The time on the eve of the change of generations of desktop platforms is a difficult period for those who decided to assemble a powerful desktop system for games or work tasks. Some components are just coming on sale and cost more than expected, some are still on the way or are delayed at all with the announcement. How not to make a mistake here and wisely spend the money left over after donating to the Armed Forces and helping loved ones? As part of “PC of the month” we offer to consider three options of systems of different price categories based on current components in view of the prospects for a rapid update of key elements of the PC.
Basic gaming PC
|CPU||Intel Core i3-12100F (4/8; 3,3/4,3 GHz, 12 MB L3)||$110|
|AMD Ryzen 5 5500 (6/12; 3,6/4,2 GHz, 16 MB L3)||$125|
|Motherboard||Intel B660 (LGA1700)||$110|
|AMD B550 (Socket AM4)||$85|
|Memory||16 GB (2×8 GB) DDR4-3200||$60|
|Videocard||GeForce GTX 1660 SUPER 6 GB||$280|
|Storage||SSD, 480/500/512 GB, M.2 PCI-E||$50|
|Case and power supply unit||Middle Tower ATX, 550 W||$60|
|Total for PCs on the Intel platform||$670|
|Total for PCs on the AMD platform||$660|
Processor positions have been updated in basic game systems. For the Intel platform, we recommend considering the option with Core i3-12100F (4/8; 3.3/4.3 GHz). Admittedly, we didn’t expect to see 4-core chips in this category, but the efficiency and overall price/performance ratio of these Alder Lake processors is so good that for an entry-level platform, the $110 Core i3-12100F looks like a very attractive option.
The alternative among Intel solutions here are the basic 6-cores of previous generations – Core i5-10400F (6/12; 2.9/4.3 GHz) for $130 and Core i5-11400F (2.6/4.4 GHz) for $145. In addition to the higher cost, the latter also require binding to the LGA1200 platform, limiting the possibility of further upgrading. In terms of gaming performance, all these chips will be close, and even in heavy projects with good multithreading support. The Core i3-12100F in such cases will pull through with noticeably better single-threaded performance, while the 6-core models offer more compute units. As a result, the number of fps is always equalized by the video card used for a PC of this class.
The cost of starter motherboards with LGA1700 connector has dropped to $75, but for this price models on the Intel H610 chipset are offered with basic functionality and minimal equipment. These are devices with a pair of DIMM slots for memory modules and the simplest power subsystems without additional cooling. And although even for such models support for top chips up to Core I9-12900KS with a TDP of 150 W and a peak of 241 W was formally declared, for stable long-term work even in a pair with less energy-intensive CPUs, we would recommend choose from more equipped boards. The preferred model is Intel B660.
In addition to improved functionality, such devices allow the use of high-speed memory kits without being limited to the basic DDR4-3200 mode for 12th generation chips. The simplest devices based on Intel B660 cost from $90, but among models for $100-115 the choice is noticeably wider.
The platform on AMD components has also received a noticeable boost. Finally, with a clear conscience, we can let go of the well-deserved veteran – Ryzen 5 1600 (6/12; 3.2/3.6 GHz). In various incarnations, this chip lasted for a very long time, first establishing itself in the optimal gaming configurations, and later settling in the basic ones. Fortunately, in the last category came a worthy replacement for the long-lived. The most interesting options here are Ryzen 5 3600 (6/12; 3.6/4.2 GHz) and Ryzen 5 5500 (6/12; 3.6/4, 2 GHz). Processors that are similar in basic functional characteristics and differ in internal layout and equipment.
Ryzen 5 3600 is a bright representative of the classic Ryzen desktop chip with a chiplet layout. The younger 6-core 12-thread model of the Matisse family based on the Zen 2 architecture with 32 MB of third-level cache memory has long been the basis of optimal gaming configurations. And this means that the CPU offers a very good balance of performance and price. At its current retail price of $115, it’s a great candidate for an entry-level gaming system, and a very serious upgrade over the Ryzen 5 1600. Even though the Ryzen 5 3600 was introduced three years ago, it’s no surprise that the Ryzen 5 3600 remains one of the most popular after a significant price drop on the market.
