Download "How does Computer Memory Work? 💻🛠"
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252F7J7X7aZvMXQ%252Fmaxresdefault.jpg&w=1139&q=75)
Similar videos from our catalog
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252F16zrEPOsIcI%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FDKhxeT9ArCo%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252FDPgfed8ypp8%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FIYnwPbKsVKA%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FNuEWxWyr3vI%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FXPksSwXR0kQ%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FdgT4DGnM8p0%252Fmaxresdefault.jpg&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi_webp%252FsnpAbc94C7s%252Fmaxresdefault.webp&w=1139&q=75)
%3Aformat(webp)%2Fi.ytimg.com%252Fvi%252FtPTaxp3LLmE%252Fmaxresdefault.jpg&w=1139&q=75)
Video tags
Subtitles
00:00:00
Have you ever wondered what’s happening inside
your computer when you load a program or video 00:00:05
game? Well, millions of operations are happening,
but perhaps the most common is simply just copying 00:00:12
data from a solid-state drive or SSD into dynamic
random-access memory or DRAM. An SSD stores all 00:00:22
the programs and data for long-term storage,
but when your computer wants to use that data, 00:00:27
it has to first move the appropriate
files into DRAM, which takes time, 00:00:33
hence the loading bar. Because your CPU works
only with data after it’s been moved to DRAM, 00:00:39
it’s also called working memory or main memory.
The reason why your desktop uses both SSDs and 00:00:47
DRAM is because Solid-State Drives permanently
store data in massive 3D arrays composed of a 00:00:54
trillion or so memory cells, yielding terabytes of
storage, whereas DRAM temporarily stores data in 00:01:02
2D arrays composed of billions of tiny capacitor
memory cells yielding gigabytes of working memory. 00:01:09
Accessing any section of cells in the massive
SSD array and reading or writing data takes 00:01:16
about 50 microseconds whereas reading or
writing from any DRAM capacitor memory 00:01:22
cell takes about 17 nanoseconds, which is 3000
times faster. For comparison, a supersonic jet 00:01:30
going at Mach 3 is around 3000 times faster
than a moving tortoise. So, the speed of 00:01:36
17 nanosecond DRAM versus 50 microsecond SSD is
like comparing a supersonic jet to a tortoise. 00:01:45
However, speed is just one factor. DRAM is
limited to a 2D array and temporarily stores 00:01:52
one bit per memory cell. For example, this stick
of DRAM with 8 chips holds 16 gigabytes of data, 00:02:00
whereas a solid-state drive of a smaller
size can hold 2 terabytes of data, more 00:02:06
than 100 times that of DRAM. Additionally,
DRAM requires power to continuously store 00:02:13
and refresh the data held in its capacitors.
Therefore, computers use both SSDs and DRAM and, 00:02:21
by spending a few seconds of loading time
to copy data from the SSD to the DRAM, 00:02:27
and then prefetching, which is the process of
moving data before it’s needed, your computer can 00:02:34
store terabytes of data on the SSD and then access
the data from programs that were preemptively 00:02:40
copied into the DRAM in a few nanoseconds.
For example, many video games have a loading 00:02:46
time to start up the game itself, and then a
separate loading time to load a save file. 00:02:52
During the process of loading a save file, all
the 3D models, textures, and the environment of 00:02:59
your game state are moved from the SSD into DRAM
so any of it can be accessed in a few nanoseconds, 00:03:06
which is why video games have DRAM capacity
requirements. Just imagine, without DRAM, 00:03:12
playing a game would be 3,000 times slower.
We covered solid-state drives in other videos, 00:03:19
so in this video, we’re going to take a deep
dive into this 16-gigabyte stick of DRAM. First, 00:03:26
we’ll see exactly how the CPU communicates
and moves data from an SSD to DRAM. Then 00:03:32
we’ll open up a DRAM microchip and see how
billions of memory cells are organized into 00:03:38
banks and how data is written to and read from
groups of memory cells. In the process, we’ll 00:03:45
dive into the nanoscopic structures inside
individual memory cells and see how each 00:03:50
capacitor physically stores 1 bit of data.
Finally, we’ll explore some breakthroughs and 00:03:56
optimizations such as the burst buffer and folded
DRAM layouts that enable DRAM to move data around 00:04:04
at incredible speeds. A few quick notes.
First, you can find similar DRAM chips inside 00:04:11
GPUs, Smartphones, and many other devices, but
with different optimizations. As examples, 00:04:18
GPU DRAM or VRAM, located all around the
GPU chip, has a larger bandwidth and can 00:04:26
read and write simultaneously, but operates at
a lower frequency, and DRAM in your smartphone 00:04:32
is stacked on top of the CPU and is optimized for
smaller packaging and lower power consumption. 00:04:39
Second, this video is sponsored by
Crucial. Although they gave me this 00:04:44
stick of DRAM to model and use in the
video, the content was independently 00:04:49
researched and not influenced by them.
