Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/the-backside-is-becoming-a-battleground-between-power-and-heat-dr-moh-kolbehdari.25687/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/EmailDomainReplace] => 1000010
            [SemiWiki/HtmlMailer] => 1000200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2031270
            [XFI] => 1060170
        )

    [wordpress] => /var/www/html
)

The Backside Is Becoming a Battleground Between Power and Heat Dr. Moh Kolbehdari

moh.kolb

Member

The Backside Is Becoming a Battleground Between Power and Heat


Backside power delivery is becoming one of the most important changes in advanced semiconductor technology.


Electrically, the attraction is clear.


Move power away from the frontside signal-routing stack.


Shorten the path to the transistor.


Use larger backside conductors.


Reduce IR drop and voltage droop.


Free valuable frontside routing resources.


Intel’s PowerVia and TSMC’s backside power rail are now taking this concept into leading-edge manufacturing.


But there is another side to the backside:


Heat also wants to use it.


And that creates an interesting engineering conflict.


Electrical design would like:


**short vertical power paths


  • dense power connections
  • low resistance
  • substantial backside metal**

Thermal design would like:


**low thermal resistance


  • effective lateral spreading
  • direct access to the cooling structure
  • minimum thermal barriers between the hotspot and coolant**

These requirements do not always point toward the same physical structure.


For example, backside power architectures require substantial silicon thinning. That helps create the backside electrical path, but reduced silicon thickness can also reduce lateral heat spreading. Imec has reported simulations showing a potential temperature penalty of roughly 14°C in a studied backside-power configuration because of this effect.


At the same time, backside metal itself can potentially become part of the thermal solution.


That is where things become much more interesting.


Instead of asking:


Where should the power network go?


perhaps the real question is:


How should the backside be co-designed as both an electrical delivery structure and a thermal transport structure?

Consider what becomes possible:


Power metal can potentially contribute to heat spreading.


Vertical structures can be optimized considering both electrical resistance and thermal conductance.


Hotspot locations can influence power-via density as well as cooling architecture.


Backside metallization, dielectric layers, silicon thickness, TIM, cold plate, and even embedded cooling can be evaluated as one coupled system.


Researchers at IBM have already explored this direction by modeling backside power delivery together with manifold microfluidic cooling, placing electrical power structures and coolant channels within the same backside architecture. Their work explicitly treats it as a balance between electrical and thermal performance rather than two independent problems.


And this changes the design question.


The backside is no longer simply:


the place where power comes in.


Nor is it simply:


the surface where heat comes out.


It is becoming a shared physical resource for both.


The optimum architecture may therefore not be the one with the absolute lowest electrical resistance.


And it may not be the one with the absolute lowest thermal resistance.


It may be the geometry that gives the best combined electrical + thermal + mechanical operating point.


As power density continues increasing, I think this co-optimization will become increasingly important:


Power map → backside PDN → heat-flow map → cooling structure → temperature → resistance → power integrity


The loop closes.


Backside power delivery is not just a PDN innovation.
It is becoming an electro-thermal architecture problem.



#Semiconductors #BSPDN #PowerIntegrity #ThermalManagement #AdvancedPackaging #AIInfrastructure #HPC #ChipDesign
 

Attachments

  • Backside_PDN_Aug11_Semi.jpg
    Backside_PDN_Aug11_Semi.jpg
    230.4 KB · Views: 158
Backside Power vs. Backside Cooling: Finding the Optimum Electro-Thermal Architecture


It should cover the electrical benefit, thermal penalty/opportunity, shared backside real estate, and the need to optimize power delivery + heat removal + mechanical integrity together rather than independently.
 

The Backside Is Becoming a Battleground Between Power and Heat


Backside power delivery is becoming one of the most important changes in advanced semiconductor technology.


Electrically, the attraction is clear.


Move power away from the frontside signal-routing stack.


Shorten the path to the transistor.


Use larger backside conductors.


Reduce IR drop and voltage droop.


Free valuable frontside routing resources.


Intel’s PowerVia and TSMC’s backside power rail are now taking this concept into leading-edge manufacturing.


But there is another side to the backside:


Heat also wants to use it.


And that creates an interesting engineering conflict.


Electrical design would like:


**short vertical power paths


  • dense power connections
  • low resistance
  • substantial backside metal**

Thermal design would like:


**low thermal resistance


  • effective lateral spreading
  • direct access to the cooling structure
  • minimum thermal barriers between the hotspot and coolant**

These requirements do not always point toward the same physical structure.


For example, backside power architectures require substantial silicon thinning. That helps create the backside electrical path, but reduced silicon thickness can also reduce lateral heat spreading. Imec has reported simulations showing a potential temperature penalty of roughly 14°C in a studied backside-power configuration because of this effect.


At the same time, backside metal itself can potentially become part of the thermal solution.


That is where things become much more interesting.


Instead of asking:


Where should the power network go?


perhaps the real question is:




Consider what becomes possible:


Power metal can potentially contribute to heat spreading.


Vertical structures can be optimized considering both electrical resistance and thermal conductance.


Hotspot locations can influence power-via density as well as cooling architecture.


Backside metallization, dielectric layers, silicon thickness, TIM, cold plate, and even embedded cooling can be evaluated as one coupled system.


Researchers at IBM have already explored this direction by modeling backside power delivery together with manifold microfluidic cooling, placing electrical power structures and coolant channels within the same backside architecture. Their work explicitly treats it as a balance between electrical and thermal performance rather than two independent problems.


And this changes the design question.


The backside is no longer simply:


the place where power comes in.


Nor is it simply:


the surface where heat comes out.


It is becoming a shared physical resource for both.


The optimum architecture may therefore not be the one with the absolute lowest electrical resistance.


And it may not be the one with the absolute lowest thermal resistance.


It may be the geometry that gives the best combined electrical + thermal + mechanical operating point.


As power density continues increasing, I think this co-optimization will become increasingly important:


Power map → backside PDN → heat-flow map → cooling structure → temperature → resistance → power integrity


The loop closes.


Backside power delivery is not just a PDN innovation.
It is becoming an electro-thermal architecture problem.
sports games online


#Semiconductors #BSPDN #PowerIntegrity #ThermalManagement #AdvancedPackaging #AIInfrastructure #HPC #ChipDesign
Interesting perspective. I agree that backside power delivery shouldn’t really be treated as a purely electrical problem anymore. Once silicon thinning, backside metal, TIM, and cooling structures are all interacting, optimizing one parameter in isolation could easily create a problem somewhere else.

I especially like the idea of using the power/thermal maps together rather than designing the PDN first and dealing with heat afterward. As power densities keep rising, that kind of electro-thermal co-design seems like it will become increasingly important.
 
Back
Top