What about the Ryzen 5 5500? Its appearance is due to the lack of interesting offers from AMD in the segment up to $150. The company became interested in chips of the highest category with an increased margin, for a while ceding the niche to the main opponent with an aggressive pricing policy. A few years ago, it was difficult to imagine such a situation, but times are changing. Ryzen 5 5500 belongs to the Cezanne family of chips, that is, to the branch of hybrid processors with a monolithic crystal, on which integrated graphics are also located. However, in the case of the Ryzen 5 5500, the video core is disabled.
In fact, it is the familiar Ryzen 5 5600G, but with deactivated graphics and slightly lower operating frequencies. Among the features of the chips of this line, it is worth noting the presence of 16 MB L3 and support for PCI-E 3.0. The reduced volume of the third-level cache memory is an unfortunate relief that will be felt in various applications, while the limitation in the form of PCI-E 3.0 is rather critical only when using video cards with four bus lines (Radeon RX 6400/6500 XT) and the fastest SSDs.
At the same time, Ryzen 5 5500 has obvious advantages. This is a 6-core 12-thread chip with Zen 3 architecture and a memory controller that is located on a single chip with computing cores. In addition, it is a very economical CPU.
As for performance, judging by the results of numerous tests, in most cases the Ryzen 5 5500 is 5-10% faster than the Ryzen 5 3600. There are exceptional situations in which 32 MB L3 put the predecessor on first place, but they are very rare. In games, the Ryzen 5 5500 usually has a minimal advantage, although there are exceptions.
Given the small price difference between the Ryzen 5 3600 ($115) and the Ryzen 5 5500 ($125), we gave the latter a place in the base gaming configuration. But it is rather “tasty”, both options are worthy.
When compared to the Core i3-12100F, both solutions from AMD will have a noticeable advantage (15-30%) in multi-threaded tasks, while in games the chip from Intel will offer slightly more fps on average. Although this is not a dogma, especially in the case of projects that use more than eight threads.
Taking care of its position in the segment of affordable chips, in addition to Ryzen 5 5500, AMD also offered two more processors – Ryzen 5 4500 (6/12; 3.6/4.1 GHz) and Ryzen 3 4100 (4/8; 3.8/ 4.0 GHz). The manufacturer rightly added them to the Ryzen 4000 line. Despite the fact that the Ryzen 5 4500 is actually the most affordable 6-core AMD with multithreading support, this chip belongs to the Renoir family with the Zen 2 architecture. In addition, the processor has 8 MB of L3 cache. In general, the Ryzen 5 4500 for $105 cannot be called a failure at all, but with the availability and current retail price of the Ryzen 5 3600/5500, the entry-level 6-core simply has no chance of being part of the basic gaming system. The case when it makes sense to add $10-20 even with a limited budget.
Ryzen 3 4100 also belongs to the Renoir family with monolithic crystal and Zen 2 architecture. Only this time it is a 4-core 8-thread chip with only 4 MB of L3. The $90 price tag certainly doesn’t appeal when there are such powerful older models that now cost only a little more.
After the next wave of cryptocurrency mining with the help of video cards broke out, the purchase of the latter stopped causing nervous tremors among gamers. The best options for entry-level gaming systems would be GeForce GTX 1660 SUPER 6 GB or GeForce GTX 1660 Ti 6 GB. Both models have similar performance and are currently offered at prices starting at $280. The first was released almost half a year after the “titanium” modification and is equipped with GDDR6 memory chips with an increased effective frequency (14,000 MHz vs. 12,000 MHz). It is for this reason that in most cases it manages to compensate for the smaller number of computers activated in the TU116 GPU – 1408 vs 1536.
The entry-level models of the GeForce GTX 1660 6GB series are practically non-existent, and those that are are offered at the same price as the SUPER/Ti versions, which are ~15% more powerful. At a cost of $ 230-240, these would be good options for basic gaming configurations, but, alas.