Third, there are faster memory structures 00:04:54
in your CPU called cache memory and even faster
registers. All these types of memory create a 00:05:01
memory hierarchy, with the main trade-off
being speed versus capacity while keeping 00:05:06
prices affordable to consumers and optimizing
the size of each microchip for manufacturing. 00:05:12
Fourth, you can see how much of
your DRAM is being utilized by 00:05:17
each program by opening your computer’s
resource monitor and clicking on memory. 00:05:22
Fifth, there are different generations of DRAM,
and we’ll explore DDR5. Many of the key concepts 00:05:29
that we explain apply to prior generations,
although the numbers may be different. 00:05:34
Sixth, 17 nanoseconds is incredibly fast!
Electricity travels at around 1 foot per 00:05:41
nanosecond, and 17 nanoseconds is about the
time it takes for light to travel across a room. 00:05:47
Finally, this video is rather long as it covers
a lot of what there is to know around DRAM. We 00:05:54
recommend watching it first at one point two
five times speed, and then a second time at 00:06:00
one and a half speed to fully comprehend this
complex technology. Stick around because this 00:06:06
is going to be an incredibly detailed video.
To start, a stick of DRAM is also called a Dual 00:06:14
Inline Memory Module or DIMM and there are 8
DRAM chips on this particular DIMM. On the 00:06:21
motherboard, there are 4 DRAM slots, and when
plugged in, the DRAM is directly connected to 00:06:28
the CPU via 2 memory channels that run through
the motherboard. Note that the left two DRAM 00:06:34
slots share these memory channels, and the right
two share a separate channel. Let’s move to 00:06:40
look inside the CPU at the processor. Along
with numerous cores and many other elements, 00:06:46
we find the memory controller which manages
and communicates with the DRAM. There’s also 00:06:52
a separate section for communicating with SSDs
plugged into the M2 slots and with SSDs and 00:06:58
hard drives plugged into SATA connectors. Using
these sections, along with data mapping tables, 00:07:04
the CPU manages the flow of data from
the SSD to DRAM, as well as from DRAM 00:07:10
to cache memory for processing by the cores.
Let’s move back to see the memory channels. 00:07:16
For DDR5 each memory channel is divided into two
parts, Channel A and Channel B. These two memory 00:07:25
channels A and B independently transfer 32 bits at
a time using 32 data wires. Using 21 additional 00:07:35
wires each memory channel carries an address
specifying where to read or write data and, using 00:07:42
7 control signal wires, commands are relayed.
The addresses and commands are sent to and shared 00:07:49
by all 4 chips on the memory channel which
work in parallel. However, the 32-bit data 00:07:55
lines are divided among the chips and thus each
chip only reads or writes 8 bits at a time. 00:08:01
Additionally, power for DRAM is
supplied by the motherboard and 00:08:07
managed by these chips on the stick itself.
Next, let’s open and look inside one of these 00:08:13
DRAM microchips. Inside the exterior packaging,
we find an interconnection matrix that connects 00:08:20
the ball grid array at the bottom with the die
which is the main part of this microchip. This 2 00:08:26
gigabyte DRAM die is organized into 8 bank groups
composed of 4 banks each, totaling 32 banks. 00:08:34
Within each bank is a massive array, 65,536 memory
cells tall by 8192 cells across, essentially rows 00:08:47
and columns in a grid, with tens of thousands of
wires, and supporting circuitry running outside 00:08:53
each bank. Instead of looking at this die, we’re
going to transition to a functional diagram, 00:08:59
and then reorganize the banks and bank groups.
In order to access 17 billion memory cells, 00:09:06
we need a 31-bit address. 3 bits are used to
select the appropriate bank group, then 2 bits 00:09:13
to select the bank. Next 16 bits of the address
are used to determine the exact row out of 65 00:09:21
thousand. Because this chip reads or writes 8
bits at a time, the 8192 columns are grouped by 00:09:30
8 memory cells, all read or written at a time,
or ‘by 8’, and thus only 10 bits are needed for 00:09:37
the column address. One optimization is that
this 31-bit address is separated into two parts 00:09:44
and sent using only 21 wires. First, the bank
group, bank, and row address are sent, and then 00:09:52
after that the column address. Next, we’ll look
inside these physical memory cells, but first, 00:09:58
let’s briefly talk about how these structures are
manufactured as well as this video’s sponsor. 00:10:03
This incredibly complicated die,
also called an integrated circuit, 00:10:09
is manufactured on 300-millimeter silicon wafers,
2500ish dies at a time. On each die are billions 00:10:18
of nanoscopic memory cells that are fabricated
using dozens of tools and hundreds of steps in 00:10:24
a semiconductor fabrication plant or fab. This
one was made by Micron which manufactures around 00:10:31
a quarter of the world’s DRAM, including both
Nvidia’s and AMD’s VRAM in their GPUs Micron also 00:10:39
has its own product line of DRAM and SSDs under
the brand Crucial which, as mentioned earlier, 00:10:46
is the sponsor of this video. In addition
to DRAM, Micron is one of the world’s leading 00:10:51
suppliers of solid-state drives such as this
Crucial P5+ M2 NVME SSD. By installing your 00:11:00
operating system and video games on a Crucial
NVMe solid-state drive, you’ll be sure to have 00:11:07
incredibly fast loading times and smooth gameplay,
and if you do video editing, make sure all those 00:11:14
files are on a fast SSD like this one as well.
