Hi,

What type of processors (x86,MIPS etc) do CISCO routers use ?


Regards,
Tejas Kokje
University of Southern California

Tejas Kokje wrote:

> Hi,
>
> What type of processors (x86,MIPS etc) do CISCO routers use ?
>
>

Yes.

To go into a little more depth:

700 Series Intel 80386SL
800 Series Motorola PowerQUICC 8xx PowerPC core
1000 Series Motorola Dragonball
1600 Series Motorola Dragonball
2500 Series Motorola 680EC30
3100 Series Motorola 680EX30
IGS Motorola 68020
AGS+ Motorola 68040
7000 Series Motorola 68040 (On RSP)
3620 IDT Orion (MIPS 4700 clone)
2600 Motorola PowerQUICC PowerPC
7200 MIPS of some description
Cat 2900 PowerPC 8xx
1700 I believe PowerPC of some description
7500s R4700(RSP2)R7000(RSP8) - MIPS family


I've probably missed some that used oddballs - for example some of their
non-router devices have used some odd things such as the MicroWebServer 100
using an 80186.

As you can see, they have a love of Motorola, however they use quite a lot
of other chips around the place.

As a rule they've used 68xxx series at the lowend, PowerPC in the mid range
and MIPS at the high end. There are some exceptions such as the AGS which
is due to it being a very old box.

P.

In article <2m5jukFei1jbU1@uni-berlin.de>, Paul S. Brown
<usenet060103@geekstuff.co.uk> wrote:

> Tejas Kokje wrote:
>
> > Hi,
> >
> > What type of processors (x86,MIPS etc) do CISCO routers use ?
> >
> >
>
> Yes.
>
> To go into a little more depth:
> [snip]
> AGS+ Motorola 68040

Depended on the vintage. We have a CGS still running (CGS, MGS, AGS
and AGS+ all used the same boards) with a CSC/4 processor and that's a
68040 but I think the earlier processors were 68030 or even 68020. I
remember hearing that CSC/1 was apparently built from the same basic
CAD layout (a Multibus or Multibus II board) as the Sun 1 workstation.
I don't know if that was a 68000, 68010 or 68020.

> 3620 IDT Orion (MIPS 4700 clone)

Our 3660 reports an R527x - don't recognise that one.

> 7200 MIPS of some description

R7000s in our VXRs but might be different in the earlier or later
models.

> Cat 2900 PowerPC 8xx

Cat 2950 is an RC32300 - don't recognise that either.
Cat 3548-XL is a PowerPC403.
Cat 3550-12G/T are reported as a generic PowerPC.

> 7500s R4700(RSP2)R7000(RSP8) - MIPS family

The Cat5500 RSM is effectively an RSP2 and is therefore an R4700.

MSFC2s seem to be R7000/SR71000 (both are reported in "show hard" - I
assume they're members of the same family).

In the big Cats the switching processor isn't reported - CatOS doesn't
have the same output and when you run Supervisor IOS you're talking to
the MSFC anyway.

> As a rule they've used 68xxx series at the lowend, PowerPC in the mid range
> and MIPS at the high end. There are some exceptions such as the AGS which
> is due to it being a very old box.

Back in the days when an AGS[+] was high end the 68K series was pretty
hot and RISC chips of any kind were rare. Moore's Law left the AGS+
standing, of course...

Sam

Sam Wilson wrote:

> In article <2m5jukFei1jbU1@uni-berlin.de>, Paul S. Brown
> <usenet060103@geekstuff.co.uk> wrote:
> Depended on the vintage. We have a CGS still running (CGS, MGS, AGS
> and AGS+ all used the same boards) with a CSC/4 processor and that's a
> 68040 but I think the earlier processors were 68030 or even 68020. I
> remember hearing that CSC/1 was apparently built from the same basic
> CAD layout (a Multibus or Multibus II board) as the Sun 1 workstation.
> I don't know if that was a 68000, 68010 or 68020.

Point - my AGS has a CSC/4 and thus the 68040 - the CSC/2 was a 68020.

>
>> 3620 IDT Orion (MIPS 4700 clone)
>
> Our 3660 reports an R527x - don't recognise that one.

Probably another Orion MIPS clone.

>
>> 7200 MIPS of some description
>
> R7000s in our VXRs but might be different in the earlier or later
> models.
>
That was actually pulled from the spec sheets for the NPE200/300/400 - dunno
about the NPE-G, but I can't imagine they'll have changed that as that
would bugger up IOS compatibiltiy unneccesarily.