All that remains for those wishing to buy a more affordable video card is the GeForce GTX 1650 4 GB for $200-210. However, along with savings, buying such a model will radically reduce gaming performance. The number of fps will decrease by a total of 60-70%. Perhaps such adapters should only be considered if the overall budget for the purchase of a PC is strictly limited, but then you can consider using more affordable CPUs and motherboards.
When choosing an affordable video card, you can also come across a GeForce GTX 1630 4 GB for ~$190. This is exclusively an anti-crisis option during the absence of graphic adapters on sale. Now there is no point in paying attention to it at all. Imagine another -40-50% of the capabilities of the slow GeForce GTX 1650, and at this price. Do you really need it?
Which of AMD’s solutions can be viewed for basic gaming systems? There are practically no interesting options. The Radeon RX 6500 XT 4GB might be attractive if only at a price of $200-$210, but it is on average 30% slower than the GeForce GTX 1660 SUPER/Ti, in addition due to the limited number of PCI-E x4 lanes , the video card is very dependent on the bus standard. Connecting via PCI-E 4.0 is actually mandatory, and Ryzen 5 5500 has problems with this, as this CPU only has PCI-E 3.0 at its disposal. For this reason, the Radeon RX 6500 XT 4GB can only be considered under significant financial constraints. In addition, in the case of an AMD platform, then it is better to focus on the Ryzen 5 3600 with PCI-E 4.0 lines for connecting discrete graphics.
Considering the capabilities of motherboards for both platforms, we recommend equipping systems with 16 GB (2×8 GB) RAM of the DDR4-3600 standard. Additional RAM bandwidth in games is only welcome. The difference in cost from the conventionally basic DDR4-3200 is minimal, so you can immediately focus on modules with a higher effective frequency, without relying on success with independent overclocking.
The ultimate gaming PC
|CPU||Intel Core i5-12400F (6/12; 2.5/4.4 GHz)||$185|
|AMD Ryzen 5 5600X (6/12; 3.7/4, 6 GHz)||$210|
|Motherboard||Intel B660 (LGA1700)||$140|
|AMD B550 (Socket AM4)||$110|
|Memory||16 GB (2×8 GB) DDR4-3600||$70|
|Video card||GeForce RTX 3060 12 GB/ Radeon RX 6650 XT 8 GB||$440|
|Storage||SSD, 960 GB/1 TB, M.2 NVMe, PCI- E x4||$120|
|Case and power supply||Middle Tower ATX, 650 W||$130|
|Total for PCs on the Intel platform</ span>||$1085|
|Total for PCs on the AMD platform||$1080|
From current processors on the Intel platform, for an optimal gaming configuration, we suggest using the same Core i5-12400F (6/12; 2.5/4, 4 GHz). A 6-core chip for $185 does the job very well.
The future update of Intel’s desktop platform will surely affect optimal systems, but it is not a fact that we will end up switching here to the 13th generation model. According to some leaks, all Core i5-13000s, except for unlocked multiplier models (Core i5-13600K/KF), will be based on the Core architecture 12th gen Alder Lake, not Raptor Lake. One of the presentation slides, which is already available online, indicates that the increase in L2 cache memory is expected for Core i5K chips and above. And this is precisely one of the markers of the transition to the updated architecture. But here we will still wait for official details.
Even in the worst-case scenario, if the base Core i5s remain on the Alder Lake chips, they will receive additional acceleration in multi-threaded tasks by activating energy-efficient cores. Intel is expected to open up E-cores for all Core i5 series chips. At the same time, the starting model Core i5-13400 will have a functional formula of 6/12+4, and Core i3-13500 and Core i5-13600 will already have 6/12+8.
We remind you that in the current Core i5-12×00 line, additional energy-efficient cores (4 pcs.) with Gracemont architecture are available only in the top Core i5-12600K/KF models, so in multi-threaded tasks they have a noticeable advantage over the younger chips of the series. If the Raptor Lake design reaches only the older Core i5-13600K/KF models, then the gap between the enthusiast versions and the basic ones will become even larger.
