Review of NVIDIA GeForce GTX 1050 and GTX 1050 Ti video cards: another Pascal

On October 25, two models of NVIDIA video accelerators with Pascal architecture for budget gaming PCs went on sale. If at the time of Maxwell NVIDIA began introducing a new architecture from the lower price segment (and AMD in Polaris started from the middle), then the Pascal family unfolds in the classic sequence - from the flagship product to the most affordable.

NVIDIA's previous announcements this year provided a complete picture of what the Pascal architecture is all about. For now, NVIDIA is enjoying the lack of competition at the top end of the price range, as AMD has yet to release an equal competitor based on the next-gen process for the GeForce GTX 1070/1080. And the NVIDIA TITAN X video card allows you to imagine what the new flagship product will look like when NVIDIA deems it necessary to expand the family of gaming accelerators towards even greater performance.

The average category in terms of price and performance in the Pascal family is formed by the GeForce GTX 1060 and GTX 1060 3 GB (we consider the latter separately, since it differs not only in the amount of RAM, but also in the configuration of the graphics core). Here AMD managed to present an alternative choice in the form of Radeon RX 470 and RX 480 accelerators on the Polaris 10 chip, which have an adequate performance-to-price ratio.

Well, the latest GeForce GTX 1050 and GTX 1050 Ti are video cards for entry-level gaming PCs and HTPCs (Home Theater PCs). At first glance, they look like a modest addition to the Pascal line, not deserving of the attention that their older predecessors received. However, you should not neglect new products just because they have no value in the eyes of hardcore gamers. The GTX 1050/1050 Ti is an important phenomenon of its kind from both a practical and theoretical point of view.

Low-end video cards based on Pascal fill the niche that remained vacant in the GeForce 900 line, and from a technical point of view, they are the heirs of the GeForce GTX 750/750 Ti (GPU GM107) rather than the GeForce GTX 950. Although these video cards formally belong to the same class, NVIDIA did not release a graphics processor to replace the GM107 in the 900 line, based on the second generation Maxwell architecture (GM2xx chips). As a result, the GTX 950 based on the partially blocked GM206 chip does not match the energy efficiency of the GTX 750 Ti, which prevented the creation of equally compact video cards that lack an additional power connector. Thus, the GTX 1050/1050 Ti, unlike other 10-series models, from an architectural point of view represents a step of not one, but one and a half generations. If we add to the increased performance per watt provided by the next-generation technical process, new Pascal multimedia capabilities (HEVC hardware codec, HDMI 2.0b and DisplayPort 1.3/1.4 video interfaces), then the new products begin to look very promising.

There is a second aspect that makes the GTX 1050/1050 Ti an intriguing research subject. In the long-awaited transition of GPUs from 28 nm to 14/16 nm, we saw chip designers who in years past had consistently ordered production from TSMC's fab took different paths. NVIDIA continued to use TSMC for Pascal processors using 16 nm FinFET technology, while AMD switched to GlobalFoundries, which licensed 14 nm FinFET technology from Samsung.

NVIDIA GP107

Now the end has come for the long-term exclusive cooperation between NVIDIA and TSMC, because the documentation for the GTX 1050/1050 Ti indicates that the GP107 chip is made according to the 14 nm FinFET standard. As you know, the last time NVIDIA resorted to the services of another manufacturer was in 2003 - to produce NV3x chips from the GeForce FX family of cards. The company does not officially disclose the name of its contractor, but according to third-party sources, it was Samsung. By the way, don’t be confused by the word TAIWAN in the GPU marking (as well as the complete coincidence of the format in which the production period and stepping of the chip are encoded). Yes, Samsung's 14nm FinFET fab is located in Texas, USA, but when NVIDIA produced the NV3x series GPUs at IBM's American fab, the chips were labeled as Korean. Apparently, NVIDIA simply assembles GPUs into cases or does something else with them in Taiwan.

Be that as it may, in terms of “energy” characteristics, the GP107 may differ from its older brothers released by TSMC. In addition, the GTX 1050/1050 Ti test results will bring us closer to answering the question of why the latest AMD GPUs performed so poorly in terms of frequency potential and performance per watt. How much of the difference in these parameters between Pascal and Polaris is due to architectural features, and how much is due to manufacturing factors? However, if we make allowance for the fact that the two companies could implement the 14 nm FinFET process in different ways, then checking the GP107 in action can only remove suspicion from the technology as such, and not from the GlobalFoundries factory in New York State, USA, which is engaged in release of the corresponding microcircuits, or Advanced Micro Devices engineers.

