From solidrepellent at gmail.com  Wed May  2 04:10:40 2012
From: solidrepellent at gmail.com (Solid Repel)
Date: Wed, 2 May 2012 14:10:40 +0300
Subject: [Hotspot] interface material thermal conductivity?
Message-ID: <CAAYqiTuV=DM0CmuygS3mLqnCzcV8U+C=7=Bfar1v9O-RBMT7Yg@mail.gmail.com>

Hi Hotspot,

What is the reason that the interface material thermal conductivity in the
default hotspot config file (also attached to this mail) is set to 4
W/(m.K)? I feel that it is too low. What is the material that has been used
in this case?

Best Regards,
Naveen.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120502/c6138ae4/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hotspot.config
Type: application/octet-stream
Size: 4506 bytes
Desc: not available
Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120502/c6138ae4/attachment.obj 

From wh6p at virginia.edu  Wed May  2 07:37:11 2012
From: wh6p at virginia.edu (Wei Huang)
Date: Wed, 2 May 2012 09:37:11 -0500
Subject: [Hotspot] interface material thermal conductivity?
In-Reply-To: <CAAYqiTuV=DM0CmuygS3mLqnCzcV8U+C=7=Bfar1v9O-RBMT7Yg@mail.gmail.com>
References: <CAAYqiTuV=DM0CmuygS3mLqnCzcV8U+C=7=Bfar1v9O-RBMT7Yg@mail.gmail.com>
Message-ID: <CANYeXBpCTsxSnK4dhkEx5_ahFxbdRPsoiQMrgJDen08h6dyLKA@mail.gmail.com>

The TIM has to be some type of gel-like material (basically thermal paste)
that can fill the tiny rough surface bumps between the silicon and the heat
spreader (and the spreader and heatsink, too). Without it, the thermal
resistance from silicon to spreader would be much worse. This type of
materials usually have relatively low thermal conductivity compared to
metals and silicon, that's why they have to be very thin. The default value
in HotSpot is based on our best knowledge. If you can find a better TIM in
practical use, please do let us know, we'd be happy to update the info.

-Wei

On Wed, May 2, 2012 at 6:10 AM, Solid Repel <solidrepellent at gmail.com>wrote:

> Hi Hotspot,
>
> What is the reason that the interface material thermal conductivity in the
> default hotspot config file (also attached to this mail) is set to 4
> W/(m.K)? I feel that it is too low. What is the material that has been used
> in this case?
>
> Best Regards,
> Naveen.
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120502/697ceea9/attachment.html 

From rz3vg at virginia.edu  Wed May  2 08:29:10 2012
From: rz3vg at virginia.edu (Runjie Zhang)
Date: Wed, 2 May 2012 11:29:10 -0400
Subject: [Hotspot] interface material thermal conductivity?
In-Reply-To: <CAAYqiTuV=DM0CmuygS3mLqnCzcV8U+C=7=Bfar1v9O-RBMT7Yg@mail.gmail.com>
References: <CAAYqiTuV=DM0CmuygS3mLqnCzcV8U+C=7=Bfar1v9O-RBMT7Yg@mail.gmail.com>
Message-ID: <CAHxZ6YC5+a8OTh5OkgJfDtk7AP9Bu0Gbq_tOtM+YaVr2O4fBpg@mail.gmail.com>

Hi, Naveen

  According to our literature study, the range of commercial TIM's thermal
conductivity is 2~12. So default hotspot config file selects a reasonable
yet not too expensive compound.

Runjie

On Wed, May 2, 2012 at 7:10 AM, Solid Repel <solidrepellent at gmail.com>wrote:

> Hi Hotspot,
>
> What is the reason that the interface material thermal conductivity in the
> default hotspot config file (also attached to this mail) is set to 4
> W/(m.K)? I feel that it is too low. What is the material that has been used
> in this case?
>
> Best Regards,
> Naveen.
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>


-- 
Runjie Zhang
Computer Engineering
University of Virginia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120502/3aa9ae6e/attachment.html 

From sthannirmalai at hotmail.com  Mon May  7 22:40:44 2012
From: sthannirmalai at hotmail.com (thannir malai)
Date: Tue, 8 May 2012 05:40:44 +0000
Subject: [Hotspot] multicore floorplan query
In-Reply-To: <BAY163-W12EF768AC600D1D3FE11DB2100@phx.gbl>
References: <BAY163-W12EF768AC600D1D3FE11DB2100@phx.gbl>
Message-ID: <BAY163-W605257CADB1A2D3BF3A8E1B2100@phx.gbl>