This is because the main speed bottleneck for 00:11:20
loading is predominantly limited by the speed of
the SSD or hard drive where the files are stored. 00:11:26
For example, this hard drive can only transfer
data at around 150 megabytes a second whereas 00:11:34
this Crucial NVMe SSD can transfer data at a
rate of up to 6,600 megabytes a second, which, 00:11:42
for comparison is the speed of a moving tortoise
versus a galloping horse. By using a Crucial NVMe 00:11:50
SSD, loading a video game that requires gigabytes
of DRAM is reduced from a minute or more down to 00:11:58
a couple seconds. Check out the Crucial NVMe
SSDs using the link in the description below. 00:12:09
Let’s get back to the details of how DRAM works
and zoom in to explore a single memory cell 00:12:15
situated in a massive array. This memory cell is
called a 1T1C cell and is a few dozen nanometers 00:12:24
in size. It has two parts, a capacitor to store
one bit of data in the form of electrical charges 00:12:31
or electrons and a transistor to access and read
or write data. The capacitor is shaped like a 00:12:37
deep trench dug into silicon and is composed of
two conductive surfaces separated by a dielectric 00:12:44
insulator or barrier just a few atoms thick, which
stops the flow of electrons but allows electric 00:12:50
fields to pass through. If this capacitor
is charged up with electrons to 1 volt, 00:12:56
it’s a binary 1, and if no charges are present
and it’s at 0 volts, it’s a binary 0, and thus 00:13:04
this cell only holds one bit of data. Designs
of capacitors are constantly evolving but in 00:13:11
this trench capacitor, the depth of the silicon is
utilized to allow for larger capacitive storage, 00:13:17
while taking up as little area as possible.
Next let’s look at the access transistor and 00:13:24
add in two wires. The wordline wire connects to
the gate of the transistor while the bitline wire 00:13:31
connects to the other side of the transistor’s
channel. Applying a voltage to the wordline 00:13:36
turns on the transistor, and, while it’s on,
electrons can flow through the channel thus 00:13:42
connecting the capacitor to the bitline. This
allows us to access and charge up the capacitor 00:13:47
to write a 1 or discharge the capacitor to write
a 0. Additionally, we can read the stored value 00:13:54
in the capacitor by measuring the amount of
charge. However, when the wordline is off, 00:14:00
the transistor is turned off, and the capacitor
is isolated from the bitline thus saving the 00:14:05
data or charge that was previously written. Note
that because this transistor is incredibly small, 00:14:12
only a few dozen nanometers wide, electrons slowly
leak across the channel, and thus over time the 00:14:19
capacitor needs to be refreshed to recharge
the leaked electrons. We’ll cover exactly how 00:14:25
refreshing memory cells works a little later.
As mentioned earlier, this 1T1C memory cell is 00:14:33
one of 17 billion inside this single die and is
organized into massive arrays called banks. So, 00:14:41
let’s build a small array for illustrative
purposes. In our array, each of the wordlines 00:14:47
is connected in rows, and then the bitlines are
connected in columns. Wordlines and bitlines 00:14:53
are on different vertical layers so one can
cross over the other, and they never touch. 00:15:00
Let’s simplify the visual and use symbols for the
capacitors and the transistors. Just as before, 00:15:06
the wordlines connect to each transistor’s control
gate in rows, and then all the bitlines in columns 00:15:13
connect to the channel opposite each capacitor.
As a result, when a wordline is active, 00:15:19
all the capacitors in only that row are
connected to their corresponding bitlines, 00:15:24
thereby activating all the memory cells in that
row. At any given time only one wordline is 00:15:31
active because, if more than one wordline were
active, then multiple capacitors in a column 00:15:37
would be connected to the bitline and the data
storage functionalities of these capacitors would 00:15:42
interfere with one another, making them useless.
As mentioned earlier, within a single bank there 00:15:48
are 65,536 rows and 8,192 columns and the 31-bit
address is used to activate a group of just 8 00:15:59
memory cells. The first 5 bits select the bank,
and the next 16-bits are sent to a row decoder 00:16:06
to activate a single row. For example, this
binary number turns on the wordline row 27,524, 00:16:15
thus turning on all transistors in that row and
connecting the 8,192 capacitors to their bitlines, 00:16:23
while at the same time the other 65
thousandish wordlines are all off. 00:16:29
Here’s the logic diagram for a simple decoder.
