PC of the Month (March 2025)

February and March turned out to be full of various events and news in the segment of components for desktop PCs. They were not always unambiguous. However, the announcement of the results of testing new video cards and even the appearance of certain models on sale add certainty, simplifying the choice of optimal gaming configurations. There are still many interesting announcements ahead, but for now we will try to choose balanced platforms for the continuation of the spring gaming season within the three price categories of PC of the Month.

Basic gaming PC of the month ($600–750)

Gaming monitor for the Basic Gaming PC from our partner Acer

24.5-inch VA monitor with Full HD resolution, up to 180Hz refresh rate and 1ms (VRB) response time. Supports AMD FreeSync, covers 99% sRGB and has an anti-glare coating for comfortable gaming without straining your eyes.

Acer Nitro KG251QP3 Gaming Monitor

Processors. For entry-level gaming configurations on the Intel platform, the 6-core Core i5-12400F chips (6/12; 2.5/4.4 GHz) remain relevant. If we ignore the fact that this model was introduced in early 2022, and evaluate it only by its actual capabilities, it still offers a decent level of performance for the specified category of systems. However, of course, choosing a 3-year-old chip in early 2025, it is not so easy to get rid of the feeling that the advantage is given to a somewhat outdated solution. If this is the only caveat, then for an entry-level gaming PC on the Intel platform it is difficult to offer anything better for ~$120.

Intel does not yet offer alternatives on the new LGA1851. Last time we already mentioned the strange situation with the cancellation of the announcement of the Core Ultra 3 processors, which were potentially supposed to become the first 8-core CPUs in the basic category. After the release of the specifications, the developers hastily removed the mention of the Core Ultra 3 205/205F models from the corporate website. Since then, there have been no more messages about these chips. So it is not yet clear whether they were canceled at all, or whether the timing of their appearance on sale has been changed. Frankly, the LGA1851 ecosystem is not yet ready for the appearance of affordable CPUs. Motherboards on the Intel H810 are almost not available on sale, and the most affordable models on the Intel H860 are offered from ~$150. So far, the combination of the Core i5-12400F processor and the Intel B760 board remains a working option for PCs on the Intel platform.

The other extreme of limited choice is multivariability. This is exactly what we observe when trying to choose the best option for a basic configuration on AMD. The current number of suitable options here is simply overwhelming.

Let's start with the same Ryzen 5 8400F (6/12; 4.2/4.7 GHz; 16 MB L3), which has been the "default" option in basic configurations for several months in a row. A six-core model in a tray configuration can now be purchased for only the equivalent of $110. Another ~$15–20 will have to be spent on a cooler, but even in this case, the Ryzen 5 8400F remains the most affordable chip for Socket AM5.

The most desirable CPU for entry-level gaming platforms is the Ryzen 5 7500F (6/12; 3.7/5.0 GHz; 32 MB). But even in the version without a cooler, this model still costs ~$175. Too expensive for basic gaming PCs. Recently, there has been a glimmer of hope for a more affordable processor with 32 MB of cache for this category of systems. In January, AMD introduced the Ryzen 5 7400F model (6/12; 3.7/4.7 GHz; 32 MB L3), which differs from the older model in terms of specifications only by a 300 MHz lower maximum acceleration frequency. However, as it turned out later, this is not the only difference. The thing is that thermal paste is used under the heat sink of the Ryzen 5 7400F instead of solder. This reduces the cost of manufacturing the CPU, but affects the speed of heat transfer and, accordingly, the ability to effectively cool the chip. The actual impact of replacing the thermal interface on the temperature regime of operation still requires additional research. According to the first reviews, the reviews are diametrically opposed - it is "all gone/burned" to "there is no difference". So for now, let's wait with the final conclusions.

Ryzen 5 7400F appeared on sale quite quickly. In Ukraine, processors in tray-configuration are already offered for ~$160. Of course, at this price they do not arouse any curiosity. Here it makes sense to either pay $10–15 extra for a faster Ryzen 5 7500F with solder under the cover and a boost of +300 MHz, or save $50, looking towards the Ryzen 5 8400F (from $110). By the way, the latest chips also have thermal paste under the heat distribution cover, but these are single-crystal Phoenix. But from a CPU of a multi-chiplet formation, you still expect a more effective thermal interface. However, as already noted, let's not rush to conclusions and wait for the price to balance.

