Some poor friends may have a rough classification of laptops: those who focus on business are called thin and light books, and those who focus on performance are called game books. In the case of processor performance is similar, I like to divide from the graphics card.
With the continuous exploitation of manufacturers' technologies, today's thin and light books have also begun to embark on the road of high performance. The stereotype that light and thin cannot play games has been broken, such as the light and thin version of the GTX 1650 and RTX 3050 editions that are very popular this year...
But in real terms, the laptop's RTX 1650 graphics card is also comparable to the game entry, 3050 although with light chase, but the effect is really a bit of a pull.
So some time ago, graphics card giant NVIDIA quietly released three notebook graphics cards, namely: RTX 2050, MX 550 and MX 570.
Then I'll be strong. As we all know, in NVIDIA's N card camp, the performance of RTX 2060 is higher than that of 3050, so as a member of the 20 series, how to look forward to next year's high-performance thin and light cost performance to a higher level.
Thin and light performance is about to take off?
After reading the general parameters of the GeForce RTX 2050 graphics card, I just want to say one thing to NVIDIA: "Worthy of you!" Leather knife guest Yellow boss! ”
This year, the knife technique has improved a lot. Cutting out half of the RTX 3050's video memory is a new product, so can I also say that it will become the new "IQ test card"?
We are not rushing to conclusions, and the messengers listen to me slowly talk about the doorway here. From the parameter table, the CUDA core size of the "RTX 2050" looks larger than that of the 2060...
This is also a place where many people are surprised, if you know nothing about graphics cards, then it is likely that people will play you like a monkey.
The larger the core size, the stronger the graphics card performance. There is nothing wrong with this conclusion per se, but there is a prerequisite for knowing its core architecture.
N cards generally have two architectures, and the previous 20 series graphics cards used turing architecture. However, the "RTX 2050" card uses the same Ampere architecture as the 30 series graphics card.
Lao Huang did not clearly mark this in the parameter table, just to play a "word game" with everyone. So what is the impact of these two architectures on graphics cards?
Ampere architecture ▼
Turing architecture ▼
The biggest difference between these two different architectures is the arrangement of the core units. The ampere architecture is in the form of "FP32 + INT32", and the two units of "FP32" and "INT32" in the Turing architecture are separated.
To clarify the meaning and relationship between these two units is slightly more complicated, you can simply understand it as two different job responsibilities.
For example, in the case of "Editing + Operation", the Turing architecture is equivalent to the editing and operation of the department to perform their respective duties; the ampere architecture is that some people must be responsible for both editing and specific operations.
Then, when counting how many people there are in this department, Boss Huang used a very cryptic way: counting and accumulating the number of people in the editorial work and the operation work separately, and replacing the actual number of people with the number of positions.
That is, the number of CUDA cores that mainly determines the performance of the graphics card, and a portion of the ampere architecture is calculated twice. Then the CUDA core size of "RTX 2050" should be much lower than that of the 2060 graphics card of the same series.
More interestingly, if you look up the RTX 3050 graphics card previously released by Lao Huang, you can find that in addition to the memory width of the "RTX 3050" is 128 bits, which is twice as high as the 2050, and the change in the frequency of the memory, its core scale, architecture and other aspects are almost exactly the same as the new card 2050.
Therefore, this graphics card is basically not comparable to 2060, and it can almost be said to be completely hanging in terms of performance.
Given the 3050's very weak light tracking performance, I don't think bad friends have to expect too much from the light tracking effect of the new product. If you don't turn on light chasing, it is also a "GTX 1650" level, which is not much different from this year's product.
In terms of parameters, RTX 2050 should have been held wrong when it was born, it should be called RTX 3040... Or according to the difference between GTX and GT, this new card called RT 3040 is more accurate. （ bushi
Unfortunately, there have been no "xxx xx 40" graphics cards in recent years.
What about the other two cards, the MX 550 and MX 570?
The MX 550 card is obviously an upgrade of the previous generation MX 450, the difference is that the previous generation supports GDDR5 and GDDR6 memory specifications, and this generation has unified into the GDDR6 form.
