From liujun at ict.ac.cn  Fri Dec  5 07:35:22 2008
From: liujun at ict.ac.cn (Jun Liu)
Date: Fri, 5 Dec 2008 23:35:22 +0800
Subject: [Hotspot] about thermal resistance
Message-ID: <200812052335224680802@ict.ac.cn>


Dear all:
  I have a question.  There is an approximate formula between power consumtion and temperature:
T=Tambient+(P/A)*R    (1)   is the amibient temperature . A is the chip area and R is the thermal resistance to the heat sink. In my thermal simulation,it is the default value 45.

I need to obtain the R value for the hotspot tool default configuration. I know the formular is correct if the thermal is uniform distribution,so I use a core in the floorplan file. I do some thermal simulation to calculate the value The part of corresponding code to show the simulate type is show here:
 strcpy(config.model_type, GRID_MODEL_STR);
  strcpy(config.grid_layer_file, "example.lcf");   // "example.lcf " is my layer file, it is made up of the default configeruation: layer0 : sillicon   lyaer1: thermal interface material
               strcpy(config.grid_map_mode, GRID_AVG_STR);
    model = alloc_RC_model(&config, flp);
   temp = hotspot_vector(model);
      power = hotspot_vector(model);
 read_power(model, power, power_file);
  populate_R_model(model, flp);
       steady_state_temp(model, power, temp);
 dump_temp(model, temp, steady_temp_file)


In the first simulation, the floorplan file is: 
    Unit1 0.006000   0.006000    0.000000    0.000000  
The simulation power value is 200.  The final steady temp is :
layer_0_Unit1377.97
layer_1_Unit1369.64

I use formular (1) to obtain the value of R: 
  
   377.97=45+273.15 +(200/(0.6*0.6))*R   the final value is R=0.107676.
   Using the layer_1_unit1 temperature 369.64, the final value of R is: R=0.092682

In the second simulatin, the floorplan file is:
      Unit1 0.008000   0.008000    0.000000    0.000000  
  The power value is 200.  
   The simulation result is:
   layer_0_Unit1365.68
layer_1_Unit1360.99

  Using the formular , I get the value of R is 0.152096 and 0.137088  corresponding to the layer 0 and layer1 temperature. 

So the value of R in the two simulations is different.  I do not know how to get the R value of the hotspot default configure. Is the formula(1) too coarse to fit the condition in my simulation?
 Is there a value of R in hotspot default configuration that can satisfy the formular(1)  approximityly.? 

Looking forward to your reply. Thanks in advance.



Jun Liu 
2008-12-05 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20081205/b5d2c2a0/attachment-0001.html 

From wh6p at cms.mail.virginia.edu  Fri Dec  5 08:06:14 2008
From: wh6p at cms.mail.virginia.edu (Wei Huang)
Date: Fri, 05 Dec 2008 11:06:14 -0500
Subject: [Hotspot] about thermal resistance
In-Reply-To: <200812052335224680802@ict.ac.cn>
References: <200812052335224680802@ict.ac.cn>
Message-ID: <9AB5DD8EF7A656E672347273@mstu1>

Hi,

The formula is not correct. it should be
T=T_ambient+P*R, basically "area" is already included in 
R=(1/k)*thickness/area, where k is the thermal conductivity.

-Wei

--On Friday, December 05, 2008 11:35 PM +0800 Jun Liu 
<liujun at ict.ac.cn> wrote:

