretical parameters, but sometimes slight discrepancies occur, especially for certain GPUs. Based on the texturing test results, the performance of the full version of the Navi 32 GPU was lower than the previous generation RX 6800 XT, which may be a little surprising.

A comparison of the new AMD video card with competitors from Nvidia showed that Radeon is inferior, although previous generations of their GPUs performed well in this test. The effective texturing speed of the new Radeon family has decreased slightly, which we already noticed in our tests of the RX 7000. This was reflected in the performance of mid-priced models based on the RDNA3 architecture, with the older one showing results lower than the competing RTX 4070 graphics card. However, it is worth remembering that the RTX The 4070 is the more expensive option.

Feature Test 2: Color Fill

As for Feature Test 2: Color Fill, this test evaluates fill speed. It uses a simple, non-performance-limiting pixel shader to write an interpolated color value to an off-screen buffer using alpha blending. The test uses the 16-bit FP16 off-screen buffer, often used in HDR-rendered games, making this test a relevant and modern benchmark for benchmarking.

The results of the second subtest of 3DMark Vantage show the performance of ROP units without taking into account the amount of video memory bandwidth. This test specifically measures the performance of the ROP subsystem, where the impact of bandwidth usually does not play a clear role. The Radeon RX 7800 XT and RX 7700 XT video cards have almost the same speed of the ROP subsystem, and the younger model is even a little faster (theoretically). But unfortunately, both cards are again inferior to similar models of the previous generation, such as the RX 6800 XT.

However, both GeForce RTX 40 graphics cards presented in the comparison showed noticeably lower performance, so the RX 7800 XT outperformed both of them without much effort. The GeForce RTX 4060 Ti is almost half behind, and the more expensive RTX 4070 is significantly inferior to the new AMD model. However, Nvidia graphics cards have always been less efficient in terms of peak fill rate, so these results in this test are not a surprise.

  Feature Test 3: Parallax Occlusion Mapping

As for Feature Test 3: Parallax Occlusion Mapping, this is one of the most interesting feature tests, since a similar technique has been used in games for a long time. This test draws a single quadrilateral using the Parallax Occlusion Mapping technique to simulate complex geometry. Resource-intensive ray tracing and high-resolution depth map operations are used. This test is challenging for a pixel shader GPU because it involves a lot of ray tracing texture sampling, dynamic branching, and complex Strauss lighting calculations.

The results of this test from the 3DMark Vantage package depend not only on the speed of mathematical calculations, the efficiency of branch execution or the speed of texture fetches, but also on several parameters simultaneously. To achieve high speed in this task, the correct balance of the GPU is important, as well as the efficiency of executing complex shaders. This test is useful because its results often correlate with the results obtained in gaming tests.

In this test, the new model of the Radeon RX 7800 XT video card showed a very good result, outperforming the RX 6800 XT from the previous generation, a video card of the same price positioning. The lower-end model of the current family lagged behind about as expected, given the importance of memory bandwidth in this test. It’s also good that the main (albeit not direct) competitor, the GeForce RTX 4070 video card, was also left behind. Despite the slight difference in performance, AMD's new product still outperformed this Nvidia video card.

  Feature Test 4: GPU Cloth

As for Feature Test 4: GPU Cloth, it is interesting because it simulates physical interactions (cloth simulation) using the GPU. The test uses vertex simulation, a combination of vertex and geometry shaders, as well as multiple passes and stream out to transfer vertices between simulation passes. This test evaluates vertex and geometry shader execution performance, as well as data streaming speed.

The rendering speed in this test should also depend on several parameters at once, and the main influencing factors should be geometry processing performance and the efficiency of geometry shaders. The strengths of Nvidia chips should have emerged, but we have been getting clearly incorrect results in this test for a long time, so there is simply no point in taking into account the results of all GeForce video cards here, they are incorrect.

Over time, Radeon video cards also found themselves in exactly the same situation — newer AMD drivers show poor results, and both new models of the Radeon RX 7000 family, created on the basis of the Navi 32 graphics processor, were inferior to the old RDNA2 architecture video card. So almost all the results in this test are incorrect and not explained by theory — the issue is clearly in the drivers, which no one has optimized for such an ancient test package for a long time.