In addition, the GeForce MX 550 has only a little more core count than its predecessor (128 more than the MX450).
It's really squeezing toothpaste.
Take a look at the poor situation of this year's MX 450 against the GTX 1650 and RTX 3050. My advice to you is that next year's MX 550 product can basically not be considered, because the performance improvement is really very limited.
Another new graphics card, the MX 570, is different. It belongs to a more embarrassing situation: it is very "chicken ribs" to be taken out alone, but if it is combined with products, it is not impossible to consider.
The graphics card 570 uses the same architecture as the RTX 2050, and the core size is exactly the same.
The main difference between the two is that the former is 2G video memory, and the latter 4G video memory and supports light tracking (this is also a good look). So, when you talk about 570 alone, you see that it doesn't seem to have any advantages other than lower power consumption.
I prefer to call it: "2G RTX 2050 Edition".
However, if there is a conscientious manufacturer next year, willing to match this graphics card with a good processor. Then in the light and thin product of about 5,000 yuan, there is still a chance to replace the MX 450 as the new "leek harvester".
In fact, the MX series is in the graphics array of notebook computers, which is a product between the core graphics and the GTX 10 series graphics cards.
So in recent years, the introductory thin and light ben likes to use this series, and then take the unique display as a selling point, but they play games and really struggle. According to this year's situation, as long as the budget is increased by about 1,000 yuan, the computer can be upgraded to the "GTX 1650" high-performance thin and light game book.
From the current point of view, it is difficult for a leather-clad knifeman to perform any exquisite knife techniques after the "MX 570", unless the "GTX 1650" on the notebook is completely eliminated...
In order not to make the bad friends feel messy, I would like to summarize:
"RTX 2050" is covered with 20 Series skin, but it is not an order of magnitude like 2060, and it is almost the same as GTX 1650, with only one more light chase /DLSS;
The "MX 570" can be seen as a castrated version of the "RTX 2050";
"MX 550" belongs to the traditional squeezed toothpaste.
This is the "true face" of NVIDIA's three new graphics cards this time.
Someone may ask, "So what level is the GTX 1650?" ”
Let me just say that it's not a problem to play with this graphics card, pubic, LOL, Dota2 or whatever.
If you have a stand-alone masterpiece, you can meet the needs of the vast majority of 1080P picture quality, and a small number of them may need to be converted to medium and high quality or reduce special effects. In addition, if you force smooth operation in 2K image quality, there are some new AAA works today, which is a bit difficult.
If you want to talk about the positioning of these three graphics cards, in my opinion, this is to create conditions for consumption upgrades for next year's thin and light notebooks, and the main product must be "RTX 2050".
In the world of Nvidia discrete graphics, GTX represents mid-to-high-end gaming graphics, MX is entry-level graphics, and RTX is the strongest king.
Because the three words "RTX" represent "light chase". In the eyes of game enthusiasts, whether the same game can open the light chase is the difference between the two worlds.
Even if the 2050 graphics card is not much better than the "GTX 1650" in terms of performance, the "R standard" is the last stubbornness of some people, representing the performance of the computer.
For the sake of this bit of face, it is not impossible to pay a few hundred dollars more. This is also a kind of old Yellow playing a heart eye... Really a marketing guru.
Then again, electronic products are," after all, "buy new, not buy old."
If two laptops with equal processors are placed in front of me, one labeled GTX 1650 and the other with RTX 2050 written on it.
What does it matter what performance is not performance, the products in 2022 do not want who wants 2019 ah? Again, the phrase" "yes or no" and "no use" are always two different things.
In fact, NVIDIA's marketing methods are not wrong, but the knife method and play word games are a bit unbearable...
Old Huang, or you can!
Author: Shiyu Editor: Surface Line
Image source:
NVIDIA website
JD Mall
Zhihu: 2021 laptop graphics card ladder chart
Zhihu asks: What do you think of the three notebook graphics cards released by Nvidia on December 17