The remaining 10 bits of the address are sent 00:16:35
to the column multiplexer. This multiplexer
takes in the 8192 bitlines on the top, and, 00:16:42
depending on the 10-bit address, connects a
specific group of 8 bitlines to the 8 input 00:16:48
and output IO wires at the bottom. For example,
if the 10-bit address we this, then only the 00:16:55
bitlines 4,784 through 4,791 would be connected
to the IO wires, and the rest of the 8000ish 00:17:05
bitlines would be connected to nothing. Here’s
the logic diagram for a simple multiplexer. 00:17:11
We now have the means of accessing any
memory cell in this massive array; however, 00:17:16
to understand the three basic operations,
reading, writing, and refreshing let’s add 00:17:22
two elements to our layout: A sense amplifier
at the bottom of each bitline, and a read and 00:17:28
write driver outside of the column multiplexer.
Let’s look at reading from a group of memory 00:17:34
cells. First the read command and 31-bit address
are sent from the CPU to the DRAM. The first 5 00:17:42
bits select a specific bank. The next step is
to turn off all the wordlines in that bank, 00:17:48
thereby isolating all the capacitors, and then
precharge all 8000ish bitlines to .5 volts. Next 00:17:57
the 16-bit row address turns on a row, and all
the capacitors in that row are connected to their 00:18:03
bitlines. If an individual capacitor holds a 1
and is charged to 1 volt, then some charge flows 00:18:10
from the capacitor onto the .5-volt bitline, and
the voltage on the bitline increases. The sense 00:18:17
amplifier then detects this slight change
or perturbation of voltage on the bitline, 00:18:22
amplifies the change, and pushes the voltage on
the bitline all the way up to 1 volt. However, 00:18:28
if a 0 is stored in the capacitor, charge
flows from the bitline into the capacitor, 00:18:35
and the .5-volt bitline decreases in voltage.
The sense amplifier then sees this change, 00:18:41
amplifies it and drives the bitline voltage down
to 0 volts or ground. The sense amplifier is 00:18:49
necessary because the capacitor is so small,
and the bitline is rather long, and thus the 00:18:54
capacitor needs to have an additional component
to sense and amplify whatever value is stored. 00:19:00
Now, all 8000ish bitlines are driven to 1
volt or 0 volts corresponding to the stored 00:19:08
charge in the capacitors of the activated
row, and this row is now considered open. 00:19:12
Next, the column select multiplexer uses
the 10-bit column address to connect the 00:19:19
corresponding 8 bitlines to the read
driver which then sends these 8 values 00:19:24
and voltages over the 8 data wires to the CPU.
Writing data to these memory cells is similar 00:19:31
to reading, however with a few key differences.
First the write command, address, and 8 bits to 00:19:39
be written are sent to the DRAM chip. Next, just
like before the bank is selected, the capacitors 00:19:46
are isolated, and the bitlines are precharged
to .5 volts. Then, using a 16-bit address, 00:19:54
a single row is activated, the capacitors perturb
the bitline, and the sense amplifiers sense this 00:20:01
and drive the bitlines to a 1 or 0 thus opening
the row. Next the column address goes to the 00:20:09
multiplexer, but, this time, because a write
command was sent, the multiplexer connects the 00:20:15
specific 8 bitlines to the write driver which
contains the 8 bits that the CPU had sent along 00:20:20
the data wires and requested to write. These
write drivers are much stronger than the sense 00:20:26
amplifier and thus they override whatever voltage
was previously on the bitline, and drive each of 00:20:32
the 8 bitlines to 1 volt for a 1 to be written,
or 0 volts for a 0. This new bitline voltage 00:20:39
overrides the previously stored charges or values
in each of the 8 capacitors in the open row, 00:20:45
thereby writing 8 bits of data to the memory
cells corresponding to the 31-bit address. 00:20:52
Three quick notes. First, as a reminder, writing
and reading happens concurrently with all the 4 00:20:58
chips in the shared memory channel, using
the same 31-bit address and command wires, 00:21:03
but with different data wires for each chip.
