Prime95 beta version 28.3 - Page 3

LaurV · 2014-01-17, 02:37

If you are going to invest time into it, and if we are allowed to vote, we will also vote for "number of minutes". It makes better sense for us to say (in case of a crash/restart) that "we lost the last 30 minutes of work, a risk which we assumed when we set the respective parameter", and not "we lost the last 194753 iterations, which multiplied by 3.7692 milliseconds, meant we lost about... (make the calculus) time"...

Thanks a lot in advance!

axn · 2014-01-17, 03:46

Quote:

Originally Posted by LaurV

If you are going to invest time into it, and if we are allowed to vote, we will also vote for "number of minutes". It makes better sense for us to say (in case of a crash/restart) that "we lost the last 30 minutes of work, a risk which we assumed when we set the respective parameter", and not "we lost the last 194753 iterations, which multiplied by 3.7692 milliseconds, meant we lost about... (make the calculus) time"...

Thanks a lot in advance!

Isn't it already like that? Save files are written every <x> minutes. The previous request was for screen output. Or am I missing something in your description?

TheMawn · 2014-01-17, 04:29

Here's some benchmarks. I've added my current running version as well.

Prime95 27.7 and Prime95 28.3.

I'm going to run a 24-hour stress test and then I'll run a dozen DC's.

petrw1 · 2014-01-17, 15:04

Quote:

Originally Posted by axn

Isn't it already like that? Save files are written every <x> minutes. The previous request was for screen output. Or am I missing something in your description?

Correct. I am referring to screen output.

kracker · 2014-01-17, 15:45

OMG... this version is already showing up on other overclocking/hardware forums...

EDIT: Just google prime95 28.3... LOL

TheMawn · 2014-01-17, 19:44

The people on overclock.net are having kittens over the ~1-2% improvement in Ivy Bridge and Sandy Bridge.

I really don't understand people. Using a beta version of software to stress test. Around here we're still stress testing the stress test. .

TheMawn · 2014-01-18, 01:50

I've started on my DC's. My stress-test went smoothly. HWMonitor recorded max temps about 2C above where they usually are after a day of full load, so this is probably the smaller FFT's.

Is the memory bottleneck something that occurs at all FFT sizes (beyond the hyper small ones which fit in the cache?)

kracker · 2014-01-18, 02:06

Quote:

Originally Posted by TheMawn

I've started on my DC's. My stress-test went smoothly. HWMonitor recorded max temps about 2C above where they usually are after a day of full load, so this is probably the smaller FFT's.

Is the memory bottleneck something that occurs at all FFT sizes (beyond the hyper small ones which fit in the cache?)

Hmm, yes I believe so. You could test that out though

rainchill · 2014-01-21, 16:45

As I suggested a few years ago having a different setting for number of iterations in the screen output and iterations in the title bar of each worker window would be great. This way you could be kept more up to date where the workers are and the performance while keeping the worker windows relatively clear of clutter.

Also, allowing decimal iterations in the CPU window of how many hours per day the program will run would be great. I find that 24 isn't too accurate but neither is 23 on my new 4770k. Being able to do 23.8 hours might be nice.

TheMawn · 2014-01-22, 04:47

Iteration: 20361987/32479871, POSSIBLE ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Mon Jan 20 03:36:04 2014]
Disregard last error. Result is reproducible and thus not a hardware problem.
For added safety, redoing iteration using a slower, more reliable method.
Continuing from last save file.
[Tue Jan 21 12:30:54 2014]

petrw1 · 2014-01-22, 15:24

Quote:

Originally Posted by TheMawn

Iteration: 20361987/32479871, POSSIBLE ERROR: ROUND OFF (0.40625) > 0.40
Continuing from last save file.
[Mon Jan 20 03:36:04 2014]
Disregard last error. Result is reproducible and thus not a hardware problem.
For added safety, redoing iteration using a slower, more reliable method.
Continuing from last save file.
[Tue Jan 21 12:30:54 2014]

For the record I've seen this in previous versions.

2014-01-17, 15:45	#27
kracker "Mr. Meeseeks" Jan 2012 California, USA 2³·271 Posts	OMG... this version is already showing up on other overclocking/hardware forums... EDIT: Just google prime95 28.3... LOL Last fiddled with by kracker on 2014-01-17 at 15:56

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LLR beta Version 3.8.13 (deprecated)	Jean Penné	Software	111	2015-01-26 21:41
Prime95 beta version 28.4	Prime95	Software	20	2014-03-02 02:51
Prime95 version 27.1 early preview, not-even-close-to-beta release	Prime95	Software	126	2012-02-09 16:17
Beta version 24.12 available	Prime95	Software	33	2005-06-14 13:19
Beta version of PRP	Prime95	PSearch	15	2004-09-17 19:21

2014-01-17, 02:37	#23
LaurV Romulan Interpreter Jun 2011 Thailand 7·1,373 Posts	If you are going to invest time into it, and if we are allowed to vote, we will also vote for "number of minutes". It makes better sense for us to say (in case of a crash/restart) that "we lost the last 30 minutes of work, a risk which we assumed when we set the respective parameter", and not "we lost the last 194753 iterations, which multiplied by 3.7692 milliseconds, meant we lost about... (make the calculus) time"... Thanks a lot in advance!

2014-01-17, 19:44	#28
TheMawn May 2013 East. Always East. 3277₈ Posts	The people on overclock.net are having kittens over the ~1-2% improvement in Ivy Bridge and Sandy Bridge. I really don't understand people. Using a beta version of software to stress test. Around here we're still stress testing the stress test. .

2014-01-18, 01:50	#29
TheMawn May 2013 East. Always East. 11·157 Posts	I've started on my DC's. My stress-test went smoothly. HWMonitor recorded max temps about 2C above where they usually are after a day of full load, so this is probably the smaller FFT's. Is the memory bottleneck something that occurs at all FFT sizes (beyond the hyper small ones which fit in the cache?)

2014-01-21, 16:45	#31
rainchill Apr 2005 DFW, tx 33₁₀ Posts	As I suggested a few years ago having a different setting for number of iterations in the screen output and iterations in the title bar of each worker window would be great. This way you could be kept more up to date where the workers are and the performance while keeping the worker windows relatively clear of clutter. Also, allowing decimal iterations in the CPU window of how many hours per day the program will run would be great. I find that 24 isn't too accurate but neither is 23 on my new 4770k. Being able to do 23.8 hours might be nice.

2014-01-22, 04:47	#32
TheMawn May 2013 East. Always East. 11·157 Posts	Iteration: 20361987/32479871, POSSIBLE ERROR: ROUND OFF (0.40625) > 0.40 Continuing from last save file. [Mon Jan 20 03:36:04 2014] Disregard last error. Result is reproducible and thus not a hardware problem. For added safety, redoing iteration using a slower, more reliable method. Continuing from last save file. [Tue Jan 21 12:30:54 2014]