CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > FloEFD, FloWorks & FloTHERM

Flotherm vs Flotherm XT vs FloEFD

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree3Likes

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   January 10, 2019, 14:56
Smile Smart Part, electronics
  #21
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
Boris,
You are correct about electronic component smart parts. I was referring to the PCB smart part. FloEFD 17.3.1 does not create this automatically as FloTHERM XT 3.2, 3.1 does. Instead EFD creates individual materials for each layer and puts them in the project. This is extraneous bookkeeping for the analyst. A single layer material is unique and not useful for any other pwb so it isn't necessary to have the part in the engineering library or project. Just creates extra work when switching the PCB type from compact, detailed or explicit. Additionally if you start with a detailed pwb (layers each with lumped bulk properties) and switch to compact pwb (single 1 layer lumped parameter body) you have no pwb component. Have to make one. Then with two versions of the pwb SolidWorks mates become problematic and have to be adjusted as well. The approach I use is after importing immediately take the %Cu for each layer and create a smart part PWB with it. The %Cu is in the name of the material after importing the EDA file.

I use both XT 3.2 and EFD 17.3.1 because of the model and user preferences and needs as Boris mentions. One limitation is even though both standalone versions use SolidWorks they are not compatible. XT uses SW 2017. After EFD V17.1 EFD uses SW 2018. Not backwardly compatible. So if you make or change the model in EFD 17.2+ you can no longer use it in XT. If you need to use both XT and EFD have to stay with EFD 17.1 (in which the parametric processor has useability glitches, fixed in later versions)

Another used asked about differences in the EDA. Although they are all called FloEDA Bridge they have differences. FloTHERM classic's EDA Bridge can plot power density (power versus area of a component). Neither XT or EFD do that. A very useful feature for an initial survey of critical components.

Another difference. The mesher in FT classic is different and has localized grids that behave differently than XT and EFD. They cannot touch each other. On a part such as an inclined block this takes up a lot of real estate and causes modeling difficulties for high packaging density. Say in a 2U-24 disk drive server for instance. Both XT and EFD have capability to have mesh refinement directly adjacent.

Even though XT and EFD are based on the same engine the meshing is different. in XT a 2 x 3 inch 12 layer pwb, using thermal territory explicit for a couple of BGA's, didn't finish geometry transfer after 18 hours....didn't even get to meshing stage and it was only using 1 core. This is with Dual XEON Gold 3+GHz processors on turbo (24 cores), 256Gb RDIMM memory using all channels, and hyper threading turned off. EFD mesher is better. Import geometry checker is better also.
DB

P.S.
Difference: Classic and XT sums all applied powers. EFD doesn't. Leaves out 2-R parts and network assemblies. Have to add them up manually.
Attached Images
File Type: jpg Smart PCB.jpg (169.1 KB, 9 views)
DB99 is offline   Reply With Quote

Old   July 8, 2020, 14:53
Default
  #22
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
Boris,
You are correct about electronic component smart parts. I was referring to the PCB smart part. FloEFD 17.3.1 does not create this automatically as FloTHERM XT 3.2, 3.1 does. Instead EFD creates individual materials for each layer and puts them in the project. This is extraneous bookkeeping for the analyst. A single layer material is unique and not useful for any other pwb so it isn't necessary to have the part in the engineering library or project. Just creates extra work when switching the PCB type from compact, detailed or explicit. Additionally if you start with a detailed pwb (layers each with lumped bulk properties) and switch to compact pwb (single 1 layer lumped parameter body) you have no pwb component. Have to make one. Then with two versions of the pwb SolidWorks mates become problematic and have to be adjusted as well. The approach I use is after importing immediately take the %Cu for each layer and create a smart part PWB with it. The %Cu is in the name of the material after importing the EDA file.

I use both XT 3.2 and EFD 17.3.1 because of the model and user preferences and needs as Boris mentions. One limitation is even though both standalone versions use SolidWorks they are not compatible. XT uses SW 2017. After EFD V17.1 EFD uses SW 2018. Not backwardly compatible. So if you make or change the model in EFD 17.2+ you can no longer use it in XT. If you need to use both XT and EFD have to stay with EFD 17.1 (in which the parametric processor has useability glitches, fixed in later versions)

Another used asked about differences in the EDA. Although they are all called FloEDA Bridge they have differences. FloTHERM classic's EDA Bridge can plot power density (power versus area of a component). Neither XT or EFD do that. A very useful feature for an initial survey of critical components.