⇡#NVIDIA GP107

GP107 fully embodies the principles laid down in the Pascal architecture in relation to gaming-class GPUs (unlike GP100, which has its own characteristics). It differs from older chips - GP106, GP104, GP102 - only in the number of computational units and the size of caches for various purposes.

The GP107 logic circuit includes two GPCs (Graphics Processing Cluster), each of which consists of a Raster Engine block that performs the initial stages of rendering (determination of polygon edges, projection and clipping of invisible pixels) and three SMs (Stream Multiprocessor). Each of the latter, in turn, contains the PolyMorph Engine geometry engine, 128 32-bit CUDA cores, 4 64-bit CUDA cores, 8 texture units and an L1 cache section. Like other gaming GPUs in the Pascal family, GP107 differs from Maxwell in its increased L1 capacity (from 24 to 48 KB).

Thus, in terms of shader and texture throughput per megahertz frequency, the full-featured GP107 has taken a step forward relative to the GM107 and corresponds to the “chopped” GM206 in the GeForce GTX 950. The back-end of the processor is represented by an L2 cache, a ROP array and four 32-bit controllers memory, there are also obvious changes.

Due to the fact that the Pascal architecture has progressive color compression, and the controllers allow working with memory at a significantly higher frequency than in the GM107 (7000 instead of 5400 Mbit/s per pin), the developers needed to double the number of ROPs compared to the GM107. At the same time, it became possible to save transistor budget on second-level cache memory (1 MB versus 2 MB in GM107).

Block diagram of NVIDIA GP107

Nevertheless, the processor turned out to be quite large for its category: the number of transistors in the GP107 is 3.3 billion. Of course, part of this volume was spent on the implementation of numerous new functions presented in the Pascal architecture, therefore, in terms of the number of transistors, the GP107 not only “swelled” significantly relative to GM107 (1.87 billion), but also ahead of GM206 (2.94 billion). Here is a short list of innovations in the Pascal architecture:

• improved color compression with ratios up to 8:1;
• the Simultaneous Multi-Projection function of the PolyMorph Engine geometric engine, which allows you to create up to 16 projections of scene geometry in one pass (for VR and systems with multiple displays in the NVIDIA Surround configuration);
• the ability to interrupt (preemption) during the execution of a draw call (during rendering) and the command flow (during calculations), which, together with the dynamic distribution of GPU computing resources, provides full support for asynchronous computing (Async Compute) - an additional source of performance in games running the DirectX 12 API reduced latency in VR;
• hardware decoding and encoding of video in H.264, H.265 (HEVC) and VP9 formats;
• display controller compatible with DisplayPort 1.3/1.4 and HDMI 2.b interfaces. High dynamic range (HDR) support;
• SLI bus with increased bandwidth (the corresponding blocks are most likely excluded from the GP107 due to the lack of SLI support).

⇡#GeForce GTX 1050/1050 Ti: technical specifications, prices

Based on the GP107 processor, NVIDIA has released two video cards, of which the GTX 1050 Ti is equipped with a fully functional core and 4 GB of RAM. The GTX 1050 has one of the six SMs disabled, and the standard video memory is 2 GB.

The first thing that catches your eye when you look at the specifications of the video cards is the conservative clock speeds compared to those of the GTX 1060 and older models. Until NVIDIA released detailed documentation for the GTX 1050/1050 Ti, it caused some confusion in the press. It's not hard to jump to the premature conclusion that Samsung's shift to 14nm FinFET production line infected NVIDIA GPUs with the same disease that plagued Polaris. However, the developer claims that the limiting factor was the thermal package, set at 75 W, so that NVIDIA partners could release video cards that do not require additional power. In the empirical part of the review we will see what frequencies the GP107 can actually reach.

It is also interesting that the GTX 1050 has increased GPU frequencies compared to the “titanium” version. According to NVIDIA, the younger of the two new products has three times the performance in games than the GeForce GTX 650, and is slightly more than 50% faster than the GeForce GTX 750 Ti.

Official prices for the GTX 1050 and GTX 1050 Ti are set at $109 and $139, respectively (US market, excluding sales tax). Russian - 8,490 rubles. and 10,490 rub. respectively.