With the Ryzen 5 5600 (6/12; 3.5/4.4 GHz) available for $95 and the Ryzen 7 5700X (8/16; 3.4/4.6 GHz) for $125 (in tray-packages), it's hard to completely abandon the mention of the Socket AM4 platform. Very attractive price tags for processors, budget motherboards, and DDR4 memory leave such options "on the table." Although these are still more situational solutions, if you want/need maximum savings. AMD is still trying to maintain interest in Socket AM4. Recently, the company again offered "new" chips - Ryzen 3 5305G/GE , Ryzen 5 5605G/GE, and Ryzen 7 5705G/GE. However, these are rather options for economical systems, especially in the case of GE versions of processors with a TDP of 35 W.

Video cards. A significant event in the video card segment was the appearance of new models from Intel on sale in Ukraine - Intel Arc B580 12 GB and Intel Arc B570 10 GB. It seems that this time we are finally talking about official deliveries. In fact, so far only adapters from Sparkle are available in "commercial quantities", but we hope that this is just the "first swallow". The older model in a three-fan configuration is offered for the equivalent of $360, the more compact younger one with a two-fan cooler design - ~$300. Subjectively, this is a bit too expensive, considering the recommended price tags - $249/219. Of course, these values are relevant for the American market, however, in Ukraine the GeForce RTX 4060 8 GB models start at $330, and the Intel Arc B580/B570 are positioned as more affordable alternatives that have advantages in 1440p.

In any case, the appearance of Intel graphics cards in our territory can only be welcomed. The price is a variable indicator that changes over time. In the niche "around $300" healthy competition is needed and Intel Arc B580/B570 can support it here perfectly. As for specific recommendations, this is an option that requires an individual approach and a preliminary detailed study of the capabilities of new graphics cards in a certain pool of projects that are a priority for the potential owner. According to averaged indicators, the Intel Arc B580 12 GB is ~10% faster than the GeForce RTX 4060 8 GB in 1440p mode, but there are nuances here. With a decrease in resolution to Full HD, which is generally more typical for graphics cards of this class, the difference is reduced, the advantage of additional memory is reduced, and no one has canceled NVIDIA's active partnership with developers and the corresponding optimizations.

Another factor that will soon have to be taken into account is the approaching announcement of the GeForce RTX 5060. The presentation of this model is expected on April 15, although it will go on sale with a certain delay - in May. According to preliminary information, the series should receive a significant increase in performance compared to its predecessor, the GeForce RTX 4060. If we evaluate only quantitative indicators, the number of active compute units will increase by 25% - from 3072 to 3840. The memory capacity will remain at 8 GB, and the bus - 128 bits, but thanks to the use of new GDDR7 chips with an effective frequency of 28,000 MHz, instead of GDDR6 with 17,000 MHz in the RTX 4060, the bandwidth will increase radically - from 272 GB/s to 448 GB/s (+65%). Even without taking into account the architectural changes, it is obvious that the GeForce RTX 5060 will receive a significant acceleration. We expect +25–30%. A side effect here will be an increase in power consumption, which will initially increase to 150 W instead of 115 W. Recommended prices have not been announced, we will learn about them on the eve of the official announcement.

According to previous leaks, NVIDIA is also preparing a desktop model GeForce RTX 5050, which will receive a more compact GB207-300 graphics processor with 2560 active Blackwell compute units, a 128-bit bus, 8 GB of GDDR6 memory and a TGP of 130 W. Apparently, this will be the most affordable video card in the new line, which should increase competition in the $250 niche. It will probably be an opponent for Intel Arc B580/B570 video cards, although the latter will retain the advantage in the form of a larger memory capacity.

As for AMD's plans in this segment, there is not much additional information yet. The company is already planning to announce the Radeon RX 9060 XT, which will probably be offered in versions with 8 GB and 16 GB of GDDR6 memory with effective 19,500 - 20,000 MHz. The GPU configuration is not known for certain, but for new models in this class, developers will probably rely on a 128-bit bus. Mentions of the certification of the Radeon RX 9060 XT have already appeared in several regions, so we expect a quick presentation of the corresponding models. However, the Radeon RX 9060 XT will rather compete with the older GeForce RTX 5060 Ti models, which will also be presented in versions with 8/16 GB of memory.

As for the base Radeon RX 9060 or even a possible Radeon RX 9050 model, there is no additional information yet. But AMD is probably preparing its answer to the upcoming GeForce RTX 5060 and will not ignore the appearance of the GeForce RTX 5050.