Second, with DDR5 for a binary 1 the voltage 00:21:11
is actually 1.1 volts, for DDR4 it’s 1.2 volts,
and prior generations had even higher voltages, 00:21:19
with the bitline precharge voltages being
half of these voltages. However, for DDR5, 00:21:26
when writing or refreshing a higher voltage,
around 1.4 volts is applied and stored in each 00:21:33
capacitor for a binary 1 because charge leaks
out over time. However, for simplicity, we’re 00:21:39
going to stick with 1 and 0. Third, the number
of bank groups, banks, bitlines and wordlines 00:21:46
varies widely between different generations
and capacities but is always in powers of 2. 00:21:53
Let’s move on and discuss the third operation
which is refreshing the memory cells in a bank. 00:21:59
As mentioned earlier, the transistors used to
isolate the capacitors are incredibly small, 00:22:05
and thus charges leak across the channel. The
refresh operation is rather simple and is a 00:22:11
sequence of closing all the rows, precharging
the bitlines to .5 volts, and opening a row. 00:22:17
To refresh, just as before, the capacitors perturb
the bitlines and then the sense amplifiers drive 00:22:24
the bitlines and capacitors of the open row fully
up to 1 volt or down to 0 volts depending on the 00:22:31
stored value of the capacitor, thereby refilling
the leaked charge. This process of row closing, 00:22:38
precharging, opening, and sense amplifying happens
row after row, taking 50 nanoseconds for each row, 00:22:46
until all 65 thousandish rows are refreshed
taking a total of 3 milliseconds or so to 00:22:53
complete. The refresh operation occurs
once every 64 milliseconds for each bank, 00:22:58
because that’s statistically below the
worst-case time it takes for a memory 00:23:03
cell to leak too much charge to make a stored 1
turn into a 0, thus resulting in a loss of data. 00:23:12
Let’s take a step back and consider the
incredible amount of data that is moved 00:23:17
through DRAM memory cells. These banks of memory
cells handle up to 4 thousand 8 hundred million 00:23:24
requests to read and write data every second
while refreshing every memory cell in each 00:23:30
bank row by row around 16 times a second.
That’s a staggering amount of data movement 00:23:37
and illustrates the true strength of computers.
Yes, they do simple things like comparisons, 00:23:44
arithmetic, and moving data around, but
at a rate of billions of times a second. 00:23:50
Now, you might wonder why computers
need to do so much data movement. Well, 00:23:55
take this video game for example. You have obvious
calculations like the movement of your character 00:24:01
and the horse. But then there are individual
grasses, trees, rocks, and animals whose 00:24:07
positions and geometries are stored in DRAM.
And then the environment such as the lighting 00:24:14
and shadows change the colors and textures of the
environment in order to create a realistic world. 00:24:21
Next, we’re going to explore breakthroughs and
optimizations that allow DRAM to be incredibly 00:24:28
fast. But, before we get into all those
details, we would greatly appreciate it 00:24:33
if you could take a second to hit that like
button, subscribe if you haven’t already, 00:24:37
and type up a quick comment below, as it helps get
this video out to others. Also, we have a Patreon 00:24:45
and would appreciate any support. This is our
longest and most detailed video by far, and we’re 00:24:51
planning more videos that get into the inner
details of how computers work. We can’t do it 00:24:57
without your help, so thank you for watching and
doing these three quick things. It helps a ton. 00:25:07
The first complex topic which we’ll explore
is why there are 32 banks, as well as what the 00:25:14
parameters on the packaging of DRAM are.
After that, we’ll explore burst buffers, 00:25:19
sub-arrays, and folded DRAM architecture
and what’s inside the sense amplifier. 00:25:25
Let’s take a look at the banks. As
mentioned earlier opening a single 00:25:30
row within a bank requires all these
steps and this process takes time.00:25:34
However, if a row were already open, we
could read or write to any section of 00:25:39
8 memory cells using only the 10-bit
column address and the column select 00:25:44
multiplexer. When the CPU sends a read or
write command to a row that’s already open, 00:25:51
it’s called a row hit or page hit, and this
can happen over and over. With a row hit, 00:25:56
we skip all the steps required to open a row, and
just use the 10-bit column address to multiplex a 00:26:03
different set of 8 columns or bitlines, connecting
them to the read or write driver, thereby saving 00:26:09
a considerable amount of time. A row miss is
when the next address is for a different row, 00:26:14
which requires the DRAM to close and isolate the
currently open row, and then open the new row. 00:26:20
On a package of DRAM there are typically 4 numbers
specifying timing parameters regarding row hits, 00:26:27
precharging, and row misses. The first number
refers to the time it takes between sending an 00:26:33
address with a row open, thus a row hit, to
receiving the data stored in those columns. 00:26:38
The next number is the time it takes to open
a row if all the lines are isolated and the 00:26:44
bitlines are precharged. Then the next number
is the time it takes to precharge the bitlines 00:26:49
before opening a row, and the last number is
the time it takes between a row activation and 00:26:55
the following precharge. Note that these
numbers are measured in clock cycles. 00:27:00
Row hits are also the reason why the address is
sent in two sections, first the bank selection and 00:27:07
row address called RAS and then the column address
called CAS. If the first part, the bank selection 00:27:14
and row address, matches a currently open row,
then it’s a row hit, and all the DRAM needs is the 00:27:20
column address and the new command, and then the
multiplexer simply moves around the open row. 00:27:26
Because of the time saving in accessing an
open row, the CPU memory controller, programs, 00:27:32
and compilers are optimized for increasing the
number of subsequent row hits. The opposite, 00:27:38
called thrashing, is when a program jumps around
from one row to a different row over and over, 00:27:44
and is obviously incredibly inefficient
both in terms of energy and time. 00:27:49
Additionally, DDR5 DRAM has 32 banks for
this reason. Each bank’s rows, columns, 00:27:57
sense amplifiers and row decoders operate
independently of one another, and thus multiple 00:28:03
rows from different banks can be open all at the
same time, increasing the likelihood of a row hit, 00:28:09
and reducing the average time it takes for the CPU
to access data. Furthermore, by having multiple 00:28:16
bank groups, the CPU can refresh one bank in each
bank group at a time while using the other three, 00:28:22
thus reducing the impact of refreshing.