However, the market niche that the new products are aimed at has already been captured by the competitor’s competing products. AMD countered the release of the GeForce GTX 1050/1050 Ti by reducing prices for modifications of the Radeon RX 460 with 2 and 4 GB of RAM - from $109/139 to $89/99. A new Russian price has been announced for the junior version of the video card - 7,530 rubles.

NVIDIA has not prepared a reference version of the GeForce GTX 1050/1050 Ti, unlike the GeForce GTX 750/750 Ti. Instead, manufacturers will begin selling video cards of their own design from the first day of sales. For testing, we were provided with samples of the MSI GeForce GTX 1050 2G OC and GeForce GTX 1050 Ti 4G OC - fairly simple implementations of the new models with slightly increased GPU clock speeds: 1,404 / 1,518 MHz and 1341 / 1455 MHz for the GTX 1050 and GTX 1050 Ti respectively. In testing, the clock speeds of MSI were brought to reference values.

So far, only the older model has been found in retail at prices of $140–150 (newegg.com) and 15,200 rubles. (market.yandex.ru).

ASUS GeForce GTX1050 Ti 4096Mb CERBERUS OC

Average in all respects - both size and performance. No additional power is needed. The core operates at 1366 MHz, video memory at 7008 MHz. There are DisplayPort, HDMI, DVI-D ports and a pair of medium-sized coolers.

The device is generally unremarkable, even in design, which is why I included it in today’s review. This, one might say, is the quintessence of this model.

If you are too lazy to choose which brand of video card to buy and scrupulously compare the numbers, but you need the GTX1050 Ti, take this modification - you definitely won’t go wrong.

⇡#MSI GeForce GTX 1050 2G OC and GTX 1050 Ti 4G OC: design

In appearance, the two MSI video cards are exactly the same. The compact board is equipped with a cooling system of the simplest design: a monolithic milled radiator and a single large impeller.

In this case, it makes no sense to describe in more detail the appearance of the device and the design of the cooling system - the photographs speak for themselves.

How much does one GTX 1050 Ti video card bring?

The information in this block is approximate and may be out of date by the time you read it. What is written here is relevant at the end of 2022. Then, due to changes in the difficulty of mining, power consumption, the cost of cryptocurrencies on the market, and so on, the situation may change.

Today, you can squeeze out about 30 net cents per day from this video card (based on the cost of electricity $0.05 per kWh). Of the dirty income, electricity costs will be about 25%.

⇡#MSI GeForce GTX 1050 2G OC and GTX 1050 Ti 4G OC: board

The only thing that distinguishes the GeForce GTX 1050 Ti from the GTX 1050 in this series of video cards is RAM chips with twice the capacity. In other respects, the devices are identical, including a four-phase power system: three phases for the GPU and one for the RAM chips.

The memory chips are manufactured by Samsung and are designed for an effective frequency of 7 GHz.

⇡#Test stand, testing methodology

What kind of power supply does the RTX 3080 need?

According to tests by experts from Tom's Hardware, the Nvidia GeForce RTX 3080

has a consumption of 335 watts.
You will need power supply
if you plan to use this video card in conjunction with an Intel Core i5 or i7 processor, or AMD analogs such as Ryzen 5 and 7 processors.

Interesting materials:

How many grades are needed for certification in 10th grade? How many marks are needed for a quarter? How much screening is in 1m3? How much paragraph indent in Word? How much to retreat from the fence when building a gazebo? How much release clause does Messi have? How many push-ups per set? How many toes does a German Shepherd have on its hind feet? How many games must a team win to win? How many games must a team win to win the match?

⇡#Performance: 3DMark

The 3DMark results, which predict the balance of power in further gaming tests, allow us to draw several conclusions.

1. The two GP107-based video cards are not that different - the GTX 1050 Ti has a 14% advantage.

2. In addition, the GTX 1050 demonstrated the same performance as the GTX 950 (the new one is 4% faster). Compared to the GTX 750 Ti, the performance increase was 54%.

3. But for the Radeon RX 460 there is bad news: the GTX 1050 was 13% faster, and the GTX 1050 Ti was 28% faster. The DirectX 12 API in Time Spy gives the AMD graphics card a marginal advantage compared to DX11 tests.

Note: Fire Strike Extreme data is excluded from percentage calculations because 2160p resolution is not an adequate use case for entry-level graphics cards, and results for graphics cards with 2GB RAM are limited by memory capacity, not GPU performance.

Is it possible to use a more powerful power supply?

