Die drei großen Grafikkartenhersteller aktualisieren gemeinsam ihre Einsteigerprodukte: AMD schneidet am schlechtesten ab, hat Intel sich nun endlich aufraffen können?
This year's PC market can truly be described as "lively," especially in the graphics card segment. Apart from a plethora of mid - to high - end graphics cards, NVIDIA, AMD, and Intel have all launched entry - level desktop graphics cards again after several years. This has brought rare popularity to a market that is often overlooked by mainstream public opinion.
Although many gamers don't think highly of entry - level graphics cards, in terms of sales, this level of graphics cards actually has a stable group of buyers, such as corporate purchasers, online game players, and content creators. Moreover, with the overall improvement of graphics card performance, the performance of entry - level graphics cards has also changed significantly. They are no longer the kind of devices that struggle to run even small games.
Since not all players have enough budget to get a sweet - spot graphics card like the RTX 5060, Lei Technology intends to have a good chat with you today. Is it worth buying an entry - level graphics card? And what are the differences in performance and actual performance among different brands?
The three giants make their moves. Who is the strongest under $2000?
Let's first briefly go through the real entry - level graphics cards of each brand. First up is NVIDIA's RTX 5050. It has been more than three years since the last desktop release of this model (RTX 3050).
In these three years, NVIDIA has been "milking" the RTX 3050 for all it's worth, releasing four different versions: RTX 3050 4 GB (Ampere, GA107), RTX 3050 8 GB (Ampere, GA106), RTX 3050 8 GB (Ampere, GA107), and RTX 3050 6 GB (Ampere, GA107). There is even an alleged RTX 3050 A that will use the new - generation AD106 core. I wonder if the RTX 5050 will follow in the footsteps of the 3050 and create a new family series later. Well, anyway.
RTX 5050: Demonstrating NVIDIA's precise cuts
Image source: NVIDIA (NVIDIA was even too lazy to create a more detailed model for the RTX 5050)
Looking back at the RTX 5050, as the current lowest - end model in the 50 series, it still uses TSMC's 4N process + Blackwell architecture. It has 2560 CUDA cores, 8GB of DDR6 video memory, a 128 - bit bandwidth, a transfer rate of approximately 320GB/s, a base clock frequency of 2.31GHz, and an accelerated clock frequency of up to 2.57GHz. It is also equipped with the fourth - generation RT cores, which significantly improve ray - tracing performance, and it also supports DLSS4.
In addition, the RTX 5050 is equipped with the fifth - generation Tensor cores, with an AI computing power of up to 421 TOPS. It is also equipped with the ninth - generation NVENC (hardware encoding) and the sixth - generation NVDEC (hardware decoding). The output interfaces include HDMI 2.1b and DisplayPort 2.1b, which can support a maximum 8K resolution output. The rated TDP is 130W.
Judging from the hardware parameters, the RTX 5050 still has some highlights. For example, its AI computing power reaches 421 TOPS, which is similar to that of the RTX 4070. However, the 8GB of video memory limits the performance of this graphics card in the AI field. Additionally, the hardware codecs are the latest models in the 50 series, and their performance in rendering and editing acceleration is also better than that of the 40 - series graphics cards.
However, those who are familiar with the parameters of the previous - generation RTX 3050 can probably spot something amiss at a glance. How come the number of CUDA cores and the video memory specifications are still the same for the RTX 5050? I can only say that this is NVIDIA's precise strategy. The performance improvement brought by the new - generation core is just right. Adding more CUDA cores would result in "excessive performance."
RX 7400: So poor in performance that gamers can ignore it
Image source: AMD
Now, let's take a look at its main competitor, the RX 7400. This graphics card is based on the RDNA 3 architecture, which is the old RX 7000 series. It has only 28 CU units (1792 in total), 8GB of GDDR6 video memory, and a 128 - bit bandwidth with a transfer rate of 173 GB/s.
Actually, it's not hard to tell from the hardware specifications that the performance of the RX 7400 is very average. It's basically a cheap online - game graphics card. If you have gaming needs, it's recommended to just ignore it. The highlight of the RX 7400 lies in its extremely low price. It is speculated that the cost of a single card is within $200, but it supports interfaces such as HDMI 2.1, which can provide good audio - visual entertainment support.
Moreover, the single - card power consumption of the RX 7400 is only 55W, which means it can be used directly without an external PCIe power supply. Coupled with the low heat - dissipation pressure brought by the low power consumption, it is very suitable for use in mini - PCs and can provide users with a certain level of GPU performance for auxiliary rendering, editing, and other needs.
