We’re already used to the fact that the next generation of discrete graphics brings significant acceleration, improved functionality and an additional margin of performance for the most demanding games. The NVIDIA company traditionally starts updating the graphics line with top solutions, so it is not surprising that the transition to devices with GPUs based on the ADA Lovelace architecture also starts with older models. Following the flagship GeForce RTX 4090 24 GB, the developers offered a slightly more affordable GeForce RTX 4080 16 GB. Today we are reviewing an extraordinary representative of the new series – MSI GeForce RTX 4080 16GB SUPRIM X.
NVIDIA GeForce RTX 4080
Unlike the top model of the RTX 4000 series with AD102 chips, GeForce RTX 4080 adapters are equipped with AD103 graphics processors, which also have the ADA Lovelace architecture and are manufactured by TSMC using the most advanced 4-nanometer process. At the same time, the GPU crystal has an area of 379 mm² and contains 45.9 billion transistors (AD102 – 608 mm²; 76.3 billion). For comparison, the 8-nanometer chip of the previous generation GA102, used in the GeForce RTX 3080, contains 28.3 billion transistors, having an area of 628 mm².
In the case of AD103-300, the processor has 9728 calculators, 304 texture blocks and 112 rasterizers. In addition, the GPU includes 304 4th generation tensor cores for matrix calculations and 76 3rd generation RT units for ray tracing processing. One of the features of the new architecture is a capacious cache buffer of the second level. In this case, L2 is increased to 64 MB.
The transition to a new GPU manufacturing process allowed developers not only to radically increase the density of placement of elements, but also to significantly increase the operating frequencies of the graphics processor. Thus, the base frequency of AD103-300 is as much as 2205 MHz, and the average value of dynamic acceleration is 2505 MHz. Again, feel the difference with previous GPU benchmarks for the RTX 3080 – 1440/1710 MHz. The theoretical computing power of the GeForce RTX 4080 is 48.74 TFLOPS, while the RTX 3080 has 29.77 TFLOPS. At the same time, it should be noted that 82.58 TFLOPS are claimed for the RTX 4090.
|GeForce RTX 4090
|GeForce RTX 4080
|GeForce RTX 3080 Ti
|GeForce RTX 3080 12 GB
|GeForce RTX 3080 10 GB
|512 (4 generation)
|304 (4 generation)
|320 (3 generation)
|280 (3 generation)
|272 (3 generation)
|128 (3 generation)
|76 (3 generation)
|80 (2 generation)
|70 (2 generation)
|68 (2 generation)
|GPU clock speeds
|24 GB GDDR6X
|16 GB GDDR6X
|12 GB GDDR6X
|12 GB GDDR6X
|10 GB GDDR6X
|Effective memory frequency
|21 000 MHz
|23 000 MHz
|19 000 MHz
|19 000 MHz
|19 000 MHz
All GeForce RTX 4080 graphics cards are equipped with 16 GB of memory. Initially, according to the plan of the developers, a version with 12 GB had to also appear, but after mass criticism, NVIDIA abandoned such intentions, planning to use a similar modification for another line of video cards. So, the GeForce RTX 4080 uses GDDR6X chips with an effective frequency of 22,400 MHz. The adapters received a 256-bit bus, so the total memory bandwidth is 716.8 GB/s. As you can see, there is virtually no progress in this component. The predecessor GeForce RTX 3080 even has a slightly higher memory bandwidth of 760 GB/s, which is achieved thanks to a 320-bit bus and 19,000 MHz GDDR6X chips. Moreover, the GeForce RTX 3080 12 GB modification presented at the beginning of 2022 even received a 384-bit bus, which made it possible to obtain a memory bandwidth of 912.4 GB/s, which is striking for this class.
However, in this case, it is difficult to compare the solutions of different generations, because more efficient data compression algorithms are usually used for the new ones, which reduces the dependence of the actual data transfer rate on the total memory bandwidth.
