We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.
[nextpage title=”Introduction”]
One of the features found on the most recent CPUs, chipsets, motherboards, and video cards is the PCI Express 3.0 connection. Nevertheless, does it offer an actual performance improvement over the PCI Express 2.0 standard? Let’s find out!
PCI Express 3.0 connection was specified in 2010, with a maximum theoretical transfer rate per lane of almost 1 GiB/s (actually, 984.6 MiB/s), twice the rate of the PCI Express 2.0 standard that offers 500 GiB/s per lane. Thus, a PCI Express 2.0 x16 slot offers a maximum theoretical bandwidth of 8 GiB/s, while a PCI Express 3.0 x16 slot reaches 16 GiB/s.
| Preview | Product | |
|---|---|---|
|
USB PCIe Card, 4 Port USB 3.0 to PCI Express Card Expansion card, PCI-E to USB 3.0 4 Port Hub... | Buy on Amazon |
Let us keep in mind that those are the maximum speeds this connection supports, which does not mean the video card will transfer data at these speeds. For more detailed technical information, read our “Everything you need to know about the PCI Express” tutorial.
Regarding to video cards, all current models are compatible with PCI Express 3.0; the first NVIDIA chips compatible with this standard were from GeForce GT/GTX 6xx generation, while the AMD models use it since Radeon HD 7xxx models.
On the other side, in most cases, is the CPU that supports PCI Express 3.0, not the chipset. However, it is necessary for the motherboard also to be compatible with the standard. Intel CPUs support PCI Express 3.0 since the third-generation Core I (“Ivy Bridge”) processors. AMD A-series CPUs (aka APUs) support the standard on all FM2+ models. FX processors, on the other hand, do not support PCI Express 3.0, because, on this platform, the PCI Express lanes are generated by the chipset, and even the most high-end model, 990FX, supports only PCI Express 2.0.
Besides the big difference on the maximum theoretical bandwidth between the CPU and GPU, we were still curious about the real-life performance impact on games by using a PCI Express 3.0 against a 2.0 connection. So, we ran 3D benchmarks using 3DMark and some recent games, using a high-end video card (which, in theory, demands more bandwidth than a mainstream or value one), first with the slot configured as PCI Express 3.0 x16, then with the same slot configured as PCI Express 2.0 x16. Finally, we tested with the same video card, but this time using a slot that supports up to PCI Express 3.0 x4, to check if an ever lower bandwidth could affect performance.
Figures 1, 2, and 3 show the configuration of the video card slot, on the “Bus Interface” field at GPU-Z screen.
Figure 1: using PCI Express 3.0 x16
Figure 2: using PCI Express 2.0 x16
Figure 3: using PCI Express 3.0 x4
We will list the configuration we used on our tests on the next page.
[nextpage title=”How We Tested”]
During our benchmarking sessions, we used the configuration listed below. Between our benchmarking sessions, the only differences were the video card PCI Express configuration.
Hardware Configuration
- CPU: Core i7-5775C
- Motherboard: ASRock Z97 Extreme4
- CPU Cooler: Intel
- Memory: 16 GiB DDR3-2133, four G.Skill Ripjaws F3-17000CL9Q-16GBZH 4 GiB memory modules configured at 1,600 MHz
- Boot drive: Kingston HyperX Savage 480 GB
- Video Card: GeForce GTX 980 Ti
- Video Monitor: Philips 236VL
- Power Supply: Corsair CX750
Operating System Configuration
- Windows 7 Home Premium 64-bit
- NTFS
- Video resolution: 1920 x 1080 60 Hz
Driver Versions
- NVIDIA driver version: 361.43
Software Used
Error Margin
We adopted a 3% error margin. Thus, differences below 3% cannot be considered relevant. In other words, products with a performance difference below 3% should be considered as having similar performance.
[nextpage title=”3DMark”]
3DMark is a program with a set of several 3D benchmarks. Fire Strike benchmark measures DirectX 11 performance and is targeted to high-end gaming computers. Sky Diver also measures DirectX 11 performance and is aimed at average computers. The Cloud Gate benchmark measures DirectX 10 performance. The Ice Storm measures DirectX 9 performance and is targeted to entry-level computers, so we did not include it on our comparison.
