Quantcast
Channel: Raspberry Pi Forums
Viewing all articles
Browse latest Browse all 8037

General discussion • Re: A Pi Pie Chart

$
0
0
Apparently eight-socket Lenovo X3950 X6 scale-up servers are plentiful on eBay and one can get a discount by purchasing more than one. As nice as they look, I prefer unsweetened hot chocolate. Unfortunately, kittens find chocolate toxic.
Some Pi chart runs for a Lenovo X3950 with 18-core E7-8880v3 processors appear in

viewtopic.php?p=2340370#p2340370

Although that scale-up server has 144 cores, I only added a 4-core run to the Pi rate chart. For comparison similar tests for the Xeon E5-1650v3 and E5-2620v3 processors yielded

Code:

 Sieve   Merge  Fourier  Lorenz Processor3356.56 1029.77 3211.73 36220.4 E7-8880v3 4C/8T3373.06 1014.66 2312.66 37366.4 E5-2620v3 4C/8T4381.47 1349.37 2679.85 46807.2 E5-1650v3 4C/8T
Given the high price of DDR5 RAM and the low price Xeon CPUs on eBay, it was tempting to swap out the 6-core E5-1650 for an 18-core E5-2699. The result was

Code:

$ ./pichart-openmp # Xeon E5-2699 v3 (18C/36T)pichart -- Raspberry Pi Performance OPENMP version 40Prime Sieve          P=14630843 Workers=72 Sec=0.0614703 Mops=15199.7Merge Sort           N=16777216 Workers=72 Sec=0.0942698 Mops=4271.29Fourier Transform    N=4194304 Workers=72 Sec=0.0436742 Mflops=10564Lorenz 96            N=32768 K=16384 Workers=72 Sec=0.0333084 Mflops=96709My Computer has Raspberry Pi ratio=450.624Making pie charts...done.
instead of

Code:

$ ./pichart-openmp # Xeon E5-1650 v3 (6C/12T)pichart -- Raspberry Pi Performance OPENMP version 40Prime Sieve          P=14630843 Workers=24 Sec=0.142508 Mops=6556.3Merge Sort           N=16777216 Workers=24 Sec=0.204932 Mops=1964.81Fourier Transform    N=4194304 Workers=24 Sec=0.133122 Mflops=3465.8Lorenz 96            N=32768 K=16384 Workers=12 Sec=0.0543474 Mflops=59271My Computer has Raspberry Pi ratio=201.395Making pie charts...done.
The corresponding 4C/8T results obtained using taskset are

Code:

$ taskset -c 0-3,18-21 ./pichart-openmp # Xeon E5-2699 v3 (4C/8T)pichart -- Raspberry Pi Performance OPENMP version 40Prime Sieve          P=14630843 Workers=16 Sec=0.235288 Mops=3971Merge Sort           N=16777216 Workers=16 Sec=0.334613 Mops=1203.34Fourier Transform    N=4194304 Workers=16 Sec=0.123439 Mflops=3737.66Lorenz 96            N=32768 K=16384 Workers=8 Sec=0.0840567 Mflops=38322My Computer has Raspberry Pi ratio=143.624Making pie charts...done.
This shows better performance with the E5-2699 on the Fourier transform than the E5-1650 but slower performance on Lorenz 96. As the CPU architectures are similar, the difference might result from a larger cache but slower clocks.

According to Scratchy, the 4-core results indicate a 10-year-old Intel processor purchased for $12 on eBay is twice faster on a core-for-core basis than a Pi 5. On the other hand, energy consumption is likely more.

Statistics: Posted by ejolson — Wed Feb 25, 2026 2:10 am



Viewing all articles
Browse latest Browse all 8037

Trending Articles