Do PCI-e riser cards introduce a lot of latency?

ViRGE

Elite Member, Moderator Emeritus
Oct 9, 1999
31,516
167
106
Since they're just mechanical devices, I doubt it.
 

Keysplayr

Elite Member
Jan 16, 2003
21,209
50
91
As long as the riser card has only 1 device installed, it should not introduce any latency. But if you perhaps were to install two or more devices into the riser (some PCI risers I used to see had 3 slots available in them for low profile (3U) rack servers), then I could see latency being introduced as they would need to share the bandwidth of one slot.

Here is a little bit about it:

Link
 

Zap

Elite Member
Oct 13, 1999
22,377
2
81
Originally posted by: keysplayr2003
Here is a little bit about it:

Link

Interesting link. With converting from one interface to another, besides losing some performance I'd be a bit more worried about the bandwidth of the slowest interface. For instance, I have benchmarked the same GPU using both PCIe and PCI interfaces (8800 GS, PCI version uses a bridge chip) and the PCI version was definately slower, but I believe most (but not all) of that performance loss due to the PCI bus (versus native PCIe) and not from any introduced latencies.
 

thilanliyan

Lifer
Jun 21, 2005
11,868
2,075
126
Originally posted by: keysplayr2003
As long as the riser card has only 1 device installed, it should not introduce any latency. But if you perhaps were to install two or more devices into the riser (some PCI risers I used to see had 3 slots available in them for low profile (3U) rack servers), then I could see latency being introduced as they would need to share the bandwidth of one slot.

Here is a little bit about it:

Link

Thanks for the link.
 

Keysplayr

Elite Member
Jan 16, 2003
21,209
50
91
Originally posted by: Zap
Originally posted by: keysplayr2003
Here is a little bit about it:

Link

Interesting link. With converting from one interface to another, besides losing some performance I'd be a bit more worried about the bandwidth of the slowest interface. For instance, I have benchmarked the same GPU using both PCIe and PCI interfaces (8800 GS, PCI version uses a bridge chip) and the PCI version was definately slower, but I believe most (but not all) of that performance loss due to the PCI bus (versus native PCIe) and not from any introduced latencies.

Oh absolutely. Any of these modern cards would be starved with the old PCI standard. Makes perfect sense that it would be slower than PCI-e, or even an AGP variant.
 

Keysplayr

Elite Member
Jan 16, 2003
21,209
50
91
Originally posted by: thilan29
Originally posted by: keysplayr2003
As long as the riser card has only 1 device installed, it should not introduce any latency. But if you perhaps were to install two or more devices into the riser (some PCI risers I used to see had 3 slots available in them for low profile (3U) rack servers), then I could see latency being introduced as they would need to share the bandwidth of one slot.

Here is a little bit about it:

Link

Thanks for the link.

:thumbsup: