Практика Shadowing: How do Graphics Cards Work? Exploring GPU Architecture - Изучайте разговорный английский по видео

Выберите режим практики

Каждый режим открывает полноэкранную комнату практики.

279 предложений
Если предложения слишком короткие или длинные, нажмите Edit, чтобы их изменить.
1
How many calculations do you think your graphics card performs every second while running video games with incredibly realistic graphics?
2
Maybe a hundred million?
3
Well, a hundred million calculations a second is what's required to run Mario 64 from 1996.
4
We need more power.
5
Maybe a hundred billion calculations a second?
6
Well, then you would have a computer that could run
7
Minecraft back in 2011.
8
In order to run the most realistic video games,
9
such as Cyberpunk 2077, you need a graphics card that can perform around 36 trillion calculations a second.
10
This is an unimaginably large number, so let's take a second to try to conceptualize it.
11
Imagine doing a long multiplication problem once every second.
12
Now let's say everyone on the planet does a similar type of calculation but with different numbers.
13
To reach the equivalent computational power of this graphics card and its 36 trillion calculations a second,
14
we would need about 4 ,400 Earths filled with people,
15
all working together and completing one calculation each, every second.
16
It's rather mind -boggling to think that a device can manage all these calculations,
17
So, in this video, we'll see how graphics cards work in two parts.
18
First, we'll open up this graphics card and explore the different components inside, as well as the physical design and architecture of the GPU,
19
or graphics processing unit.
20
Second,
21
we'll explore the computational architecture and see how GPUs process mountains of data and why they're ideal for running video game graphics,
22
bitcoin mining, neural networks, and AI.
23
So stick around, and let's jump right in.
24
This video is sponsored by Micron, which manufactures the graphics memory inside this graphics card.
25
Before we dive into all the parts of the GPU, let's first understand the differences between GPUs and CPUs.
26
Inside this graphics card, the graphics processing unit or GPU has over 10 ,000 cores.
27
However, when we look at the CPU or central processing unit that's mounted to the motherboard,
28
we find an integrated circuit or chip with only 24 cores.
29
So, which one is more powerful?
30
10 ,000 is a lot more than 24, so you would think the GPU is more powerful.
31
However, it's more complicated than that.
32
A useful analogy is to think of a GPU as a massive cargo ship and a CPU as a jumbo jet airplane.
33
The amount of cargo capacity is the amount of calculations and data that can be processed,
34
and the speed of the ship or airplane is the rate at which how quickly those calculations and data are being processed.
35
Essentially, it's a trade -off between a massive number of calculations
36
that are executed at a slower rate versus a few calculations that can be performed at a much faster rate.
37
Another key difference is that airplanes are a lot more flexible since they can carry passengers, packages or containers and can take off
38
and land at any one of tens of thousands of airports likewise CPUs are flexible in
39
that they can run a variety of programs
40
and instructions however giant cargo ships carry only containers with bulk contents inside
41
and are limited to traveling between ports similarly GPUs are a lot less flexible than CPUs
42
and can only run simple instructions like basic arithmetic.
43
Additionally, GPUs can't run operating systems or interface with input devices or networks.
44
This analogy isn't perfect, but it helps to answer the question of which is faster, a CPU or a GPU?
45
Essentially, if you want to perform a set of calculations across mountains of data, then a GPU will be faster at completing the task.
46
However, if you have a lot less data
47
that needs to be evaluated quickly then a cpu will be faster furthermore
48
if you need to run an operating system
49
or support network connections in a wide range of different applications
50
and hardware then you'll want a cpu we're planning a separate video on cpu architecture
51
so make sure to subscribe so you don't miss it
52
but let's now dive into this graphics card and see how it works.
53
In the center of this graphics card is the printed circuit board, or PCB, with all the various components mounted on it.
54
And we'll start by exploring the brains, which is the graphics processing unit, or GPU.
55
When we open it up, we find a large chip, or die, named GA102, built from 28 .3 billion transistors.
56
The majority of the area of the chip is taken up by the processing cores,
57
which have a hierarchical organization specifically the chip is divided into seven graphics processing clusters
58
or GPCs and within each processing cluster are 12 streaming multi processors
59
or SMs next inside each of these streaming multi processors are four warps
60
and one ray tracing core and then inside each warp are 32 CUDA or shading cores and one tensor core.
61
Across the entire GPU are 10 ,752 CUDA cores,
62
336 tensor cores, and 84 ray tracing cores.
63
These three types of cores execute all the calculations of the GPU, and each has a different function.
64
CUDA cores can be thought of as simple binary calculators with an addition button, a multiply button, and a few others,
65
and are used the most when running video games.
66
Tensor cores are matrix multiplication and addition calculators and are used for geometric transformations and working with neural networks and AI.
67
And ray tracing cores are the largest but the fewest and are used to execute ray tracing algorithms.
68
Now that we understand the computational resources inside this chip, one rather interesting fact is that the 3080,
69
3090, 3080 Ti and 3090
70
Ti graphics cards all use the same GA102 chip design for their GPU.
71
This might be counterintuitive because they have different prices
72
and were released in different years but it's true
73