As for power consumption, the GeForce RTX 4080 is much more economical than the flagship – 320 W vs. 450 W. In fact, the declared TDP is identical to that of the predecessor GeForce RTX 3080 10 GB, while the novelty should consume less energy than the RTX 3080 modification with 12 GB of onboard memory (350 W).
NVIDIA has offered a GeForce RTX 4080 Founders Edition reference model, which is visually very similar to the reference version of the GeForce RTX 4090. The adapters use the same cooling system, while the PCB is expected to have some differences in layout. MSRP for the GeForce RTX 4080 16GB is $1,199 for the US market, excluding tax. The declared cost caused considerable complaints. Apparently, NVIDIA is trying to partially use the recent drastic increase in the prices of video cards to somewhat smooth out the feeling of “abnormality” of such a price tag for the RTX xx80 series adapter. We will remind that GeForce RTX 3080 10 GB was announced two years ago with a MSRP of $699, and almost a year ago the accelerated modification of GeForce RTX 3080 12 GB was estimated at $799. Usually, the one-and-a-half-fold increase in the price of adapters within the series raises many questions, even taking into account the general inflationary processes and the increase in the price of individual components. In addition, official price tags are initial values that do not limit partners at all. Modified versions of the GeForce RTX 4080 can cost significantly more than the basic models.
However, thinking about the pricing and comparing the indicators with those of the previous generation solutions, we can also mention the GeForce RTX 3080 Ti 12 GB, the recommended price of which at the start (May 2021) was also $1199.
MSI GeForce RTX 4080 16GB SUPRIM X
Simultaneously with the announcement of GeForce RTX 4080, MSI presented six models of the new series – two each in the SUPRIM, TRIO and VENTUS lines. Today we are reviewing the top modification MSI GeForce RTX 4080 16GB SUPRIM X.
The graphics card is delivered in a fairly large cardboard box, decorated in light colors with a large photo of the device on the front side of the package. Inside the box, the adapter is additionally fixed using a molded damper. An adapter from three 8-pin connectors to one 12VHPWR is supplied with the device for connecting additional power. Considering the weight of the graphics card, an additional metal support of variable length for fixation does not seem superfluous. But the fabric mouse pad with the MSI Suprim logo can already be considered a bonus for the owner of the MSI GeForce RTX 4080 16GB SUPRIM X.
The tuned graphics card received factory overclocking. The GPU frequency formula has been changed from the recommended 2205/2505 MHz to 2205/2625 MHz.
The MSI GeForce RTX 4080 16GB SUPRIM X catches the eye right after you take the graphics card out of the box. This is a rather massive graphics adapter, which is generally expected for a model of this level. The dimensions of the device are 336 x 142 x 78 mm. That is, we are talking about a three-slot model, and in fact – a four-slot one, which also weighs 2364 grams. This must be taken into account when designing the system, and also do not consider the complete additional support as an unnecessary accessory, especially if you plan to move the system unit.
By the corporate visual design with sharp lines, beveled edges and polygonal casing formation, it is easy to recognize that the adapter belongs to the SUPRIM line. The composite aluminum panel has a matte finish with directional sanding marks. An octagonal frame can be seen around the fans, which adds a geometric style. Additional accents are added by polished chamfers on the edges of the surfaces.
The MSI GeForce RTX 4080 16GB SUPRIM X uses the updated TRI-FROZR 3S cooling system, which has quite a few features. First of all, it is necessary to note the very large radiator unit. Although it is two-sectioned, the specific design placement of the heat pipes made it possible to further increase the total effective scattering area, and the set of plates as one single array. In addition, the radiator plates have a wave-shaped edge placed in antiphase (WAVE-CURVED 3.0), which theoretically further increases efficiency.
To accelerate the transfer of heat to remote areas of the radiator, the design of the cooler uses two 8-mm and six 6-mm heat pipes. But as the main heat sink, a large evaporation chamber is used, which covers not only the GPU, but also the memory chips.