Hi,
I created a multicore floorplan (2-core) with ev6.flp provided in the hotspot.  The power trace for the multicore floorplan was obtained by duplicating the already existing gcc.ptrace.
When I run the single core ev6.flp with gcc.ptrace, the peak temperature observed was 372 K.When I run the multicore ev6.flp with the same power trace, the peak temperature observed was 407 K. That is an increase of 35 K. I am not sure whether this is the expected behavior.Please help me figure out what is happening.
I have attached the following files:
Single core:ev6.flp (default one provided)gcc.ptrace (default one provided)hotspot.config (default one provided)
Multi core:mcoreev6.flp (2 core ev6 floorplan)gcc-mcore.ptrace (duplicating the values for the additional core)hotspot-mcore.config (changing the side length of both heat sink and heat spreader)
Thanks,-Malai


 		 	   		   		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120508/cd6553a0/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: files.tar.gz
Type: application/x-gzip
Size: 9821 bytes
Desc: not available
Url : http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120508/cd6553a0/attachment-0001.bin 

From nastaran_motavali at yahoo.com  Wed May  9 04:21:58 2012
From: nastaran_motavali at yahoo.com (nastaran motavali)
Date: Wed, 9 May 2012 04:21:58 -0700 (PDT)
Subject: [Hotspot] Technology node in PTScalar
Message-ID: <1336562518.71089.YahooMailNeo@web120503.mail.ne1.yahoo.com>



?
?
?Hi,
?I need to use PTScalar and Hotspot? as my power and temperature simulator respectively
and for this reason, I have to make full consistency between these two
simulators such as the processor floorplan. Fortunately, both baseline
processors are Alpha-21264 but the problem is the technology node. Hotspot
scaled the area of the EV6 processor from 350nm down to 130nm. On the
other hand, PTScalar asserts that the technology node used in this
simulator(PTScalar) is 65nm; but the area of all functional blocks (and
therefore, the area of whole processor) in both simulators are the same.
Is there any problem?
I would be grateful if you help me in this problem.
Thanks and best regards.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120509/6e90460f/attachment.html 

From lgliyang at yeah.net  Thu May 10 23:49:20 2012
From: lgliyang at yeah.net (=?GBK?B?wO7R9A==?=)
Date: Fri, 11 May 2012 14:49:20 +0800 (CST)
Subject: [Hotspot] How does HotSpot calculate temperature
Message-ID: <78c056db.1dd1.1373aa8a84f.Coremail.lgliyang@yeah.net>

Hi?


I'm just wondering how HotSpot calculate the temperature.  The document says that it takes a  power trace file and a floorplan file as inputs and outputs the corresponding transient temperatures onto a temperature trace file. I think the floorplan is used to generate the RC circuit, then how does HotSpot calculate the temperature using the RC circuit and the power trace file? What is the relation between thermal resistance, thermal capacitance, power and temperature? In the paper "Temperature-aware microarchitecture" a differential equation is mentioned, so can you please show me what exact the differential equation is ?


Best regards,


Li Yang


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120511/49cc4cd0/attachment.html 

From wh6p at virginia.edu  Tue May 15 07:41:01 2012
From: wh6p at virginia.edu (Wei Huang)
Date: Tue, 15 May 2012 09:41:01 -0500
Subject: [Hotspot] How does HotSpot calculate temperature
In-Reply-To: <4FB25E5E.5040102@cs.virginia.edu>
References: <78c056db.1dd1.1373aa8a84f.Coremail.lgliyang@yeah.net>
	<4FB25E5E.5040102@cs.virginia.edu>
Message-ID: <CANYeXBqfFDhZa5f2OMw2UuaPeT39q-Pd0k5OjzAe=-R+pk1RKg@mail.gmail.com>

Hi,

Detailed explanation of heat conduction can be found at
http://en.wikipedia.org/wiki/**Heat_conduction<http://en.wikipedia.org/wiki/Heat_conduction>

In HotSpot, we use numerical methods to solve the heat conduction
equations. Specifically, for steady-state solutions, we just use matrix
inversion. For transient, we use Runge-Kutta and finite difference method.
There are also more efficient thermal solvers out there in the literature.
The source code itself is quite self explanatory.

Hope this helps.