A question you may have had earlier is why 00:28:28
are banks significantly taller than they are
wide? Well, by combining all the banks together 00:28:34
one next to the other you can think of this chip
as actually being 65 thousand rows tall by 262 00:28:44
thousand columns wide. And, by adding 31 equally
spaced divisions between the columns, thus 00:28:50
creating banks, we allow for much more flexibility
and efficiency in reading, writing and refreshing. 00:28:58
Also, note that on the DRAM packaging are
its capacity in Gigabytes, the number of 00:29:04
millions of data transfers per second, which
is two times the clock frequency, and the peak 00:29:10
data transfer rate in Megabytes per second.
The next design optimization we’ll explore 00:29:16
is the burst buffer and burst length. Let’s add a
128-bit read and write temporary storage location, 00:29:23
called a burst buffer to our functional diagram.
Instead of 8 wires coming out of the multiplexer, 00:29:30
we’re going to have 128 wires that connect
to these 128-bit buffer locations. Next 00:29:38
the 10-bit column address is broken into two
parts, 6 bits are used for the multiplexer, 00:29:44
and 4 bits are for the burst buffer.
Let’s explore a reading command. With 00:29:49
our burst buffer in place, 128 memory cells and
bitlines are connected to the burst buffer using 00:29:56
the 6 column bits, thereby temporarily loading,
or caching 128 values into the burst buffer. 00:30:04
Using the 4 bits for the buffer, 8 quickly
accessed data locations in the burst buffer 00:30:10
are connected to the read drivers and the data is
sent to the CPU. By cycling through these 4 bits, 00:30:16
all 16 sets of 8 bits are read out, and thus the
burst length is 16. After that a new set of 128 00:30:25
bitlines and values are connected and loaded
into the burst buffer. There’s also a write 00:30:31
burst buffer which operates in a similar way.
The benefit of this design is that 16 sets of 00:30:37
8 bits per microchip, totaling 1024 bits, can be
accessed and read or written extremely quickly, 00:30:45
as long as the data is all next to one
another, but at the same time we still 00:30:49
have the granularity and ability to access any
set of 8 bits if our data requests jump around. 00:30:56
The next design optimization is that this bank
of 65536 rows by 8192 columns is rather massive, 00:31:07
and results in extremely long wordlines and
bitlines, especially when compared to the size of 00:31:14
each trench capacitor memory cell. Therefore,
the massive array is broken up into smaller 00:31:20
blocks 1,024 by 1,024, with intermediate
sense amplifiers below each subarray, 00:31:27
and subdividing wordlines and using a hierarchical
row decoding scheme. By subdividing the bitlines, 00:31:34
the distance and amount of wire that each tiny
capacitor is connected to as it perturbs the 00:31:40
bitline to the sense amplifier is reduced, and
thus the capacitor doesn’t have to be as big. By 00:31:47
subdividing the wordlines the capacitive load from
eight thousandish transistor gates and channels is 00:31:53
decreased, and thus the time it takes to turn on
all the access transistors in a row is decreased. 00:31:59
The final topic we’re going to talk about is
the most complicated. Remember how we had 00:32:05
a sense amplifier connected to the bottom of
each bitline? Well, this optimization has two 00:32:10
bitlines per column going to each sense amplifier
and alternating rows of memory cells connected to 00:32:17
the left and right bitlines, thus doubling the
number of bitlines. When one row is active, 00:32:22
half of the bitlines are active while the other
half are passive and vice versa when the next row 00:32:28
is active. Moving down to see inside the sense
amplifier we find a cross-coupled inverter. How 00:32:34
does this work? Well, when the active bitline is
a 1, the passive bitline will be driven by this 00:32:41
cross-coupled inverter to the opposite value
of 0, and when the active is a 0, the passive 00:32:46
becomes a 1. Note that the inverted passive
bitline isn’t connected to any memory cells, 00:32:52
and thus it doesn’t mess up any stored data. The
cross-coupled inverter makes it such that these 00:32:58
two bitlines are always going to be opposite
one another, and they’re called a differential 00:33:03
pair. There are three benefits to this design.
First, during the precharge step, we want to bring 00:33:09
all the bitlines to .5 volts and, by having a
differential pair of active and passive bitlines, 00:33:15
the easiest solution is to disconnect the cross
coupled inverters and open a channel between the 00:33:22
two using a transistor. The charge easily
flows from the 1 bitline to the 0, and they 00:33:28
both average out and settle at .5 volts.
