Speed testing -x16 vs. -x1 with PCI Express video cards

1. Introduction

Our xprs-px-x16 product allows using of PCI Express -x16 cards, outside of computer systems through cables. For details on the way of use, you can check this article .

Number of our customers, interested in using this product, asked technical details. The xprs-px-x16 allows using of a PCI Express -x16 card, through -x1 link and cables.

In general, one -x16 card, has 16 transmit and 16 receive pairs/lines. One -x1 card, has 1 transmit and 1 receive pair/line. Theoretically a -x16 card has to be 16 times faster than one -x1 card.

PCI Express interface allows the work of a card, using the smaller number of transmit/receive pairs.

This article compares the speed difference when the same -x16 card works using all 16 transmit/receive pairs, and when the card works using 1 transmit/receive pair.


2. Way of testing

PCI Express -x16 card normally is mounted on a desktop motherboard.

-x16 on a motherboard

(click on image to enlarge)



The above image is a very simplified representation of mounting. A desktop motherboard is normally mounted in a box, hooked to a power supply ... etc.

The image below shows the same video card mounted on our xprs-px-x16 product, connected through 1m cables, to xprs-host-dt mounted on the same motherboard.

-x16 on a xprs-px-x16

The above image is a also simplified representation of mounting. The xprs-px-x16 product has to be connected to a power supply in order to power on the PCI Express -x16 card.


3. Cards and tests

The tests wre performed on 2 video cards, both of them based on NVIDIA chips:
- geforce 8400 card, 512mb memory (card on right side)
- geforce GT 430 card, 1gb memory (card on left side)

2 -x16 cards

As a result of a web search for doing a benchmark test on a video card, we picked 2 tests:
- by Freestone-Group - free download
by Passmark Software - 30 days free for evaluation, $24 cost

freestone-group test

The Freestone-Group test software (image above) is only for benchmarking video cards and gives a single score as the end result.

The Passmark test software (image below) is for general benchmarking of a system, includes 2D and 3D tests for video cards, and gives separate scores for the each of the sub-tests.

passmark test


4. Test results

The Freestone-Group test gives these scores:

 
 Card        -x16        -x1     Difference
'8400'        44          40      9%
'gt 430'      238         231     2.9%



The Freestone-Group mix of tests like the above show a very small difference between -x16 and -x1 mode.


The Passmark test gives these scores for '8400' card:

   Test                           -x16    -x1    Difference
Graphics 2D - Solid Vectors:       0.9     0.9    0
Graphics 2D - Transparent Vectors: 0.9     0.8   11.1%
Graphics 2D - Complex Vectors:     87.1   84.1    3.4%
Graphics 2D - Fonts and Text:      36.3   28.8   20.6$
Graphics 2D - Windows Interface:   14.0   11.2   20.0%
Graphics 2D - Image Filters:       92.3   88.4    4.2%
Graphics 2D - Image Rendering:    202.2  141.5   30.0%
Graphics 3D - Simple:             130.5   83.1   36.3%
Graphics 3D - Medium:              50.6   34.4   32.0%
Graphics 3D - Complex:             14.7    2.3   84.3% !!!
Graphics 3D - DirectX 10:           0.9    0.9    0 
2D Graphics Mark:                 157.7  135.7   13.9%
3D Graphics Mark:                 117.9   65.7   42.2%



The Passmark test gives these scores for 'gt 430' card:

   Test                              -x16     -x1    Difference
Graphics 2D - Solid Vectors:          1.0     1.0    0
Graphics 2D - Transparent Vectors:    1.0     1.0    0
Graphics 2D - Complex Vectors:       91.4    77.4   15.3%
Graphics 2D - Fonts and Text:        36.1    35.5    1.6%
Graphics 2D - Windows Interface:     14.0    14.0    0
Graphics 2D - Image Filters:         92.6    92.4   0.1%
Graphics 2D - Image Rendering:      203.7   203.1   0.3%
Graphics 3D - Simple:               406.5   396.9   2.3%
Graphics 3D - Medium:               271.3   204.5  24.6%
Graphics 3D - Complex:               16.0     4.0  75.0% !!!
Graphics 3D - DirectX 10:             7.4     4.9  33.7%
2D Graphics Mark:                   162.1   156.0   3.7%
3D Graphics Mark:                   460.0   337.5  26.6%


The tests show a very small difference for the 2D Graphics tests.

A large difference appears for the Complex 3D Graphics test. However the larger the video memory is and the better/faster the GPU - the gap between -x16 and -x1 modes gets smaller.

Conclusion - a PCI Express -x16 card placed in '-x1' mode is getting most of the performance, and may be good enough solution in case of using a peripheral card outside of a computer system, through a cable.


5. DIY (do-it-yourself) testing

While we did test with 2 video cards, they are most likely different your own video card(s).

If you are curious about what the difference is between -x16 and -x1 modes for your video cards, there is a way to find this.

The PCI Express connectors have compatible pin-outs - with the signals on the pins on the -x1 connector identical to the respective pins for the part of the -x16 connector.

In practice this means that you can plug one -x1 card into a -x16 connector an it will still work.

-x1, -x16 cards and a tape

The opposite is also true - using a tape, you can cover the additional pins of one -x16 card, and leave exposed only pins matching -x1 connector.

forcing -x1 mode

When you plug a -x16 video card prepared in such way in a motherboard, it will be forced to operate in -x1 mode.

Then you can run some of the above tests or any other test you like.

After testing, when the tape is removed, please make sure the pins on the -x16 card are cleaned with alcohol or other solutions.