-Wei

On Tue, May 15, 2012 at 8:47 AM, Kevin Skadron <skadron at cs.virginia.edu>wrote:

> Wei, you're probably a better judge of what level of detail to provide.
>  I'm inclined to simply point him to
> http://en.wikipedia.org/wiki/**Heat_conduction<http://en.wikipedia.org/wiki/Heat_conduction>
>
> /K
>
> -------- Original Message --------
> Subject:        [Hotspot] How does HotSpot calculate temperature
> Date:   Fri, 11 May 2012 14:49:20 +0800 (CST)
> From:   ?? <lgliyang at yeah.net>
> To:     hotspot at mail.cs.virginia.edu
>
>
>
> Hi??
>
> I'm just wondering how HotSpot calculate the temperature. The document
> says that it takes a power trace file and a floorplan file as inputs and
> outputs the corresponding transient temperatures onto a temperature
> trace file. I think the floorplan is used to generate the RC circuit,
> then how does HotSpot calculate the temperature using the RC circuit and
> the power trace file? What is the relation between thermal resistance,
> thermal capacitance, power and temperature? In the paper
> "Temperature-aware microarchitecture" a differential equation is
> mentioned, so can you please show me what exact the differential
> equation is ?
>
> Best regards,
>
> Li Yang
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120515/8ae1ad0f/attachment.html 

From skadron at cs.virginia.edu  Tue May 15 13:29:19 2012
From: skadron at cs.virginia.edu (Kevin Skadron)
Date: Tue, 15 May 2012 16:29:19 -0400
Subject: [Hotspot] multicore floorplan query
In-Reply-To: <BAY163-W605257CADB1A2D3BF3A8E1B2100@phx.gbl>
References: <BAY163-W12EF768AC600D1D3FE11DB2100@phx.gbl>
	<BAY163-W605257CADB1A2D3BF3A8E1B2100@phx.gbl>
Message-ID: <4FB2BC9F.4060907@cs.virginia.edu>

Temperature certainly should increase; more cores => more power => more 
heat. Specifically, with a convection resistance of 0.8K/W, as in your 
case, if the additional power due to duplicating all blocks is about 40W 
on average, an increase in temperature by 35K (~0.8*40) is reasonable.

/K

On 5/8/12 1:40 AM, thannir malai wrote:
> Hi,
>
> I created a multicore floorplan (2-core) with ev6.flp provided in the
> hotspot. The power trace for the multicore floorplan was obtained by
> duplicating the already existing gcc.ptrace.
>
> When I run the single core ev6.flp with gcc.ptrace, the peak temperature
> observed was 372 K.
> When I run the multicore ev6.flp with the same power trace, the peak
> temperature observed was 407 K. That is an increase of 35 K. I am not
> sure whether this is the expected behavior.
> Please help me figure out what is happening.
>
> I have attached the following files:
>
> Single core:
> ev6.flp (default one provided)
> gcc.ptrace (default one provided)
> hotspot.config (default one provided)
>
> Multi core:
> mcoreev6.flp (2 core ev6 floorplan)
> gcc-mcore.ptrace (duplicating the values for the additional core! )
> hotspot-mcore.config (changing the side length of both heat sink and
> heat spreader)
>
> Thanks,
> -Malai
>
>
>
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot

From sthannirmalai at hotmail.com  Sun May 27 22:53:09 2012
From: sthannirmalai at hotmail.com (thannir malai)
Date: Mon, 28 May 2012 05:53:09 +0000
Subject: [Hotspot] multicore floorplan query
In-Reply-To: <4FB2BC9F.4060907@cs.virginia.edu>
References: <BAY163-W12EF768AC600D1D3FE11DB2100@phx.gbl>
	<BAY163-W605257CADB1A2D3BF3A8E1B2100@phx.gbl>,
	<4FB2BC9F.4060907@cs.virginia.edu>
Message-ID: <BAY163-W54FBEA180FE706C2A5EFDCB2040@phx.gbl>


Thanks for your reply. But the average power density still remains constant. So, the temperature shouldnt increase that much right?

