Today we will tell you about the high-speed PCIexpress bus. What is it, how does it work and why is it needed?
What is PCI? Parsing
We all hear about PCI-Express lanes all the time. There are 8 of them here, there are 16 of them. Here is PCI-Express 3.0, here 4.0, etc. But what does it all mean? And why is it now almost as important to have a fresh PCI Express as a powerful processor or video card?
So today we will understand the technology? Let's find out what a tire is.
Compare gigatransactions and gigabytes. And also find out why PCI Express is the fastest way to get you to work.
What?
First, let's figure out what is PCI-Express and why is it needed?
And for this, let's solve a logical puzzle.
Here is a 4K monitor, a powerful graphics card and a fast SSD drive. What do all these devices have in common? In addition, all this now costs incredible money.
Let's discuss. All these devices need to exchange data with the central processor for normal operation, and in huge volumes and at high speeds. Therefore, all these devices can be connected to the CPU via the high-speed PCI-Express bus.
PCI Express (Peripheral Component Interconnect Express)
What is a bus and why is it high speed?
Well, look, in order to combine something with something inside a computer, we firstly need a physical connection, and secondly, some kind of protocol, that is, the rules by which data will be transmitted.
And so both that and another in one bottle is and there is a tire. And PCI-Express is one of the fastest buses. But how fast?
Here I have a laptop, this is MSI GP66 Leopard 11UH-229RU. It supports the current PCIe version 4.0.
Theoretically, if you put a fast SSD with PCIe Gen 4 support here, you can get speed 7, maybe even 7.5 GB / s. Imagine 7 GB per second! And I have seen such figures in practice!
And since there is a slot for a second drive, if you combine two drives in a RAID, you can double this figure.
But in my configuration, not the fastest Kingston SSD with QLC memory and only PCIe 3.0 support is installed. Therefore, it will not be possible to test the bus's capabilities on it.
But where do these speeds come from? What are these lines? And what PCIe is really capable of, let's talk about this.
How does PCIe work?
So why is PCIe so fast. First of all, it's about architecture.
Imagine that you need to drive from home to an office that is 10 km away from you. Under what conditions will the journey take the least time? It should be straight, with no speed limit, and preferably no other cars at all.
So, just PCI (without express) worked like a regular city road network, all devices were connected in parallel to a common bus, that is, their data ran along the same roads. Therefore, the speed of work could vary depending on the traffic.
But the PCIe standard has paved an individual path for each device, which directly connects it to a central hub located either on the same chip with the central processor, or separately on the chipset.Therefore, PCIe guarantees a stable maximum connection speed for each device.
And such an individual road is bilateral. Therefore, the data can be transmitted to one, and in the other side simultaneously at full speed. This data transfer mode is called duplex.
Each such road from the device to the switch is called the PCIE line. And such lines for each device can be spent several: 1, 2, 4, 8 or 16, depending on the needs. Therefore, the more lines, the greater the speed, that is, 2 lines are 2 times faster than one, and 16 lines 16 times faster. And no loss.
I think that the general principle is clear, but now let's go deep and wonder how the data is transmitted.
Gigatrazakia.
The speed affects not only the architecture, but also the method of data transfer.
At the electrical level, each connection uses low-voltage differential signal transmission or LVDS – Low-Voltage Differedial Signaling.
We will not deepen in it. But in general terms, in the LVDS, the signal is encoded by feeding different voltage levels of a very small amplitude of the signal.
Such an approach avoids interference outside and transmit data at high frequencies with a cheap copper twisted pair. In general, the benefits of mass.
Data in computer tires are transmitted by packages. Each packet of transmitted data is called a transaction. Therefore, the data transfer rate in the tires is not measured in gigabit or gigabytes per second and in gigatransactions per second (well, or gigatransfers, so also say). What it is?
In fact, the transaction per second is not the volume of transmitted data, but the frequency with which this data is transmitted. And than this frequency is higher, the higher the bandwidth.
It looks like a CPU, the more Gigahertz, theoretically faster.
For example, for example, in PCI-E 1.0, each line operates with a frequency of 2.5 GT / s, that is, 2.5 billion transactions per second. And in version 2.0. The frequency is 2 times higher than 5 Gt / s, therefore the interface bandwidth is 2 times higher.
One PCIE 1.0 line passed 0.25 GB / s, and 2.0 already 0.5 GB / s.
But not only the tire frequency affects the bandwidth.
In PCIe 3.0, the frequency has grown out from 5 GT / s to 8, and it seems not much. But at the same time the bandwidth rose almost twice! How so?
The fact is that the tire speed also affects the method of encoding information.
When transmitting data, it is important for us to be sure that everything is accomplished safe and maintained. Therefore, special integrity checks are added to each package with data and other service information that also occupy a place.
For example, in the first and second versions of PCIE for every 8 bits of useful information, there were 2 official bits. Such an encoding is indicated as 8 bits divided by 10 bits. or 8b / 10b.