so why is this well during the manufacturing process sometimes patterning errors dust particles
74
or other manufacturing issues cause damage
75
and create defective areas of the circuit instead of throwing out the entire chip
76
because of a small defect engineers find the defective region
77
and permanently isolate and deactivate the nearby circuitry by By having a GPU with a highly repetitive design,
78
a small defect in one core only damages that particular streaming multiprocessor circuit and doesn't affect the other areas of the chip.
79
As a result, these chips are tested and categorized, or binned, according to the number of defects.
80
The 3090 Ti graphics card have flawless GA -102 chips,
81
with all 10 ,752 CUDA cores working properly.
82
The 3090 has 10 ,496 cores working.
83
The 3080 Ti has 10 ,240, and the 3080 has 8 ,704 CUDA cores working,
84
which is equivalent to having 16 damaged and deactivated streaming multiprocessors.
85
Additionally, different graphics cards differ by their maximum clock speed and the quantity and generation of graphics memory that supports the GPU,
86
which we'll explore in a little bit.
87
Because we've been focusing on the physical architecture of this GA -102 GPU chip,
88
let's zoom into one of these CUDA cores and see what it looks like.
89
Inside this simple calculator is a layout of approximately 410 ,000 transistors.
90
This section of 50 ,000 transistors performs the operation of A times B plus C,
91
which is called fused multiply and add or FMA and is the most common operation performed by graphics cards.
92
Half of the CUDA cores execute FMA using 32 -bit floating point numbers
93
which is essentially scientific notation
94
and the other half of the cores use either 32 -bit integers or 32 -bit floating point numbers.
95
Other sections of this core accommodate negative numbers
96
and perform other simple functions like bit shifting
97
and bit masking as well as collecting and cueing the incoming instructions and operands and then accumulating and outputting the results.
98
As a result, this single core is just a simple calculator with a limited number of functions.
99
This calculator completes one multiply and one add operation each clock cycle,
100
and therefore with this 3090 graphics card and its 10 ,496 cores and 1 .7 GHz clock,
101
we get 35 .6 trillion calculations a second.
102
However, if you're wondering how the GPU handles more complicated operations like division, square root and trigonometric functions, Well,
103
these calculator operations are performed by the special function units
104
which are far fewer as only four of them can be found in each streaming multiprocessor.
105
Now that we have an understanding of what's inside a single core,
106
let's zoom out and take a look at the other sections of the GA102 chip.
107
Around the edge, we find 12 graphics memory controllers, the NVLink controllers and the PCIe interface.
108
On the bottom is a 6 megabyte level 2 SRAM memory cache.
109
And here's the GigaThread engine, which manages all the graphics processing clusters and streaming multiprocessors inside.
110
Now that we've explored this GA102 GPU's physical architecture, let's zoom out and take a look at the other parts inside the graphics card.
111
On this side are the various ports for the displays to be plugged into.
112
On the other side is the incoming 12 -volt power connector.
113
And then, here are the PCIe pins that plug into the motherboard.
114
On the PCB, the majority of the smaller components constitute the voltage regulator module,
115
which takes the incoming 12 volts and converts it to 1 .1 volts
116
and supplies hundreds of watts of power to the GPU.
117
Because all this power heats up the GPU,
118
most of the weight of the graphics card is in the form of a heat sink with four heat pipes
119
that carry heat from the GPU and memory chips to the radiator fins where fans then help to remove the heat.
120
Perhaps some of the most important components, aside from the GPU, are the 24 gigabytes of graphics memory chips
121
which are technically called GDDR6X SDRAM and were manufactured by Micron
122
which is the sponsor of this video whenever you start up a video game
123
or wait for a loading screen the time
124
it takes to load is mostly spent moving all the 3d models of a particular scene
125
or environment from the solid state drive into these graphics memory
126
chips as mentioned earlier the gpu has a small amount of
127
data storage in its six megabyte shared level 2 cache
128
which can hold the equivalent of about this much of the video game's environment.
129
Therefore, in order to render a video game, different chunks of scene are continuously being transferred between the graphics memory and the GPU.
130
Because the cores are constantly performing tens of trillions of calculations a second,
131
GPUs are data -hungry machines and need to be continuously fed terabytes upon terabytes of data
132
and thus these graphics memory chips are designed kind of like
133
multiple cranes loading a cargo ship at the same time specifically
134
these 24 chips transfer a combined 384 bits at a time
135
which is called the bus with and the total data
136
that can be transferred or the bandwidth is about 1 .15 terabytes a second. In contrast,
137
the sticks of DRAM that support the CPU only have a 64 -bit bus width
138
and a maximum bandwidth closer to 64 gigabytes a second.
139
One rather interesting thing is that you may think that computers only work using binary ones and zeros.
140
However, in order to increase static transfer rates, GDDR6X and the latest graphics memory GDDR7 send
141
and receive data across the bus wires using multiple voltage levels beyond just zero and one.
142
For example, GDDR7 uses three different encoding schemes to combine binary bits into ternary digits or PAM3 symbols with voltages of zero,
143
one and negative one.
144
Here's the encoding scheme on how three binary bits are encoded into two ternary digits
145
and this scheme is combined with an 11 bit to 7
146