In addition to developing new solutions, the manufacturer obviously wants to extend the relevance of models on RDNA 3. Recently, AMD introduced the Radeon RX 7650 GRE 8 GB for the Chinese market. If such video cards later reach Ukraine, it is useful to know that they are actually no different from the well-known Radeon RX 7600 8 GB, except for a slightly higher GPU operating frequency - 2.69 GHz vs. 2.62 GHz.

In the pool of relevant solutions for basic gaming configurations, the GeForce RTX 3060 12 GB remains. Despite the warning about the urgent curtailment of sales of this line, the corresponding models are still presented on virtual shelves. Although the increase in prices has to some extent affected these models, there is a shift in user preferences towards the newer GeForce RTX 4060. This was confirmed by another Steam statistics, where the GeForce RTX 3060 models lost popularity to the GeForce RTX 4060 for the first time.

Optimal gaming PC of the Month ($1200–1400)

Gaming monitor for the “Optimal Gaming PC” from our partner Acer

27-inch Agile-Splendor IPS monitor with 2560×1440 resolution, up to 180Hz refresh rate and up to 0.5ms response time. Supports AMD FreeSync and displays 95% of the DCI-P3 color space for deep, rich colors.

Acer Nitro VG272UV3bmiipx gaming monitor

Processors. For a mid-range platform on Intel, the optimal choice remains the Core i5-12600K (6/12+4; 3.7/4.9 GHz + 2.8/3.6 GHz) for $180, or the version without integrated graphics Core i5-12600KF , which can be purchased for $170. Here, the only competition is the Core i5-14400F (6/12+10; 2.5/4.7 GHz + 1.8/3.5 GHz), which have significantly decreased in price over the past few months. The most affordable version in a blister (tray) is already sold for the equivalent of $140, a retail modification with a simple cooler in the kit can be purchased for $150.

So, the Core i5-12600K/KF will be slightly faster due to higher operating frequencies, but more expensive and require a more efficient cooler for cooling, which is not offered in the kit at all. In turn, the Core i5-14400F is generally more affordable, and in case of savings, allows you to do without a standard cooler.

The most affordable processor for the LGA1851 platform is still the Core Ultra 5 225F (6+4; 3.3/4.9 GHz + 2.7/4.4 GHz), which is offered for $275, while the Core Ultra 5 225 version with activated graphics costs ~$300. So even at the price within the optimal configuration, such offers are still "limitedly suitable". Not that this situation generally weakens Intel's position in this category - the Core i5-12600K is still holding its own, but here again you have to choose from models from three years ago.

As for the new CPUs, there are still many questions. Apparently, the launch of the new platform went "not according to plan". Despite the fact that Arrow Lake-S is more economical than its predecessors, has better IPC, accelerated graphics and even a hardware NPU, minimal changes in gaming performance, where at times you can even observe a certain regression, have become an unpleasant surprise for many potential buyers. The lack of relatively affordable CPUs for LGA1851 on sale also does not add anything positive, slowing down the overall spread of the platform. It is not surprising that in such conditions the share of the competitor is significantly increasing . Especially when it comes to the "self-assembly" segment (DIY, Do it yourself).

As for the AMD platform, as in the case of basic configurations, there are even too many acceptable options. Starting from the Ryzen 5 7500F (6/12; 3.7/5.0 GHz; 32 MB) and Ryzen 5 7600 (6/12; 3.8/5.1 GHz; 32 MB) available for this category, and ending with 8-core chips. Recently, the cost of the Ryzen 5 9600X (6/12; 3.9/5.4 GHz; 32 MB) has also decreased, which can now be purchased for $220 (tray). However, given the potential for the future, a rational choice here may be the Ryzen 7 7700 (8/16; 3.8/5.3 GHz; 32 MB L3), which in the version without a cooler can already be obtained for the equivalent of $230.

The increased number of computing cores eliminates the situation when, due to a 6-core configuration, it will be possible to get stators with unpleasant random "jerking" of the picture. So far, such situations have not occurred (if there are any - report them in the comments), but this cannot be ruled out. A certain decrease in operating frequencies is quite easy to fix, having an unlocked multiplier for increase. In addition, motherboards even on AMD B650/B850 allow you to experiment with CPU overclocking.