Another difference. The mesher in FT classic is different and has localized grids that behave differently than XT and EFD. They cannot touch each other. On a part such as an inclined block this takes up a lot of real estate and causes modeling difficulties for high packaging density. Say in a 2U-24 disk drive server for instance. Both XT and EFD have capability to have mesh refinement directly adjacent.

Even though XT and EFD are based on the same engine the meshing is different. in XT a 2 x 3 inch 12 layer pwb, using thermal territory explicit for a couple of BGA's, didn't finish geometry transfer after 18 hours....didn't even get to meshing stage and it was only using 1 core. This is with Dual XEON Gold 3+GHz processors on turbo (24 cores), 256Gb RDIMM memory using all channels, and hyper threading turned off. EFD mesher is better. Import geometry checker is better also.
DB

P.S.
Difference: Classic and XT sums all applied powers. EFD doesn't. Leaves out 2-R parts and network assemblies. Have to add them up manually.
Hi DB

You emphasized the differences between XT and EFD and you obviously prefer to work with both of them. Not many people/companies can afford that, so if you have to select only one of them for your application (servers if I got it correctly) which one would you choose?
CFDfan is offline   Reply With Quote

Old   July 9, 2020, 04:48
Smile FloTHERM, XT, EFD, which one for servers.
  #23
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
Aahh,

That's a complicated question. Some time has passed since my initial post. FloEFD has since implemented a smart PCB I believe. I think it was initially in beta some time ago. I have not used it yet. I recall also a new addition of some sort of network reduction of a local area. ie: resistance network replacement of cells. I do not recall the details of this either, only the notice about the new feature. Maybe Boris can enlighten us on this and save me trying to find the email or notice on it.
If the new feature smart pcb has the same features as XT, which are providing the option to switch back and forth easily from compact to detailed, and providing the calculated bulk Kxy and Kz (useable for other manual PCBs) for this feature it would be a toss up. As it stand now EFD generates a material for every layer. This is difficult to suppress and unsuppress. Too much bookkeepping within a project and between projects, and the materials database. The materials are only associated with the geometry the first initial import of the EDA. After that it is up to you to manage it. Time consuming.
Another thing to consider. In XT I have not yet had a successful mesh generated in an explicit PCB model, nor with only the "thermal territory" feature. The mesher runs differently in XT than EFD (better). On a simple 2" x 3" 12 layer PCB the XT mesher ran for 16 hours and didn't get anywhere (Dual Xeon Gold, 256Gb main memory). I think it never got to the mesher, geometry translator too slow is my guess.


***********************
Without considering the above if you can live with compact and detailed I prefer XT for PCB work. I can do them pretty fast in XT. Most of the time is spent working up the PCB component details. If you do your PCB's first and alone, and only do main components at the system level, then this isn't an issue. That might change if the Smart PCB upgrade in EFD has the features of XT's Smart PCB. EFD has some more bells and whistles available to the user for fans and things but at times it slows things down.....too many dialog box options, used infrequently, is adverse to being fast at it. I also do not like having to add up the thermals manually in eFD. Prone to errors in a fast paced environent and not convenient for quick checks during quick turn around power level what-ifs.
***********************



I might have mentioned it earlier but in my version the Electronics Module in EFD does not have a Smart Heat Sink Part. That alone is enough to make me us XT for electronic work. Even some housing heat sink work. You can make a parametric heat sink in Solidworks (Mentor's suggestion) but it's a pain to debug it and get the parametric scenarios to work correctly, especially if it's not a textbook simple heat sink configuration.

If there are narrow channels to consider FlotHERM Classic is to cumbersome in meshing to mesh them quickly. It takes a lot of manual follow up meshing. Both XT and EFD to a nice quick job on narrow channels.