ternary digit encoding scheme resulting in sending 276 binary bits using only 176 ternary digits the previous generation gdd r6x
147
which is the memory in this 3090 graphics card used a
148
different encoding scheme called PAM -4 to send two bits of data using four different voltage levels.
149
However, engineers in the graphics memory industry agreed to switch to PAM3
150
for future generations of graphics chips in order to reduce encoder complexity, improve the signal -to -noise ratio, and improve power efficiency.
151
Micron delivers consistent innovation to push the boundaries on how much data can be transferred every second
152
and to design cutting -edge memory chips another advancement by Micron is the development of HBM
153
or the high bandwidth memory
154
that surrounds AI chips HBM is built from stacks of DRAM memory chips
155
and uses TSVs or through silicon vias to connect the stack into a single chip essentially forming a cube of AI memory
156
For the latest generation of high -bandwidth memory,
157
which is HBM3e, a single cube can have up to 24 to 36 gigabytes of memory,
158
thus yielding 192 gigabytes of high -speed memory around the AI chip.
159
Next time you buy an AI accelerator system, make sure it uses Micron's HBM3e,
160
which uses 30 % less power than the competitive products.
161
However, unless you're building an AI data center, you're likely not in the market to buy one of these systems
162
which cost between $25 ,000 to $40 ,000 and are on backorder for a few years.
163
If you're curious about high bandwidth memory or Micron's next generation of graphics memory, take a look at one of these links in the description.
164
Alternatively, if designing the next generation of memory chips interests you, Micron is always looking for talented scientists
165
and engineers to help innovate on cutting -edge chips and you can find out more about working from Micron using this link.
166
Now that we've explored many of the physical components inside this graphics card and GPU,
167
let's next explore the computational architecture and see how applications like video game graphics
168
and Bitcoin mining run what's called embarrassingly parallel operations.
169
Although it may sound like a silly name, embarrassingly parallel is actually a technical classification of computer problems where little
170
or no effort is needed to divide the problem into parallel tasks, and video game rendering and Bitcoin mining easily fall into this category.
171
Essentially GPUs solve embarrassingly parallel problems using a principle called SIMD
172
which stands for single instruction multiple data where the same instructions or steps are repeated across thousands to millions of different numbers.
173
Let's see an example of how SIMD or single instruction multiple data is used to create this 3d video game environment.
174
As you may know already, this cowboy hat on the table is composed of approximately 28 ,000 triangles,
175
built by connecting together around 14 ,000 vertices, each with x, y, and z coordinates.
176
These vertex coordinates are built using a coordinate system called model space, with the origin of 000 being at the center of the hat.
177
To build a 3D world, we place hundreds of objects, each with their own model space, into the world environment.
178
And, in order for the camera to be able to tell where each object is relative to other objects,
179
we have to convert or transform all the vertices from each separate model space into the shared world coordinate system, or world space.
180
So, as an example, how do we convert the 14 ,000 vertices of the cowboy hat from model space into world space?
181
Well, we use a single instruction
182
which adds the position of the origin of the hat in world space to the corresponding x,
183
y and z coordinate of a single vertex in model space.
184
Next, we copy this instruction to multiple data, which is all the remaining X,
185
Y, and Z coordinates of the other thousands of vertices that are used to build the hat.
186
Next, we do the same for the table and the rest of the hundreds of other objects in the scene,
187
each time using the same instructions but with the different objects' coordinates in world space, and each object's thousands of vertices in model space.
188
As a result, all the vertices and triangles of all the objects are converted to a common world space coordinate system,
189
and the camera can now determine which objects are in front and which are behind.
190
This example illustrates the power of SIMD, or Single Instruction Multiple Data,
191
and how a single instruction is applied to 5 ,629 different
192
objects with a total of 8 .3 million vertices within the scene, resulting in 25 million addition calculations.
193
The key to SIMD in embarrassingly parallel programs is
194
that every one of these millions of calculations has no dependency on any other calculation,
195
and thus all these calculations can be distributed to the thousands of cores of the GPU
196
and completed in parallel with one another.
197
It's important to note
198
that vertex transformation from model space to world space is just
199
one of the first steps of a rather complicated video game graphics rendering pipeline,
200
and we have a separate video that delves deeper into each of these other steps.
201
Also, we skipped over the transformations for the rotation and scale of each object,
202
but factoring in these values is a similar process that requires additional SIMD calculations.
203
Now that we have a simple understanding of SIMD, let's discuss how this computational architecture matches up with the physical architecture.
204
Essentially, each instruction is completed by a thread and this thread is matched to a single CUDA core.
205
Threads are bundled into groups of 32 called warps
206
and the same sequence of instructions is issued to all the threads in a warp.
207
Next warps are grouped into thread blocks which are handled by the streaming multiprocessor
208
and then finally thread blocks are grouped into grids which are computed across the overall GPU.
209