If you are still attracted to models based on the Zen 5 architecture, the new 6-core Ryzen 5 9600 (6/12; 3.8/5.2 GHz) have recently gone on sale. The starting price tags of ~$240 certainly look unattractive, but with a price drop to ~$200, these chips can be taken into account. Especially considering the fact that, unlike the X-modifications, the basic versions in the BOX delivery are already offered with a cooler included.

Another interesting model that has appeared on our market is the Ryzen 5 7600X3D (6/12; 4.1/4.7 GHz; 96 MB L3). The already familiar "X3D" designation in the name is a distinction that the chip is equipped with an additional 3D V-Cache buffer. Initially, the processor was exclusively presented for the American MicroCenter network, and last fall it became available in German retail outlets. So the appearance of this model in Ukraine was only a matter of time. An additional 64 MB of L3 cache usually works wonders in games, significantly increasing the number of fps. Almost an ideal candidate for an optimal configuration. But there is one "but". The starting price tag of the Ryzen 5 7600X3D at ~$390 is frankly not pleasing. After all, we are talking about a 6-core chip with the Zen 4 architecture, which does not support additional overclocking. Especially if you remember that for such money only half a year ago in Ukraine you could freely buy an 8-core Ryzen 7 7800X3D. Of course, this is a very interesting chip for experiments, but still too expensive to recommend. If the high price does not stop you – add the CPU to the list of possible purchases.

Video cards. When choosing a video card for optimal configurations, this time we act as pragmatically as possible. A certain transition period forces us to do just that. The starting cards for systems of this class can be considered the GeForce RTX 4070 12 GB and Radeon RX 7800 XT 16 GB . There are actually a minimum of questions about the model with an AMD graphics processor. Predicted performance, 16 GB of memory and a fairly moderate price tag (from $550). As for the NVIDIA solution, recently we usually relied on the GeForce RTX 4070 SUPER 12 GB , but the cost of these models has increased significantly, so $850+ for a video card for a mid-range gaming platform is already too much. And there are only a few representatives of the RTX 4070 SUPER series left on sale. But the basic GeForce RTX 4070 12 GB can be purchased for ~$620. Yes, they are somewhat slower (10–15%) than SUPER modifications, but currently offer a significantly better price/performance ratio.

Of course, at the junction of generational changes, these are more likely temporary or transitional options for cases when the platform is needed urgently and there is no time to wait for new models, as well as the mandatory stage of price normalization. Moreover, this can be a conscious, balanced choice, when preference is given to proven solutions with objectively reasonable price tags.

In early March, NVIDIA officially introduced the GeForce RTX 5070 12 GB on the GB205 graphics processor in a configuration with 6144 compute units. The video card received a 192-bit bus and 12 GB of memory. At the same time, the adapter is equipped with new GDDR7 chips with an effective operating frequency of 28,000 MHz, which allows you to get a bandwidth of 672 GB/s. This is 33% more than the RAM in the GeForce RTX 4070 (504 GB/s.) The power consumption of the new product is 250 W versus 200 W in its predecessor.

According to the test results, the gaming performance of the new model is 20–25% higher than that of its predecessor, the GeForce RTX 4070 12 GB. The GeForce RTX 5070's performance is mostly close to the fps that the GeForce RTX 4070 SUPER and even the GeForce RTX 4070 Ti are capable of offering. At the same time, the RTX 5070 of course has full support for DLSS 4 with Multi Frame Generation. Even despite the presence of only 12 GB of memory, the recommended price tag of $549 makes the GeForce RTX 5070 potentially a very good option for optimal gaming configurations equipped with 1080p/1440p monitors. However, such a recommendation will have to wait. The starting retail price of GeForce RTX 5070 modifications starts at ~$830. The days are not so long ago when we offered video cards of a similar price for advanced systems. For mid-level gaming PCs, this is already an obvious overkill.

The same applies to the new Radeon RX 9070 16 GB . Which, with an identical recommended price of $549, hit retail shelves in Ukraine with price tags starting at ~$820. A very worthy competitor to the GeForce RTX 5070 with a larger memory capacity and, as the test results showed, even a certain advantage in pure raster rendering, but the actual prices of the new products still stop us from recommending them.

In mid-April , GeForce RTX 5060 Ti graphics cards with 8/16 GB of memory should be introduced. These models will receive a GB206 processor with 4608 compute units, a 128-bit bus on GDDR7 (28,000 MHz) with a total RAM of 448 GB/s. Practical tests will show whether this model can be considered for a mid-level gaming system. If we are talking about +10–15% to the performance of its predecessor, then this will obviously not be enough. In current conditions, the GeForce RTX 4070, which is the starting point for this category of systems, is ahead of the GeForce RTX 4060 Ti by ~30%. It is unlikely that the GeForce RTX 5060 Ti will be able to compensate for this difference.