> Date: Tue, 15 May 2012 16:29:19 -0400
> From: skadron at cs.virginia.edu
> To: sthannirmalai at hotmail.com
> CC: hotspot at cs.virginia.edu
> Subject: Re: [Hotspot] multicore floorplan query
> 
> Temperature certainly should increase; more cores => more power => more 
> heat. Specifically, with a convection resistance of 0.8K/W, as in your 
> case, if the additional power due to duplicating all blocks is about 40W 
> on average, an increase in temperature by 35K (~0.8*40) is reasonable.
> 
> /K
> 
> On 5/8/12 1:40 AM, thannir malai wrote:
> > Hi,
> >
> > I created a multicore floorplan (2-core) with ev6.flp provided in the
> > hotspot. The power trace for the multicore floorplan was obtained by
> > duplicating the already existing gcc.ptrace.
> >
> > When I run the single core ev6.flp with gcc.ptrace, the peak temperature
> > observed was 372 K.
> > When I run the multicore ev6.flp with the same power trace, the peak
> > temperature observed was 407 K. That is an increase of 35 K. I am not
> > sure whether this is the expected behavior.
> > Please help me figure out what is happening.
> >
> > I have attached the following files:
> >
> > Single core:
> > ev6.flp (default one provided)
> > gcc.ptrace (default one provided)
> > hotspot.config (default one provided)
> >
> > Multi core:
> > mcoreev6.flp (2 core ev6 floorplan)
> > gcc-mcore.ptrace (duplicating the values for the additional core! )
> > hotspot-mcore.config (changing the side length of both heat sink and
> > heat spreader)
> >
> > Thanks,
> > -Malai
> >
> >
> >
> >
> >
> > _______________________________________________
> > HotSpot mailing list
> > HotSpot at mail.cs.virginia.edu
> > http://www.cs.virginia.edu/mailman/listinfo/hotspot
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120528/42ed4155/attachment.html 

From nedahpg at gmail.com  Mon May 28 02:37:15 2012
From: nedahpg at gmail.com (Neda Hassanpour)
Date: Mon, 28 May 2012 14:07:15 +0430
Subject: [Hotspot] Hotspot Configuration
Message-ID: <CAPa2X=S20-w88PqFqx7qU=F9BOAoOX+sek9PMC1y44TC--4eow@mail.gmail.com>

Hi,

I have a problem in adjusting some parameters in hotspot.config and .lcf
files.

Since I used Hotspot 3D stacking feature, I defined my own lcf file and
adjusted the parameters like resistivity and thickness for silicon and TIM
layers. I also set the value of silicon thermal conductivity in
hotspot.config but I am confused about the parameters relate to silicon
layer like silicon thermal conductivity, its resistivity and thickness.

Does what we name as ?silicon layer?in lcf and config files, contain ILD
and metal layers too? In different technologies the number of metal layers
in each layer of a 3D IC can be different. Regarding to the possible
effects of the thickness and thermal conductivity of these layers on
resulted temperatures, it seems they should be considered in hotspot
configuration. But I cannot find a place to set the parameters relate to
metal and ILD layers in hotspot.config file.

I would be thankful if you help to solve this problem.

Regards

Neda
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120528/1b8f4f95/attachment-0001.html 

From skadron at cs.virginia.edu  Mon May 28 08:01:57 2012
From: skadron at cs.virginia.edu (Kevin Skadron)
Date: Mon, 28 May 2012 11:01:57 -0400
Subject: [Hotspot] multicore floorplan query
In-Reply-To: <BAY163-W54FBEA180FE706C2A5EFDCB2040@phx.gbl>
References: <BAY163-W12EF768AC600D1D3FE11DB2100@phx.gbl>
	<BAY163-W605257CADB1A2D3BF3A8E1B2100@phx.gbl>,
	<4FB2BC9F.4060907@cs.virginia.edu>
	<BAY163-W54FBEA180FE706C2A5EFDCB2040@phx.gbl>
Message-ID: <4FC39365.2010805@cs.virginia.edu>

Even though power density is the same, total power (hence total heat) 
increases a lot.

To give an extreme example, imagine two chips, both with the same power 
density.  But one is small (or has only a small area active), so total 
power is 1 W.  The other is a large, high-performance multicore chip at 
100 W.  For the same cooling solution, the latter will be much hotter!

Another way to look at this is to use Ohm's Law for the equivalent 
steady-state circuit.  V = IR, or deltaT = P*theta, where P = total 
power and theta = thermal resistance.  Obviously this is an 
approximation when power densities are non-uniform and there are 
boundary conditions, but it shows how we did our reasoning below.