>
>
> Dear all:
>   I have a question.  There is an approximate formula
> between power consumtion and temperature:
>
> T=Tambient+(P/A)*R    (1)   is the amibient temperature .
> A is the chip area and R is the thermal resistance to the
> heat sink. In my thermal simulation,it is the default
> value 45.
>
>
>
> I need to obtain the R value for the hotspot tool default
> configuration. I know the formular is correct if the
> thermal is uniform distribution,so I use a core in the
> floorplan file. I do some thermal simulation to calculate
> the value The part of corresponding code to show the
> simulate type is show here:
>
>  strcpy(config.model_type, GRID_MODEL_STR);
>
>   strcpy(config.grid_layer_file, "example.lcf");   //
> "example.lcf " is my layer file, it is made up of the
> default configeruation: layer0 : sillicon   lyaer1:
> thermal interface material
>                strcpy(config.grid_map_mode, GRID_AVG_STR);
>     model = alloc_RC_model(&config, flp);
>    temp = hotspot_vector(model);
>       power = hotspot_vector(model);
>
>  read_power(model, power, power_file);
>   populate_R_model(model, flp);
>        steady_state_temp(model, power, temp);
>
>  dump_temp(model, temp, steady_temp_file)
>
>
>
>
>
> In the first simulation, the floorplan file is:
>
>     Unit1 0.006000   0.006000    0.000000    0.000000
>
> The simulation power value is 200.  The final steady temp
> is :
>
> layer_0_Unit1 377.97
>
> layer_1_Unit1 369.64
>
>
> I use formular (1) to obtain the value of R:
>
>
>
>    377.97=45+273.15 +(200/(0.6*0.6))*R   the final value
> is R=0.107676.
>
>    Using the layer_1_unit1 temperature 369.64, the final
> value of R is: R=0.092682
>
>
>
> In the second simulatin, the floorplan file is:
>
>       Unit1 0.008000   0.008000    0.000000    0.000000
>
>   The power value is 200.
>
>    The simulation result is:
>
>
> layer_0_Unit1 365.68
>
> layer_1_Unit1 360.99
>
>
>   Using the formular , I get the value of R is 0.152096
> and 0.137088  corresponding to the layer 0 and layer1
> temperature.
>
>
>
> So the value of R in the two simulations is different.  I
> do not know how to get the R value of the hotspot default
> configure. Is the formula(1) too coarse to fit the
> condition in my simulation?
>
>  Is there a value of R in hotspot default configuration
> that can satisfy the formular(1)  approximityly.?
>
>
>
> Looking forward to your reply. Thanks in advance.
> __________________________________________________
>
> Jun Liu
> 2008-12-05





From liujun at ict.ac.cn  Fri Dec  5 17:06:46 2008
From: liujun at ict.ac.cn (Jun Liu)
Date: Sat, 6 Dec 2008 09:06:46 +0800
Subject: [Hotspot] about thermal resistance
References: <200812052335224680802@ict.ac.cn>, <9AB5DD8EF7A656E672347273@mstu1>
Message-ID: <200812060906457811774@ict.ac.cn>

Hi, 
  In the formula T=Tambient+(P/A)*R    (1) . where R=(1/k)*thickness. So I think it should be a constant for the default configuration of hotspot. But in my simulation I can't the constant.  
   I have a question to the formula R=(1/k)*thickness/area.  Is the thickness the chip thickness ?  Is area the chip area? In my first simulation, Is it right area=06*0.6?
 Thanks in advance. Looking forward to your reply


2008-12-06 



Jun Liu 



???? Wei Huang 
????? 2008-12-06  00:07:00 
???? Jun Liu; hotspot at cs.virginia.edu 
??? 
??? Re: [Hotspot] about thermal resistance 
 
Hi,

The formula is not correct. it should be
T=T_ambient+P*R, basically "area" is already included in 
R=(1/k)*thickness/area, where k is the thermal conductivity.