Feature Test 5: GPU Particles

A test of physical simulation of effects based on particle systems calculated using a graphics processor. A vertex simulation is used, where each vertex represents a single particle. Stream out is used for the same purpose as in the previous test. Several hundred thousand particles are calculated, all are animated separately, and their collisions with the height map are also calculated. Particles are rendered using a geometry shader that creates four vertices from each point to form the particle. Most of all, shader units are loaded with vertex calculations; stream out is also tested.

Here we see almost the same thing — and in the second geometric test from 3DMark Vantage, the results of the new product were far from theoretical, although they are already a little closer to the truth than in the previous subtest of the same benchmark. If we consider the results to be correct, then the video card in question was inferior to everyone except its younger sister — they are close, since the frequencies are approximately equal and the number of blocks is not much different, and bandwidth does not affect the result. What’s much sadder is that the RX 7800 XT lost to both of the competitor’s video cards presented in the comparison, with the RTX 4070 being one and a half times faster, which can only be explained by poor driver optimization.

Feature Test 6: Perlin Noise

Vantage's latest feature test is a math-intensive GPU test that calculates several octaves of Perlin noise in a pixel shader. Each color channel uses its own noise function to put more stress on the video chip. Perlin noise is a standard algorithm often used in procedural texturing and uses a lot of math.

In this mathematical test, the performance of all solutions, although not entirely consistent with theory, is usually close to the peak performance of video chips in extreme tasks. The test uses floating point operations, and the new Ada Lovelace and RDNA3 architectures should have revealed some of their unique capabilities by running the corresponding commands twice, but this test is already quite outdated and is not fully capable of showing the new capabilities of modern GPUs, judging by the comparative results.

The eldest model of a pair of mid-price video cards of the Radeon RX 7000 family showed the expected result against the background of its younger sister, significantly ahead of it — the RX 7700 XT is still very lacking in bandwidth. As for the RX 6800 XT of a similar positioning from the previous generation, the new product is noticeably faster, so the RDNA3 optimizations worked here, the new product is clearly better than the previous generation solution. The RX 7800 XT also outperformed its main rival, the RTX 4070, which is also more expensive. But let's see what happens in more modern synthetic tests that use increased GPU load.

Direct 3D 12 tests

We decided to remove examples from Microsoft's DirectX SDK and from AMD's SDK that use the Direct3D12 graphics API from our tests, since they have long shown incorrect results in most cases. And the well-known Time Spy benchmark from 3DMark remains in this section as the only computing test with Direct3D12 support. In it, we are interested not only in the general comparison of GPUs in terms of power, but also in the difference in performance with the asynchronous computing capabilities that appeared in DirectX 12 enabled and disabled. To be sure, we tested the video cards in two graphics tests at once.

In this test, all Radeons always look clearly better than the GeForces competing with them at the same price, this must be taken into account. This time, too, the performance of the Radeon RX 7800 XT video card being reviewed today was clearly superior not only to a competitor’s solution that was close in price, but also to a more expensive model — the GeForce RTX 4070, which turned out to be almost equal to the younger RX 7700 XT, especially in the second test.

And compared to the only previous generation Radeon video card presented today, the new model also turned out to be noticeably more powerful — the RX 6800 XT was left behind. This suggests that the mid-price video card we are considering will also be quite good in games — without taking into account ray tracing, which we will now consider separately.

Ray tracing tests

One of the first tests of ray tracing performance is the Port Royal benchmark from the creators of the famous 3DMark series tests. This test works on all GPUs that support the DirectX Raytracing API. We tested several video cards at a resolution of 2560x1440 at various settings, when reflections are calculated using ray tracing in two modes, as well as the traditional rasterization method.

Subtests related to ray tracing through the DirectX Raytracing API are usually new rendering technologies that use hardware acceleration to render reflections and shadows. These tests may demonstrate differences in ray tracing support between graphics card manufacturers such as AMD and Nvidia.

When it comes to the results, they often show which graphics cards handle ray tracing scenes more efficiently. For example, the DXR test in 3DMark, which does not use rasterization but only ray tracing, makes it possible to evaluate GPU performance specifically in the area of hardware tracing acceleration. This can be important for assessing a graphics card's ability to handle new display technologies and for comparing performance between different models.

Your posts compare different graphics cards in these tests and show which models may be more efficient under certain ray tracing conditions. Such tests allow us to evaluate how much hardware acceleration of ray tracing affects the performance of video cards from different manufacturers and generations.