You can buy a power supply with more power than your original one, the main thing is not less. That is, if you have a current of 3.42 A, then you can take a charger with a current of 4.74 A or 6.3 A - it will only be better. Another important feature of the block is the connector.

Interesting materials:

Is it possible to cover a bathroom with plastic panels? Is it possible to apply for a pension at a location other than your place of registration? Is it possible to cool a room using a refrigerator? Is it possible to leave water in the Steam Cleaner? Is it possible to refuse alimony in Belarus? Can coconut oil be washed off? Is it possible to wash a pen from clothes? Is it possible to boil champignons? Is it possible to transfer units from kcell to asset? Is it possible to move objects with the power of thought?

⇡#Performance: games (1920 × 1080, 2560 × 1440)

In gaming tests, the different amounts of RAM in the GTX 1050 and GTX 1050 Ti create a more extreme situation. Even at 1080p resolution without full-screen antialiasing, the GeForce GTX 1050's lack of RAM makes itself felt. On average, frame rates on the GTX 1050 Ti are 32% and 43% higher than the smaller model at 1080p and 1440p, respectively. In extreme cases (DOOM), the GTX 1050 can be 2.56 to 5 times slower than the GTX 1050 Ti!

As in synthetics, the GTX 1050 is equivalent to the GTX 950, boasting a 6% and 3% advantage at 1080p and 1440p, respectively. The difference with the results of the GTX 750 Ti is 40 and 37%.

As for the Radeon RX 460, the GeForce GTX 1050 Ti, which has the same amount of RAM, has an advantage of 25 and 27% at 1080p and 1440p resolutions, respectively. Unfortunately, during the testing period we did not have a sample of the RX 460 with 2 GB of memory to compare with the GTX 1050 on equal terms, but the fact is that the RX 460 4 GB is on average only 5 and 13% faster than the younger new product from NVIDIA , does not speak in favor of the AMD product. Games that are content with 2 GB of memory also prefer the GeForce GTX 1050 separately.

Note: DirectX 12 in Total War: Warhammer is not supported on GeForce GTX 750 Ti.

⇡#Performance: video decoding

GP107 is equipped with exactly the same video decoding unit as the older GPUs of the Pascal family, adjusted for slightly lower performance due to lower frequencies. However, the hardware H.264/H.265 decoder in NVIDIA chips is still the best in its class.

Note: since decoders within the same GPU line usually do not differ, the diagrams show one device from each family (or more if this rule is violated).

⇡#Performance: Computing

GeForce GTX 1050 and GTX 1050 Ti took the top lines of the charts in most tasks. The exception was Sony Vegas 13 Pro, where AMD GPUs traditionally lead. Additionally, the Radeon RX 460 performs double-precision operations at 1/16th of FP32 (rather than NVIDIA's 1/32nd), which earned it a win in the corresponding Sandra test.

Sapphire Radeon RX 460 2GB: for the budget conscious

AMD also offers modern video cards without an additional power connector. However, here you should be a little more careful: for example, the Radeon RX 460 comes with or without a 6-pin power connector. A cheap (about 8,000 rubles) Sapphire Radeon RX 460 with 2 GB of video memory requires only 75 W, so simply connecting through the slot will be enough and no additional power cord will be needed.

The card offers 896 shader units, 56 TMUs and 16 ROPs. With its 2 GB, it is only conditionally suitable for games, which reflects its position in our corresponding rating. At the same time, the Polaris architecture from AMD, on which the RX 460 is based, has great potential for working with DirectX 12. Therefore, most likely, this model will show full performance only when a sufficient number of games with DX12 support appear on the market. However, even until then it is only a conditional alternative to the GeForce GTX 1050 (Ti).

Sergey Plotnikov

⇡#Clock speeds, power consumption, temperature, overclocking

As it turns out, the clock frequencies of the new video cards did not suffer at all from the transition to the Samsung 14 nm FinFET process technology. In Crysis 3, the GeForce GTX 1050 maintains a stable frequency of 1633 MHz, and a peak frequency of 1709 MHz. For the GTX 1050 Ti, the values ​​are 1595 and 1696 MHz, respectively. However, take into account that MSI has increased the frequency range of video cards by unlocking additional steps in the GPU Boost table: the difference between Boost Clock and Base Clock in the GTX 1050 and GTX 1050 Ti compared to the reference values ​​is 4 and 12 MHz.

