Looking for Linux systems with more than 4 cpus
Hi group,
I am working on a real time performance program and I need someone with access to a machine with more tahn 4 cpus to execute a few command for me and post the output. The commands I need are:
Thanks for your time. |
Member Response
Hi,
This from Intel i7 Code:
~# cat /proc/stat | wc -c Have fun & enjoy! :hattip: |
Then you're looking for Intel i7-5xxx and AMD FX-6xxx and FX-8xxx CPU's. The >4 core Intels aren't cheap, but are massively more powerful and use less powerful than the AMD ones. However, the AMD's can be had cheaply (3.5 GHz octo-core 8320 can be had for $140) so that a heavily threaded machine can be built on a budget.
Unless you're ok with Hyperthreading (virtual cores). In which case, any Core-i7-4xxx or better will have 4 cores with 4 virtual cores. |
Quote:
Code:
[bbausch@athlon64 ~]$ cat /proc/stat | wc -c |
[Solved] Looking for Linux systems with more than 4 cpus
Thanks guys, that was the exact data I needed.:D
|
My system with 4 physical CPUs, 8-cores each, currently is running with hyperthreading off. If hyperthreading were on, there would be twice as many apparent processors:
Code:
$ cat /proc/stat | wc -c |
Quote:
|
Member Response
Hi,
Penguin count will be of no use to the OP's request for data output for the commands presented. |
Quote:
|
@johnsfine
Wow that's a hefty machine, also I beleive that was the exact answer to the OP's question. It's obviously a server, are hosting other websites? Also what would be the purpose of not running hyperthreading, is it just to reduce heat output? |
Quote:
|
Quote:
It was set up as a test system for very large simulations, to verify algorithms that depend on effective NUMA behavior in the underlying system. As such it was a dismal failure. Its performance on giant examples is terrible, while other NUMA systems act as expected. Its primary use now is running large numbers of small tests of the simulation software. (A developer makes an enhancement, and before checking it in verifies there are not any unintended consequences to the wide range of product features that were not intended to be affected by that enhancement). Our simulation software (and a wide range of other simulation software) runs much better with hyperthreading turned off. Running 16 single threaded tasks in parallel with hyperthreading off takes significantly less than half the total time of 32 single threaded tasks in parallel with hyperthreading on. The opposite is true of the build side. With hyperthreading on, running 32 single threaded compiles in parallel takes less than double the time of 16 with hyperthreading off. So turning on hyperthreading and doubling the number of compiles launched in parallel reduces the total project build time. At the moment, on this system, the performance loss in testing from hyperthreading would be greater than the performance gain in building with hyperthreading, so we have it off. |
AMD64 FX8450 8-core
Quote:
|
Wow, I guess being retired has really made me dumber. I need to rethink my program to accommodate the greater number of possible CPU counts. Right now I use a 1024 byte buffer, which should hold about 12 CPU record?
Thanks again fpr your time. |
Now a new question:Are this numbers
'unsigned ling int' or 'unsigned long long int'? And does that change with the machine ( 32 or 64 bit )? |
All times are GMT -5. The time now is 01:58 AM. |