Should you wait for the presentation of new Intel chips if you are now planning to assemble a mid-range gaming system? More likely than not. Almost certainly, Raptor Lake will be presented at the Intel Innovation event scheduled for September 27. However, there is still no confirmation of the timing of the appearance of chips on sale, as well as the sequence of product releases.
Perhaps only older models from Core i5K will be offered at first, and more affordable ones will appear a little later. In any case, the situation will be clarified already at the presentation. Then there will be specifics on the architectural nuances of all CPUs, the terms of the start of sales, as well as recommended prices. The latter is important, taking into account the fact that Intel was going to increase the cost of products in various categories up to 20% in the fourth quarter of this year. Wouldn’t want it to be the Raptor Lake chips, but not out of the question.
In the case of the AMD platform, we bet on the proven Ryzen 5 5600X (6/12; 3.7/4.6 GHz). The older 6-core series is on offer now for $210 and is also a very good investment. The performance of chips from Intel and AMD is generally very similar, with some advantages of one or the other processor in different tasks.
Expanding the model range of inexpensive chips, AMD also offered another 6-core processor that can be part of the optimal configuration – Ryzen 5 5600 (6/12; 3.5/4.4 GHz). It differs from the X-modification of the same name only with a reduced operating frequency of 200 MHz. Now it is this chip that is basic in the Vermeer line. More recently, it retailed for $190, and the $20 price difference with the older model is not a very convincing argument. But at the time of publishing the material, Ryzen 5 5600 began to be sold for $170 on several online sites. If there is an opportunity to buy a processor at such a price, you should think about saving. Although there is a difference in performance with the Ryzen 5 5600X, it is minimal.
Here it is important not to get carried away in the pursuit of value and mistakenly buy Ryzen 5 5500. The names of the models are similar, but, as we mentioned earlier, this is a model from a completely different family. It is relevant for the basic configuration, but the optimal one requires a more powerful CPU.
In this case, it may be better to focus on the presented Ryzen 7000? It would be good, but at the start of sales, such a kit will still be expensive for optimal configurations. The base model of the line was the 6-core Ryzen 5 7600X with a recommended price of $299. We can assume that at the beginning the retail price will be much higher – $350-400. And AMD B650 motherboards will go on sale in October at best, and they will also be more expensive than models for current-generation processors. The announced “from $125” for basic models should be treated with a certain amount of skepticism and not count on such price tags at the start of sales.
Here we can also mention DDR5 standard RAM, which is mandatory for Ryzen 7000. Fortunately, the difference in the cost of DDR5 and DDR4 kits is not as scary as before, but high-speed modules are still more expensive than the previous generation kits.
As a result, all the mentioned costs lie on one side of the scale, while on the other side there is an attractive prospect of getting a progressive platform with a long lifespan. And there are also promises that the Ryzen 6 7600X will be able to compete with the Core i9-12900 in games … Well, at least it was so in AMD’s presentation slides. Let’s look at the results of independent sources and then we will decide. The sale of Ryzen 7000 and AMD X670 motherboards will start on September 27. At the same time, you will be able to familiarize yourself with performance measurements in various applications and games.
For optimal gaming configurations, the GeForce RTX 3060 16 GB and Radeon RX 6650 XT 8 GB video cards can be considered the reference. The first will better understand situations with activated ray tracing, the second is 10-15% more productive in games with classic rendering. About $430–450 will be needed to buy these models. Previously, we identified the Radeon RX 6600 XT 8 GB as an opponent of the GeForce RTX 3060 16 GB. The market still offers these models, but the current price tags do not seem to have gotten rid of the train of the next mining round. As a result, an absurd situation is created, when the new accelerated versions of the Radeon RX 6650 XT 8 GB are almost cheaper than the basic Radeon RX 6600 XT 8 GB. Here you just need to take this point into account.
