Integrity Servers
cancel
Showing results for 
Search instead for 
Did you mean: 

RX3600/RX6600 I/O slot speed

 
mits
Respected Contributor

RX3600/RX6600 I/O slot speed

All,

This is just FYI.

Current RX3600/RX6600 quick specs have inacurate I/O slot speed description below.

++++++
The two slots at the outside edge of the backplane (slots #9 and #10) each have their own dedicated 64 bit 266 MHz PCI X bus and their own independent 2.1 GB/s I/O channel. These slots should be used for high performance PCI X cards, such as clustering interconnects or multi port storage adapters. The next two slots (slots #7 and #8) each have their own dedicated 64 bit 133 MHz PCI X bus and their own independent 1.0 GB/s I/O channel. The independent buses and I/O channels provide improved performance and error containment. Independence protects each I/O card from bus hangs or extended latencies due to the failure or high bandwidth demands of other I/O cards. Independence also ensures that each I/O card can achieve maximum throughput. Both of these slots support hot plug operations in HP UX. Hot plug is not supported with the Linux or OpenVMS operating systems.

The remaining four slots (#3 through #6) share two 64 bit 66 MHz PCI X buses, with two slots allocated to each bus. Each pair of slots shares a 0.5 GB/s I/O channel. If a 33 MHz card shares a bus with a 66 MHz card, the faster card will downgrade to 33 MHz. All four of these slots support hot plug operations in HP UX and Windows. Hot plug is not supported with the Linux or OpenVMS operating systems.
+++++

As you can see, it says slot 9 and 10 have 266MHz PCI-X bus. This is wrong. Here is a command "olrad -q" output.

# olrad -q
Driver(s)
Capable
Slot Path Bus Max Spd Pwr Occu Susp OLAR OLD Max Mode
Num Spd Mode
3 0/7/1 35 266 266 Off No N/A N/A N/A PCI-X PCI-X
4 0/3/1 15 266 266 Off No N/A N/A N/A PCI-X PCI-X
5 0/6/1 30 133 133 Off No N/A N/A N/A PCI-X PCI-X
6 0/2/1 10 133 133 Off No N/A N/A N/A PCI-X PCI-X
7 0/5/1 21 66 66 Off No N/A N/A N/A PCI-X PCI-X
8 0/5/2 21 66 66 Off No N/A N/A N/A PCI-X PCI-X
9 0/1/1 1 66 66 Off No N/A N/A N/A PCI-X PCI-X
10 0/1/2 1 66 66 On Yes No Yes N/A PCI-X PCI-X

RX3600/RX6600 service guides have accurate information. I have sent my feedback on this document issue to the webmaster. If you have any I/O performace issue with RX3600/RX6600, you may need to check your I/O slot configuration.
4 REPLIES 4
mits
Respected Contributor

Re: RX3600/RX6600 I/O slot speed

Closed
rick jones
Honored Contributor

Re: RX3600/RX6600 I/O slot speed

I too came across this when doing some 10G NIC netperf work on an rx6600 under Linux. I contacted a number of folks I know and fixes to the quickspecs should be forthcoming RSN. If not, please let me know so I can go knock some heads.

If someone with an rx4640 and HP-UX can do the same sort of sanity check that would be good. Some of the shared wording between the quickspecs makes me worry, and I've no rx4640 with which to check.
there is no rest for the wicked yet the virtuous have no pillows
Torsten.
Acclaimed Contributor

Re: RX3600/RX6600 I/O slot speed

Hi Rick,

this is what the rx4640 reports:

#:olrad -q
Driver(s)
Capable
Slot Path Bus Max Spd Pwr Occu Susp OLAR OLD Max Mode
Num Spd Mode
3 0/3/1 128 66 N/A On Yes No Yes N/A PCI-X PCI
4 0/3/2 128 66 N/A Off No N/A N/A N/A PCI-X PCI
5 0/4/1 160 66 66 On Yes No Yes N/A PCI-X PCI-X
6 0/4/2 160 66 66 Off No N/A N/A N/A PCI-X PCI-X
7 0/2/1 64 133 66 On Yes No Yes N/A PCI-X PCI
8 0/5/1 192 133 66 On Yes No Yes N/A PCI-X PCI

Hope this helps!
Regards
Torsten.

__________________________________________________
There are only 10 types of people in the world -
those who understand binary, and those who don't.

__________________________________________________
No support by private messages. Please ask the forum!

If you feel this was helpful please click the KUDOS! thumb below!   
rick jones
Honored Contributor

Re: RX3600/RX6600 I/O slot speed

Torsten - thanks - therein I suspect resides "the problem" - I think the rx3600 and rx6600 QuickSpecs were copy-and-edit jobs from the rx4640's and I'll bet euros to berliners that whomever did the editing "assumed" that the order of fast to slow slots was the same in the rx3600 and rx6600 as it was in the rx4640, when we can see that it is reversed.
there is no rest for the wicked yet the virtuous have no pillows