All these computations are managed or scheduled by the giga thread engine which efficiently maps thread blocks to the available streaming multiprocessors.
210
One important distinction is that within SIMD architecture,
211
all 32 threads in a warp follow the same instructions and are in lockstep with each other, kind of like a phalanx of soldiers moving together.
212
This lockstep execution applied to GPUs up until around 2016.
213
However, newer GPUs follow a SIMD architecture or single instruction multiple threads.
214
The difference between SIMD and SIMT is that while both send the same set of instructions to each thread,
215
with SIMT, the individual threads don't need to be in lockstep with each other and can progress at different rates.
216
In technical jargon, each thread is given its own program counter.
217
Additionally, with SIMT, all the threads within a streaming multiprocessor use a shared 128 kilobyte L1 cache
218
and thus data that's output by one thread can be subsequently used by a separate thread.
219
This improvement from SIMD to SIMT
220
allows for more flexibility when encountering warp divergence via data -dependent conditional branching
221
and easier reconvergence for the threads to reach the barrier synchronization.
222
Essentially, newer architectures of GPUs are more flexible and efficient, especially when encountering branches in code.
223
One additional note is that although you may think that the term warp is derived from warp drives, it actually comes from weaving,
224
and specifically, the jacquard loom.
225
This loom from 1804 used programmable punch cards to select specific threads out of a set to weave together intricate patterns.
226
As fascinating as looms are, let's move on.
227
The final topics we'll explore are Bitcoin mining, tensor cores, and neural networks.
228
But first, we'd like to ask you to like this video, write a quick comment below, share it with a colleague, friend, or on social media,
229
and subscribe if you haven't already.
230
The dream of branch education is to make free and accessible,
231
visually engaging educational videos that dive deeply into a variety of topics on science, engineering, and how technology works,
232
and then to combine multiple videos into an entirely free engineering curriculum for high school and college students.
233
Taking a few seconds to like, subscribe, and comment below helps us a ton.
234
Additionally, we have a Patreon page with AMAs and behind -the -scenes footage
235
and if you find what we do useful we would appreciate any support thank you so now
236
that we've explored how single instruction multiple threads is used in
237
video games let's briefly discuss why gpus were initially used for mining bitcoin
238
we're not going to get too far into the algorithm behind the blockchain
239
and we'll save it for a separate episode but But essentially, to create a block on the blockchain,
240
the Shor 256 hashing algorithm is run on a set of data that includes transactions,
241
a timestamp, additional data, and a random number called a nonce.
242
After feeding these values through the SHA -256 hashing algorithm, a random 256 -bit value is output.
243
You can kind of think of this algorithm as a lottery ticket generator, where you can't pick the lottery number, but based on the input data,
244
the SHA -256 algorithm generates a random lottery ticket number therefore
245
if you change the nonce value
246
and keep the rest of the transaction data the same you'll generate a new random lottery ticket number the winner of this
247
Bitcoin mining lottery is the first randomly generated lottery number to have the first 80 bits all zeros
248
while the rest of the 176 values don't matter and once a winning
249
Bitcoin lottery ticket is found the reward is 3
250
Bitcoin coin, and the lottery resets with a new set of transactions and input values.
251
So why were graphics cards used?
252
Well, GPUs ran thousands of iterations of the SHA -256 algorithm with the same transactions,
253
timestamp, other data, but with different nonce values.
254
As a result, a graphics card like this one could generate around 95 million SHA -256 hashes,
255
or 95 million randomly numbered lottery tickets every second.
256
And hopefully, one of those lottery numbers would have the first 80 digits as all zeros.
257
However, nowadays, computers filled with ASICs, or application -specific integrated circuits,
258
perform 250 trillion hashes a second, or the equivalent of 2 ,600 graphics cards,
259
thereby making graphics cards look like a spoon when mining
260
Bitcoin next to an excavator
261
that is an ASIC mining computer let's next discuss the design
262
of the tensor cores it'll take multiple full -length videos to cover generative AI in neural networks
263
so we'll focus on the exact matrix math
264
that tensor cores solve essentially tensor cores take three matrices
265
and multiply the first two add in the third and then output the result.
266
Let's look at one value of the output.
267
This value is equal to the sum of values of the
268
first row of the first matrix multiplied by the values from the first column of the second matrix
269
and then the corresponding value of the third matrix is added in.
270
Because all the values of the three input matrices are ready at the same time,
271
the tensor cores complete all of the matrix multiplication and addition calculations concurrently.
272
Neural networks and generative AI require trillions to quadrillions of matrix multiplication and addition operations and typically uses much larger matrices.
273
Finally, there are ray tracing cores which we explored in a separate video that's already been released.
274
That's pretty much it for graphics cards.
275
We're thankful to all our Patreon and YouTube membership sponsors for supporting our videos.
276
If you want to financially support our work, you can find the links in the description below.
277
This is Branch Education, and we create 3D animations that dive deeply into the technology that drives our modern world.
278
Watch another Branch video by clicking one of these cards, or click here to subscribe.
279
Thanks for watching to the end.

Скачать приложение

Всё для свободной речи

ИИ-оценка произношенияОценка каждой фразы
Практика IPAОсвойте каждый звук
СловарьСоберите свой словарный запас
Игра со словамиУчитесь играя

Тот же канал

FOR YOU

Рекомендуемое видео

Лексика и советы по произношению к этому уроку

Этот урок разговорной практики построен на видео «How do Graphics Cards Work? Exploring GPU Architecture». Чаще всего повторяются слова: graphics, Gpu, core, chip, card. В этом видео 18 предложений и 3809 слов для шедоуинга. Речь длится 28:29. Говорящий держит ровный темп — около 134 слов в минуту, удобный для шедоуинга. Только 72% слов входят в 3000 самых частых слов английского языка, поэтому лексика сложная. Вопросов здесь 7 из 18 предложений, так что это хорошая тренировка вопросительной интонации и коротких ответов.

Ключевая лексика этого видео

15 слов из видео, которые стоит выучить, с произношением и значением:

СловоПроизношениеЗначение
graphics существительное/ˈɡɹæfɪks/проекти́рование, черче́ние
chip существительное/t͡ʃɪp/ще́пка, лучи́на
calculation существительное/ˌkælkjuˈleɪʃn̩/вычисле́ние, исчисле́ние
thread существительное/ˈθɹɛd/нить, ни́тка
instruction существительное/ɪnˈstɹʌkʃən/инстру́кция, инструкта́ж
architecture существительное/ˈɑː.kɪˌtɛk.t͡ʃə/архитекту́ра, зо́дчество
lottery существительное/ˈlɑtɚi/лотере́я
essentially наречие/ɪˈsɛnʃəli/по существу́, суще́ственно
explore глагол/ɪkˈsploɹ/иссле́довать
bitcoin существительное/ˈbɪtˌkɔɪn/битко́йн, битко́ин
warp существительное/ˈwɔːp/деформа́ция, искривле́ние
coordinate глагол/koʊˈɔɹ.dəˌneɪt/координи́ровать, скоордини́ровать
parallel прилагательное/ˈpæ.ɹəˌlɛl/паралле́льный
algorithm существительное/ˈælɡəɹɪðm̩/алгори́тм
calculator существительное/kæl.kjə.leɪ.tɚ/калькуля́тор, вычисли́тельная маши́на

