Prime 95

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

VirtualLarry

No Lifer
Aug 25, 2001
56,587
10,225
126
Originally posted by: myocardia
Originally posted by: JAG87
Oh wise myocardia, I took courses in university on operating systems and cpu scheduling, you really think I need to learn this from your or any links you give me? And don't think I am joking because I have transcripts to back me up.

You obviously do, since you mistakenly think that Windows is able to thread schedule itself.;)
Uhm, it is, and it does. Kernel threads are not special, they are treated identically by the scheduler. With an SMP OS like XP, things like kernel interrupts can happen on any available CPU, from what I've read.

 

bsobel

Moderator Emeritus<br>Elite Member
Dec 9, 2001
13,346
0
0
Originally posted by: myocardia
Originally posted by: JAG87
Oh wise myocardia, I took courses in university on operating systems and cpu scheduling, you really think I need to learn this from your or any links you give me? And don't think I am joking because I have transcripts to back me up.

You obviously do, since you mistakenly think that Windows is able to thread schedule itself.;)

The other posters are 100% correct and you are mistaken. There is no such thing as 'windows dlls run on core 0'. Threads are threads and are scheduled as such. There is no scheduling mechanism at the dll level, and nothing special about windows threads and processes that require them to have their affinity set for any core. I'm not sure where you heard this originally, but your completely off base.

 

jzodda

Senior member
Apr 12, 2000
824
0
0
Didn't mean to start an argument :) Good information to read though in any event.

Anyway it just finished a 24 hour Prime run without me interfering and they were on the same test when I stopped it. This is off topic to this thread but the new motherboard Gigabyte GA-EP45-UD3P is something special. On auto settings it booted into windows at 605 FSB, (I know crazy volts-I only did it one time)

467x9 (4.2Ghz) is not too shabby for 24/7 though on my E8400 @ 1.33v in bios, 1.29v in windows LLC enabled.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Originally posted by: DanDaMan315
Is Prime95 loaded up with spyware? I have a feeling it is.

Yes, yes it is. That's why everyone uses it, habitually, because we luvs us some spyware.

Intuition and gut feelings of unjustified paranoia FTW.
 

ch33zw1z

Lifer
Nov 4, 2004
39,831
20,428
146
Originally posted by: JAG87
Originally posted by: myocardia
Originally posted by: JAG87
Thread scheduling or no thread scheduling the point is that threads in p95 are not 1:1 with the cores unless you set affinity. If you are worried that other processes are using up cpu cycles you can even set p95 processes to the highest priority. If you still have discrepancy between threads, there is something wrong, and in the OP's case its all the shit he is doing in the background. Period, thats it, thats all there is.

I never said thread scheduling didn't take place with Prime95. As a matter of fact, if you had learned reading comprehension in college, you would have seen where I said that you are correct about Prime95, it does not run one thread per core, unless forced to do so:

Originally posted by: myocardia
And yes, you are right about Prime95. Prime95 is thread scheduled, like all applications within Windows.

Reading comprehension, FTW. What I said in my very first post is that one core is running Windows processes + Prime95, while the other core is only running Prime95, since Windows DLL's only run on core #0, which you then argued with. So, where was this big misunderstanding on my part?:confused:



No, I never once argued with that. Find me where I contradicted you and said windows processes are thread scheduled. Talk about reading comprehension...

At this point what bugs me is that you are still clinging on to these so called Windows DLLs or System processes as I like to call them, which you damn very well know do not use enough resources to cause a major discrepancy in the threads. A major discrepancy would only be cause by an owner process that is consuming a lot of cpu cycles (Firefox comes to mind from what the OP said), OR by throttling.


Edit.
Ok nevermind, I see where I contradicted you. Sorry, Im lacking sleep, couldn't remember what I wrote yesterday.

So basically the only point of my argument remaining is that p95 threads don't run 1:1 with the cores, and thats it. You still need to set affinity to find which core is throttling. I'm sure you agree with me on this.

Prime95 screenshot, it shows the affinity of worker threads assigned to separate logical cpus. Is Prime95 doing what you are saying needs to be done?
 

VirtualLarry

No Lifer
Aug 25, 2001
56,587
10,225
126
Originally posted by: ch33zw1z
Prime95 screenshot, it shows the affinity of worker threads assigned to separate logical cpus. Is Prime95 doing what you are saying needs to be done?
Thanks for that picture. It appears that Prime95 25.7 is explicit in how it sets the affinity for the worker threads. As far as I've experienced, it has done that all along with 25.x versions (the ones that work with multi-core automatically).

In fact I'm curious how Jag was able to select the number of worker threads, since my experience is that it automatically configures them.

 

JAG87

Diamond Member
Jan 3, 2006
3,921
3
76
ch33zw1z, thanks for pointing that out. it appears that what we were arguing about was finally implemented in 25.7

I have 25.6 right here in front of me and it does not set affinity to the worker threads.

VirtualLarry, its very simple, when you launch P95, on the "run a torture test" window (which is the very first window that pops up) you just change the number of threads to whatever you want.