In the niche of mid-range solutions, a quick transition to next-generation devices is not expected. Both NVIDIA and AMD plan to start upgrading discrete graphics from the top-end, expensive solutions, so we’re unlikely to see GeForce RTX 4000 and Radeon RX 7000 models in this category in the next few months.
Moreover, it is expected that NVIDIA does not yet plan to end the story with a device based on chips with the Ampere architecture. In the fall, the developer is expected to present updated mid-range models – GeForce RTX 3060 8 GB and GeForce RTX 3060 Ti 8 GB with GDDR6X memory. In the first case, to reduce the amount of memory from the standard 12 GB for this line to 8 GB, it will also be necessary to reduce the bus bit rate from 192 to 128 bits. If the operating frequencies of the GDDR6 chips do not change, the bandwidth of the memory subsystem will decrease from 360 GB/s to 240 GB/s. Only practical tests will show how much this will affect performance. One way or another, NVIDIA will surely have the opportunity to balance the offer on the price.
In the case of converting the GeForce RTX 3060 Ti 8 GB to GDDR6X and increasing the effective frequency from 14,000 MHz to 19,000 MHz, the memory bandwidth will increase from the base 448 GB/s for this line to 608 GB/s.
The activity of Californians in the middle class is not accidental and is related to the fact that Intel will soon join the game in the segment of discrete graphics. Despite the fact that the initial dates for the launch of video cards with their own GPUs have been repeatedly postponed, this time the long-awaited event is approaching. Judging by the fact that the manufacturer no longer hides the key technical characteristics and also openly talks about speed code Intel Arc, the official launch is just around the corner. Let’s assume that the announcement can take place at the same Intel Innovation conference on September 27.
At the start, we expect four models at once – Intel Arc A770, Intel Arc A750, Intel Arc A580 and Intel Arc A380. The latter is an entry-level solution and has been available on the Chinese market for some time. As for older adapters, they are all based on the ACM-G10 graphics processor, produced by TSMC on a 6-nanometer process. Based on Intel’s internal test results, the Arc A750 competes with the GeForce RTX 3060 in terms of performance. In this case, it can be assumed that the top model Intel Arc A770 will be able to compete with the GeForce RTX 3060 Ti, and the younger Arc A580 can be a threat to the RTX 3050 .
Of course, video cards from Intel are thoroughly late. But it’s better to be late. The market sorely lacks a third strong player capable of influencing the overall landscape and development of the discrete graphics segment. Intel still has a lot to do here to gain user trust, but that’s the way to go. Let’s hope that initial difficulties and delays in the release of products will not be the reason for abandoning the development of this direction. Despite the fact that opponents are already approaching the launch of new generation products, there is still time for maneuvers in the niche of mid-range devices.
Progressive Gaming PC
|CPU||Intel Core i7-12700KF (8/16+4; 3,6/5,0 GHz, 2,7/3,8 GHz)||$410|
|AMD Ryzen 7 5800X (8/16; 3,8/4,7 GHz)||$330|
|Motherboard||Intel Z690 (LGA1700)||$210|
|AMD B550/X570S (Socket AM4)||$160|
|Memory||32 GB (2×16 GB) DDR4-3600||$130|
|Videocard||GeForce RTX 3080 10 GB / Radeon RX 6800 XT 16 GB||$850|
|Storage||SSD 2 TB, M.2 PCI-E 4.0||$250|
|Case and power supply unit||Middle Tower ATX, 750 W||$210|
|Total for PCs on the Intel platform||$2060|
|Total for PCs on the AMD platform||$1930|
With processors for progressive gaming configurations, everything is easy and difficult at the same time. As of now, the working options are easy to determine: Core i7-12700KF (8/16+8, 3.6/5.0 GHz and 2.7/3.8 GHz) for $410 and Ryzen 7 5800X (8/16; 3.8/4.7 GHz), which has dropped sharply, for $330. But in this case, the desire to use new platforms right from the start may well coincide with financial opportunities. For Intel-based systems, it may be some Core i7-13700KF or even Core i7-13700F.