For compact you also should know that in order to insure three cells through the PCB, especially if it is not aligned with the orthogonal mesh you need to add a "1/3 PCB". This is easy. It is a copy of the original PCB outline, cut down to one third the thickness, mated so as to be at the center of the main PCB, overlapping it geometrically, and beneath it in the model tree so it overwrites the main PCB. This should ensure keypoints of the mesh are generated at the 1/3 and 2/3 through plane points so as to force three cells through the PCB. This is unnecessary with Detailed PCB.
It also depends on the overall size of your server system. If you're aware of the mesh cell ration limit then in my experience sometimes XT has difficulty resolving small/thin geometry if the domain is large. Even when local meshes are applied. Some mesh settings in XT are built in and hidden in attempts to make it easier for a CAD person to run an analysis. EFD gives you more mesh controls. If it is a single rack not a problem. If it is a whole cabinet this situation might arise, especially if you use contact resistances, model thin PCB perimeter EMI strips (0.05mm thick) or such small features. The quick solution is to increase mesh density of the domain (smallest resolvable feature is hence reduced in size) and live with the longer run time.

Sorry for the long text. Guess I"m a little over tired at this hour (2:45 am).
DB99 is offline   Reply With Quote

Old   July 9, 2020, 04:50
Cool supplement
  #24
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
XT Smart PCB: bulk Kxy, Kz and density and Cp also.
It also shows the stack up in a separate window for screen shot capturing.
DB99 is offline   Reply With Quote

Old   July 9, 2020, 05:18
Smile supplement
  #25
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
I work on more than just servers. Consider that for electronics, that is what XT was built and specialized for. Many of the features of Classic were incorporated into XT. So for electronics in my opinion and use XT is more thorough and faster model building than EFD. EFD has more versatility outside of that. Don' recall but I think the included electronics component starter pack library is much more complete in XT. Preloaded with 2R models in the library for a lot of packages. Would have to check tomorrow at work.
XT is less expensive. "XT Ultra" includes EDA bridge. With EFD you have to add the electronics module and the EDA module to the price. The EFD electronics module does not get you up to snuff with XT.
DB99 is offline   Reply With Quote

Old   July 9, 2020, 10:56
Default
  #26
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
I work on more than just servers. Consider that for electronics, that is what XT was built and specialized for. Many of the features of Classic were incorporated into XT. So for electronics in my opinion and use XT is more thorough and faster model building than EFD. EFD has more versatility outside of that. Don' recall but I think the included electronics component starter pack library is much more complete in XT. Preloaded with 2R models in the library for a lot of packages. Would have to check tomorrow at work.
XT is less expensive. "XT Ultra" includes EDA bridge. With EFD you have to add the electronics module and the EDA module to the price. The EFD electronics module does not get you up to snuff with XT.
I work in electronic cooling area as well and have used solidworks flow simulation (SWFS), that is Floefd embedded in solidworks, for a very long time. Was happy with it till a very large enclosure with a lot of details came along and SWFS simply refused to simulate it (with 128gB of RAM). After a lot of research we bought scStream. It was considerably more expensive than SWFS but very powerful and stable and we have never looked back. Some people in this forum were advocating for 6sigmaDC because of its accuracy. It is another code designed explicitly for electronic cooling. I have no opinion about its capabilities, knowing only that its price tag is the highest.
CFDfan is offline   Reply With Quote

Old   July 9, 2020, 13:36
Default
  #27
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
Both XT and EFD and most likely SWFS use only one core for geometry transfer. This is the bottleneck I see. With 24 cores it only uses 4% of CPU as shown in Resource Monitor or Task Manager - Performance tab. The mesher, once it gets to it is somewhat better because I think MS windows is getting it to use multiple cores. the mesher itself is not multi-threaded last I heard.
How many cells do you expect it might be? I could try meshing it in XT or EFD if it wasn't proprietary in nature and the SW version was compatible. Or how big is a .stp file?
In XT there is a green progress bar in the lower left corner of GUI. If it doesn't get past 25% (rough number) it's not in the mesher yet, it's still in the geometry transfer/translator stage.
Maybe try reducing the domain mesh to very light just to see what happens. No narrow channels, no fluid to solid refinement, no solid refinement, no fluid refinement. Does SWFS have these options available?
DB99 is offline   Reply With Quote