The efficiency of the unit is also increased with the help of additional corner deflectors placed in certain areas of the plates to emphasize the air flows on the heat pipes.
Three 100 mm axial fans are used to blow the radiator unit. In this case, Torx Fan 5.0 fans with an interesting blade design are used. Each fan has nine blades, with additional arcs on the outer radius connecting every three blades. The blades themselves are turned at an angle of 22 degrees. According to the developers, this structural form allows you to maintain high static pressure even at low revolutions, and the specific triad combination increases the air flow compared to the classic configuration. Double ball bearings are used for fans to increase durability.
Of course, the Zero Frozr technology is implemented for the graphics card, which involves stopping the fans at low load and, accordingly, heating the GPU.
For general strengthening of the structure, as well as additional cooling of the elements, the graphics card is equipped with a cast frame-bracket.
It is interesting that the TRI-FROZR 3S cooler is used in the identical configuration for the top MSI GeForce RTX 4090 24GB SUPRIM X. And if it copes with a much more energy-intensive graphics card, then in the case of the RTX 4080 the efficiency of the cooler should be more than enough.
We did not have the opportunity to look under the radiator, so we will use the help of colleagues to quickly assess the layout of the printed circuit board. The PCBs of the models from partners have nothing to do with the reference GeForce RTX 4080 boards from NVIDIA, so every engineering school had an opportunity to show their abilities here.
In the electrical part, the MSI GeForce RTX 4080 16GB SUPRIM X is thoroughly charged. The GPU power unit includes 18 power phases based on OnSemi NCP302150 DrMOS elements with an operating current of up to 50 A. UPI uP9512R controller is provided for phase management. Powering the memory chips has a three-phase circuit with NCP3102150 DrMOS assemblies (50 A) and a uPI uP9529Q control controller.
Eight GDDR6X microcircuits are arranged in a U-shaped manner almost close to the textolite substrate of the GPU. In this case, Micron chips marked D8BZF (MT61K512M32KPA-24) are used, which according to the specification are designed for frequencies up to 1500 MHz (24,000 MHz effective). Among the technical features of the adapter, we also note the use of carbonyl chokes (HCI) and the presence of fuses in the power circuit.
To connect additional power, a progressive 12VHPWR connector is used, a possible increased heating of which recently attracted a lot of attention of owners and potential buyers of top GeForce RTX 4090 video cards. As for GeForce RTX 4080, no such cases were registered. Obviously, the much lower consumption of the younger model forgives even the cases of the connector not being firmly fixed. However, of course, you should not neglect the general recommendations of the developer and make sure that the connectors are in good contact. To connect the power, we used an additional complete adapter for three 8-pin connectors and this did not cause any difficulties.
The MSI GeForce RTX 4080 16GB SUPRIM X has two BIOS chips with different modes. The switch at the top edge of the PCB allows you to use either the standard mode (Gaming) or to switch to the Silent mode, which involves the algorithm of the cooling system with a slower fan speed.
The reverse side of the adapter is covered by an aluminum plate with a complex profile, which performs several functions at once. First and foremost, it definitely beautifies the graphics card, and also protects the elements from accidental mechanical damage. In addition, in the GPU area, heat-conducting pads are provided, which partially divert heat to the plate. The photo also shows an area with additional perforation. Through blowing allows to increase the overall efficiency of the cooler, also the holes prevent the formation of zones of increased heating behind the plate.
On the interface panel of the graphics card there are four video outputs – one HDMI 2.1a and three DisplayPort 1.4a. After the announcement of the new line of video cards, NVIDIA received many complaints that the developers did not foresee and implement support for ports with a higher bandwidth for connecting ultra-fast 4K panels and 8K screens with a refresh rate of more than 60 Hz. However, for 4K@144–165 MHz, the possibilities are quite sufficient, and high-speed displays are still artificial solutions.