Фразовые глаголы, которые вы услышите

СловоПроизношениеЗначение
find out глаголузнава́ть, узна́ть
heat up глаголразогрева́ть, разогре́ть
start up глагол/stɑɹˈtʌp/вывалиться, взя́ться
take off глаголснима́ть, снять
take up глаголподнимать

Грамматика в этом видео

Конструкции, которые говорящий использует чаще всего, с точными словами из видео:

КонструкцияВ видео
Пассивный залог be + причастие прошедшего времени — важно, что происходит, а не кто это делаетis sponsored · can be processed · being processed
Present Perfect have/has + причастие прошедшего времени — прошлое действие, важное сейчасwe've explored · that's already been released
Придаточные определительные who / which + предложение — уточнение о человеке или предметеbrains, which is · data, which is · blocks which are

Произношение, на которое стоит обратить внимание

Говорящий использует 31 сокращённых и редуцированных форм, например we'll, we've, you're. Произносите их коротко, так, как слышите.

  • Звуки «th»: thread /ˈθɹɛd/, algorithm /ˈælɡəɹɪðm̩/, bandwidth /ˈbændwɪdθ/, width /ˈwɪtθ/, math /mæθ/
  • Звуки «sh» и «zh»: calculation /ˌkælkjuˈleɪʃn̩/, instruction /ɪnˈstɹʌkʃən/, essentially /ɪˈsɛnʃəli/, computational /ˌkɑmpjʊˈteɪʃənəl/, transaction /tɹænˈzækʃən/
  • Длинные слова — следите за ударением: architecture /ˈɑː.kɪˌtɛk.t͡ʃə/, essentially /ɪˈsɛnʃəli/, coordinate /koʊˈɔɹ.dəˌneɪt/, calculator /kæl.kjə.leɪ.tɚ/, computational /ˌkɑmpjʊˈteɪʃənəl/

Звуки, трудные для русскоязычных:

  • /w/ — губы округлены, это не /v/: warp /ˈwɔːp/, bandwidth /ˈbændwɪdθ/, width /ˈwɪtθ/, sequence /ˈsiː.kwəns/, reward /ɹɪˈwɔɹd/
  • /h/ — лёгкий выдох, а не русское «х»: hash /ˈhæʃ/, hungry /ˈhʌŋ.ɡɹi/, hardware /ˈhɑɹdˌwɛɹ/
  • Звонкие согласные в конце слова — не оглушайте их: thread /ˈθɹɛd/, dive /daɪv/, subscribe /səbˈskɹaɪb/, complicated /ˈkɑm.plɪˌkeɪ.tɪd/, voltage /ˈvəʊltɪd͡ʒ/

Как заниматься с этим видео

  1. Прослушайте всё видео один раз молча и выпишите незнакомые слова.
  2. Повторяйте предложение за предложением на обычной скорости, пока ваш ритм не совпадёт с ритмом говорящего.
  3. Запишите себя и сравните с оригиналом, обращая внимание на такие слова, как graphics, chip, calculation.

Что такое техника Shadowing?

Shadowing — это научно обоснованная техника изучения языка, изначально разработанная для подготовки профессиональных переводчиков и популяризированная полиглотом доктором Александром Аргуэльесом. Метод прост, но эффективен: вы слушаете аудио на английском от носителей языка и немедленно повторяете вслух — как тень, следующая за говорящим с задержкой в 1–2 секунды. В отличие от пассивного прослушивания или грамматических упражнений, Shadowing заставляет мозг и мышцы рта одновременно обрабатывать и воспроизводить реальные речевые паттерны. Исследования показывают, что это значительно улучшает точность произношения, интонацию, ритм, связную речь, понимание на слух и беглость речи — что делает его одним из самых эффективных методов для подготовки к IELTS Speaking и реального общения на английском.

Техника шедоуинга: читать полное пошаговое руководство →