Old   July 9, 2020, 13:48
Default
  #28
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
I'm not familiar with scStream but tried out the PICL demo version, ok for quick idea, what it is.
DB99 is offline   Reply With Quote

Old   July 9, 2020, 22:58
Default
  #29
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
Both XT and EFD and most likely SWFS use only one core for geometry transfer. This is the bottleneck I see. With 24 cores it only uses 4% of CPU as shown in Resource Monitor or Task Manager - Performance tab. The mesher, once it gets to it is somewhat better because I think MS windows is getting it to use multiple cores. the mesher itself is not multi-threaded last I heard.
How many cells do you expect it might be? I could try meshing it in XT or EFD if it wasn't proprietary in nature and the SW version was compatible. Or how big is a .stp file?
In XT there is a green progress bar in the lower left corner of GUI. If it doesn't get past 25% (rough number) it's not in the mesher yet, it's still in the geometry transfer/translator stage.
Maybe try reducing the domain mesh to very light just to see what happens. No narrow channels, no fluid to solid refinement, no solid refinement, no fluid refinement. Does SWFS have these options available?
as I said SWFS is Floefd embedded in Solidworks, so whatever Floefd has SWFS has it as well (including the options you listed)
CFDfan is offline   Reply With Quote

Old   July 9, 2020, 23:54
Default
  #30
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
I'm not familiar with scStream but tried out the PICL demo version, ok for quick idea, what it is.
below is a brochure of all Cradles's products including scStream. Scstream has a module that is dedicated to electronics called HeatDesigner, but i prefer the full blown scstream because it has more features.
Very briefly - structured mesh, a lot of freedom for mesh adjustment, excellent handling of the memory - with 128gB of ram one can simulate 100-150 milion of cells (volumes), so very large and complex models can be simulated on a single PC. Excellent post-processing of the results, much more advanced than EFD or XT. Has a module called HeatPathView showing the path of the thermal flux between components and one can get an idea what is heating what and to what extend and how the heat dissipation within a component is distributed and if something can be done to reduce the hot spot temperatures
Very stable if the aspect ratio of the cells is kept below a threshold, yet that ratio is known after the meshing and before the solving, so one can refine/remesh the areas with bad aspect ratios (using so called multiblocks which are basically local meshes). The meshing is very fast, no more than a couple of minutes for say 100mil of cells. Tolerates overlapping of solids, the same way EFD does but in addition generates a message which parts are overlapping, so one can fix that if necessary. It is a standalone code that imports the geometry in Parasolid format. The geometry import doesn't take more than a minute.

Has much more sophisticated fan models (than EFD). Also has a lot of turbulence models (compared with the single one of EFD and XT), so one can get an idea how realistic the results are (by running with different turbulence models). Relatively fast solver that uses the CPU resources efficiently (if one has a license for all the cores).
Widely used by the electronic companies in Japan, Korea and China (scstream is a Japanese product bought by MSC Inc and then by Hexagon).
Excellent product support - updates are available every month and the bugs get fixed very quickly.

After these positive words I have to say that I don't have any association with scstream. I am just an user who worked in electronic cooling area for a very long time using a few CFD codes. The only thing I don't like about scstream is its price (but you get what you paid for).

https://media.mscsoftware.com/cdn/fa...e_a4_email.pdf
CFDfan is offline   Reply With Quote

Old   July 10, 2020, 04:46
Wink
  #31
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
I just recalled something. Several years ago I was on a project doing stress analysis. The thermal-fluid person was using SWFS. He ran a complicated cabinet/rack simulation model ok. I was surprised how quickly he turned around one re-sim after changing fans and fan curves. It had some bends in the inlet to the cabinet, a distribution plenum, and several racks. Based on the geometry and detail I saw it would have been in the many millions of mesh cells. Maybe in the tens of millions I seem to recall.

I was running femap/Nastran stress and vibration sims. To take a shortcut on one minor aspect of the project with some simpler geometry I tried to run it in SW simulation instead of exporting/importing a stp/x_t file. It ran ok until I added contact. Then with a minimal amount of contact surfaces it never finished pre-processing. Had to go back to Nastran for it. I seem to recall something about SW contact "manager" being the problem.