-Wei

--On Friday, December 05, 2008 11:35 PM +0800 Jun Liu 
<liujun at ict.ac.cn > wrote:

>
>
> Dear all:
>   I have a question.  There is an approximate formula
> between power consumtion and temperature:
>
> T=Tambient+(P/A)*R    (1)   is the amibient temperature .
> A is the chip area and R is the thermal resistance to the
> heat sink. In my thermal simulation,it is the default
> value 45.
>
>
>
> I need to obtain the R value for the hotspot tool default
> configuration. I know the formular is correct if the
> thermal is uniform distribution,so I use a core in the
> floorplan file. I do some thermal simulation to calculate
> the value The part of corresponding code to show the
> simulate type is show here:
>
>  strcpy(config.model_type, GRID_MODEL_STR);
>
>   strcpy(config.grid_layer_file, "example.lcf");   //
> "example.lcf " is my layer file, it is made up of the
> default configeruation: layer0 : sillicon   lyaer1:
> thermal interface material
>                strcpy(config.grid_map_mode, GRID_AVG_STR);
>     model = alloc_RC_model(&config, flp);
>    temp = hotspot_vector(model);
>       power = hotspot_vector(model);
>
>  read_power(model, power, power_file);
>   populate_R_model(model, flp);
>        steady_state_temp(model, power, temp);
>
>  dump_temp(model, temp, steady_temp_file)
>
>
>
>
>
> In the first simulation, the floorplan file is:
>
>     Unit1 0.006000   0.006000    0.000000    0.000000
>
> The simulation power value is 200.  The final steady temp
> is :
>
> layer_0_Unit1 377.97
>
> layer_1_Unit1 369.64
>
>
> I use formular (1) to obtain the value of R:
>
>
>
>    377.97=45+273.15 +(200/(0.6*0.6))*R   the final value
> is R=0.107676.
>
>    Using the layer_1_unit1 temperature 369.64, the final
> value of R is: R=0.092682
>
>
>
> In the second simulatin, the floorplan file is:
>
>       Unit1 0.008000   0.008000    0.000000    0.000000
>
>   The power value is 200.
>
>    The simulation result is:
>
>
> layer_0_Unit1 365.68
>
> layer_1_Unit1 360.99
>
>
>   Using the formular , I get the value of R is 0.152096
> and 0.137088  corresponding to the layer 0 and layer1
> temperature.
>
>
>
> So the value of R in the two simulations is different.  I
> do not know how to get the R value of the hotspot default
> configure. Is the formula(1) too coarse to fit the
> condition in my simulation?
>
>  Is there a value of R in hotspot default configuration
> that can satisfy the formular(1)  approximityly.?
>
>
>
> Looking forward to your reply. Thanks in advance.
> __________________________________________________
>
> Jun Liu
> 2008-12-05
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20081206/4d0585fc/attachment.html 

From wh6p at cms.mail.virginia.edu  Sat Dec  6 08:51:27 2008
From: wh6p at cms.mail.virginia.edu (Wei Huang)
Date: Sat, 06 Dec 2008 11:51:27 -0500
Subject: [Hotspot] about thermal resistance
In-Reply-To: <200812060906457811774@ict.ac.cn>
References: <200812052335224680802@ict.ac.cn>, <9AB5DD8EF7A656E672347273@mstu1>
	<200812060906457811774@ict.ac.cn>
Message-ID: <383B6A331CA3682A33C73F9F@[192.168.1.101]>

Hi,

(I am trying to answer your questions in your recent two 
emails...)
I see what you want to find is an "equivalent" thermal 
resistance from one block to the ambient. Depending on the 
size and power of the block,  this value will not be 
constant, whether you include area in the formula of R or 
not. This is because to use R=(1/k)*thickness or 
R=(1/k)*thickness/area, the underlying assumption is that, 
at the "thickness", the temperature has to be isothermal 
(that is, equal temperatures everywhere at that depth). 
This makes the "thickness" hard to find analytically. For 
example, for a large hot block, this thickness is deeper 
because you need more depth to spread the heat and get 
isotherm. For cool block, it is the other way around. 
Because of this difficulty in accurate close-form 
analytical temperature calculations, we have to use 
simulators such as ANSYS or HotSpot.