Despite the fact that Raptor Lake is based on the Alder Lake architecture, the chips provide a number of architectural optimizations at the level of ring bus operation and cache memory. At the same time, L2 cache volumes will be increased both for the productive Raptor Cove cores (from 1.25 to 2 MB per core) and for the economical Gracemont (from 2 to 4 MB per 4-core cluster). The number of the latter will also be increased for each line. If earlier Core i7 received four E-cores, now the chips of this line will receive 8 such computers. The amount of cache memory of the third level will also increase and the operating clock frequencies will be significantly increased (+400–600 MHz).
As a result, according to the leaks, the specification of the same Core i7-13700KF is 8/16(3.4/5.4 GHz) + 8(2.5/4.2 GHz), 24 MB L2, 30 MB L3, while the key specs formula Core i7-12700KF – 8/16(3.6/5.0GHz) + 4(2.7/3.8GHz), 12MB L2, 25MB L3.
Even the quantitative difference is noticeable. Through optimizations, an increase in the number of E-cores and increased frequencies, the manufacturer promises an increase (SPECint_2017 test) of single-threaded performance – up to 15%, and multi-threaded – up to 41% compared to Alder Lake chips. If such statements, made as part of the Intel Tech Tour held recently in Israel, will be confirmed during practical application tests, then such an update will completely change the generation of processors. In stagnant years, 5–10% was enough for this.
During the last technical tour, the developers also emphasized that Raptor Lake will be the first processors in the world with a standard boost frequency of up to 6 GHz. Apparently, it was a limited version of Core i9-13900KS, which will be presented later to the main group. Because the Core i9-13900K available in the starting specification has a peak of 5.8 GHz. The megahertz race is getting another round. In light of this, Intel considers it appropriate to note that Raptor Lake will be able to accelerate up to 8 GHz. But there are no details about the configuration of active cores and the necessary volumes of liquid nitrogen to obtain the desired CPU-Z screenshot.
With all that, the frequency flexibility of the new Intel chips is surprising, especially if you consider that the Intel 7 (10 nm) technological process is used for the manufacture of Raptor Lake. Here it is worth paying tribute to the skill of technologists who are able to squeeze everything out of silicon wafers. But you will have to pay for high frequencies by increasing energy consumption. The basic TDP of the top chips will remain at the previous level – 125 W, but the Turbo Power (PL2) will increase slightly – from 241 to 253 W. When buying such processors, you should immediately take care of an effective cooling system.
Among the features of Raptor Lake, you can highlight the standard support for memory DDR5-5600, instead of DDR5-4800 in Alder Lake. At the same time, compatibility with DDR4 is preserved. Therefore, the 13th generation Core can be used with any motherboards equipped with the LGA1700 connector, regardless of the type of memory. You only need to update the BIOS first.
At the time of Alder Lake’s launch, most models were offered with slots for DDR4, but obviously with the release of new motherboards on Intel 700 chipsets, the emphasis will shift towards DDR5.
Along with the Raptor Lake family of processors, Intel is expected to offer new sets of logic, although this time the changes in the functionality of the chipsets are symbolic. For example, the differences between the Intel Z790 and the Intel Z690 can be singled out except for more internal lines of the PCI-E 4.0 bus, as well as support for up to five USB 3.2 Gen2x2 ports (20 GB/s), instead of four in the predecessor. Frankly, it is not enough to attribute PCH to the next generation, so the appearance of Intel 700 chipsets is rather a kind of bonus for motherboard manufacturers who have already prepared a scattering of their devices on updated chipsets. Here it will be necessary to evaluate the general equipment of specific models and compare their cost with those for similar devices on Intel 600.
When collecting a progressive configuration on AMD components, it is also worth waiting for the results of Ryzen 7000 performance tests from independent sources. The sale of new CPUs will start on September 27. There is very little left to endure.