In this test it is already clearer that RDNA3 works more efficiently when tracing rays compared to RDNA2, since even the RX 7700 XT turned out to be faster than the RX 6800 XT, although as expected it lagged behind the RX 7800 XT. Today’s new product could not beat the nominally competing model of the Ada Lovelace family — the RTX 4070 is almost one and a half times faster, but this is a purely synthetic test, in games this will only happen in Portal RTX, Quake II RTX and the special mode of Cyberpunk 2077.

In most games using ray tracing, the load on the RT units is noticeably lower, and the Radeon RX 7800 XT will not be so sad. But in this test, Nvidia video cards continue to have a clear advantage. Nvidia's dedicated RT cores do most of the work and are more versatile; they do not lose performance when tracing is enabled as much as the competitor's Ray Accelerator cores + regular SIMD cores.

With the release of new generations of Nvidia and AMD graphics processors last year, another test with a fairly serious load specifically on ray tracing was added to the 3DMark package — Speed Way. In terms of its load on various GPU blocks, it seems more similar to common gaming projects that actively use ray tracing, and therefore is very interesting for us.

The comment mentions the performance of various video cards in the Boundary benchmark, which is a Chinese game project using DXR and DLSS. This benchmark puts a heavy load on the GPU, using ray tracing for complex reflections, soft shadows and global illumination.

AMD graphics cards do not support DLSS technology, which is used in this test, so test results may be limited by the capabilities of AMD as opposed to NVIDIA, which supports this technology. In addition, in this case, it may be difficult to adequately compare the performance of video cards between AMD and NVIDIA due to the difference in the technologies used and their support in the test.

However, the mentioned results show that the Radeon RX 7800 XT delivered comparable performance to the previous top of the line RX 6950 XT in the Boundary test. This could indicate improvements in performance and efficiency of the new RDNA3 architecture, allowing the newer graphics card to perform on par with AMD's previous top-end model despite differences in the technologies used.

Without resolution scaling technologies, even in Full HD resolution, only sufficiently powerful video cards can work acceptably, which includes the model we are considering today — it showed a frame rate much higher than the minimum limit of playability (we are not talking about 4K resolution, which even the RX 7900 cannot cope with XT). The new model of the RX 7800 XT video card, although it lags behind both GeForce video cards presented in the comparison, including the RTX 4060 Ti, which costs less, is at the same time noticeably faster than its direct predecessor, the RX 6800 XT. AMD video cards are still far from even GeForce RTX 30 and Intel solutions, not to mention comparison with video cards of the RTX 40 family, but compared to their previous generation, the result of the new product turned out to be quite good.

Computational tests

We continue to search for benchmarks that use OpenCL for current computing tasks to include in our synthetic test suite. For now, this section remains a rather old and not very well optimized ray tracing test (not hardware) — LuxMark 3.1. This cross-platform test is based on LuxRender and uses OpenCL.

The Radeon RX 7800 XT video card, created on the full version of the Navi 32 graphics processor of the RDNA3 architecture, showed a weak result compared to the RX 6800 XT from the previous generation — the advantages of the new architecture do not work in this test, and the new product was inferior to the previous generation GPU in two out of three tests. But in the most difficult test it turned out to be noticeably faster. The younger RX 7700 XT is noticeably slower in the same tests, which means it is severely limited by memory bandwidth. As for the conditional price competitor, the result of the latest AMD video card in all subtests turned out to be lower than that of the RTX 4070. However, this solution is sold at a higher price and is not a direct competitor to the new product.

Unfortunately, another test of GPU ray tracing performance without hardware acceleration, V-Ray Benchmark, does not work on new AMD graphics cards. But most recently, Maxon released Cinebench 2024 — a new version of the popular 3D rendering benchmark, which allows you to evaluate the hardware capabilities of the processor and video card by rendering a photorealistic scene. Cinebench 2024 is based on the 3D graphics and animation program Cinema 4D and its Redshift rendering engine. It is important that the same algorithms and scenes are used for testing CPU and GPU, and their results can be compared.

First, let's look only at AMD video cards. Today's new Radeon RX 7800 XT is very much ahead of its younger sister on the same graphics processor, but greatly reduced in memory and cache, and turns out to be very fast compared to the RX 6800 XT from the previous generation. True, if we add both GeForce video cards to the comparison, their advantage is immediately impressive — even the RTX 4060 Ti turned out to be faster than the RX 7800 XT in this test, and the RTX 4070 showed almost twice the result. It seems that doubling the instruction rate of the new RDNA3 graphics architecture did not work in this test, and the overall performance did not increase very much, and with a high peak theoretical performance, the computing speed of the new architecture solutions in real-world tasks is not much different from the RX 6000.

DLSS/XeSS/FSR technology tests

XeSS (Xe Super Sampling) is a performance enhancement technology that uses lower resolution rendering techniques and then upscales the image to a higher resolution. This method, similar to NVIDIA's DLSS 2.0, was developed by Intel. XeSS also includes the use of artificial intelligence to improve image quality and restore detail in the frame.

It is important to note that XeSS, unlike DLSS, provides the ability to be used on all modern GPUs, including not only video cards from a particular developer, but also devices from other manufacturers, although its effectiveness on Intel’s own solutions may be more noticeable. To test the performance of this technology, a specialized benchmark from the 3DMark package was used.

XeSS technology, despite its versatility, provides a significant increase in frame rate in the benchmark, exceeding twice or more. It has certain advantages and disadvantages compared to DLSS and FSR. For example, Nvidia's DLSS provides more advanced capabilities, but is limited only to that company's products. FSR, on the other hand, is a universal method, but its functionality and quality do not reach the level of DLSS, and it also does not yet make full use of specialized blocks to speed up the process. XeSS is versatile, but is currently inferior to DLSS in both quality and functionality.

The performance of the new Radeon RX 7800 XT in this test is comparable to previous generation video cards, although results for the RX 6800 XT are not available. Overall, the RX 7800 XT is close in performance to the RX 7700 XT and even the RX 7900 XT, which is a bit surprising. Among Nvidia video cards, the results are also close to AMD solutions, with the exception of Intel, whose solutions are slightly more efficient thanks to specialized units. The RX 7800 XT in this test turned out to be more productive than the RTX 4060 Ti, but inferior to the RTX 4070.

AMD's FSR 2.0 technology, which is one of the representatives of rendering scaling, was the latest to appear on the list of 3DMark subtests. The scenes of different upscaling technologies differ, making direct comparisons difficult. Performance gains need to be taken into account as well as actual rendering resolution and quality differences for a full comparison.

Since FSR is another universal technology, it works approximately the same on different GPUs, and we did not find any special revelations in the FSR 2.0 tests — all GPUs were positioned approximately as we expected. The new Radeon RX 7800 XT model is this time much faster than the RTX 4060 Ti in all subtests, as it should be in theory, but the difference in speed with the RTX 4070 is no longer in favor of AMD’s solution, especially in modes with FSR enabled. At the same time, the new product also outperformed the RX 6800 (not XT) in this test, which is also quite expected from a theoretical point of view. The difference between the RX 7700 XT and RX 7800 XT is small, which is also understandable.

In general, we didn’t see anything particularly interesting or unusual in the tests of scaling technologies, but the new product definitely showed good results compared to its competitors. Let's move on to testing the new AMD video card in real gaming tests to confirm or refute all our conclusions drawn from synthetic tests.

Testing: gaming tests

Test bench configuration

  • Computer based on Intel Core i9-13900K processor (Socket LGA1700):
  • Platform:
  • Intel Core i9-13900K processor (overclocked to 5.4 GHz on all cores);
  • ZhSO Cougar Helor 360;
  • Asus ROG Strix Z790-A Gaming WiFi D4 motherboard based on the Intel Z790 chipset;
  • RAM TeamGroup Xtreem ARGB White (TF13D416G5333HC22ADC01, CL22-32-32-52) 32 GB (2×16) DDR4 5333 MHz;
  • SSD Intel 760p NVMe 1 TB PCIe;
  • SSD Intel 860p NVMe 2 TB PCIe;
  • ThermalTake Toughpower GF3 1000W power supply;
  • Thermaltake Level20 XT case;
  • operating system Windows 11 Pro 64-bit;
  • TV LG 55Nano956 (55″ 8K HDR, HDMI 2.1);
  • AMD drivers version 23.11.1/2;
  • Nvidia drivers version 546.01;
  • Intel drivers version 101.4925;
  • VSync is disabled.

