MPC8313 performance evaluation

MPC8313 performance evaluation

[Lutz Jaenicke]

Hi!

I am currently evaluating the performance of the MPC8313E for firewall
(and VPN) purposes.
The results obtained do not match my expectations so I would be interested
in some discussion.

Test setup:
-----------
A linux 2.6.27 based firewall software is used. The software is the
same among
To be replaced: IXP42x custom made board (266MHz and 533MHz)
* MPC8343 custom made board (400MHz)
* MPC8313E-RDB Rev. C (333/333)
As a software firewall the throughput is determined by looking into
the routing information and IP headers, hence the performance is best
characterized by "frames/s" which is constant over the frame size
(as long as no limit is hit like the 100Mbit/s limit on the IXP42x).

With the MPC8343@400MHz I get a throughput of approx. 24500 frames/s
using the predefined firewall rules.
With the MPC8313 I get a significantly lower value:
MPC8313@250MHz          12500fps
MPC8313@333MHz          14500fps
MPC8313@416MHz          15500fps      (333MHz type, overclocked)
Kernel 2.6.27 with SKB recycling from later versions and the
e300c2/c3/c4 TLB errata workaround. Some pure computational benchmark
with small footprint (namely "openssl speed") which should fit into
the cache size indicates the perfect 3/4/5 ratio used.

Some discussion with the the freescale rep. lead to the CSB frequency
of the 8313 (166MHz) being significantly lower than that of the 8343.
Is the CSB the critical point here?
Note: the IXP42x uses an internal bus speed of 133MHz and operates
at frame rates similar to the 8343...

Does anybody else have numbers for a firewall scenario using 8313
and/or other processors of the family?

Best regards,
	Lutz
PS: Yes, I know that absolute numbers will be hard to compare with
other scenarios but relative values would be quite interesting as
well.
-- 
Dr.-Ing. Lutz Jänicke
CTO
Innominate Security Technologies AG  /protecting industrial networks/
tel: +49.30.921028-200
fax: +49.30.921028-020
Rudower Chaussee 13
D-12489 Berlin, Germany
www.innominate.com

Register Court: AG Charlottenburg, HR B 81603
Management Board: Dirk Seewald
Chairman of the Supervisory Board: Volker Bibelhausen

Re: MPC8313 performance evaluation

[Kumar Gala]

On Aug 6, 2009, at 1:00 PM, Lutz Jaenicke wrote:

> Hi!
>
> I am currently evaluating the performance of the MPC8313E for firewall
> (and VPN) purposes.
> The results obtained do not match my expectations so I would be  
> interested
> in some discussion.
>
> Test setup:
> -----------
> A linux 2.6.27 based firewall software is used. The software is the
> same among
> To be replaced: IXP42x custom made board (266MHz and 533MHz)
> * MPC8343 custom made board (400MHz)
> * MPC8313E-RDB Rev. C (333/333)
> As a software firewall the throughput is determined by looking into
> the routing information and IP headers, hence the performance is best
> characterized by "frames/s" which is constant over the frame size
> (as long as no limit is hit like the 100Mbit/s limit on the IXP42x).
>
> With the MPC8343@400MHz I get a throughput of approx. 24500 frames/s
> using the predefined firewall rules.
> With the MPC8313 I get a significantly lower value:
> MPC8313@250MHz          12500fps
> MPC8313@333MHz          14500fps
> MPC8313@416MHz          15500fps      (333MHz type, overclocked)
> Kernel 2.6.27 with SKB recycling from later versions and the
> e300c2/c3/c4 TLB errata workaround. Some pure computational benchmark
> with small footprint (namely "openssl speed") which should fit into
> the cache size indicates the perfect 3/4/5 ratio used.
>
> Some discussion with the the freescale rep. lead to the CSB frequency
> of the 8313 (166MHz) being significantly lower than that of the 8343.
> Is the CSB the critical point here?
> Note: the IXP42x uses an internal bus speed of 133MHz and operates
> at frame rates similar to the 8343...
>
> Does anybody else have numbers for a firewall scenario using 8313
> and/or other processors of the family?
>
> Best regards,
> 	Lutz
> PS: Yes, I know that absolute numbers will be hard to compare with
> other scenarios but relative values would be quite interesting as
> well.

What DDR frequencies (and width) are you running the 8343 vs 8313 at.   
This can have a significant impact on performance.

- k

Re: MPC8313 performance evaluation

[Lutz Jaenicke]

On Thu, Aug 06, 2009 at 02:16:55PM -0500, Kumar Gala wrote:
>
> On Aug 6, 2009, at 1:00 PM, Lutz Jaenicke wrote:
>> With the MPC8343@400MHz I get a throughput of approx. 24500 frames/s
>> using the predefined firewall rules.

>> With the MPC8313 I get a significantly lower value:
>> MPC8313@250MHz          12500fps
>> MPC8313@333MHz          14500fps
>> MPC8313@416MHz          15500fps      (333MHz type, overclocked)

> What DDR frequencies (and width) are you running the 8343 vs 8313 at.   
> This can have a significant impact on performance.

The 8343 is running DDR2 32bit at 266MHz (CSB 266MHz)
The 8313 is running DDR2 32bit at 333MHz (CSB 166MHz)

The test were performed with 128byte frames so that the overall
bandwidth needed is far below even 100Mbit/s.