These two issues suggest to me a Solidworks issue in implementing simulations of either type. SW may be set up for only simpler jobs. At the time I think SW used the Cosmos stress solver which was pretty good. So I wouldn't blame EFD in SWFS for your long time issues.
DB99 is offline   Reply With Quote

Old   July 10, 2020, 05:04
Unhappy
  #32
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
Unlike Ansys and others EFD and XT doesn't limit how many cores you use. You can use as many as you like. This keeps cost down compared to programs that charge for core usage. That makes large models too expensive and prohibitive.

I sense a sales pitch here. I think that is an inappropriate use of a technical forum, and quite frankly wasting my time where I am trying to help another analyst on my personal time. Somewhat of a bait and switch gambit too. For what it's worth the PICL demo was extremely limited in it's features and somewhat short sighted for PCB analysis. Only a simple not very useful case except for an inexperienced Engineer. Anyone familiar with Steinbergor a few years of PCB experience would't need PICL for the simple things it did. No component case, TIM, to housing capability. Hope the stream isn't the same way.

You mentioned cost. How much does a comparable stream program cost, as long as you're selling?
Pardon my boldness if I am in error.


PS. Since EFD can do supersonic and Mach 20 with plasma dissociation and recombination it must have a decent turbulence capability. If I have it correct it also can use a shear at the wall approach to handle the boundary layer appropriately. You can get the powerpoint explanation from the EFD Product Manager.


Electronic cooling isn't aerodynamics. Since the geometry details are usually abbreviated, say for several packaged components on a PWB, that are block models wthout pins or legs, extra detail in the flow solution is fantasy. For electronics cooling Spaulding developed what was needed for the typical characteristic dimensions. Since the manufacturers thermal data is mostly fantasy as well reading too much into it is more fantasy. ie: Theta_jc, Theta_jb, Delphi models are commonly poorly derived. Some of the thermal reports I've received from IC manufacturers don't even apply or get the JEDEC definitions right!
DB99 is offline   Reply With Quote

Old   July 10, 2020, 05:05
Thumbs up
  #33
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
I"ve found that the physics capability of EFD is superb and the scientific code developers are first rate.
DB99 is offline   Reply With Quote

Old   July 10, 2020, 05:55
Default
  #34
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
I"ve found that the physics capability of EFD is superb and the scientific code developers are first rate.
Well, if you think that of EFD(and I agree with you), read about the founders of 6SigmaDC - graduates from the 3 best technical universities of USA, UK and Russia.

I believe the Japanese scientists are world class too, we just know too little about them

Last edited by CFDfan; July 10, 2020 at 07:40.
CFDfan is offline   Reply With Quote

Old   July 10, 2020, 06:13
Default
  #35
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
I just recalled something. Several years ago I was on a project doing stress analysis. The thermal-fluid person was using SWFS. He ran a complicated cabinet/rack simulation model ok. I was surprised how quickly he turned around one re-sim after changing fans and fan curves. It had some bends in the inlet to the cabinet, a distribution plenum, and several racks. Based on the geometry and detail I saw it would have been in the many millions of mesh cells. Maybe in the tens of millions I seem to recall.

I was running femap/Nastran stress and vibration sims. To take a shortcut on one minor aspect of the project with some simpler geometry I tried to run it in SW simulation instead of exporting/importing a stp/x_t file. It ran ok until I added contact. Then with a minimal amount of contact surfaces it never finished pre-processing. Had to go back to Nastran for it. I seem to recall something about SW contact "manager" being the problem.

These two issues suggest to me a Solidworks issue in implementing simulations of either type. SW may be set up for only simpler jobs. At the time I think SW used the Cosmos stress solver which was pretty good. So I wouldn't blame EFD in SWFS for your long time issues.
Don't get me wrong - I used SWFS for more than 20 years when it was still a part of the German's NIKA (before Flomerics, Mentor and now Siemens) and was very familiar with it. I knew one of the developers too and used to get first rate suggestions from him when facing an issue (the local technical support wasn't of much help). It was and is excellent software and that's the reason I stuck to it for so long. It is just that I found scstream to be better.
CFDfan is offline   Reply With Quote

Old   July 10, 2020, 06:43
Default
  #36
Member
 
Join Date: Jun 2011
Posts: 79
Rep Power: 11
CFDfan is on a distinguished road
Quote:
Originally Posted by DB99 View Post
Unlike Ansys and others EFD and XT doesn't limit how many cores you use. You can use as many as you like. This keeps cost down compared to programs that charge for core usage. That makes large models too expensive and prohibitive.