The other two benefits are noise immunity, 00:33:34
and a reduction in parasitic capacitance of the
bitline. These benefits are related to that fact 00:33:39
that by creating two oppositely charged electric
wires with electric fields going from one to 00:33:45
the other we reduce the amount of electric fields
emitted in stray directions and relatedly increase 00:33:51
the ability of the sense amplifier to amplify
one bitline to 1 volt and the other to 0 volts. 00:33:58
One final note is that when discussing DRAM,
one major topic is the timing of addresses, 00:34:04
command signals and data, and the related
acronyms DDR or double data rate, and SDRAM, 00:34:12
or Synchronous DRAM. These topics were omitted
from this video because it would have taken an 00:34:17
additional 15 minutes to properly explore.
That’s 00:34:24
pretty much it for the DRAM, and we are grateful
you made it this far into the video. We believe 00:34:30
the future will require a strong emphasis on
engineering education and we’re thankful to all 00:34:36
our Patreon and YouTube Membership Sponsors
for supporting this dream. If you want to 00:34:41
support us on YouTube Memberships, or Patreon,
you can find the links in the description. 00:34:47
A huge thanks goes to the Nathan, Peter, and
Jacob who are doctoral students at the Florida 00:34:53
Institute for Cybersecurity Research for helping
to research and review this video’s content! They 00:35:00
do foundational research on finding the weak
points in device security and whether hardware 00:35:05
is compromised. If you want to learn more about
the FICS graduate program or their work, check out 00:35:11
the website using the link in the description.
This is Branch Education, and we create 3D 00:35:18
animations that dive deep into the technology that
drives our modern world. Watch another Branch 00:35:24
video by clicking one of these cards or click here
to subscribe. Thanks for watching to the end!Description:
Check out Crucial NVMe SSDs Here: https://www.crucial.com/ Have you ever wondered why it takes time for computers to load programs or video games? Also, ever wonder why your computer uses both DRAM as well as SSDs when they both are used to store data? Well, most of that time is spent moving data from a hard drive or SSD into DRAM or Dynamic Random Access Memory, which is the working memory inside your computer. In this video, we're going to take a very deep dive into DRAM. We'll see how it connects to other parts of your computer, and then we'll explore how DRAM can store gigabytes of data in nanoscopic capacitors. After that, we'll cover the three main operations of DRAM: Reading, Writing, and Refreshing. And finally, we'll dive deep into some more complex aspects of DRAM that make it so amazingly fast such as folded DRAM architecture. We'll also learn what burst buffers are, and why there are so many banks of DRAM memory cells. Do you want to support in-depth engineering and technology education? Support us at: https://www.patreon.com/brancheducation Website: https://www.branch.education/ On Facebook: https://www.facebook.com/unsupportedbrowser On Twitter: https://twitter.com/TeddyTablante On Insta: https://www.facebook.com/unsupportedbrowser Thanks to Nathan, Peter, and Jacob for helping research and review this video! They're doctoral students at the Florida Institute for Cybersecurity Research, and you can learn more about their program here: https://fics.institute.ufl.edu/ Table of Contents: 00:00 - Intro to Computer Memory 00:47 - DRAM vs SSD 02:23 - Loading a Video Game 03:25 - Parts of this Video 04:07 - Notes 06:10 - Intro to DRAM, DIMMs & Memory Channels 10:43 - Crucial Sponsorship 12:09 - Inside a DRAM Memory Cell 15:28 - An Small Array of Memory Cells 17:41 - Reading from DRAM 19:38 - Writing to DRAM 21:55 - Refreshing DRAM 23:16 - Why DRAM Speed is Critical 25:06 - Complicated DRAM Topics: Row Hits 26:21 - DRAM Timing Parameters 27:51 - Why 32 DRAM Banks? 29:17 - DRAM Burst Buffers 30:58 - Subarrays 32:02 - Inside DRAM Sense Amplifiers 34:24 - Outro to DRAM Key Branches from this video are: How do Solid State Drives Work? Erratum: At 10m 08s : Cicruit || Should be Circuit At 21m 54s : 32 Bank Groups || Should be 32 Banks. Script, Modeling, Animation: Teddy Tablante Twitter: @teddytablante Animation: Mike Radjabov Modeling: Prakash Kakadiya Voice Over: Phil Lee Sound Design: www.drilu.mx Music Editing: Luis Zuleta Sound Effects: Paulo de los Cobos Supervising Sound Editor and Mixer: Luis Huesca Animation built using Blender 3.1.2 https://www.blender.org/ Post with Adobe Premiere Pro References: DDR5 SDRAM. JEDEC Standard. JESD79-5 July 2020 Dr. Cutress, Ian. "Insights into DDR5 Sub-Timings and Latencies". Oct 6th, 2020. Dr. El-Maleh, Aiman. "Functions and Functional Blocks: Digital Logic Design" College of Computer Sciences and Engineering. King Fahd University of Petroleum and Minerals. Hajimiri, Ali. Et al. "Design Issues in Cross-Coupled inverter Sense Amplifier". IEEE. Stanford University 1998 IBM. Understanding DRAM Operation. IBM 1996. Jacob, Bruce. NG, Spencer W. Wang, David T. "Memory Systems: Cache, DRAM, Disk." Elsevier Inc. 2008 Keeth, Brent. Baker, R Jacob. Johnson, Brian. Lin, Feng. "DRAM Circuit Design: Fundamental and High-Speed Topics." IEE Press 2008. Kim, Yoongu et. Al. "A Case for Exploiting Subarray-Level Parallelism in DRAM". Carnegie Mellon University Lee, Donghuk et.al. "Tiered-Latency DRAM: A Low Latency and Low Cost DRAM Architecture. Carnegie Mellon University Micron. "DDR4 SDRAM. MT40A4G4. MT40A2G8. MT40A1G16. 