A large part of the panel has perforation with large holes for the removal of heated air outside the system unit.
Traditionally, RGB lighting adds a few points to any gaming device, and video cards are no exception. Additional backlight with backlight of the logo and the name of the SUPRIM series are located on the upper panel of the adapter. In addition, corner segments of frosted glass are provided near each fan. In the matter of decorative lighting, the developers were guided by the principle of reasonable sufficiency. Here it is important to be able to restrain yourself, because it is very easy to turn appropriate and effective accents into bad taste. Subjectively, the designers managed to maintain a balance here.
Naturally, the graphics card supports Mystic Light, allowing you to synchronize the backlight with other components in the MSI Center application.
When there is no sensory load, the cooler fans stop, and the video card accordingly works absolutely silently. In this mode, the temperature of the GPU is kept at the level of 44C. At the same time, the memory chips heat up to 48C.
Automatic stop technology, similar to Zero Frozr, is a real find for lovers of silence. It’s always nice to note, especially when it comes to super-powered gaming models.
Under the game load in normal mode (Gaming), the graphics processor warmed up to 59C. The temperature at the hottest point (Hot Spot) was about 10C higher – about 70C. At the same time, the GDDR6X chips heated up to only 60C.
Under load, the cooler usually worked in active mode, however, in the conditions of the open stand, the speed of rotation of the fans did not exceed 1350 rpm. At the same time, the video card worked quite quietly.
As can be seen on the utility screen, the frequency of the graphics processor increased to 2880 MHz, however, the long-term constant values under load were somewhat more modest – 2650-2800 MHz.
After changing the BIOS switch to Silent mode, the algorithm of the cooling system becomes even more “softer”. In this case, the speed of rotation of the fans barely exceeded 1000 rpm. The temperatures of the GPU and memory chips rose a little – up to 65C and 68C, respectively. Given that the frequency characteristics remained unchanged, the performance also remained constant. The practical benefit of switching to Silent is not so much, however, for those who prefer silence, it may make sense. In our case, the CPU cooler worked louder, so the difference remained imperceptible.
Test bench configuration
- Processor: Intel Core i5-13600KF (6/12+8; 3.5/5.1 GHz + 2.6/3.9 GHz)
- Cooling system: MSI MPG CORELIQUID K360
- RAM: 2×16 GB DDR4-3600 Kingston FURY Beast DDR4 RGB (KF436C18BBAK2/32)
- Drive: MSI SPATIUM M480 PCIe 4.0 NVMe M.2 (2 TB)
As a test platform, we used a system with a Core i5-13600KF processor (6/12+8; 3.5/5.1 GHz; 2.6/3.9 GHz). With the appearance of the 13th generation Core chips, the capabilities of the top “six cores” from Intel have significantly increased, now they are quite appropriate even in the composition of progressive gaming configurations. Additional processors, increased operating frequencies and performance per clock (IPC) and still moderate prices allow us to talk about them as balanced options precisely in the context of powerful gaming PCs. Perhaps in the case of the GeForce RTX 4080/4090, a model with even higher peak frequencies would not hurt to reveal the potential in Full HD mode, but now we are still more interested in the capabilities of the video card in 4K. In this case, the Core i5-13600KF will be quite enough.
To evaluate the performance of the MSI GeForce RTX 4080 16GB SUPRIM X as a base model from the family of discrete graphics of the previous generation, we used the MSI GeForce RTX 3080 SUPRIM X 12G video card. The predecessor has many similarities in appearance, but the novelty is much larger. In addition, even at first glance, you can appreciate the design advantages of the RTX 4080 cooler.