However, that's about all the highlights of this graphics card. Perhaps AMD also thought that this graphics card wouldn't perform well in the consumer market. Since its release, the RX 7400 has only been sold as part of an OEM system.
B380: It's a good deal if it costs less than $200
Image source: Twitter
Finally, there's Intel's B380. From the currently announced information, it will most likely have 16 Xe cores and be paired with 8GB of video memory, with a price within $200. Although there isn't much exposed information, combining the performance of the B570 (18 Xe cores) and B580 (20 Xe cores), we can have some expectations for the B380.
Taking the B580 as an example, the performance of this graphics card is 17% higher than that of the RTX 4060. It also supports the new - generation XeSS multi - frame generation. In supported games, the actual frame rate can be increased by more than 20% compared to the RTX 4060, which means it can meet the running requirements of most games at a 1080P resolution.
Based on the performance of the B580, the performance of the B380 should probably be similar to that of the RTX 4060. If its price is within $200 (about 1436 RMB), then its cost - performance ratio is quite prominent. After all, the price of a second - hand RTX 3060 on Xianyu is still stable at around 1300 RMB. One is a brand - new card, and the other is a card that has been used for three or four years and has worse performance. It's not hard to make a choice.
Overall, there is an obvious divide among the RTX 5050, RX 7400, and B380. The price of the RTX 5050 is more than $50 (25%) higher than the latter two. For ordinary users, since they are considering entry - level graphics cards, this price difference has a significant impact.
However, for those with limited budgets, are these graphics cards really the best choice?
Don't buy entry - level for gaming, and don't game with entry - level
Although the three major graphics card manufacturers have rarely updated their entry - level graphics cards, the market response has not been very enthusiastic. In Xiaolei's opinion, only the RTX 5050 may have good sales, followed by the B380, and the RX 7400 will most likely be "non - existent" in the consumer - grade market.
Therefore, the following analysis will mainly focus on the RTX 5050 and B380. First, the current price of the RTX 5050 is around 2000 RMB, which is 300 - 400 RMB more than the official price, further reducing its cost - performance ratio.
However, as the leading product in the current graphics card market, the RTX 5050, with NVIDIA's DLSS 4 and the new - generation AI cores, has an "undefeatable" position in the entry - level market in advance. Graphics cards cheaper than it have worse AI performance, and if you can afford a more expensive one, well, there's not much to say.
Image source: 4Gamers (If budget allows, who wouldn't like the 5090?)
At least in the $2000 price range, if you're looking for a graphics card with relatively balanced performance in all aspects, the RTX 5050 is actually worth considering, provided that you really don't have an extra 500 RMB to buy an RTX 5060. On the other hand, the price of the RTX 5050 should remain stable in the future. Even if you use it as a transitional graphics card, you won't lose much when selling it second - hand.
As for the B380, the biggest problem with Intel's graphics cards is their low popularity. They seem to have the strength but don't know where to use it. To be fair, Intel's B - series graphics cards are all cost - effective. Whether you use them for gaming or creation, the experience is not bad. Moreover, Intel is very generous with the video memory, so it's also a good choice for "AI - art generation" (running AI image - generation software).
If your budget is less than 1500 RMB, then the upcoming B380 may be your only choice, or if you're willing to accept a second - hand graphics card, then the RTX 3070 and 3070Ti are also good options. Moreover, Intel's graphics cards are basically not over - priced, and you can probably buy them at the official price, which makes them more cost - effective compared to other "over - priced" graphics cards.
Image source: NVIDIA
To be honest, it's basically impossible to find a "highly cost - effective" graphics card at this stage. Intel and AMD have basically shifted their focus to AI professional cards, and the competition in the consumer - grade graphics cards is mainly concentrated on mid - to high - end models. Looking around, actually, only the RX 9070XT this year has been a pleasant surprise.
For most users, there's no need to be overly concerned about "buying new instead of old." DLSS 4 is not a panacea. Based on Xiaolei's own experience, if the DLSS multiplier is set too high, problems such as latency are very obvious, and even single - player games can be a bit of a struggle.
Therefore, Xiaolei still recommends that gamers give priority to second - hand graphics cards. For example, AMD's RX 7000 series is a good choice. Although its ray - tracing performance is poor, its rasterization performance is very strong, and with the support of large - capacity video memory, it is more friendly to 4K gamers.
As for these entry - level graphics cards, it's still not recommended for gamers to consider them. They are more suitable for audio - visual entertainment users and content creators and are at least qualified as hardware accelerators.
This article is from the WeChat official account "Lei Technology," and is published by 36Kr with authorization.