From previous information, we note that the GeForce RTX 5060 Ti will receive a TGP of 180 W, and certain models may be equipped with a 16-pin 12VHPWR power connector. As we can see, all updated models on the Blackwell architecture GPU have a slightly higher power consumption level than the corresponding solutions of the previous generation. The increase in computing units and GPU operating frequencies is noticeable, considering that this time the graphics processors are manufactured using the same process technology (TSMC 4N, 5 nm) as the chips of the previous generation.

Progressive Gaming PC of the Month ($1800+)

Gaming monitor for the “Progressive Gaming PC” from our partner Acer

34-inch curved VA monitor with 3440×1440 resolution, up to 165Hz refresh rate, and up to 1ms response time. Supports AMD FreeSync, features USB-C (65W), KVM switch, stereo speakers, and 90% sRGB coverage.

Acer Nitro XV345CURVBMIPHUZFX Gaming Monitor

Processors. When configuring powerful gaming systems on the Intel platform, you still have to rely on chips from previous generations. For progressive PCs, we recommend the Core i5-14600K (6/12+8; 3.5/5.3 GHz + 2.6/4.0 GHz), the price tag of which is ~$250 and the overall price/performance ratio dispels prejudices, justifying the appearance of this model in systems of this level in early 2025. If you need more computing power for work tasks, you can consider the Core i7-14700K (8/16+12; 3.4/5.6 GHz + 2.5/4.3 GHz; ~$400), although in this case you will have to pay special attention to the efficiency of the cooling system. Under high load, the top Raptor Lake processors have considerable energy appetites.

If professional tasks generally take priority over entertainment, it makes sense to take a closer look at the LGA1851 platform and the new Arrow Lake-S chips – Core Ultra 5 245K (6+8; 4.2/5.2 GHz + 3.6/4.6 GHz; ~$340) and Core Ultra 7 265K (8+12; 3.9/5.5 GHz + 3.3/4.6 GHz; ~$420). Architectural changes have significantly increased performance per clock (IPC), compensating even for the lack of Hyper-Threading support. At the same time, they are more economical, have radically faster integrated graphics and a hardware NPU unit. Prices for new CPUs have already stabilized at a reasonable level, so if in certain applications included in your work tasks, the new Intel chips have an advantage over their predecessors, or at least consume less power and heat up in appropriate conditions – LGA1851 can be considered a platform for powerful PCs. Of course, it's unpleasant to get a conditional 180 fps instead of 200 fps, but such annoying inconveniences can be tolerated if there are practical advantages in other cases.

The Arrow Lake Refresh (Core Ultra 300) processors, which according to preliminary information may appear in the second half of this year, may additionally improve the situation with the Intel platform. At the same time, the modernization will only concern enthusiast models of the K/KF series. If it is only about increasing operating frequencies, then this is not a serious update. Perhaps future CPUs will receive a much more powerful NPU, for example, from Lunar Lake chips, which will outline Intel's commitment to the development of a local AI ecosystem. As for the fundamentally new Nova Lake-S desktop chips, these models will appear no earlier than 2026. But against the background of global changes at Intel, the company's current plans may be significantly adjusted.

The shortage of AMD chips with an additional 3D V-Cache cache seems to have been gradually overcome. There are plenty of suitable offers on the market, and prices are gradually decreasing. For this reason, we are returning to progressive gaming configurations the Ryzen 7 7800X3D (8/16; 4.2/5.0 GHz; 96 MB L3), which can already be purchased in a tray configuration for $480. Yes, it is still too expensive, but it would be at least frivolous to refuse an option that directly affects the number of fps when assembling a powerful gaming PC. Of course, when the cost of this model increased to ~$650–700, the internal "ratio" mode was subconsciously turned on, but under current conditions, the additional costs are bearable. Especially against the background of price arbitrariness in the video card segment, which we will return to later.

