McPat output question

[Digant Desai]

Hi all,

I am using default (gainstown) configuration and dumping stats periodically. 

I am using mcpat.py file to generate power output but I am getting unusually 

large leakage power values.

Processor: 

  Area = 242.105 mm^2

  Peak Power = 54.2357 W

  Total Leakage = 19.0729 W

  Peak Dynamic = 35.1628 W

  Subthreshold Leakage = 17.1829 W

  Gate Leakage = 1.88992 W

  Runtime Dynamic = 3.9931 W

  Total Cores: 4 cores 

  Device Type= ITRS high performance device type

    Area = 145.089 mm^2

    Peak Dynamic = 33.7228 W

    Subthreshold Leakage = 12.9466 W

    Gate Leakage = 1.80216 W

    Runtime Dynamic = 3.98253 W

It this normal for 45 nm technology node ? 

Thanks

Digant

[Trevor E. Carlson]

Digant,

Although we did not develop the McPAT models, we did find a good correlation between the peak power and the real hardware (See our PACT paper [1]). Additionally, these numbers look to be in line with the original results from the McPAT paper from 2009. In that paper, their Xeon Tulsa validation (and published data comparisons) show the processor to have almost 30% leakage power (33.93 W out of 116.09 W total for the McPAT simulations), which is similar to what you are seeing here.

-Trevor

[1] http://snipersim.org/w/Paper:Pact2012Heirman

--
--
--
You received this message because you are subscribed to the Google
Groups "Sniper simulator" group.
To post to this group, send email to snip...@googlegroups.com
To unsubscribe from this group, send email to
snipersim+...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/snipersim?hl=en
 
---
You received this message because you are subscribed to the Google Groups "Sniper simulator" group.
To unsubscribe from this group and stop receiving emails from it, send an email to snipersim+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Guangshuo Liu]

Hi Trevor, 

I am also have some question understanding the McPAT output. Which are exactly Peak Dynamic and Runtime Dynamic? If I am only interested in the average dynamic power, which should I use?

Thanks,

Guangshuo

[Wim Heirman]

Hi Guangshuo,

You should read up on the McPAT paper [1]. If I remember correctly,
they had better results (compared to real hardware) with Peak Dynamic
so that would be considered more accurate; this is also the value we
used for the comparisons we made in our PACT paper [2].

Regards,
Wim

[1] Li, Sheng, et al. "McPAT: an integrated power, area, and timing
modeling framework for multicore and manycore architectures."
Microarchitecture, 2009. MICRO-42. 42nd Annual IEEE/ACM International
Symposium on. IEEE, 2009.

[2] Heirman, Wim, et al. "Power-Aware Multi-Core Simulation for Early
Design Stage Hardware/Software Co-Optimization." International
Conference on Parallel Architectures and Compilation Techniques
(PACT), 2012.

[Shamik Saha]

Hi,

I have been running Asim benchmarks and have got the same values of peak dynamic power for all benchmarks. However, there was variation in runtime dynamic power. Is this expected?

[Wim Heirman]

I thought I've seen variations in peak power as well, I'm not sure how McPAT is modeling it exactly and how it depends on the workload -- but in theory peak power is just a property of the machine (the maximum power it can attain for the worst-case workload, such as a power virus). Runtime dynamic on the other hand is the dynamic power consumed by *this* workload.

-Wim

For more options, visit https://groups.google.com/d/optout.

[Dan Petrisko]

I too am seeing variations in peak power on the order of several watts.  Does anyone know why this would be the case in Sniper?  Colleagues who have used McPAT with other simulators have not experienced this issue.

[Wim Heirman]

Hi Dan,

Could you compare the .xml files that generate the different peak powers? Normally the parameters that determine peak power should only be dependent on the architecture, not the workload. If those do change that could indicate a bug in Sniper's mcpat.py script.

Regards,

Wim

snipersim+unsubscribe@googlegroups.com

For more options, visit this group at
http://groups.google.com/group/snipersim?hl=en

---
You received this message because you are subscribed to the Google Groups "Sniper simulator" group.

To unsubscribe from this group and stop receiving emails from it, send an email to snipersim+unsubscribe@googlegroups.com.