Both cards have a peak GPU supply voltage of 1.1V, which is slightly higher than the other Pascal chips on the reference boards (1.05-1.062V).

Although the GTX 1050 and GTX 1050 Ti technically have the same TDP, the latter in practice develops slightly more power. Both cards are slightly superior to the GeForce GTX 750 Ti in this regard, judging by measurements in Crysis 3. However, FurMark, which isolates the GPU as the main power consumer from the central processor, shows that Pascal, on the contrary, has lower power consumption compared to the GTX 750 Ti . And of course, the GTX 1050/1050 Ti compares favorably with the GTX 950 and Radeon RX 460, which allowed the new products to do without an additional power connector.

The temperature of the video cards under load does not exceed 64–66 °C, which also indicates the energy efficiency of the GP107 chip, since the test samples from MSI have a fairly simple cooling system. For comparison, the Radeon RX 460 manufactured by SAPPHIRE requires a more serious cooler, and heats up much more.

The prospect of overclocking the GP107 without additional power can be viewed with skepticism. After all, the BIOS of video cards does not allow you to increase the power limit one iota (although the supply voltage is allowed to be increased by 100 mV). Nevertheless, we managed to increase the base frequency of the GTX 1050 and GTX 1050 Ti by 220 and 260 MHz, respectively, up to 1574 and 1550 MHz.

Under load, the peak frequencies of overclocked video cards reached 1911 MHz, but stable frequencies are in the range of 1811–1835 and 1747–1797 MHz for the GTX 1050 and GTX 1050 Ti, respectively. Thus, good base clock overclocking and high peak clocks will not improve application performance that much. The whole point, of course, is the power limit of 75 watts. Even in normal mode, monitoring regularly signals that the GPU frequency is hitting the TDP limit, so with an additional power connector we might have achieved more impressive results. At the same time, the maximum GPU supply voltage on the GTX 1050 increased to 1.112 V, while on the GTX 1050 Ti, on the contrary, it decreased to 1.05 V (apparently, the power limit worked).

The overclocking utilities we used (MSI Afterburner and ASUS GPU Tweak II) in current versions do not allow us to set the effective RAM frequency above 8008 MHz on the tested GTX 1050/1050 Ti samples, but at this frequency the video cards worked absolutely stably.

Overclocking had the most insignificant effect on the system power, because stable frequencies under load did not increase as significantly as the base ones. But the cooling system installed by MSI is clearly not adapted to the increased load - even at full rotation speed of the impellers, the GPU temperature after overclocking decreased by only 6-9%.

Review and testing of the Nvidia GeForce GTX 1050 video card (page 2)

Detailed data

To evaluate the behavior of video cards in nominal operating mode, we will turn to the more than hour-long 3DMark Stress test, in parallel we will collect data from the graphics core, as well as values ​​of memory frequencies, power consumption and temperatures. We combine the obtained data into single graphs and see how closely the factory characteristics correspond to real figures, and what happens to important indicators during overclocking.

Core frequency.

The declared operating frequencies of the graphics core must fall within the range between 1354 and 1455 MHz. Without overclocking in games, frequencies fluctuate between 1493 and 1645 MHz, reaching an average of 1609 MHz.

Let me remind you that the previously tested GeForce GTX 1050 Ti, whose frequency range is shifted down, showed an average of 1538 MHz. For MSI video cards, the functions of increasing voltage and TDP are blocked, so the overclocking process comes down to simply increasing the frequency in the hope that GPU Boost itself will adjust the remaining parameters.

You shouldn’t dream of a simple 2 GHz; due to strict energy savings, the GP107 will more often be overclocked at a frequency of about 1700-1800 MHz, without a significant deterioration in power consumption. Our sample showed a minimum frequency in tests of 1620 MHz, a maximum of 1785 MHz, and an average of 1749 MHz.

Memory frequency.

The video card is equipped with 2 GB of GDDR5 SDRAM memory, located in four chips on the front side of the printed circuit board.

SK Hynix microcircuits are designed for a nominal operating frequency of 1750 MHz, but they can easily overclock to 2 GHz. For now, the BIOS has a built-in limitation on the maximum memory overclocking at exactly 2 GHz. An attempt to set a higher value leads to instability of the video card, even in 2D.

Typically, such memory can be overclocked to 2000-2400 MHz. Considering the 128-bit data exchange bus, games with high bandwidth requirements will see good dividends.