> The Cat5500 RSM is effectively an RSP2 and is therefore an R4700.
>
On the other hand there's the RSM blade for the Cat 3000 series which is a
2503 and uses a 68030. There was also an earlier RSM blade for the 5500s
which was a 4500M on a stick - don't know what CPU it used.

> MSFC2s seem to be R7000/SR71000 (both are reported in "show hard" - I
> assume they're members of the same family).
>
> In the big Cats the switching processor isn't reported - CatOS doesn't
> have the same output and when you run Supervisor IOS you're talking to
> the MSFC anyway.

The switching processor in the big Cats is a custom chip - in the 5500s it
was EARL, dunno about the 6500s and later.

>
>> As a rule they've used 68xxx series at the lowend, PowerPC in the mid
>> range and MIPS at the high end. There are some exceptions such as the AGS
>> which is due to it being a very old box.
>
> Back in the days when an AGS[+] was high end the 68K series was pretty
> hot and RISC chips of any kind were rare. Moore's Law left the AGS+
> standing, of course...
>

I wonder what would have happeneed if Cisco had decided to implement an M88k
based router and try to max it out - lots of FDDI everywhere I suspect.

It's interesting to compare Cisco to Extreme and Juniper - the Extreme
Summit 48i has a Pentium 2 in it as does the Juniper M5 (at least the ones
I played with). The Pentium is used simply to run the configuration
interface - in the Juniper it's actually running FreeBSD. All the funky
stuff is done in ASICs.

Cisco are somewhat atypical in having a good few of their routers do pure
software forwarding dependant on the CPU. This has changed with the
later/bigger boxes starting with the SSP on the 7000s and working down the
range to the VIPs on 7200s. The 2600s appear not to have any custom
forwarding hardware. Nor do the 1700s. No idea about the 3700s as I've not
had a chance to play with them yet.


P.

[Sam Wilson]

| > As a rule they've used 68xxx series at the lowend, PowerPC in the mid range
| > and MIPS at the high end. There are some exceptions such as the AGS which
| > is due to it being a very old box.
|
| Back in the days when an AGS[+] was high end the 68K series was pretty
| hot and RISC chips of any kind were rare. Moore's Law left the AGS+
| standing, of course...

It's certainly the impression that more and more of the high-end boxes
are moving towards PowerPC based processors, e.g. PRP-2 for the 12000
series:

http://www.cisco.com/en/US/products/hw/routers/ps167/products_data_sheet0900aecd800f414a.html

and the new CRS-1:

http://www.cisco.com/en/US/products/ps5763/products_data_sheet09186a008022d5f1.html

Steinar Haug, Nethelp consulting, sthaug@nethelp.no

Paul S. Brown wrote:
> There was also an earlier RSM blade for the 5500s
> which was a 4500M on a stick - don't know what CPU it used.

The 4500M used a 100MHz IDT Orion (= MIPS R4400).

> later/bigger boxes starting with the SSP on the 7000s and working down the
> range to the VIPs on 7200s.

7500s, of course. 7200s don't have VIPs.

Regards,

Marco.

["Paul S. Brown"]

| >> 7200 MIPS of some description
| >
| > R7000s in our VXRs but might be different in the earlier or later
| > models.
| >
| That was actually pulled from the spec sheets for the NPE200/300/400 - dunno
| about the NPE-G, but I can't imagine they'll have changed that as that
| would bugger up IOS compatibiltiy unneccesarily.

NPE-G1 is still MIPS based.

| > In the big Cats the switching processor isn't reported - CatOS doesn't
| > have the same output and when you run Supervisor IOS you're talking to
| > the MSFC anyway.
|
| The switching processor in the big Cats is a custom chip - in the 5500s it
| was EARL, dunno about the 6500s and later.

6500s and later supervisor has a MIPS based processor *and* several ASICs.

| It's interesting to compare Cisco to Extreme and Juniper - the Extreme
| Summit 48i has a Pentium 2 in it as does the Juniper M5 (at least the ones
| I played with). The Pentium is used simply to run the configuration
| interface - in the Juniper it's actually running FreeBSD. All the funky
| stuff is done in ASICs.

Note that the Junipers also have a PowerPC chip in there, the so called
"exception processor".

| Cisco are somewhat atypical in having a good few of their routers do pure
| software forwarding dependant on the CPU. This has changed with the
| later/bigger boxes starting with the SSP on the 7000s and working down the
| range to the VIPs on 7200s.