I sense a sales pitch here. I think that is an inappropriate use of a technical forum, and quite frankly wasting my time where I am trying to help another analyst on my personal time. Somewhat of a bait and switch gambit too. For what it's worth the PICL demo was extremely limited in it's features and somewhat short sighted for PCB analysis. Only a simple not very useful case except for an inexperienced Engineer. Anyone familiar with Steinbergor a few years of PCB experience would't need PICL for the simple things it did. No component case, TIM, to housing capability. Hope the stream isn't the same way.

You mentioned cost. How much does a comparable stream program cost, as long as you're selling?
Pardon my boldness if I am in error.


PS. Since EFD can do supersonic and Mach 20 with plasma dissociation and recombination it must have a decent turbulence capability. If I have it correct it also can use a shear at the wall approach to handle the boundary layer appropriately. You can get the powerpoint explanation from the EFD Product Manager.


Electronic cooling isn't aerodynamics. Since the geometry details are usually abbreviated, say for several packaged components on a PWB, that are block models wthout pins or legs, extra detail in the flow solution is fantasy. For electronics cooling Spaulding developed what was needed for the typical characteristic dimensions. Since the manufacturers thermal data is mostly fantasy as well reading too much into it is more fantasy. ie: Theta_jc, Theta_jb, Delphi models are commonly poorly derived. Some of the thermal reports I've received from IC manufacturers don't even apply or get the JEDEC definitions right!
I was afraid I would sound as a salesmen and that's why I made the statement, but believe it or not I have nothing to do with scStream.
The only thing I could tell you about its price is that it was about 2.5x the price of SWFS, bit less expensive than 6sigma. In addition, as you probably know, the prices of the CFD codes can vary significantly from user to user, so these figures may not be the market prices.

I agree completely that "Electronic cooling isn't aerodynamics" and what you said after that. I would just add that EFD originated in Russia and was developed by Russian scientists in aerodynamics, so it wasn't targeting too much the electronic cooling. In addition handling Mach 20 and the natural convection cooling with a single (or two) turbulent model, regardless how good it is, doesn't seem to come without limitations but I may be wrong about that.
Look, I don't want to spent more time on this discussion. I was expressing my personal opinion and you can agree or disagree. Again, EFD is an excellent CFD code and does an excellent job in vast majority of electronic cooling applications.
CFDfan is offline   Reply With Quote

Old   July 10, 2020, 15:33
Default FloTHERM XT sample project
  #37
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
I've heard of 6Sigma and looked it's brochures over but not ever been with company that had it.


The attached should suffice to confirm capabilities. Whoops file is too big. I can send it if you like. You might search the web for it. It is a Siemens unrestricted presentation.


Liquid Cooling System for a 17kW Artificial Intelligence Module-Ver2.0.pdf


It's 17.6 kW on a single wafer ~ 20cm with 84 cores on it. Largest chip ever built known to the public. There are several layers of complicated inlet manifolds to give impingement cooling on each core.


I'm not the author or analyst of the project.
DB99 is offline   Reply With Quote

Old   July 10, 2020, 15:35
Default
  #38
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
Presented at Realize Live (Siemens)
DB99 is offline   Reply With Quote

Old   July 10, 2020, 15:38
Default
  #39
New Member
 
DB
Join Date: Jan 2019
Posts: 22
Rep Power: 3
DB99 is on a distinguished road
physics programming smarts is good and necessary. It's the user GUI experience that also counts big time and doesn't take a physics genius. Something that is unfortunately in these days left to the wayside over continuously adding more features to the programs.
DB99 is offline   Reply With Quote

Old   July 14, 2020, 09:36
Default
  #40
Senior Member
 
Boris Marovic
Join Date: Jul 2009
Posts: 600
Rep Power: 20
Boris_M will become famous soon enough
Wow, one week away and there is a blast of posts 

Ok, here’s an update and some information on what has happened with Simcenter FLOEFD (as it is called now) but short FLOEFD for my hands sake  and to shorten the post at least a little.