The GP107 graphics core turned out to be very cold. The cooling system, which is elementary in design, consisting of a rectangular aluminum radiator without heat pipes and a copper core, does not allow it to heat above 65C.

The voltage during the test fluctuates between 0.925 V and 1.031 V (average 1 V). During overclocking, GPU Boost uses the remaining TDP reserve and raises the bar to 1.045V (0.962-1.093V).

Initially, I thought that the energy-efficient GPU would be the first core in history to operate at less than 1 V. Alas, my assumptions turned out to be incorrect. However, I think that passively cooled GTX 1050 video cards will indeed make do with less voltage in order to maintain a low TDP.

Electricity consumption level

Energy consumption, W Average
Enable JavaScript to see graphs

At standard frequencies, the GeForce GTX 1050 consumes about 64 W. Overclocking to 150 MHz adds another 3 W. It is noteworthy that during testing the video card never exceeded the 75 W limit.

Test results

Video card frequencies and driver versions

Test participants

The @overclockers_news Telegram channel is a convenient way to keep up with new materials on the site. With pictures, extended descriptions and no advertising. overclockers.ru

⇡#Performance: overclocking

Pascal architecture GPUs require a very significant increase in frequency for overclocking to bring a noticeable increase in frame rates. Additionally, while our GTX 1050/1050 Ti samples overclocked well at base GPU clocks, dynamic clocks remained limited by the power limit. As a result, performance increased only slightly.

The GeForce GTX 1050 gained an average of 7% in 3DMark and 9% in gaming tests. For the GTX 1050 Ti, the results are 11% in 3DMark and 11-12% in games.

⇡#Test results

Overclocking details are described separately for each of the eight models. For greater clarity, I have summarized all the results in a single table.

As expected even before testing, the best results in overclocking will be shown by GIGABYTE and MSI models, since these video cards are equipped with additional power and an improved (relative to the reference GeForce GTX 1050 Ti) VRM zone. In most cases, video cards without additional power showed similar overclocking results. The exceptions were the ASUS CERBERUS-GTX1050Ti-A4G and Palit GeForce GTX 1050 Ti Dual models. The first showed the worst results in overclocking. However, again, perhaps I was just unlucky with the sample. The Palit model has an additional 6-pin power connector, but does not have the ability to increase the power limit.

Overclocking the GPU and memory increases the performance of the GeForce GTX 1050 Ti in the 3DMark Fire Strike and Time Spy benchmarks by 4-15%, depending on the model. The average was 10%.

Finally, overclocking a video card such as the GeForce GTX 1050 Ti does not significantly affect the increase in power consumption, noise and heating of the main components of the device. The quietest models were GIGABYTE GV-N105TG1 GAMING-4GD and MSI GeForce GTX 1050 Ti GAMING X 4G. Obviously, the cooling systems of these adapters are redundant for the GP107 GPU. The noisiest model was the GeForce GTX 1050 Ti Dual. At the same time, the simple Palit video card cooler cools the GPU extremely efficiently. It turns out that we can independently adjust the operation of the fans so that the processor does not heat up too much and the noise level is reduced to a comfortable level.

In games, in all cases, there is, in my opinion, a very insignificant difference between video cards, so choosing a GeForce GTX 1050 Ti based solely on frequencies is wrong. In addition, Pascal architecture GPUs require a very significant increase in frequency for overclocking to bring a truly noticeable increase in FPS. An average increase of 10% is, of course, pleasant, but, for example, in FarCry 5, a difference of 4-6 FPS does not make a difference, alas. At the very beginning of the article, I already noted that a more effective solution would not be to overclock the video card components, but to reduce the quality of graphics in a particular application.

And yet, the most productive models were the GIGABYTE GV-N105TG1 GAMING-4GD and MSI GeForce GTX 1050 Ti GAMING X 4G.

⇡#Conclusions

GeForce GTX 1050/1050 Ti is a timely and successful addition to the existing line of NVIDIA desktop accelerators based on Pascal architecture. Being successors to the GeForce GTX 750/750 Ti rather than the GTX 950, the new products inherited the advantage of their predecessors - power consumption within 75 W, which allows NVIDIA partners to produce GP107-based video cards with a simple and compact (optionally, even passive) cooling system. And thanks to the absence of an additional power connector, the GTX 750/750 Ti is suitable for upgrading office computers that do not have an appropriate cable.