This means that the 20% of the bus bandwidth was spent on the transfer of official information.
Starting from the 3rd version of the interface began to use the encoding 128B / 130B.This means that for every 2 bits of service information, there are 128 bits of useful information. And this is only 1.5% loss. Therefore, despite the slightly increased frequencies from 5 to 8 Gigatransfers, in the third version of PCIe, the throughput increased by almost 2 times.
By the way, if you want to accurately calculate the throughput of each PCIe version, you can do it using this formula. Well, or just look at the plate on Wikipedia.
PCIe bandwidth calculation formula:
- BW (MB/s) = FR (MT/s) * EN * 1B/8b
- BW – desired transfer rate in MB / s
- FR – bus frequency in GT/s
- EN – encoding type
- Example, for PCIe 2.0:
- BW = 5000 * 8/10 * 1/8 = 5000 * 0.8 * 0.125
- BW = 500 MB/s
Why is PCIe 4.0 needed?
The current version is PCI Express 4.0. Unlike the previous version 3.0, the new bus is 2 times faster, the data transfer rate on the bus here reaches 16 billion transfers per second. And the bandwidth of one line here reaches almost 2 GB / s (1969 MB / s). And as much as 31.5 GB / s for 16 lines!
Just imagine – 31 GB per second. But why is this needed and how does it affect performance in general?
Well, look, the 3rd generation of the tire appeared already in 2010 and is still used almost everywhere. And the 4th generation was ready back in 2017, but only now it is starting to become mainstream.
Everything is developing so slowly because, until now, there have been very few tasks for which such data transfer speeds would be needed. That is, there was no point in updating before.
But now there are affordable and super fast SSD drives that are not enough for the standard 4x PCI 3.0 lanes at 3.9 GB / s. And then there are powerful graphics cards and high-resolution monitors that also need a lot of bandwidth.
Therefore, first of all, PCI-e 4.0 is needed for modern games and support for DirectStorage technology from Microsoft.
If you want games to load fast, textures to load quickly and be in high photorealistic resolution, and level transitions and location changes to be seamless, a fast SSD and PCIe 4.0 is a must.
All of these things are already standard on next generation consoles, and so all future games will also be developed with these new fast data streaming capabilities in mind.
Secondly, if you have several 4K and 5K monitors that you work with at the same time, having PCI Express version 4 may also become a necessity. But these are already more professional cases, and here people themselves know what they need.
How does it affect performance?
Okay, let's move on to practice and our test laptop MSI GP66 Leopard
MSI GP66 Leopard 11UH-229RU
- Processor Intel Core i7-11800H
- RAM 16 (2×8) GB DDR4-3200
- Nvidia GeForce RTX 3080 video subsystem
- Screen 15.6, 1920×1080, 144Hz, IPS
- 512 GB SSD + free M.2 2280 slot
- Battery 65 Wh
- Dimensions 358×271×34 mm (maximum measured by us)
- Weight without power supply 2.32 kg (measured by us)
The strengths of this laptop are the powerful 11th gen Intel 11800H processor, which is matched by 20 PCIe 4.0 lanes to better expose the Nvidia RTX 3080 graphics card.
Plus, the laptop has a rather thick case and a proprietary CoolerBoost 5 cooling system.As well as a wonderful IPS display from LG with a frequency of upgrade 144 Hz.
Well, the main drawback is a complete SSD, which clearly does not reveal the possibilities of iron. Nevertheless, you can always put the second SSD if this is not enough.
This laptop is not positioned as a game, such as this is a laptop for computing. But judging by the characteristics, as well as the RGB keyboard illumination. We seem to understand which calculations you mean MSI, so I launched Cyberpunk 2077 on it.
And the laptop showed himself in the game dough perfectly. In Cyberpunk 2077, I set the ultra settings and in the laptop the mode of extreme performance.
Also look at the test results in other games in comparison with competitors:
How much has affected the presence of PCI Express 4.0 on these results, it is difficult to say. But the interface was not exactly a bottleneck, and completely its potential can be disclosed in the future when the DirectStorage technology will appear in Windows and if you put a new SSD.
Otherwise, the laptop is very good for resource-intensive computing and just ideal for games.
What to wait in the future?
What do we expect in the future? We are waiting for PCI Express 5.0 and 6.0, each of which is twice as fast as its predecessor.
The 5th generation specification was finalized back in 2019, and the first processors with support for the new tire appeared on sale already very soon in November 2021, these will be the Intel processors of the 12th generation of ALDER LAKE.
There are no need for the 5th generation of tires yet. Almost everyone is enough of the 4th version for many years. Is that now the same speeds will be available with a smaller number of lines, which will save a little space.
As for the 6th generation. The specification was to be approved this year, but was transferred to the 22nd, and the appearance of the first devices before 25 years can not be expected.
But what can be said for sure that this year the PCIe 3.0 time, which lasts us for more than 10 years already considered. Therefore, if you take a computer or laptop, then at least with the 4th version.