A potentially interesting chip for a progressive configuration is the 8-core Ryzen 7 7700X (8/16; 4.5/5.4 GHz) with an MSRP of $399. Of course, retail prices at the start will probably be higher.
AMD promises a very serious acceleration for the Raphael family of chips. Models with the Zen 4 architecture received quite a few improvements, although the computing part is based on Zen 3. Deep optimizations of the caching subsystem and data presampling, double L2 volume, support for the AVX-512 instruction set. This is just a short list, which allowed to increase the clock performance (IPC) by ~13%. The transition to a 5-nanometer technological process and an increase in operating clock frequencies promise a final increase in single-threaded productivity of up to 29%. In multithreading, an increase of up to 50% is promised compared to the chips of the previous generation.
AMD is also paying attention to improving the power efficiency of the Ryzen 7000, but the overall consumption will still increase – the top 12/16-core chips have a TDP of 170 W, and a PPT (Package Power Target) of 230 W.
Ryzen 7000 received support for the PCI Express 5.0 bus and DDR5 memory. In addition, all chips in the series will include an integrated graphics core with RDNA 2 architecture, which is housed in the I/O die. The solution is somewhat simple, but the ability to do without a discrete video card is something that AMD’s performance processors lacked.
After such transformations, the need to update the platform is obvious. Therefore, to work with Raphael chips, you will need motherboards with Socket AM5 connector. Initially, top models based on AMD X670 chipsets will be offered with Ryzen 7000 processors /X670E, while the more affordable versions on the AMD B660/B660E should go on sale in October. At the same time, you need to be prepared for the fact that such devices will cost more than solutions for Socket AM4. A more expensive processor connector, a reinforced power subsystem, as well as a harness for PCI-E 5.0 will all affect the final price.
If you don’t want to delay building a system, the $330 Ryzen 7 5800X (8/16; 3.8/4.7 GHz; 32 GB L3) is still a work-balanced option for a productive configurations. For a system with an emphasis on games, you can look towards the Ryzen 7 5800X3D (8/16; 3.4/4.5 GHz; 96 MB L3). The benefit of a significantly increased L3 cache is often very noticeable here. Using the 1080p graphics mode, the average increase is 15–20%. In some projects, it reaches a simply phenomenal +35-40%, while in a number of games it is almost zero. In general, the solution is very interesting from an engineering point of view, but it should be taken into account that this model is now significantly more expensive than the basic one – about $450.
In addition, the processor with 3D V-Cache has lower operating frequencies (250–300 MHz), while overclocking of this version of the chip is not nominally provided. As a result, you should expect a slight drop in performance in multi-threaded tasks (rendering, video encoding, etc.). If such work applications are a priority for you, you can look towards the 12-core 24-thread Ryzen 9 5900X (12/24; 3.7/4.8 GHz; 64 MB L3), which on the eve of changes generations of the platform became cheaper to $400. Although, when you are ready to spend a similar amount on the purchase of a processor, it is very interesting to find out what the Ryzen 7000 will offer for this money.
When choosing a video card for a progressive configuration, you may also have to turn on the “waiting” mode and show even more patience than in the case of processors. Both NVIDIA and AMD plan to offer their next-generation discrete graphics by the end of the year. Californians are going to introduce GeForce RTX 4000 models based on GPUs with Ada Lovelace architecture, and AMD is preparing Radeon RX 7000 based on processors under RDNA 3. Both companies promise thorough acceleration.
Already on September 20, a separate session devoted to gaming discrete graphics will be held as part of the GTC (GPU Technology Conference). At the event with the promising name GeForce Beyond, the announcement of video cards of the GeForce RTX 4000 series will take place. It is likely to be the top model GeForce RTX 4090 24 GB, based on the AD102 processor, manufactured by TSMC according to 5-nanometer technology (TSMC 4N). If the leaks are to be believed, the video card will be able to offer 19-20,000 points in the 3DMark Time Spy Extreme test. This is about 90% more than the GeForce RTX 3090 (~10,000-10,500 points).