At the same time, the GTX 1050/1050 Ti variants, powered only by the PCI Express slot, are not able to unleash the full potential of the GP107 chip. As tests have shown, abandoning the 16 nm FinFET photolithography proven by previous Pascal processors in favor of 14 nm FinFET using Samsung technology did not harm the energy efficiency of the GPU and the potential of clock frequencies. Even on a board without additional power - such as the MSI GeForce GTX 1050 2G OC or GTX 1050 Ti 4G OC that participated in testing - the GP107 can reach frequencies above 1900 MHz when overclocked, but cannot maintain them precisely due to power limitations . Overclocking such video cards is perhaps a pleasant, but rather pointless activity from the point of view of practical “exhaust”. The GTX 1050/1050 Ti versions with additional power, which have already appeared, will certainly perform much better in this regard.

In addition to impressive energy efficiency, the GTX 1050/1050 Ti has a full set of functional innovations presented in the older models of the Pascal family, including both graphics pipeline optimizations and multimedia functions. DisplayPort 1.3/1.4 and HDMI 2.0b interfaces allow you to display images with resolutions up to 8K and high dynamic range (HDR). Of course, video content that can use new interfaces is still far from mass distribution, but in encoding and conversion tasks, the GTX 1050/1050 Ti can replace disproportionately more expensive multi-core CPUs thanks to the leading performance of the hardware codec of the H.264/HEVC/VP9 standards. However, if multimedia functions are all you require from a video card, then NVIDIA, as unofficial sources report, is preparing to release an even more compact GP108 processor, which, logically, will form the basis of the GeForce GTX 1040.

In terms of gaming performance, both of the reviewed graphics cards have an advantage over the 2GB and 4GB RAM versions of the Radeon RX 460, respectively, and the recent price cut by competitor NVIDIA reflects this quite clearly. However, as with the Radeon RX 460, memory size is critical here. Formally, video cards of this class are designed for undemanding online games, and in games with advanced graphics it is necessary to lower the settings to ensure a comfortable frame rate. Nevertheless, even among modern AAA games, some have enough power from the GP107 chip itself to run at high settings and with an acceptable frame rate, but 2 GB of RAM is absolutely not enough. Thus, choosing the youngest of the new products allows you to save $30 (a significant amount when the device costs $109/139), but you must be aware that many possibilities are already closed even for a budget video card with 2 GB of RAM.

About quality

Quality is the main characteristic of any power supply. In the dashing 90s, a good provider of all components was judged by its weight, but the cunning Chinese quickly learned to weigh down the body using bolts mounted on adhesive tape. Nowadays no one is doing such “life hacks”, and there is no need for it. For a long time now, the weight of the power supply does not mean anything. In 2016, it is very difficult to determine the build quality by external signs. "Metal box" "With a fan." “With a bunch of braided wires.” “Oh, this is where the cables come off!” In fact, only the manufacturer’s name, price and efficiency classification 80 PLUS speak about the reliability of the unit.

All three points, as you yourself understand, are interconnected. So, cheap devices from SeaSonic, ENERMAX and be quiet! You won't find it on sale. Yes, the computer power supply market is quite conservative. Therefore, the name plays a big role. The paradox is that there are not very many power supply manufacturers themselves, and companies producing ready-made solutions under their own logo are a dime a dozen. Here everything depends on the office itself, namely on how closely it monitors the production process. Therefore, sometimes sad situations happen when at first the quality of a particular model remains at a consistently high level. And she gains well-deserved popularity. But after a year or two, under the same name, they release a block on a different platform with a modified “filling” of worse quality. Below is a table of reliable manufacturers; it includes those who are more or less represented in Russian retail. Selling expensive blocks costing 10-30,000 rubles in times of crisis is difficult. Therefore, retail is predominantly filled with simpler solutions. The same ENERMAX in Russia is almost “over”. But, as Ostap Bender said, abroad will help us!

As another movie character said, shit happens. Unfortunately, not a single power supply manufacturer is insured against defects. If this or that company is not on the list, then either it is poorly represented in our country, or I do not trust it. Some companies are very widely represented in the domestic market. For example, Corsair. This company sells both cheap and very expensive elite blocks. That is why the “corsairs” fell into the second category. There are plenty of such manufacturers. For example, I recently tested the 1.2-kilowatt MasterWatt Maker 1200 from Cooler Master. This is one of the first units with an ARM processor and a removable Bluetooth module. Certified to 80 PLUS Titanium standard.