-Wei


--On Saturday, December 06, 2008 9:06 AM +0800 Jun Liu 
<liujun at ict.ac.cn> wrote:

>
> Hi,
>   In the formula T=Tambient+(P/A)*R    (1) . where
> R=(1/k)*thickness. So I think it should be a constant for
> the default configuration of hotspot. But in my
> simulation I can't the constant.
>    I have a question to the formula
> R=(1/k)*thickness/area.  Is the thickness the chip
> thickness ?  Is area the chip area? In my first
> simulation, Is it right area=06*0.6?
>  Thanks in advance. Looking forward to your reply
>
>
> 2008-12-06
> __________________________________________________
>
> Jun Liu
> __________________________________________________
>
> ???????? Wei Huang
> ?????????? 2008-12-06  00:07:00
> ???????? Jun Liu; hotspot at cs.virginia.edu
> ??????
> ?????? Re: [Hotspot] about thermal resistance
>
>
> Hi,
>
> The formula is not correct. it should be
> T=T_ambient+P*R, basically "area" is already included in
> R=(1/k)*thickness/area, where k is the thermal
> conductivity.
>
> -Wei
>
> --On Friday, December 05, 2008 11:35 PM +0800 Jun Liu
> <liujun at ict.ac.cn > wrote:
>
>>
>>
>> Dear all:
>>   I have a question.  There is an approximate formula
>> between power consumtion and temperature:
>>
>> T=Tambient+(P/A)*R    (1)   is the amibient temperature
>> .  A is the chip area and R is the thermal resistance to
>> the  heat sink. In my thermal simulation,it is the
>> default  value 45.
>>
>>
>>
>> I need to obtain the R value for the hotspot tool
>> default  configuration. I know the formular is correct
>> if the  thermal is uniform distribution,so I use a core
>> in the  floorplan file. I do some thermal simulation to
>> calculate  the value The part of corresponding code to
>> show the  simulate type is show here:
>>
>>  strcpy(config.model_type, GRID_MODEL_STR);
>>
>>   strcpy(config.grid_layer_file, "example.lcf");   //
>> "example.lcf " is my layer file, it is made up of the
>> default configeruation: layer0 : sillicon   lyaer1:
>> thermal interface material
>>                strcpy(config.grid_map_mode,
>>                GRID_AVG_STR);  model =
>>     alloc_RC_model(&config, flp);
>>    temp = hotspot_vector(model);
>>       power = hotspot_vector(model);
>>
>>  read_power(model, power, power_file);
>>   populate_R_model(model, flp);
>>        steady_state_temp(model, power, temp);
>>
>>  dump_temp(model, temp, steady_temp_file)
>>
>>
>>
>>
>>
>> In the first simulation, the floorplan file is:
>>
>>     Unit1 0.006000   0.006000    0.000000    0.000000
>>
>> The simulation power value is 200.  The final steady
>> temp  is :
>>
>> layer_0_Unit1 377.97
>>
>> layer_1_Unit1 369.64
>>
>>
>> I use formular (1) to obtain the value of R:
>>
>>
>>
>>    377.97=45+273.15 +(200/(0.6*0.6))*R   the final value
>> is R=0.107676.
>>
>>    Using the layer_1_unit1 temperature 369.64, the final
>> value of R is: R=0.092682
>>
>>
>>
>> In the second simulatin, the floorplan file is:
>>
>>       Unit1 0.008000   0.008000    0.000000    0.000000
>>
>>   The power value is 200.
>>
>>    The simulation result is:
>>
>>
>> layer_0_Unit1 365.68
>>
>> layer_1_Unit1 360.99
>>
>>
>>   Using the formular , I get the value of R is 0.152096
>> and 0.137088  corresponding to the layer 0 and layer1
>> temperature.
>>
>>
>>
>> So the value of R in the two simulations is different.
>> I  do not know how to get the R value of the hotspot
>> default  configure. Is the formula(1) too coarse to fit
>> the  condition in my simulation?
>>
>>  Is there a value of R in hotspot default configuration
>> that can satisfy the formular(1)  approximityly.?
>>
>>
>>
>> Looking forward to your reply. Thanks in advance.
>> __________________________________________________
>>
>> Jun Liu
>> 2008-12-05
>
>
>
>
>
>






