Greyguy1948
Member
Have you tested some of the programs here:
http://www.netlib.org/performance/html/PDSreports.html
I have tested some on my Ryzen 1700 (3.7 GHz).
Nsieve (max=6942, min 6001)
Heapsort (max 4537, min 3349)
Sim 6.1 sec
MM (matrix multiply 500x500 with alternatives -n,-i,-t,-u,-m,-w,-r)
This is much better on 64 bit than 32 bit
-n 0.06 sec
-i 0.05 sec
-m 0.05 sec
Some c-code can be found here:
http://home.vianetworks.nl/users/mhx/
FLOPS benchmark on 64 bit Linux on Ryzen:
MFLOPS(1)=6645
MFLOPS(2)=4320
MFLOPS(3)=4789
MFLOPS(4)=4869
Floating point is really high if you compare with a low budget ARM like ASUS Tinkerboard (1800 MHz):
MFLOPS(1)=544
MFLOPS(2)=422
MFLOPS(3)=571
MFLOPS(4)=634
Integer has not the same ratio. Compare queens -c 14:
Ryzen 1,59 sec
Tinker 3.33 sec
If we compare Giga clocks/ run
Ryzen 5.88
Tinker 5.99
AMD K6 4.11
AMD K5 3.67
Pent Pro 5.43
Pentium 4.16
PPC 601 4.16
So the pipelines are longer today but not like Pentium 4.
http://www.netlib.org/performance/html/PDSreports.html
I have tested some on my Ryzen 1700 (3.7 GHz).
Nsieve (max=6942, min 6001)
Heapsort (max 4537, min 3349)
Sim 6.1 sec
MM (matrix multiply 500x500 with alternatives -n,-i,-t,-u,-m,-w,-r)
This is much better on 64 bit than 32 bit
-n 0.06 sec
-i 0.05 sec
-m 0.05 sec
Some c-code can be found here:
http://home.vianetworks.nl/users/mhx/
FLOPS benchmark on 64 bit Linux on Ryzen:
MFLOPS(1)=6645
MFLOPS(2)=4320
MFLOPS(3)=4789
MFLOPS(4)=4869
Floating point is really high if you compare with a low budget ARM like ASUS Tinkerboard (1800 MHz):
MFLOPS(1)=544
MFLOPS(2)=422
MFLOPS(3)=571
MFLOPS(4)=634
Integer has not the same ratio. Compare queens -c 14:
Ryzen 1,59 sec
Tinker 3.33 sec
If we compare Giga clocks/ run
Ryzen 5.88
Tinker 5.99
AMD K6 4.11
AMD K5 3.67
Pent Pro 5.43
Pentium 4.16
PPC 601 4.16
So the pipelines are longer today but not like Pentium 4.
