E4500 - Can't Get 100MHz Bus
All
I have an E4500, prtdiag below. I had two CPU boards replaced with ones that are 100MHz capable, but the system still runs at 84MHz.
What is holding it back? Many thanks in advance.
System Configuration: Sun Microsystems sun4u 8-slot Sun Enterprise E4500/E5500
System clock frequency: 84 MHz
Memory size: 6656Mb
========================= CPUs =========================
RunEcacheCPUCPU
Brd CPUModuleMHzMBImpl.Mask
- ---
0003364.0US-II2.0
0113364.0US-II2.0
2403364.0US-II2.0
2513364.0US-II2.0
4803364.0US-II2.0
4913364.0US-II2.0
61203364.0US-II2.0
61313364.0US-II2.0
========================= Memory =========================
Intrlv. Intrlv.
BrdBankMBStatusCondition SpeedFactorWith
-- - - - -- - -
001024ActiveOK60ns4-wayA
011024ActiveOK60ns2-wayB
201024ActiveOK60ns4-wayA
211024ActiveOK60ns2-wayB
401024ActiveOK60ns4-wayA
41256ActiveOK60ns2-wayC
601024ActiveOK60ns4-wayA
61256ActiveOK60ns2-wayC
========================= IO Cards =========================
BusFreq
Brd Type MHzSlotName Model
- - - - --
1SBus250 QLGC,isp/sd (block)QLGC,ISP1000U
1SBus251 lpfs/sd (block)LP9002S
1SBus253 SUNW,hme
1SBus253 SUNW,fas/sd (block)
3SBus250 QLGC,isp/sd (block)QLGC,ISP1000U
3SBus251 lpfs/sd (block)LP9002S
3SBus253 SUNW,hme
3SBus253 SUNW,fas/sd (block)
========================= HW Revisions =========================
ASIC Revisions:
Brd FHC AC SBus0 SBus1 PCI0 PCI1 FEPS Board TypeAttributes
-- -- -- - - - --
015 CPU 100MHz Capable
1151122Dual-SBus-SOC+ 100MHz Capable
215 CPU 100MHz Capable
3151122Dual-SBus-SOC+ 100MHz Capable
415 CPU 100MHz Capable
615 CPU 100MHz Capable
System Board PROM revisions:
-
Board 0:OBP3.2.30 2002/10/25 14:03POST 3.9.30 2002/10/25 14:04
Board 1:FCODE 1.8.30 2002/10/25 14:02iPOST 3.4.30 2002/10/25 14:03
Board 2:OBP3.2.30 2002/10/25 14:03POST 3.9.30 2002/10/25 14:04
Board 3:FCODE 1.8.30 2002/10/25 14:02iPOST 3.4.30 2002/10/25 14:03
Board 4:OBP3.2.30 2002/10/25 14:03POST 3.9.30 2002/10/25 14:04
Board 6:OBP3.2.30 2002/10/25 14:03POST 3.9.30 2002/10/25 14:04
[2382 byte] By [
Mr_Jimbo] at [2007-11-26 9:25:39]

# 1
It took me some time, so very long ago, to eventually understand this issue.
The system bus speed in Sun's Enterprise server systems, (E3000 through the E6500)
is a result of the configuration, not an elementary root function of the configuration.
It's the reverse of what you usually see in the PC universe.
Think of it as backwards to what you might have otherwise expected.
The clockboard governs what will be the proper bus speed, per the cpu's you've installed.
For your current configuration, you have 336MHz processors,
which have an inherent design multiplier/divisor of 4:1.
So ... 336 divided by 4 == 84
If you replaced those cpus with 400MHz modules,
and use FRUs that were designed for the same 4:1 divisor,
your E4500 gigaplane bus would spontaneously be clocked at 100MHz.
There's another "gotcha" for the E6000's and E6500's.
Since their centerplane is so very big, as perceived at a computer bit-and-byte level,
their system bus speed will NEVER be 100MHz.
There isn't enough power to drive them that fast and maintain data integrity.
The E6x00 clockboards will commonly set their bus to 80MHz to avoid EMI data corruption.
(5:1 and 6:1 divisors as may be appropriate)
# 3
All eight of your current cpu modules are 336MHz and each has a 4mb L2 cache.
Your clockboard has set the gigaplane bus to 84MHz
because the cpus were designed and manufactured with a 4:1 divisor.
The clockboard is functioning perfectly.
Think of it as quickly peeking at the processors during POST, finding their clock ratio,
and then it sets the bus speed to whatever is best.
If you were to shut the system down and take one of those cpus out and examine it,
you would see that it has a part number of 501-4363.
That happens to be the only [url=http://www.sunshack.org/data/sh/2.1/infoserver.central/data/syshbk/Systems /E4500/component.cpu-module.html]336MHz part[/url] that's ever been in those systems.
A 100MHz bus speed is a theoretical maximum.
It will be that or slower.
You cannot set the bus speed. It will be, whatever it ends up as ...