The Radeon RX 7800 XT provides excellent comfort in games without ray tracing at resolutions from 1080p to 2160p (4K) inclusive. It is ahead of its predecessor, the Radeon RX 6800 XT, and could well replace both it and the Radeon RX 6900 XT. The new accelerator competes with the GeForce RTX 3080 Ti, and also, although less successfully, with the GeForce RTX 4070 Ti, being significantly faster than the GeForce RTX 4070.

Games using ray tracing and DLSS/FSR/XeSS:

In games with RT and FSR/XeSS, the Radeon RX 7800 XT will provide full comfort only at 1080p and 1440p resolutions, and quite good at 4K. At the same time, the Radeon RX 7800 XT outperforms its competitor from the previous generation Radeon RX 6900 XT, that is, it confirms the reputation of its substitute. But in terms of competition with Nvidia cards, it is only strong in comparison with the GeForce RTX 3080 and GeForce RTX 4060 Ti, even the GeForce RTX 4070 is significantly more productive due to support for DLSS 3. Of course, when games with FSR 3 support start coming out, the situation will be the same for Radeon RX The 7800 XT (and all modern AMD cards) could improve.

Conclusions and comparison of energy efficiency

The AMD Radeon RX 7800 XT (16GB) is a natural extension of the Radeon RX 7000 line of graphics cards. This new graphics card takes the position of its predecessor Radeon RX 6800 XT in the same price segment, with only a slight performance gap. In terms of performance, the Radeon RX 7700 XT made a stronger impression as it outperformed the Radeon RX 6700 XT and even caught up with the Radeon RX 6800.

In games that lack ray tracing (RT) technology, the Radeon RX 7800 XT sits between the GeForce RTX 4070 Ti and the GeForce RTX 4070 (closer to the latter). However, in games using RT and upscaling technologies, it shows a noticeable drop in performance compared to the GeForce RTX 4070, despite having 16 GB of memory (as opposed to 12 GB for the GeForce RTX 4070). It is worth noting that the Radeon RX 7800 XT has only recently appeared on the market, and as often happens when new devices are released, the initial price for it tends to be inflated.

In the fight for user attention, the fact that prices for long-released Nvidia video cards from the GeForce RTX 40 family have dropped significantly, and prices for the new GeForce RTX 40 have also dropped, also plays against AMD. This makes things tricky for the Radeon RX 7800 XT. However, stocks of both the Radeon RX 6000 and GeForce RTX 30 are quickly drying up, and as early as 2024 they will likely represent only formal competitors.

At the time of preparing our review, sales of the Radeon RX 7800 XT had already begun; we draw conclusions based on real price tags from marketplaces. As we noted above, these prices are still a little high. We expect that the introduction of FSR 3 into games will help change the balance of power.

If you look at energy efficiency, the GeForce RTX 4070 is again in the lead, and the Radeon RX 7800 XT is far from it.

The AMD Radeon RX 7800 XT, like other models in the line, is built on the modern RDNA3 architecture. The new GPU's compute units offer comparable functionality to older models, updated ray tracing cores, and the ability to accelerate artificial intelligence tasks. The video card based on the Navi 32 chip is organized according to the principle of a chiplet structure.

Significant improvements have been made to the output subsystem throughout the Radeon RX 7000 family. The new product, even as a mid-budget video card, offers DisplayPort 2.1 UHBR13.5 connectors, which leaves behind GeForce cards, including flagship models, which are limited by DisplayPort 1.4a capabilities. This allows you to transmit 4K resolution images at 240 Hz without streaming compression over a single cable. In addition, hardware capabilities for video encoding and decoding are similar to older models, including full support for the AV1 format.

The Gigabyte Radeon RX 7800 XT Gaming OC 16G (16 GB) card has standard dimensions and occupies three slots in the system unit. The installed cooler is quite effective, although it can be noisy. Power consumption can reach up to 260W and uses two standard 8-pin connectors for power.

The manufacturer provides a 4-year warranty on this card with mandatory registration on the company website.

This hardware is ideal for gaming at 1080p and 1440p resolutions at maximum graphics settings and ray tracing, and can also successfully run some games at 4K resolutions at the same settings. Games without ray tracing technology can provide great gaming fun at all three of these resolutions.