From soundara at cse.psu.edu  Mon Dec  8 08:30:37 2008
From: soundara at cse.psu.edu (Niran Soundararajan)
Date: Mon, 8 Dec 2008 11:30:37 -0500 (EST)
Subject: [Hotspot] Measuring temperature for spec2k benchmarks
Message-ID: <Pine.SOC.4.64.0812081129070.16518@invid.cse.psu.edu>


Hello,

I am looking at ways to integrate Simplescalar with Wattch and Hotspot 
(measure temperature with variation in power). In 
case someone has done it before, it would be helpful if you could provide 
me. Otherwise tips on the integration would be very helpful.

Thank you,
Niranjan


From wh6p at cms.mail.virginia.edu  Mon Dec  8 09:56:08 2008
From: wh6p at cms.mail.virginia.edu (Wei Huang)
Date: Mon, 08 Dec 2008 12:56:08 -0500
Subject: [Hotspot] Measuring temperature for spec2k benchmarks
In-Reply-To: <Pine.SOC.4.64.0812081129070.16518@invid.cse.psu.edu>
References: <Pine.SOC.4.64.0812081129070.16518@invid.cse.psu
 .edu>
Message-ID: <5F12C1F62F19AC60A04DABF3@mstu1>

Hi,

You may want to refer to "sim-template" in the HotSpot 
download for starters. It contains a template to integrate 
HotSpot with SS and Wattch. Detailed integration is up to 
you...

Hope this helps,

-Wei

--On Monday, December 08, 2008 11:30 AM -0500 Niran 
Soundararajan <soundara at cse.psu.edu> wrote:

>
> Hello,
>
> I am looking at ways to integrate Simplescalar with
> Wattch and Hotspot  (measure temperature with variation
> in power). In  case someone has done it before, it would
> be helpful if you could provide  me. Otherwise tips on
> the integration would be very helpful.
>
> Thank you,
> Niranjan
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot





From ks4kk at cs.virginia.edu  Mon Dec  8 10:02:42 2008
From: ks4kk at cs.virginia.edu (Karthik Sankaranarayanan)
Date: Mon, 8 Dec 2008 13:02:42 -0500 (EST)
Subject: [Hotspot] Measuring temperature for spec2k benchmarks
In-Reply-To: <5F12C1F62F19AC60A04DABF3@mstu1>
References: <Pine.SOC.4.64.0812081129070.16518@invid.cse.psu .edu>
	<5F12C1F62F19AC60A04DABF3@mstu1>
Message-ID: <Pine.SOC.4.64.0812081300360.23082@cobra.cs.Virginia.EDU>


Hello Niranjan,

To add to Wei's inputs, have you looked at the "Source-level 
integration with a power/performance simulator" subsection of the HotSpot 
How-To? (http://lava.cs.virginia.edu/HotSpot/HotSpot-HOWTO.htm). Also, 
here is the header file (hotspot-iface.h) referred to in the text:
http://lava.cs.virginia.edu/HotSpot/hotspot-iface.h

Hope this helps.
-karthik


On Mon, 8 Dec 2008, Wei Huang wrote:

> Hi,
>
> You may want to refer to "sim-template" in the HotSpot
> download for starters. It contains a template to integrate
> HotSpot with SS and Wattch. Detailed integration is up to
> you...
>
> Hope this helps,
>
> -Wei
>
> --On Monday, December 08, 2008 11:30 AM -0500 Niran
> Soundararajan <soundara at cse.psu.edu> wrote:
>
>>
>> Hello,
>>
>> I am looking at ways to integrate Simplescalar with
>> Wattch and Hotspot  (measure temperature with variation
>> in power). In  case someone has done it before, it would
>> be helpful if you could provide  me. Otherwise tips on
>> the integration would be very helpful.
>>
>> Thank you,
>> Niranjan
>>
>> _______________________________________________
>> HotSpot mailing list
>> HotSpot at mail.cs.virginia.edu
>> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>