Keep in mind that we used a GeForce GTX 980 Ti video card on all tests.
We ran every benchmark three times, and each number in the graphs is the arithmetical mean between the obtained scores.
On Fire Strike, the performance was the same on all the configurations.
On Sky Diver, there was no difference between the PCI Express 3.0 x16 and PCI Express 2.0 x16 benchmarks. On the other hand, on the test using PCI Express 3.0 x4, the performance was 5.5% below the first test.
On the Cloud Gate benchmark, there was also no difference between PCI Express 2.0 and 3.0, but the test using PCI Express 3.0 x4 led to a 7.9% lower performance.
[nextpage title=”Gaming Performance”]
Battlefield 4
Battlefield 4 is one of the most popular games of the Battlefield franchise, being released in 2013. It is based on the Frostbite 3 engine, which is DirectX 11. In order to measure performance using this game, we walked our way through the first mission, measuring the number of frames per second (fps) three times using FRAPS. We ran this game at Full HD, setting overall image quality at “high.”
The results below are expressed in fps and they are the mean between the three collected results.
On Battlefield 4, there was no significant performance difference on the three tests.
Dirt Rally
Dirt Rally is an off-road racing game released in April 2015, using Ego engine. To measure performance using this game, we ran the performance test included in the game, in Full HD resolution and image quality configured as “high” and MSAA x2.
The results below are expressed in frames per second (fps).
On the Dirt Rally, the performance with PCI Express 3.0 x16 and PCI Express 2.0 x16 was the same. However, using PCI Express 3.0 x4, the performance dropped by 14%.
Dying Light
Dying Light is an open-world horror game launched in January 2015, using the Chrome Engine 6. We tested the performance at this game with quality options at “high” and Full HD resolution, measuring three times the frame rate using FRAPS.
The results below are expressed in fps and they are the mean between the three collected results.
In this game, the performance with PCI Express 3.0 x16 was 7% superior to the test with PCI Express 2.0 x16. On the PCI Express 3.0 x4 test, the performance was similar to the PCI Express 2.0 x16.
Grand Theft Auto V
Grand Theft Auto V, or simply GTA V, is an open-world action game released for PCs in April of 2015, using the RAGE engine. In order to measure the performance on this game, we ran the performance test of the game (the part the camera follows the plane), measuring the frame rate with FRAPS. We ran GTA V at Full HD, with image quality set to the maximum in options referring to textures and to the minimum regarding post-processing.
The results below are expressed in frames per second.
On GTA V, there was a performance improvement of 3.6% when we used PCI Express 3.0 x16, compared to the test with PCI Express 2.0 x16. The test with PCI Express 3.0 x4 ran 13% slower than the first test.
Mad Max
Mad Max is an open-world action game launched in September of 2015, using the Avalanche engine. In order to measure the performance using this game, we ran its intro, measuring the frame rate with FRAPS three times. We ran the game at Full HD, with image quality set as “high.”
The results below are expressed in fps and they are the mean between the three collected results.
On Mad Max, the result was the same on the three tests.
[nextpage title=”Conclusions”]
Our tests showed a clear result: on only one game (Dying Light) there was a significant performance gain by using PCI Express 3.0 instead of PCI Express 2.0. Even at it, the improvement was of only 7%. On the other hand, when we used the PCI Express 3.0 x4 slot, there was a loss of performance on most of the games.
Remember that we used a high-end video card, in a high-end system. Even we had not run tests on a more basic system, it is logical to think that, on a system with a lower video card performance, the impact of the bandwidth between the CPU and the GPU should be even smaller.
So, if you are looking for a video card, CPU or motherboard to buy, the presence or absence of the PCI Express 3.0 connection should not be a crucial factor do decide what to buy, at least on a single-GPU system.
Last update on 2021-06-16 at 18:07 / Affiliate links / Images from Amazon Product Advertising API