/K

On 5/28/2012 1:53 AM, thannir malai wrote:
> Thanks for your reply. But the average power density still remains
> constant. So, the temperature shouldnt increase that much right?
>
>  > Date: Tue, 15 May 2012 16:29:19 -0400
>  > From: skadron at cs.virginia.edu
>  > To: sthannirmalai at hotmail.com
>  > CC: hotspot at cs.virginia.edu
>  > Subject: Re: [Hotspot] multicore floorplan query
>  >
>  > Temperature certainly should increase; more cores => more power => more
>  > heat. Specifically, with a convection resistance of 0.8K/W, as in your
>  > case, if the additional power due to duplicating all blocks is about 40W
>  > on average, an increase in temperature by 35K (~0.8*40) is reasonable.
>  >
>  > /K
>  >
>  > On 5/8/12 1:40 AM, thannir malai wrote:
>  > > Hi,
>  > >
>  > > I created a multicore floorplan (2-core) with ev6.flp provided in the
>  > > hotspot. The power trace for the multicore floorplan w! as obtained by
>  > > duplicating the already existing gcc.ptrace.
>  > >
>  > > When I run the single core ev6.flp with gcc.ptrace, the peak
> temperature
>  > > observed was 372 K.
>  > > When I run the multicore ev6.flp with the same power trace, the peak
>  > > temperature observed was 407 K. That is an increase of 35 K. I am not
>  > > sure whether this is the expected behavior.
>  > > Please help me figure out what is happening.
>  > >
>  > > I have attached the following files:
>  > >
>  > > Single core:
>  > > ev6.flp (default one provided)
>  > > gcc.ptrace (default one provided)
>  > > hotspot.config (default one provided)
>  > >
>  > > Multi core:
>  > > mcoreev6.flp (2 core ev6 floorplan)
>  > > gcc-mcore.ptrace (duplicating the values for the additional core! )
>  > > hotspot-mcore.config (changing the side length of both heat sink and
>  > &! gt; heat spreader)
>  > >
>  > > Thanks,
>  > > -Mal ai
>  > >
>  > >
>  > >
>  > >
>  > >
>  > > _______________________________________________
>  > > 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

From wh6p at virginia.edu  Mon May 28 15:39:43 2012
From: wh6p at virginia.edu (Wei Huang)
Date: Mon, 28 May 2012 17:39:43 -0500
Subject: [Hotspot] Hotspot Configuration
In-Reply-To: <CAPa2X=S20-w88PqFqx7qU=F9BOAoOX+sek9PMC1y44TC--4eow@mail.gmail.com>
References: <CAPa2X=S20-w88PqFqx7qU=F9BOAoOX+sek9PMC1y44TC--4eow@mail.gmail.com>
Message-ID: <CANYeXBoz5+TyKHbfa6OECpfRXgtWbyD_rMr46_2wdv_KeR11AQ@mail.gmail.com>

Hi,

on-chip metal layers are usually much thinner than silicon, therefore no
affect thermal conductivity significantly. For 3D chips which has very thin
silicon layers, this might not be entirely true. However, you can always
treat metal layers as individual layers in the layer config file. Keep in
mind this will increase the solving time as more layers are added.

-Wei

On Mon, May 28, 2012 at 4:37 AM, Neda Hassanpour <nedahpg at gmail.com> wrote:

> Hi,
>
> I have a problem in adjusting some parameters in hotspot.config and .lcf
> files.
>
> Since I used Hotspot 3D stacking feature, I defined my own lcf file and
> adjusted the parameters like resistivity and thickness for silicon and TIM
> layers. I also set the value of silicon thermal conductivity in
> hotspot.config but I am confused about the parameters relate to silicon
> layer like silicon thermal conductivity, its resistivity and thickness.
>
> Does what we name as ?silicon layer?in lcf and config files, contain ILD
> and metal layers too? In different technologies the number of metal layers
> in each layer of a 3D IC can be different. Regarding to the possible
> effects of the thickness and thermal conductivity of these layers on
> resulted temperatures, it seems they should be considered in hotspot
> configuration. But I cannot find a place to set the parameters relate to
> metal and ILD layers in hotspot.config file.
>
> I would be thankful if you help to solve this problem.
>
> Regards
>
> Neda
>
>
> _______________________________________________
> HotSpot mailing list
> HotSpot at mail.cs.virginia.edu
> http://www.cs.virginia.edu/mailman/listinfo/hotspot
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/hotspot/attachments/20120528/d6568f36/attachment.html