16Gb: x4, x8, x16 DDR4 SDRAM Features" Micron Technologies 2018 Micron. "DDR5 SDRAM Product Core Data Sheet DDR5SDRAM Features." Micron Technologies 2020 Ryan, Kevin J. Morzano, Christopher K. Li, Wen. "Write Data Masking for Higher Speed DRAMS" US Patent 6532180 B2 Mar. 11 2003. Shilov, Anton. "SK Hynix Details DDR5-6400". ANANDTECH. Feb 26th, 2019. Sunami, Hideo. "Dimension Increase in Metal-Oxide Semiconductor Memories and Transistors". From intechopen.com. From "Advances in Solid State Circuit Technologies". Apr 2010. Wikipedia contributors. "CAS Latency". "DDR5 SDRAM". "Dynamics Random-Access Memory". "Memory Timing". "Synchronous Dynamic Random-Access Memory". Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, Visited Nov 2022
Mediafile available in formats
* — If the video is playing in a new tab, go to it, then right-click on the video and select "Save video as..."
** — Link intended for online playback in specialized players
Questions about downloading video
How can I download "How does Computer Memory Work? 💻🛠" video?
http://univideos.ru/ website is the best way to download a video or a separate audio track if you want to do without installing programs and extensions.
The UDL Helper extension is a convenient button that is seamlessly integrated into YouTube, Instagram and OK.ru sites for fast content download.
UDL Client program (for Windows) is the most powerful solution that supports more than 900 websites, social networks and video hosting sites, as well as any video quality that is available in the source.
UDL Lite is a really convenient way to access a website from your mobile device. With its help, you can easily download videos directly to your smartphone.
Which format of "How does Computer Memory Work? 💻🛠" video should I choose?
The best quality formats are FullHD (1080p), 2K (1440p), 4K (2160p) and 8K (4320p). The higher the resolution of your screen, the higher the video quality should be. However, there are other factors to consider: download speed, amount of free space, and device performance during playback.
Why does my computer freeze when loading a "How does Computer Memory Work? 💻🛠" video?
The browser/computer should not freeze completely! If this happens, please report it with a link to the video. Sometimes videos cannot be downloaded directly in a suitable format, so we have added the ability to convert the file to the desired format. In some cases, this process may actively use computer resources.
How can I download "How does Computer Memory Work? 💻🛠" video to my phone?
You can download a video to your smartphone using the website or the PWA application UDL Lite. It is also possible to send a download link via QR code using the UDL Helper extension.
How can I download an audio track (music) to MP3 "How does Computer Memory Work? 💻🛠"?
The most convenient way is to use the UDL Client program, which supports converting video to MP3 format. In some cases, MP3 can also be downloaded through the UDL Helper extension.
How can I save a frame from a video "How does Computer Memory Work? 💻🛠"?
This feature is available in the UDL Helper extension. Make sure that "Show the video snapshot button" is checked in the settings. A camera icon should appear in the lower right corner of the player to the left of the "Settings" icon. When you click on it, the current frame from the video will be saved to your computer in JPEG format.
How do I play and download streaming video?
- in video formats, hover your mouse over "Streaming Video**";
- right-click on "Copy link";
- open VLC-player;
- select Media - Open Network Stream - Network in the menu;
- paste the copied link into the input field;
- click "Play".
- copy the video address (URL);
- select "Open Network Stream" in the "Media" item of VLC player and paste the link to the video into the input field;
- click on the arrow on the "Play" button and select "Convert" in the list;
- select "Video - H.264 + MP3 (MP4)" in the "Profile" line;
- click the "Browse" button to select a folder to save the converted video and click the "Start" button;
- conversion speed depends on the resolution and duration of the video.
For this purpose you need VLC-player, which can be downloaded for free from the official website https://www.videolan.org/vlc/.
How to play streaming video through VLC player:
To download streaming video via VLC player, you need to convert it:
Warning: this download method no longer works with most YouTube videos.
What's the price of all this stuff?
It costs nothing. Our services are absolutely free for all users. There are no PRO subscriptions, no restrictions on the number or maximum length of downloaded videos.