Serverworks chip sets
Steffen Persvold
sp at scali.no
Mon Dec 11 09:14:20 PST 2000
Ray Muno wrote:
>
> We are currently looking at different ServerWorks based boards for dual
> CPU cluster nodes.
>
> Is the STREAM info available for different boards?
>
> We have access to 3 different ServerWorks machines right now, a 370DER HE-SL,
> an Intel STL2 (LE) and a Dell Poweredge 1400 (SE).
I don't think there is anything called a Serverworks SE chipset, DELL
must have misstyped it. The types I know of is HE, HE-SL, LE and WS
(which I actually think is the HE-SL),
>
> They are dual CPU machines. At the moment, we do not find the the HE-SL
> buys us anything with the codes we have run (actual CFD codes we run locally).
>
> If I want to run the STREAM benchmark, does anyone have any pointers?
> We only have access to the 370DER machine for a few days so we would like
> to gather as much info as possible.
Stream result on a 370DER with dual PIII 800EB (133MHz) 512MB ECC RAM
One CPU:
# OMP_NUM_THREADS=1 ./stream_d.x86.omp
----------------------------------------------
Double precision appears to have 16 digits of accuracy
Assuming 8 bytes per DOUBLE PRECISION word
----------------------------------------------
Array size = 2000000
Offset = 0
The total memory requirement is 45 MB
You are running each test 20 times
--
The *best* time for each test is used
*EXCLUDING* the first and last iterations
----------------------------------------------------
Your clock granularity appears to be less than one microsecond
Your clock granularity/precision appears to be 1 microseconds
----------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 343.4030 0.0932 0.0932 0.0933
Scale: 373.7880 0.0857 0.0856 0.0857
Add: 438.1443 0.1096 0.1096 0.1096
Triad: 437.6808 0.1097 0.1097 0.1097
----------------------------------------------------
Solution Validates!
----------------------------------------------------
Two CPU's:
# OMP_NUM_THREADS=2 ./stream_d.x86.omp
----------------------------------------------
Double precision appears to have 16 digits of accuracy
Assuming 8 bytes per DOUBLE PRECISION word
----------------------------------------------
Array size = 2000000
Offset = 0
The total memory requirement is 45 MB
You are running each test 20 times
--
The *best* time for each test is used
*EXCLUDING* the first and last iterations
----------------------------------------------------
Your clock granularity appears to be less than one microsecond
Your clock granularity/precision appears to be 1 microseconds
----------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 448.5438 0.0720 0.0713 0.0735
Scale: 461.8001 0.0699 0.0693 0.0711
Add: 530.2986 0.0911 0.0905 0.0922
Triad: 533.1437 0.0907 0.0900 0.0919
----------------------------------------------------
Solution Validates!
----------------------------------------------------
>
> As an aside, does anyone have any comments on the Intel STL2. For all the
> ServerWorks traffic, I have not seen much mention of this board. It is
> interesting to us since it appears to have the ability to control it from
> the serial port (EMP port in Intels docs). I would like to have access
> to the BIOS over the serial port and this is not something I see mention
> of from other manufacturers.
The 370 DER board from Supermicro has the 'BIOS over serial port'
feature but it's not EMP (well atleast I don't think so).
There are also other Serverworks based mainboards which have this
feature (e.g the Tyan 2510). Actually I think it's up to the BIOS
manufacturer to implement this feature, and both Tyan and Serverworks
use AMI bios.
One feature I would like to have on the PC platform is the ability to
drop the whole menu driven BIOS thing and use a more command line like
interface (Like on Alphas and SPARC). If the motherboards could
autodetect (wich I already think it does) that there is no VGA adapter,
it could switch over to the serial port interface. Wouldn't that be a
cluster friendly BIOS ?
Best regards,
--
Steffen Persvold Systems Engineer
Email : mailto:sp at scali.no Scali AS (http://www.scali.com)
Tlf : (+47) 22 62 89 50 Olaf Helsets vei 6
Fax : (+47) 22 62 89 51 N-0621 Oslo, Norway
More information about the Beowulf
mailing list