Best regards,
	Lutz
-- 
Dr.-Ing. Lutz Jänicke
CTO
Innominate Security Technologies AG  /protecting industrial networks/
tel: +49.30.921028-200
fax: +49.30.921028-020
Rudower Chaussee 13
D-12489 Berlin, Germany
www.innominate.com

Register Court: AG Charlottenburg, HR B 81603
Management Board: Dirk Seewald
Chairman of the Supervisory Board: Volker Bibelhausen

RE: MPC8313 performance evaluation

[Liu Dave-R63238]

> > On Aug 6, 2009, at 1:00 PM, Lutz Jaenicke wrote:
> >> With the MPC8343@400MHz I get a throughput of approx.=20
> 24500 frames/s
> >> using the predefined firewall rules.
>=20
> >> With the MPC8313 I get a significantly lower value:
> >> MPC8313@250MHz          12500fps
> >> MPC8313@333MHz          14500fps
> >> MPC8313@416MHz          15500fps      (333MHz type, overclocked)
>=20
> > What DDR frequencies (and width) are you running the 8343=20
> vs 8313 at.  =20
> > This can have a significant impact on performance.
>=20
> The 8343 is running DDR2 32bit at 266MHz (CSB 266MHz)
> The 8313 is running DDR2 32bit at 333MHz (CSB 166MHz)
>=20
> The test were performed with 128byte frames so that the overall
> bandwidth needed is far below even 100Mbit/s.

The CSB bus freq is key for the throught.
CSB freq of 8313 is lower than 8343, it will cause the performance
degrade.

RE: MPC8313 performance evaluation

[Liu Dave-R63238]

> Some discussion with the the freescale rep. lead to the CSB frequency
> of the 8313 (166MHz) being significantly lower than that of the 8343.
> Is the CSB the critical point here?

I believe the CSB is critical point here. They are right.

> Note: the IXP42x uses an internal bus speed of 133MHz and operates
> at frame rates similar to the 8343...

It is possible, IXP42x has the differenet SoC architecture with 83xx.

Re: MPC8313 performance evaluation

[Lutz Jaenicke]

On Fri, Aug 07, 2009 at 04:08:50PM +0800, Liu Dave-R63238 wrote:
> 
> > Some discussion with the the freescale rep. lead to the CSB frequency
> > of the 8313 (166MHz) being significantly lower than that of the 8343.
> > Is the CSB the critical point here?
> 
> I believe the CSB is critical point here. They are right.

I have performed some additional measurements with other multiplier/divider
settings 

Previous values with CSB=166MHz
> >> With the MPC8313 I get a significantly lower value:
> >> MPC8313@250MHz          12500fps
> >> MPC8313@333MHz          14500fps
> >> MPC8313@416MHz          15500fps      (333MHz type, overclocked)

New value with CSB=200MHz (overclocked)
     MPC8313@400MHz          17500fps

This indeed indicates that the CSB is the limiting factor.
Until a few days ago I have not even been aware of the CSB being a
performance critical component. All of the nice powerpoints explaining
the processors and used for comparing different families shown by the
Freescale Rep include the core frequencies and the DRAM interface
and frequency but do not even mention the CSB...

> > Note: the IXP42x uses an internal bus speed of 133MHz and operates
> > at frame rates similar to the 8343...
> 
> It is possible, IXP42x has the differenet SoC architecture with 83xx.

That is very true indeed, then XScale (ARM) based IXP42x does have
a completely different implementation.

Best regards,
	Lutz
-- 
Dr.-Ing. Lutz Jänicke
CTO
Innominate Security Technologies AG  /protecting industrial networks/
tel: +49.30.921028-200
fax: +49.30.921028-020
Rudower Chaussee 13
D-12489 Berlin, Germany
www.innominate.com

Register Court: AG Charlottenburg, HR B 81603
Management Board: Dirk Seewald
Chairman of the Supervisory Board: Volker Bibelhausen

Re: MPC8313 performance evaluation

[Lutz Jaenicke]

On Fri, Aug 07, 2009 at 12:56:54PM +0200, Lutz Jaenicke wrote:
> On Fri, Aug 07, 2009 at 04:08:50PM +0800, Liu Dave-R63238 wrote:
> > 
> > > Some discussion with the the freescale rep. lead to the CSB frequency
> > > of the 8313 (166MHz) being significantly lower than that of the 8343.
> > > Is the CSB the critical point here?
> > 
> > I believe the CSB is critical point here. They are right.
> 
> This indeed indicates that the CSB is the limiting factor.
> Until a few days ago I have not even been aware of the CSB being a
> performance critical component. All of the nice powerpoints explaining
> the processors and used for comparing different families shown by the
> Freescale Rep include the core frequencies and the DRAM interface
> and frequency but do not even mention the CSB...

Having this said, is there any good white paper to be read about it?
For firewall usage there are different influence factors:
* Ethernet interfaces (DMA to/from DRAM via CSB!?)
* CPU processing for the firewall rules (code/data to/from DRAM closely
  related to cache size or misses)
Hence I would like to understand better the impact of the different
components.
(If only available under NDA I can also contact my Freescale Rep but
having public source always makes things easier.)

Best regards,
	Lutz
-- 
Dr.-Ing. Lutz Jänicke
CTO
Innominate Security Technologies AG  /protecting industrial networks/
tel: +49.30.921028-200
fax: +49.30.921028-020
Rudower Chaussee 13
D-12489 Berlin, Germany
www.innominate.com

Register Court: AG Charlottenburg, HR B 81603
Management Board: Dirk Seewald
Chairman of the Supervisory Board: Volker Bibelhausen