It is important to note that the MSI GeForce RTX 3080 SUPRIM X 12G belongs to the GeForce RTX 3080 12GB family. That is, this is a version with an increased number of GPU computers, memory capacity, as well as bus bit rate (384 bits). In addition, we are talking about a tuned version with factory GPU overclocking – 1260/189 0 MHz instead of the recommended 1260/1710 MHz. That is, it has little in common with the base model GeForce RTX 3080 10 GB, and offers ~10% higher performance. That is, the performance of the MSI GeForce RTX 3080 SUPRIM X 12G is close to those of the GeForce RTX 3080 Ti 12 GB and the GeForce RTX 3090 24 GB within the margin of error. This should be taken into account when evaluating and comparing the obtained indicators.
We start the performance evaluation with the unchanged synthesis – 3DMark. To compare indicators, we chose three stages – Time Spy Extreme, Speed Way and Port Royal.
We run all tests first on a platform with MSI GeForce RTX 4080 16GB SUPRIM X and save screenshots with the results.
We repeat the sequence with the MSI GeForce RTX 3080 SUPRIM X 12G and begin to be surprised by the results. At the Time Spy Extreme stage, the RTX 4080 model has an advantage of 47.3% – 14,057 against 9,544 points. The difference in Speed Way is somewhat more modest, but also noticeable – 39%. The Port Royal test using ray tracing again increases the advantage of the novelty to 45%. The indicators are pleasantly surprising, especially considering that the confrontation is not with the basic RTX 3080 10 GB, but with the forced RTX 3080 12 GB, which is much closer in terms of indicators to the RTX 3090/3080 Ti. So, the RTX 3090 Ti still remains in the last generation lineup, but given the significant advantage of the MSI GeForce RTX 4080 16GB SUPRIM X, we can safely say that the RTX 4080 has a higher performance than any gaming video card on a GPU with the NVIDIA Ampere architecture.
However, let’s not hurry. Still, synthetics are mostly intended for comparing the number of “parrots” and evaluating the theoretical potential, while real games return to the mundane in practical application.
We tested ten quite resource-intensive projects using the mode with a resolution of 3840×2160 (4K) and settings with maximum image quality. The obtained results demonstrate a significant advantage of the MSI GeForce RTX 4080 16GB SUPRIM X over its predecessor. Although the difference in performance is slightly smaller than in the 3DMark tests, it is quite noticeable – 22-43%. On average, the advantage of the novelty was almost a third (32%).
Here, we emphasize once again that the comparison is not with the GeForce RTX 3080 10 GB, but with the forced GeForce RTX 3080 12 GB, which is much faster than the stock version and is almost not inferior to the GeForce RTX 3090/3080 Ti. It may also be appropriate to recall the recommended prices for these models – GeForce RTX 3080 Ti – $1199, GeForce RTX 3090 – $1999. This is in no way an exhortation to justify NVIDIA, but rather an attempt to understand the logic of pricing for RTX 4xxx graphics cards. There must be one. Although, unfortunately, no one canceled the insane thirst for profit.
The overall fps rates in games show that the time has come for 4K screens with higher refresh rates. It may be a coincidence, but the range of gaming monitors capable of providing 4K@144+ has expanded significantly recently.
Although GeForce RTX 4080 16 GB video cards are obviously primarily interesting to owners of screens with increased resolution, we also evaluated the capabilities of the new product in 1920×1080 mode. In Full HD, the performance of both video cards is expected to be much closer. On average, RTX 4080 16 GB was 15% more productive than RTX 3080 12 GB. But this is exactly the case when it is difficult to generalize. The fact is that in light projects the difference was only 2-8%, while in “heavy” games the advantage of a more powerful video card was much higher. For example, in Assassin’s Creed Valhalla, the MSI GeForce RTX 4080 16GB SUPRIM X video card had an advantage of 22%, in Dirt 2 – 26%, and in Forza Horizon 5 even around 30%. It would seem that there is already much more, but even in cases where we are talking about 150+ fps, there is still room for growth.
