BurnIn Test data Repeatability

[Farhan Khan]

Hi All,
I've performed CPU, RAM, DISKs, Network, USB and Serial Port Test on a computer with multiple runs. The standard deviation of RAM test is higher than all other test results.

Operations-->	1st Run	2nd Run	3rd Run	4th Run	Avg	Std Dev
CPU (Trillion)	10.283	10.401	10.343	11.16	10.54675	0.41166198
Memory (RAM) (Billion)	113	110	108	28.486	89.8715	40.9752208
2D Graphics	11366	10264	9254	9757	10160.25	903.418831
3D Graphics	10536	10753	11159	11343	10947.75	368.923998
USB3 (Billion)	93.818	95.558	97.304	99.377	96.51425	2.38070163
USB2 (Million)	495	504	502	507	502	5.09901951
Sound (Million)	7.199	6.549	7.464	6.67	6.9705	0.43348087
Serial Port 1 (Million)	17.231	17.076	16.972	16.979	17.0645	0.12072144
Serial Port 2 (Million)	17.251	17.103	17.006	16.989	17.08725	0.12016204
Disk (C (Billion)	81	86.168	80.866	90.194	84.557	4.49617689
Disk (F (Billion)	110	107	109	101	106.75	4.03112887
Network1	655440	653840	659760	646400	653860	5566.51896

The standard deviation in RAM, 2D graphics and 3D graphics test is more. What could be the possible reason for large variation in data. The test duration is 15 minutes in all test runs; and all settings are same between each Run. The machine was not connected to network and only OS is installed; with no other software installed in machine.

Regards

[Tim (PassMark)]

Were these test run consecutively or was there a reboot between them?
Does the RAM test continue to get slower on the system each time?

Do you have the log files from these runs?
here may be some information in them to indicate what might be causing the behaviour.

[Farhan Khan]

It is consecutive run with no reboot between them.
The behavior of RAM Test is random and it gets slower randomly between each runs.
I ran "Multi Process Torture Test" with test settings of 4 multi process and 22% of RAM each. Test pattern is cyclic. Log files are attached.

[Tim (PassMark)]

We haven't been able to reproduce any similar behaviour here and the results are very consistent over multiple runs.

Do you see the same behaviour if you are just running the RAM test by itself?

[David (PassMark)]

Maybe on the 4th run the system was slightly low of RAM and there was a bit of use of the paging file. Paging file is typically around 100x slower than normal ram. So you don't need much to dramatically lower the result.

If you want to benchmark the hardware, consider using our PerformanceTest software.

[Farhan Khan]

Hi,
If I run RAM test by itself then the variation is within confidence interval to prove equivalence.
When the system is low on RAM, it uses paging and might result in increased CPU usage with decreased RAM usage. The C disk operations are increased with Paging operation in process. Could it be due to RAM has been over allocated?
I'll try with performance test software too. Thanks for helping.

Regards.