Yes, you are right, an explicit PCB import and then getting it to the meshing can take quite a while to mesh as it is creating the full detail of the PCB geometry. So it depends on the amount of copper traces and the through connects etc. that needs to be created and then handled by the mesher. But as smart as the FLOEFD developers are (and this is a unique feature to FLOEFD only at the moment), they developed the SmartPCB. This has nothing to do with smart parts or libraries in any way. It is a thermal network assembly model of a PCB. With that, the PCB becomes mesh independent as the PCB is solved with a thermal network model and does not need a fine mesh to resolve the traces etc. The components on the PCB are excluded from this model, this only is for the PCB itself. The accuracy of the PCB representation can easily be defined in nodes which can be visualized similar to a pixelated image of the PCBs copper traces. It is extremely fast with very little accuracy loss from this approach.
This way, your meshing should not take 16 hours anymore for a complex PCB as the mesh is completely irrelevant for the PCB. ;-)

Yes, smart parts are a nice thing and that’s the foundation XT and Flotherm are based on, while FLOEFD is a CAD and feature based tool. So geometry typically comes from CAD designs. As you said, such libraries can be created inside the CAD system and I have done that in the past and with some scripting work you can get them nice and easy to use, this is all part of the XT and Flotherm package for their purpose of course.

As for the modules and things, there is the new “Electronics Cooling Center” module which brings a whole load of new electronics cooling features into FLOEFD in a single package. It combines several modules (which you can still get separately if you don’t need all of them):

- Electronics Cooling Center Module
---- EDA Bridge Module
---- Electronics cooling Module
---- T3STER Automatic Calibration Module
---- BCI-ROM + Package Creator Module
- Power Electrification Module
- LED Module
- HVAC Module
- Extended Design Exploration Module
- Advanced Module

With the Electronics Cooling Module it is about equally powered as XT Ultra with the usual differences such as smart parts etc. It received a lot of the electronics cooling knowhow and technology of Flotherm, but is still meant for CAD designers as a core user focus, hence the UI.
And no, Flotherm will not go away, it has a different user focus. If you consider that automotive companies also don’t just have one car model that suits all.

The CAD system is the bottleneck as it is not parallelized. Once the geometry is exported for the mesher, the mesher is parallelized just like the solver. The mesher is parallelized since version 10 or 11 I think, I can’t remember exactly but that’s almost 10 years now.

Yes, SWFS has the basic functions of FLOEFD mostly with some minor differences, the bigger differences come from the modules as SWFS only has the HVAC and Electronics cooling module.

By the way, FLOEFD also has the heat path viewer.

SWFS and SW Simulation are completely separate and have nothing to do with each other. Except using pressure and temperature loads being possible.
I cannot tell about the structural tools but know that FLOEFD for example can export to other structural tools, this is one of the basic version differences that SWFS cannot do.

Actually FLOEFD can do up to Mach 30 


This all is not meant as a sales pitch, after all it’s up to you if you buy or not. I just saw the messages and thought I’ll add my 2 cent on some of the things mentioned to avoid any misunderstanding or just to update on the latest news.

Regards,
Boris

Last edited by Boris_M; July 15, 2020 at 06:25.
Boris_M is offline   Reply With Quote

Reply

Tags
floefd, flotherm, flotherm xt

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
FloEFD for axial and radial blower CmdrFALCO FloEFD, FloWorks & FloTHERM 3 September 9, 2015 03:31
Fluent Vs Solid-works Flow Simulation Mortezab Main CFD Forum 16 July 5, 2013 03:47
FloEFD and FloTHERM Blogs @ Mentor.com RobinB FloEFD, FloWorks & FloTHERM 0 July 17, 2009 16:59
about data produced by flotherm Adam Main CFD Forum 0 December 14, 2005 21:50
Flotherm-Icepak comparison for electronic cooling Masoud Ameli FLUENT 0 February 28, 2005 10:00


All times are GMT -4. The time now is 20:35.