In addition, the Core i5-13600KF with a peak of 5.1 GHz could be a certain limiter here. Also, some undershooting of maximum frames/c in Full HD may have occurred due to the use of DDR4-3600 memory. In the presence of DDR5-6000+ with small delays in a number of projects, one could count on an additional +5% to productivity. But these are nuances that do not affect the general assessment of the state of affairs and the ratio of the strengths of both video cards.
To test the capabilities of graphics cards when ray tracing technology is activated, we used two projects Cyberpunk 2077 and Metro Exodus Enhanced Edition with maximum RT quality settings. In both cases, the advantage of the MSI GeForce RTX 4080 16GB SUPRIM X was about 45%, but as we can see, in 4K mode with similar settings, even the RTX 4080 requires additional acceleration in certain projects.
The performance situation improves significantly when DLSS is enabled. In Full HD mode, the advantage of the RTX 4080 is 8-30%, but in 4K the average number of fps is again 45% higher in the model on Ada Lovelace. Although the quantitative difference of tensor cores and RT-blocks in test video cards is relatively small – 304/76 vs. 280/70, new generation solutions provide greater productivity.
In addition, GeForce RTX 40xx graphics cards have another “secret weapon” in the form of support for Frame Generation and the DLSS 3.0 technology implemented on its basis. The updated version of Portal with RTX seems to have become a kind of demo of the capabilities of the new development. Unfortunately, Portal with RTX does not have a built-in benchmark, and third-party utilities were unable to determine the average FPS during the game, so to compare the performance of video cards in different modes, we had to use the least accurate indicator – the current FPS in the Steam overlay. It adequately represents the difference on the same bench when used in the same scene, but it is not worth comparing our figures with those obtained by other resources in their own reviews, even with the same test bench configuration.
Without deactivating DLSS, even in Full HD mode with high image quality on the RTX 3080 12 GB we have 45 fps, while in Ultra mode the figure drops to 31 fps. GeForce RTX 4080 16 GB in the same conditions already provided 75/59 frames/c, but these are definitely not the values you expect from an almost top video card in 1920×1080 mode. Activating DLSS 2.0 immediately improves the performance of both video cards by 70-90%. On the GeForce RTX 4080, it is possible to enable DLSS 3.0, after which we already get about 200 fps with high settings and 166 fps in Ultra quality mode.
In 4K mode without AI scaling, we have a slideshow on the RTX 3080 and a better, but generally insufficient, performance on the RTX 4080. The inclusion of DLSS 2.0 improves things radically, but on the RTX 3080, 12 GB is still too small for comfortable gaming. GeForce RTX 4080 has a 60-70% advantage, and the use of DLSS 3.0 generally increases the number of frames to almost 100 fps at high picture quality settings and up to 80 fps on Ultra. So think after that Portal with RTX is not a technological demo for graphics cards of the GeForce RTX 40xx series. It’s good that it’s available for free to owners of previous versions of Portal, otherwise the total number of satisfied players would be quite limited.
In general, DLSS 3.0 is gaining momentum and a more detailed analysis of the capabilities and impact on performance deserves a separate article. Support for AI-scaling of the new generation has already been announced for almost two dozen projects, including the heavyweight Cyberpunk 2077, which will certainly not be hindered by additional acceleration.
According to the results of the tests, the MSI GeForce RTX 4080 16GB SUPRIM X was on average 32% faster than the MSI GeForce RTX 3080 SUPRIM X 12G, which is close to the GeForce RTX 3090/3080 Ti. The advantage when using modes with ray tracing and DLSS is even more convincing – 40-50%.
Obviously, the GeForce RTX 4080 is faster than any other model next generation based on the Ampere architecture. However, for the GeForce RTX 4080 16 GB, new graphics cards with AMD chips – Radeon RX 7900 XTX 24 GB – are a new challenge. Judging by the already available reviews, graphics cards based on Navi 31 with RDNA 3 architecture can be a real competitor to RTX 4080. On average, the model is even slightly more productive in modes with traditional rasterization, however, it is inferior to the latter by 15-20% when tracing is activated. Either way, it will be a very interesting and spectacular showdown.
