Computing Server for 1 week

thistle

@IonSwitch_Stan said:

This command python -m timeit '"-".join(str(n) for n in range(100))' is single threaded, while most modern processor are multiple core. Hopefully your workload is parallelized and can use multiple threads. A better test might be to use sysbench (sysbench --test=cpu --cpu-max-prime=20000 run) for a single threaded run, and sysbench (--test=cpu --cpu-max-prime=100000 --num-threads=<cpu cores>) for multiple cpu machines.

I can let you use 2 x E5-2670's for 10c / hour / each. (E5-2670, 128GB ram).

Yes but I just time the result by 4 The commands I need to do are single threaded that I am multithreading. (if that makes sense)

I would love for that offer any chance I could get more cores?

willie

IonSwitch_Stan said: Only as batch compute nodes

That works for me. Do they have some local storage and network to move data in and out? Thanks

thistle

Please let me know ASAP @IonSwitch_Stan and @GenjiSwitchPls my uni deadline is Thursday!

thistle

I forgot to say. I setup a Vultr $80 a month and it ran the python script in 30 seconds :O Absolutely awful.

willie

Thistle, something is awful there: how many loops did it run? The i7-3770 did 100k in around 6 seconds, but a $5/mo DO droplet does 10k in about 1.2 sec. I.e. the script seems to adjust the # of loops depending on the cpu speed.

If I were you I'd just spin up an i7-3770 at Hetzner for a month, and maybe even keep it around after that. Then you can use it any time you want.

IonSwitch_Stan

[root@ionswitch-vps1 ~]# cat /etc/redhat-release

CentOS Linux release 7.3.1611 (Core)

[root@ionswitch-vps1 ~]# python --version

Python 2.7.5

[root@ionswitch-vps1 ~]# python -m timeit '"-".join(str(n) for n in range(100))'

10000 loops, best of 3: 28.7 usec per loop

On an E5-2660v1. Seems like this benchmark's behavior changes quite a bit.

thistle

@willie said:
Thistle, something is awful there: how many loops did it run? The i7-3770 did 100k in around 6 seconds, but a $5/mo DO droplet does 10k in about 1.2 sec. I.e. the script seems to adjust the # of loops depending on the cpu speed.

If I were you I'd just spin up an i7-3770 at Hetzner for a month, and maybe even keep it around after that. Then you can use it any time you want.

Ahh jesus. Sorry about that. I would love to if I had any money unfortunately a student and just need power ASAP

seanho

From the timeit man page:

If -n is not given, a suitable number of loops is calculated by trying successive powers of 10 until the total time is at least 0.2 seconds.

https://docs.python.org/2/library/timeit.html

Are you sure this particular benchmark (which is just a simple example taken from the timeit documentation) is representative of your actual workload?

willie

seanho said: representative of your actual workload?

It's a proxy for overall cpu speed. Geekbench and passmark are the same. The interesting question is whether sklearn knows how to use SSE instructions, HD graphic gpu, Nvidia boards etc. At that point the Python benchmark means nothing.

wwabbit

How to cheat on this benchmark....

$ python -m timeit '"-".join(str(n) for n in range(100))'
10000 loops, best of 3: 17.7 usec per loop
$ alias python=pypy
$ python -m timeit '"-".join(str(n) for n in range(100))'
100000 loops, best of 3: 4.1 usec per loop