Re: [Ngspice-users] Performance problems with Athlon 64 X2?

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Not exactly on the subject, but for those buying AMD dual cores, they 
use two different cache sizes: 512kb and 1024kb. You get two of these 
caches per chip, i.e. each cpu has it's own cache. I opted for a 4400 
cpu to get the larger cache.

Back on the speed subject, some motherboard need to "map" the memory. 
Don't ask me what this means. It is a bios option. If you don't map the 
memory, the computer can't see all the RAM.

On the benchmark, it is really best to have one written by those who 
know the internals. Years ago, OK, decades ago, I was evaluating the 
first generation layout versus schematic software. We had just hired an 
engineer from Intel that was familiar with the problems with Intel's 
in-house version of the software. We laid out a test chip that tested 
all the known software problems, and it really flogged the vendors. Of 
the three companies tested, two were out of business in a few years, and 
we picked the one that dominated the market.

Two things come to mind in the benchmark. One would be a large circuit, 
i.e. more nodes to settle. The other would be two use a simpler circuit 
but one with high gain nodes. It may be that some architectures will do 
better with one test versus the other. It may be a ring oscillator will 
do for the high gain test circuit.

Bill Richards wrote:
> Hi, and thanks for the reply.  Could you tell me what compiler flags 
> were actually used for the x86_64 package? As I mentioned in my previous 
> post, I made a clean binary from source and found exactly the same 
> results.  I think the x86_64 package is OK-  I think my system 
> configuration is not right somehow... /proc/cpuinfo reports the following:
> 
>    processor       : 0
>    vendor_id       : AuthenticAMD
>    cpu family      : 15
>    model           : 43
>    model name      : AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
>    stepping        : 1
>    cpu MHz         : 2000.000
>    cache size      : 512 KB
>    physical id     : 0
>    siblings        : 2
>    core id         : 0
>    cpu cores       : 2
>    fpu             : yes
>    fpu_exception   : yes
>    cpuid level     : 1
>    wp              : yes
>    flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr 
> pge mca cmov pat pse36 clflush mmx fxsr sse s
> se2 ht syscall nx mmxext fxsr_opt lm 3dnowext 3dnow pni lahf_lm cmp_legacy
>    bogomips        : 4202.19
>    TLB size        : 1024 4K pages
>    clflush size    : 64
>    cache_alignment : 64
>    address sizes   : 40 bits physical, 48 bits virtual
>    power management: ts fid vid ttp
> 
>    processor       : 1
>    vendor_id       : AuthenticAMD
>    cpu family      : 15
>    model           : 43
>    model name      : AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
>    stepping        : 1
>    cpu MHz         : 2000.000
>    cache size      : 512 KB
>    physical id     : 0
>    siblings        : 2
>    core id         : 1
>    cpu cores       : 2
>    fpu             : yes
>    fpu_exception   : yes
>    cpuid level     : 1
>    wp              : yes
>    flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr 
> pge mca cmov pat pse36 clflush mmx fxsr sse s
>    se2 ht syscall nx mmxext fxsr_opt lm 3dnowext 3dnow pni lahf_lm 
> cmp_legacy
>    bogomips        : 4202.19
>    TLB size        : 1024 4K pages
>    clflush size    : 64
>    cache_alignment : 64
>    address sizes   : 40 bits physical, 48 bits virtual
>    power management: ts fid vid ttp
> 
> 
> This looks OK, and based on the bogomips values, I should have seen a 
> reasonable speedup from my old XP 1600+.  Instead I find a slowdown of a 
> factor of 2!  Very odd.  I've turned off the FC6 cpuspeed service... no 
> effect.
> 
> I'm open to any suggestions right now...
> Thanks again,
> -bill-
> 
> Chitlesh GOORAH wrote:
>> On Friday 11 May 2007 04:07:24 Bill Richards wrote:
>>   
>>> I recently upgraded my simulation system from an old XP 1600+ box to an
>>> Athlon 64 X2 3800+.  I grabbed the x86_64 ngspice package via yum, and
>>> ran some test cases. 
>>>     
>>
>>
>> Hello,
>> if you found a better solution for the x86_64 arch, please let me know to 
>> correct the ngspice package of fedora so that it won't be a problem in a 
>> future.
>>
>> However, ngspice at the Fedora Project is compiled with the compiler flags of 
>> each arch automatically.
>>
>> regards,
>> Chitlesh
>>   
>> ------------------------------------------------------------------------
>>
>> -------------------------------------------------------------------------
>> This SF.net email is sponsored by DB2 Express
>> Download DB2 Express C - the FREE version of DB2 express and take
>> control of your XML. No limits. Just data. Click to get it now.
>> http://sourceforge.net/powerbar/db2/
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Ngspice-users mailing list
>> Ngs...@li...
>> https://lists.sourceforge.net/lists/listinfo/ngspice-users
>>   
> 
> 
> ------------------------------------------------------------------------
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Ngspice-users mailing list
> Ngs...@li...
> https://lists.sourceforge.net/lists/listinfo/ngspice-users