VIP is for the 7500, and it's still software based forwarding.

| The 2600s appear not to have any custom
| forwarding hardware. Nor do the 1700s. No idea about the 3700s as I've not
| had a chance to play with them yet.

3700 is software based.

Steinar Haug, Nethelp consulting, sthaug@nethelp.no

Steinar Haug wrote:

> ["Paul S. Brown"]
>

>
> | Cisco are somewhat atypical in having a good few of their routers do
> | pure software forwarding dependant on the CPU. This has changed with
> | the later/bigger boxes starting with the SSP on the 7000s and working
> | down the range to the VIPs on 7200s.
>
> VIP is for the 7500, and it's still software based forwarding.

Spot the person who can't tell the difference between a 7513 and a 7206. You
are of course quite correct and I am wrong. I need to go and do some
studying again as I stopped being handson on this size of kit just before
dCEF was mainstreamed and have never played with PXF and its ilk.

It's interesting to see that Cisco have started building SMP based systems
with IOS-XR. I look forward in anticipation to the future, as Cisco appear
to be innovating again, which they haven't really done in a while.

>
> | The 2600s appear not to have any custom
> | forwarding hardware. Nor do the 1700s. No idea about the 3700s as I've
> | not had a chance to play with them yet.
>
> 3700 is software based.
>

Makes sense given where they are in the range and what they are being asked
to do.

P.

In article <2m72v3Fjfuk1U1@uni-berlin.de>, Paul S. Brown
<usenet060103@geekstuff.co.uk> wrote:

> Cisco are somewhat atypical in having a good few of their routers do pure
> software forwarding dependant on the CPU. This has changed with the
> later/bigger boxes starting with the SSP on the 7000s and working down the
> range to the VIPs on [7500s]. ...

Point of order, m'lud. Cisco's dedicated switch processor architecture
started with the AGS+ and the cBus controller.

Sam

In article <2m74o3Fjnt7sU1@uni-berlin.de>, Paul S. Brown
<usenet060103@geekstuff.co.uk> wrote:

> Spot the person who can't tell the difference between a 7513 and a 7206. ...

You've never dropped one on your foot, have you?

(No, neither have I, but it seemed like a good differentiator.)

Sam

Sam Wilson wrote:

> In article <2m74o3Fjnt7sU1@uni-berlin.de>, Paul S. Brown
> <usenet060103@geekstuff.co.uk> wrote:
>
>> Spot the person who can't tell the difference between a 7513 and a 7206.
>> ...
>
> You've never dropped one on your foot, have you?
>
> (No, neither have I, but it seemed like a good differentiator.)
>
No, but I have dropped a 7507 on my foot. Thank goodness for toetectors.

P.

In article <2m72v3Fjfuk1U1@uni-berlin.de>, Paul S. Brown
<usenet060103@geekstuff.co.uk> wrote:

> Point - my AGS has a CSC/4 and thus the 68040 - the CSC/2 was a 68020.

We had CSC/3s for a while but I can't remember whether they were 020s
or 030s - there was something about the CSC/n numbering not reflecting
the 680n0 numbering.

Another data point: PIXen are PC based - don't know if the bottom end
ones are Pentiums but the bigger ones are. The 6500/7600 FWSM (PIX
blade) has a couple of PIIIs[1] and one, two or three[2] IBM NP4GS3
network processors.

[1] According to documentation I have seen (don't ask) but there's only
mention of one in "show ver".
[2] The same documention showed a single NP but "show xlate" for
instance shows two NPs whereas "show console", which shows a console
log during boot, seems to refer to three NPs.

Sam

Program ended abnormally on 21/07/2004 09:52, Due to a catastrophic Sam Wilson
error:
> In article <2m72v3Fjfuk1U1@uni-berlin.de>, Paul S. Brown
> <usenet060103@geekstuff.co.uk> wrote:
>
>
>>Point - my AGS has a CSC/4 and thus the 68040 - the CSC/2 was a 68020.
>
>
> We had CSC/3s for a while but I can't remember whether they were 020s
> or 030s - there was something about the CSC/n numbering not reflecting
> the 680n0 numbering.
>
> Another data point: PIXen are PC based - don't know if the bottom end
> ones are Pentiums but the bigger ones are.

Yep. Ye olde Pix classic was a PII 200MHz.

--
Francois Labreque | The surest sign of the existence of extra-
flabreque | terrestrial intelligence is that they never
@ | bothered to come down here and visit us!
videotron.ca | - Calvin