AMD traditionally has a difficult start with new graphics, but later they gain momentum and make up for lost time by optimizing drivers and software. The battle between GeForce RTX 4080 and Radeon RX 7900 XTX has even more intrigue, given that the AMD model has a lower MSRP of $999 vs. $1199. However, the NVIDIA version has its advantages in the form of faster RT processing and DLSS 3.0 support with an already emerging ecosystem. And the plan of energy consumption has a certain advantage due to the novelty from the Californians.
It is great that AMD managed to offer a competitive solution in this segment, and to present it not much later than the main opponent. All users will benefit from this, regardless of their own preferences.
A separate item that interested us while testing the capabilities of the new video card is power consumption. The GeForce RTX 4080 16GB claims a total TBP of 320W. This is radically less value than the flagship model GeForce RTX 4090, which received a package of 450 W. The declared consumption is at the level of the base GeForce RTX 3080 10 GB, while the GeForce RTX 3080 12 GB this figure is already at the level of 350 W.
It is interesting that in the case of the MSI GeForce RTX 4080 16GB SUPRIM X, despite the fact that it is a forced version, the manufacturer leaves the 320 W declared for reference in the specification. Usually, tuned models have slightly higher indicators. However, what do we have in practice?
According to GPU-Z monitoring data, the power consumption of the MSI GeForce RTX 4080 16GB SUPRIM X peaked at 311 W during gaming battles. That is, the graphics card is inserted into the allocated TBP.
For comparison, we also monitored the performance of the MSI GeForce RTX 3080 SUPRIM X 12G test model. As you can see, this video card required up to 385 W at its peak. Approximately such indicators were expected, because for this forced version of the RTX 3080 12 GB, the power consumption declared by the manufacturer is 390 W.
Thus, MSI GeForce RTX 4080 16GB SUPRIM X turned out to be 24% more economical than its predecessor. And if you take into account how much faster the RTX 4080 is than the RTX 3080 12GB, it is clear that the overall energy efficiency of the model based on the 4nm GPU with the Ada Lovelace architecture is much better.
We also tried to evaluate the frequency potential of the MSI GeForce RTX 4080 16GB SUPRIM X in express mode. The video card allows you to increase the Power Limit parameter by 25%, which was done before the experiments.
Without additional efforts, it was possible to increase the base frequency of the GPU from 2205 MHz to 2300 MHz. At the same time, the average acceleration value increased from 2625 to 2720 MHz, respectively. The memory chips were consistently running at an effective 24,000 MHz instead of the stock 22,400 MHz.
It was possible to overclock the graphics card to a chip frequency of 2330 MHz and a memory frequency of 1550 MHz – in this case, the system passed the 3DMark Time Spy Extreme test without any artifacts. But it turned out that a much better stability test is now the game Portal with RTX – it managed to get normal operation without drops at only 2300 MHz and 1500 MHz, respectively.
After additional acceleration with the increased power limit, the maximum temperature of the GPU increased to 64C, and the memory chips heated up to 68C. The fans peaked at 1800+ rpm, but for the most part under load it stayed at 1600 rpm.
The graphics processor was accelerated up to 2970 MHz. It seems not so long ago that we were amazed at the possibility of GPU operation at 2 GHz, but the next generation of discrete graphics with new 4-nanometer and chips radically changes the perception of the frequency capabilities of crystals with tens of billions of transistors. Now the 3 GHz mark is a reality.
After increasing the Power Limit, the power consumption of the graphics card under load increased to 362 W.
As for the practical impact on performance, in relative terms, it is minimal (2-4%) and rather interesting for test sessions. For example, in the 3DMark Time Spy Extreme test, after additional overclocking, the graphics card scores 14,445 points, instead of 14,057 in standard mode.