Even without taking into account internal optimizations, the new GPUs will contain a significantly larger number of CUDA computing units (more than 18,000) than their predecessors, and the GPUs themselves will also add in operating frequencies. Another powerful booster is 96 MB of L2 cache memory instead of 6 MB in the GA102.
Although the GeForce RTX 4090 24GB is nominally included in the gaming line, it is more likely to be used for professional tasks, where a high-speed GPU and a relatively large amount of memory are required. This could already be observed with the GeForce RTX 3090 24 GB.
As for the gaming GeForce RTX 4080, it will be presented in November. Again, judging by the sources, this model will be offered at once in two versions – with 16 GB and 12 GB of memory. Both versions will differ not only in the amount of memory, but also in the number of computers activated in the GPU. We have seen a similar technique before, for example, in the GeForce GTX 1060 3/6 GB series. But then the appearance of both modifications was largely spread over time, and the GeForce RTX 4080 16/12 GB will rather be presented at the same time. In terms of performance, the older model is expected to score around 15,000 in the same 3DMark Time Spy Extreme test. Simply outstanding results for the current generation lineup.
Whether the GeForce RTX 4070 will appear this year is not yet known for sure, but if we trust the information of the insiders, the model will be based on the AD104 chip with 7680 computers and will be able to give 10-11,000 points at the Time Spy Extreme stage. This is exactly the level of the GeForce RTX 3090/3080 Ti. It is too attractive to ignore the relatively fast release of new graphics cards of this class.
Information about future graphic novelties from AMD is much less. During the past presentation of Ryzen 7000 chips and the new Socket AM5 platform, AMD head Lisa Su also briefly mentioned the preparation of video cards that will appear on sale by the end of this year. We are talking about models with graphics processors based on the RDNA 3 architecture. The solutions should also be very interesting.
Top GPUs will use a chiplet layout with a main crystal made using 5-nanometer technology technology and six accompanying buffering chips on one substrate. Thus, the AMD Infinity Cache caching technology will develop. The developers will also redesign the executive modules and optimize the graphics pipeline. The progressive technical process will allow to increase the operating clock frequencies of the GPU. There is no information about the performance of future Radeons yet, as well as the terms of the official announcement. During her talk, Lisa Su only assured that the devices will be introduced later this year. Almost certainly, AMD will also start updating the graphics line from older models.
If you do not have time for a waiting position, then in the current situation there are two branches of development for the progressive configuration. Buy a video card at the level of GeForce RTX 3070 8 GB (~$630) or Radeon RX 6750 XT 12 GB (~$600) right away, and while playing calmly wait for the launch, the official start of sales, the appearance at retail and reducing prices to an acceptable level on GeForce RTX 0 or Radeon RX 7700/7800 XT. Then, if necessary/possible, upgrade to a new generation graphics with lower losses.
The second option is to buy GeForce RTX 3080 10 GB (~$850-880) or Radeon RX 6800 XT 16 GB (~$850) and enjoy gaming for a couple of years without looking back and engaging in self-flagellation on the subject of “maybe I should have waited?”
For modern configurations, we recommend equipping platforms with 32 GB of RAM. For DDR4 systems, it is logical to take a dual-channel kit DDR4-3600, focusing on minimum timings. A set of this volume with CL18 and even RGB sparkles will cost about $130. Memory of the DDR5 standard has significantly decreased in price over the past six months, but it is still not cheap. A set of 32GB (2x16GB) DDR5-5600 will cost $200, and a set of the same amount of DDR5-5200 will cost $175. For a top platform, the surcharge is already too much.
In the September “PC of the month” we have too many offers to wait. This is purely from the desire to get the best. However, recent events show how quickly priorities, desires and opportunities can change. Therefore, the best time to do something is right now.
The article was prepared with the support of the store TELEMART.UA.
Assemble your “PC of the month” with convenient configurator that allows you to quickly choose a PC, get advice, choose one of the powerful EVOLVE computers, or create your own configuration.