Athlon vs. PIII

[Ole W. Saastad]

I have run a few tests to check the performance of
Athlon vs. PIII. 
The Atlas (tuned BLAS libs) tests suggest very high 
performance for Athlon, however in real tests like
Gaussian 98 there is little difference. 


Summary of performance Athlon vs. PIII 

- with Gaussian98 runs.

pgf77 (cache size 512k)
Athlon:  22 minutes 26.8 seconds.
PIII:    17 minutes 57.3 seconds.

pgf77 (cache size 64k)
Athlon:  23 minutes  5.8 seconds.
PIII:    17 minutes 56.3 seconds.


-with Whetstone runs

g77 (Compiled for Athlon) (See below)
Athlon: Double Whetstone KIPS  602410
PIII:   Double Whetstone KIPS  574713

g77 (Compiled for Pentium)
Athlon: Double Whetstone KIPS  602410
PIII:   Double Whetstone KIPS  581395


Detailed results:

Compiled with pgf77.

pgf77  -O2 -tp p6 -Munroll -Mvect=cachesize:524288 -c redund.f
Cache size 512k.
Included Blas code compiled. 
 
: ole at pgs1-11:g98-work; g98/g98 form.com
model name      : Pentium III (Coppermine)
stepping        : 3
cpu MHz         : 799.665
cache size      : 256 KB


 SCF Done:  E(UB+HF-LYP) =  -190.288331671     A.U. after   10 cycles
 Job cpu time:  0 days  0 hours 17 minutes 58.6 seconds.
 Job cpu time:  0 days  0 hours 17 minutes 57.3 seconds.


: ole at pc-10:g98-work; g98/g98 form.com
model name      : AMD Athlon(tm) Processor
stepping        : 1
cpu MHz         : 807.203
cache size      : 64 KB


 SCF Done:  E(UB+HF-LYP) =  -190.288331671     A.U. after   10 cycles
 Job cpu time:  0 days  0 hours 22 minutes 29.0 seconds.
 Job cpu time:  0 days  0 hours 22 minutes 26.8 seconds.


Athlon run PIII code almost as good as PIII. However, the cache
size for Athlon is not 512k as PII but 64k.


Compiled with pgf77.

pgf77 -O2 -tp p6  -Munroll -Mvect=cachesize:65520 -c unfchk.f
Cache size 64k.
Included Blas code compiled.


: ole at pgs1-11: g98-work; g98/g98 form.com
model name      : Pentium III (Coppermine)
stepping        : 3
cpu MHz         : 799.665
cache size      : 256 KB

 SCF Done:  E(UB+HF-LYP) =  -190.288331671     A.U. after   10 cycles
 Job cpu time:  0 days  0 hours 17 minutes 56.3 seconds.


: ole at pc-10:g98-work; g98/g98 form.com
model name      : AMD Athlon(tm) Processor
stepping        : 1
cpu MHz         : 807.203
cache size      : 64 KB

 SCF Done:  E(UB+HF-LYP) =  -190.288331671     A.U. after   10 cycles
 Job cpu time:  0 days  0 hours 23 minutes  5.8 seconds.



Whetstone benchmark:

pc-10:
model name      : AMD Athlon(tm) Processor
stepping        : 1
cpu MHz         : 807.203
cache size      : 64 KB

pgs1-11:
model name      : Pentium III (Coppermine)
stepping        : 3
cpu MHz         : 799.665
cache size      : 256 KB



g77 -o  whetstoned.exe -s -O3 -fomit-frame-pointer -Wall -mathlon
-mcpu=athlon -march=athlon -malign-functions=4 -funroll-loops
-fexpensive-optimizations -malign-double -fschedule-insns2
-mwide-multiply  whetstoned.f


ole at pc-10:f77;whetstoned.exe (Athlon)

 Benchmark #2 -- Double Precision Whetstone (A001)
  Double Whetstone KIPS  602410
ole at pc-10:f77; 


ole at pgs1-11:f77;whetstoned.exe (PIII)
  Double Whetstone KIPS  574713
ole at pgs1-11:f77; 


g77 -o  whetstoned.exe -s -O3 -fomit-frame-pointer -Wall -mpentium
-mcpu=pentium -march=pentium -malign-functions=4 -funroll-loops
-fexpensive-optimizations -malign-double -fschedule-insns2
-mwide-multiply  whetstoned.f

: ole at pc-10:f77;whetstoned.exe (Athlon)

 Benchmark #2 -- Double Precision Whetstone (A001)
  Double Whetstone KIPS  602410
ole at pc-10:f77;

ole at pgs1-11:f77;whetstoned.exe (PIII)
  Double Whetstone KIPS  581395
ole at pgs1-11:f77; 


gcc patched with patches from the web site:

http://www.athlonlinux.org/




-- 
Ole W. Saastad, Dr.Scient.        mailto:ole at scali.no
Scali, Scalable Linux Systems     http://www.scali.com

Olaf Helsets Vei 6                tel:    +47 22 62 89 68 (direct)
P.O.Box 70 Bogerud                fax:    +47 22 62 89 51        
N-0621 Oslo                       tel:    +47 22 95 21 45 (home)
NORWAY