How about the Ryzen 7 9800X3D (8/16; 4.7/5.2 GHz; 96 MB L3)? The ultimate option, which is better than its predecessor in all respects, except for its higher power consumption. Updated architecture, higher operating frequencies, unlocked multiplier. The only question is the price. If $600 seems like a justified expense for maximum performance, don't hold back. The advantage of the Ryzen 7 9800X3D is also that the chip has high frequencies under multi-threaded load, so in work tasks it is definitely not inferior to the basic 8-core Ryzen 7 9000 models, which makes it universal. While the Ryzen 7 7800X3D is more of a gaming CPU, with even more modest computing performance than, for example, the much more affordable Ryzen 7 7700X.

However, if you are interested in uncompromising chips for any workload, new 12- and 16-core processors with 3D V-Cache have already appeared on sale - Ryzen 9 9900X3D (12/24; 4.4/5.5 GHz) and Ryzen 9 9950X3D (16/32; 4.3/5.7 GHz). The flagship of the line has regained its status as the fastest chip for work tasks, outpacing both its predecessor Ryzen 9 7950X3D and the Core Ultra 9 285K. In games, the average advantage remained with the 8-core Ryzen 7 9800X3D, but the difference here barely exceeds the measurement error. So if you really need the fastest desktop chips and you definitely understand how to use such computing power, the Ryzen 9 9900X3D can already be purchased for $780, and for the 16-core Ryzen 9 9950X3D you will have to pay ~$940. You can ignore the recommended price ($599 and $699) - we are not in the USA, but for the start of sales, local price tags are quite moderate.

Video cards. If everything is more or less clear and predictable with processors for advanced gaming PCs, then choosing the right video card raises a lot of rhetorical questions. The period of changing generations of graphics adapters is usually accompanied by a stage of uncertainty, when models of the previous generation disappear from sale, and new products are offered in limited assortment and quantity at inflated prices. We will observe something similar in the spring of 2025.

In the second half of February , the GeForce RTX 5070 Ti 16 GB series video cards were officially introduced. Potentially a good option for progressive configurations. As practical testing has shown, the new product is 10–15% ahead of the GeForce RTX 4070 Ti SUPER , in some places demonstrating indicators at the level of the GeForce RTX 4080/4080 SUPER . Perhaps this is not quite what you usually expect from the next generational change of discrete graphics. However, the "raster" changes are rather the top of technological transformations, which this time are aimed at increasing AI capabilities and their more active practical implementation. DLSS 4 and multiple frame generation (MFG, Multi Frame Generation) are also only gaining popularity and are synchronized with Reflex 2, which allows not only to generate as many frames as needed, but also reduces actual delays.

Returning directly to the GeForce RTX 5070 Ti, we note that the recommended price of this model for the American market is $749. Unfortunately, such values are only declarative. Manufacturers initially set significantly higher price tags for their original models. And since the GeForce RTX 5070 Ti is not offered at all in the reference Founders Edition version, the actual cost of video cards "on the ground" has nothing to do with the declared one. In Ukraine, a rather limited selection of models of the new series is currently offered, and current prices start at 45,000 UAH (~$1,100), while the cost of top versions with the quietest coolers can reach ~$1,500. Yes, this is not $749 at all... Smart users can probably rightly note that for such, or even less money, it is better to buy a GeForce RTX 4080 SUPER! One could agree with this if these same GeForce RTX 4080 SUPERs were still on sale. That's all. There are no such offers anymore.

Raising the stakes, we can mention the GeForce RTX 5080 16 GB . Recently, several models of the series were also reviewed by us. Everything is very good, except for the same price. Not the recommended ($999), but the actual one. You can perceive a relatively small increase in performance, but the price tags from 63,000 UAH (~$1,500) for the basic versions of the GeForce RTX 5080 and up to 85,000 UAH (~$2,040) are frankly confusing. Moreover, we are not even talking about the flagship GeForce RTX 5090 32 GB . Let's not even mention the latter... The cost of ~$4,000 does not allow us to consider this model as a video card for games at all.

In general, the launch of the GeForce RTX 50 line was not cloudless. As it turned out, certain copies of the GeForce RTX 5090/5080/5070 Ti series are offered with a reduced number of rasterization units (ROPs, Raster Operation Pipelines) than declared in the specifications. Each series has its own "losses", but in all cases this leads to a noticeable decrease in gaming performance (4–10%). NVIDIA acknowledged the existence of a problem, guaranteeing that all purchased video cards with GPUs that have a smaller number of ROPs will be replaced by manufacturers. But this is a rather serious number. Even the developers claim that ~0.5% of video cards have a similar feature. Apparently, this is a certain manufacturing error, since in all cases the number of ROPs is reduced by a specific value for each series. That is, this is definitely not a desire to sell some defective GPUs. But of course, it is difficult to imagine how such GPUs passed the testing stages of NVIDIA and directly the video card manufacturers.

