Understanding the results

[Avaakaford]

HI Team,

I'm doing vmmark performance testing on two different hardware (Intel and AMD ) underlying infrastructure for my test is same. I got following results and i will wramp up more tiles in upcoming days.I wanna understand the results here 

	Intel-Tile1	AMD-Tile1	Intel-Tile3	AMD-tile3
Unreviewed_VMmark3_Applications_Score	1.37	1.35	4.01	4.12
Unreviewed_VMmark3_Infrastructure_Score	0.15	0.15	0.14	0.14
Unreviewed_VMmark3_Score	1.12	1.11	3.23	3.32

* Tile1 : single tile 

* Tile 3 : three tiles

1) How much difference is the big difference? 

         Here I can see some differences between Tile 1 vs Tile 3. As per single tile, Intel looks better and it is the opposite for tile 3 . I'm curious, how can we calculate this kind of situation? How much difference is the big difference ?

Note : I'm wrapping up more tiles for better visibility. 

Thanks & Regards

Ajay

[fredab2]

For what I see below both systems are doing the Infrastructure workloads (VMotion, Storage vMotion - look in document for more details on Infrastructure workloads) exactly the same. 

To read more into the Applications scores (4.01 vs 4.12)  you could look at the individual workload (look at manual for details on Application workloads) numbers in output and see if one is doing more work than the others between the 2 systems. Let's say DVD store is better on one system then the other it could mean it has a better storage sub-system or utilizes storage better than the other.  While ramping up more tiles also keep a look at cpu utilization of each host, in the end you have a maximum number of tiles that pass and a score.  You can compare this to the total number of tiles and score maxed out on the other system.

How you interpret these numbers is up to you.  Could look at it by total cost of each system per tile or maybe total available CPU frequency of each system compared to number of tiles passing and score. Or maybe you are just looking for the system that can do the most tiles period. That is up to you to determine.

Fred

[Avaakaford]

Thanks @fredab2 , that helps

Here Tile5 run of both hardware .Can i get more input please .

Thanks & Regards

Ajay

[fredab2]

All I can say is these runs look pretty similar if you look at the average of all 3 phases of test (p0, p1, p2) for both systems. Also, these are short Turbo runs so you might want to do full runs to compare.  

[Avaakaford]

@fredab2 

Here non Turbo output

	Hardware1	hardware2
Unreviewed_VMmark3_Applications_Score	11.29	12.03
Unreviewed_VMmark3_Infrastructure_Score	1.24	1.26
Unreviewed_VMmark3_Score	9.28	9.88

Back to original question: is this difference a lot or how much difference we can consider a lot?

Thanks & Regards

Ajay

[Avaakaford]

more output

HW1 vs HW2	HW1-Tile1	HW2-Tile1	HW1-Tile3	HW2-tile3	HW1-Tile5	HW2-Tile5	HW1-Tile10	HW2-Tile10	HW2-Tile11(full run)	HW1-Tile11(full run)	HW1-Tile12	HW2-Tile12	HW2-Tile15	HW1-Tile15
Unreviewed_VMmark3_Applications_Score	1.38	1.35	4.05	4.12	6.91	6.82	11.73	11.24	11.29	12.03	11.82	11.13	10.55	11.2
Unreviewed_VMmark3_Infrastructure_Score	0.15	0.15	0.14	0.14	0.13	0.14	0.13	0.13	1.24	1.26	0.13	0.13	0.12	0.13
Unreviewed_VMmark3_Score	1.13	1.11	3.27	3.32	5.55	5.49	9.41	9.02	9.28	9.88	9.49	8.93	8.46	8.98

Thanks & Regards

Ajay