Using Pi as a Benchmark for Computers
How fast can your computer compute Pi? That's the type of question that keeps me up at night because apperantly not very fast! So i'm using cruncher wich is a C++ program that uses the CPU to complete the calculations according to a set of instructions like an algorithms. I had a lot of issues with making the config file and only after multiple trails & errors it finally ran, this was my first ever run and it's very basic. I've configured it to come up with 50.000.000.000 digits and only use a maximum of 200GB of storage space on the second drive, this became an issue first and only when I moved the entrie program to the second drive it accepted it as the output location... Also this program cannot be configured to make use of the GPU to get a faster result, wich is very strange in my opinion. It does support multi-threading on the CPU wich bassically means that it can run multiple proccesses simultaniously, so that's cool I geuss... I was running this config on the fastest machine I currently own, wich has an Intel Core I9-10850K @ 3.60GHz inside, and it occasionally bumped all the cores up to 4.9GHz while expierincing heavy load. It also packs a total of 32GB and the SSD used in this test has a faster read/write time then all the other ones I did not buy at the time. Anyway, for the first 45 min. it went really smooth, it went from 0% to 10% to 12% > 24% in no time, but just after 52% it got stuck or something...? After a while I even needed to press the spacebar every few moments for it to continue the task or else it would just sit there in idle... The results are in! please remember, this was my very first test. Constant: Pi Algorithm: Chudnovsky (1988) Decimal Digits: 50,000,000,000 Hexadecimal Digits: 41,524,101,187 Computation Mode: Swap Mode Multi-Threading: Push Pool -> 20 / ? / 20 (all cores) Far Memory Config: Disk Raid 0: * Default Path * Far Memory Tuning: 1.00 MiB/seek, no parallelism Start Time: Tue Jan 14 19:13:34 2025 Working Memory... 24.4 GiB (locked, spread: 100%/1) Twiddle Tables... 13.2 MiB (locked, spread: 100%/1) I/O Buffers... 64.0 MiB (locked, spread: 100%/1) Begin Computation: Series CommonP2B3... 3,525,683,480 terms (Expansion Factor = 3.059) Time: 14561.547 seconds ( 4.045 hours ) Large Division... Time: 893.724 seconds ( 14.895 minutes ) InvSqrt(10005)... Iterating: 000: O1M ( 3,125,000,034 ) So 4+ hours to come up with 50.000.000.000 digits of Pi. first it's neccesarry to increase the ram from 32GB to at least 64GB or maybe even 128GB, and later on swapping the CPU and socket to a newer gen. In the mean time I'm also going to do some test on other devices like a server rack or Linux devices running in a cluster?
update: het programma valideert ook na afloop het resultaat of het wel klopt en of er niet geknoeid is met bestanden.
2025-01-14 23:21:55