An unexpected "curiosity" of the new RTX 50 series was also the lack of support for PhysX technology in 32-bit programs. It was in this version that the physics engine was used in a considerable number of games with additional effects. Usually these are already age-old, but still relevant releases: Borderlands 2, Batman: Arkham Asylum, BioShock Infinite, Assassin's Creed IV: Black Flag, Mafia II and others.

Shifting PhysX calculations to the CPU radically slows down such projects, forcing them to abandon the decorations that added dynamism to scenes. On video cards of previous generations, the 32-bit implementation of PhysX works, but on the GeForce RTX 50 it no longer does. You can solve the issue by transferring the processing of PhysX effects to the second video card in the system. We conducted similar successful experiments during a recent review of the GeForce RTX 5070 Ti. The option really works, since in this case you can choose which device to use for PhysX calculations in the settings, but this is already a serious "crutch" and rather an opportunity for enthusiasts to have fun.

In the case of the top RTX 50, the topic of heating the 12VHPWR connectors does not lose its relevance. The consequences of a not very successful design solution are felt even after this issue has been "solved". For some time, NVIDIA has also been unable to get rid of problems with drivers, which were usually distinguished by good optimization.

The landmark event of March was the launch of new AMD graphics cards – Radeon RX 9070 and Radeon RX 9070 XT. It seems that against the background of some accumulated negativity towards NVIDIA's latest solutions, the appearance of an interesting alternative was perceived by the active gaming community with additional excitement. Of course, the concept of "anything, but different" will not work for such demanding users. So what attracts Radeon RX 9070/9700 XT?

The models use Navi 48 graphics processors with RDNA 4 architecture, manufactured according to 4 nm standards and having a significantly higher density of elements than the GPU for RTX 50 – 151 million/ mm² versus 121 million/ mm² . GPUs for RX 9070 series video cards differ in the number of computational and functional blocks, while both models use a 256-bit bus and fairly affordable GDDR6 memory chips (20,000 MHz) with a total capacity of 16 GB. The power consumption of the Radeon RX 9070 XT is declared at 304 W, for the Radeon RX 9070 – 220 W.

Practical tests have shown that the Radeon RX 9070 XT 16 GB demonstrates performance almost at the level of the flagship of the previous generation - Radeon RX 7900 XTX 24 GB. This is a remarkable achievement, considering the simpler single-chip GPU design and lower memory bus bandwidth. Formally, the 2025 solution with a recommended price of $ 599 offers performance at the level of the top model of 2022 with a price tag of $ 999.

Among the key features of the Radeon RX 9070 series models, we should also note support for FSR 4 technology using AI algorithms to improve the quality of intelligent image scaling. At the start, more than 30 projects will work with FSR 4, and by the end of this year their number should increase to 75+. AMD has also significantly improved the performance of new graphics cards with ray tracing. Although they still have not managed to catch up with comparable models from NVIDIA, the difference with RT has decreased to 10–15%, instead of the 30 percent "gap" in the previous generation. For clarity, in games with RT, the Radeon RX 9070 TX 16 GB video card is ahead of the Radeon RX 7900 XTX 24 GB by ~15%, and has an advantage of ~30% over the Radeon RX 7900 XT 20 GB.

If we compare the performance of the new AMD and NVIDIA models, the Radeon RX 9070 XT 16 GB is on average slightly (0–5%) inferior to the GeForce RTX 5070 Ti 16 GB in modes with classic rendering. Of course, as is customary, a lot depends on the specific project. In some cases, the AMD novelty is even better. As for modes with RT, here the advantage of the RTX 5070 Ti is 15%.

There is no need to mention the recommended prices for the American market. In Ukraine, the Radeon RX 9070 XT is offered from $920. The price tag is certainly overpriced, but against the background of the cost of the same GeForce RTX 5070 Ti, even such values look acceptable. So, based on the totality of consumer characteristics and the price/performance ratio, this time for progressive gaming systems we recommend the Radeon RX 9070 XT 16 GB. If AMD solutions for some reason do not suit you (work tasks with CUDA, love for RT, etc.), then the GeForce RTX 5070 Ti 16 GB will be an alternative for this category, but the budget will have to be increased by at least $200–250.