flat assembler
Message board for the users of flat assembler.

Index > Heap > CPU die pictures

Goto page Previous  1, 2, 3  Next
Author
Thread Post new topic Reply to topic
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
I wonder why do they always make the insides in rectangular shapes? Couldn't you fit more in if you used, say triangles?

Of course you'd have to start from scratch...
Post 02 Jan 2010, 19:15
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
Huh? Triangles have less area per the required flat surface (by the standards used), the extra space would be wasted. What do you intend to put between the edges of the triangle? (I mean, where a square would have surface of course).

_________________
Previously known as The_Grey_Beast
Post 02 Jan 2010, 19:26
View user's profile Send private message Reply with quote
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
Borsuc wrote:
Huh? Triangles have less area per the required flat surface (by the standards used), the extra space would be wasted. What do you intend to put between the edges of the triangle? (I mean, where a square would have surface of course).


Triangles stack just like squares (but sometimes upside-down). I guess the die would be triangular then, no big deal? (EDIT: In cutting the die from the rod slices, a triangle would actually fit better as they could be arranged into hexagons, which match up nicer to the rod slices round shape.)

Is the lack of area space per module even actually problematic? E.g.: Can you proove it's not better that way?

LP,
Jure
Post 02 Jan 2010, 20:01
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
If the die was triangular, it means the CPU socket should be triangular or else there's wasted space. It means the motherboard has to be triangular also. And to have efficient use of space it means the computer case has to be triangularly-shaped.
Post 02 Jan 2010, 20:04
View user's profile Send private message Reply with quote
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
Borsuc wrote:
If the die was triangular, it means the CPU socket should be triangular or else there's wasted space. It means the motherboard has to be triangular also. And to have efficient use of space it means the computer case has to be triangularly-shaped.


Egh... no? Have you actually ever seen a CPU die planted in the CPU circuit board? Without the heatspreader?

AMD K7's came without heatspreaders:
Image

The black thing in the centere is the CPU die. If it were triangular the amount of wasted space would be, for all intents and purposes, pretty much the same.

EDIT: Note that modern CPU dies are often even smaller, whereas the socket size remains the same.

LP,
Jure
Post 02 Jan 2010, 20:12
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
Sorry but the shape has to match for efficient heat spreading Smile

For the record by "wasted space" I meant wasted space between the socket and the motherboard's components, not between the heat spreader and the die...

_________________
Previously known as The_Grey_Beast


Last edited by Borsuc on 02 Jan 2010, 20:21; edited 1 time in total
Post 02 Jan 2010, 20:20
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17279
Location: In your JS exploiting you and your system
revolution
Most mounting for dies already has lots of space, so changing to triangles would not make much difference to the package. It can still be a square chip carrier with a triangle chip inside. Nothing says the carrier has to be the same shape as the chip (except maybe the latest CSPs of course, but a different situation there).
Post 02 Jan 2010, 20:20
View user's profile Send private message Visit poster's website Reply with quote
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
Borsuc wrote:
Sorry but the shape has to match for efficient heat spreading Smile


BS. Why?

The P4 heatsink attaches to the CPU heatspreader using a copper cylinder, where the contact surface is round. Does that mean it's less efficient than if it were square like the die? I find the whole argument rather ridiculous.

Quote:
For the record by "wasted space" I meant wasted space between the socket and the motherboard's components, not between the heat spreader and the die...


The socket can remain rectangular even if the die is triangular.

LP,
Jure
Post 02 Jan 2010, 20:26
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17279
Location: In your JS exploiting you and your system
revolution
However, triangular chips, with corresponding triangular carriers, with balls also arranged in a triangular matrix, would make it slightly more difficult to design the PCB footprints using current software tools. Generally the pads are expected to be in a regular X/Y arrangement. But the software could change if triangular chips ever became popular.
Post 02 Jan 2010, 20:33
View user's profile Send private message Visit poster's website Reply with quote
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
revolution wrote:
However, triangular chips, with corresponding triangular carriers, with balls also arranged in a triangular matrix, would make it slightly more difficult to design the PCB footprints using current software tools. Generally the pads are expected to be in a regular X/Y arrangement. But the software could change if triangular chips ever became popular.


I thought so too. In other words, they use square shapes because their software does.

I guess then a way forward would be to create software that does not have this pre-defined requirement, and then optimize your die layouts with that. If the software had no requirement to fit everything into boxes but would instead just go for whatever worked best, it could possibly come up with better (more efficient) results.

LP,
Jure
Post 02 Jan 2010, 20:39
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17279
Location: In your JS exploiting you and your system
revolution
I don't think the die shapes follow the software capabilities. You could still use the current software, just that it would be a bit 'messier'.
Post 02 Jan 2010, 20:45
View user's profile Send private message Visit poster's website Reply with quote
LocoDelAssembly
Your code has a bug


Joined: 06 May 2005
Posts: 4633
Location: Argentina
LocoDelAssembly
Triangles would have to have larger sides than the square to have the same surface, additionally in the triangle the more you approach to a vertex the smaller the chances you have to fit something there.

I think it can complicate placement and routing very much and that you would end up wasting space inside the triangle.
Post 02 Jan 2010, 21:05
View user's profile Send private message Reply with quote
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
LocoDelAssembly wrote:
Triangles would have to have larger sides than the square to have the same surface, additionally in the triangle the more you approach to a vertex the smaller the chances you have to fit something there.

I think it can complicate placement and routing very much and that you would end up wasting space inside the triangle.


Remembering the point that eveything on a chip die, short of a transistor (which is small enough anyway), is attained using different lengths of wire, lack area space does not seem too directly problematic to me.

Of course I have not thought it trough, it is possible that a certain circuit requires a certain amount of area space to be done. However as far as I recall electronics classes, with most circuits the problem is how to get your inputs where you need them, rather than needing a bulk of area at a particular point to complete a maze. If that last assessment is right, triangles would work better.

But of course I am not insisting on triangles. I say, anything that is not necessarily square, whatever works.

LP,
Jure
Post 02 Jan 2010, 21:13
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
DustWolf wrote:
BS. Why?

The P4 heatsink attaches to the CPU heatspreader using a copper cylinder, where the contact surface is round. Does that mean it's less efficient than if it were square like the die? I find the whole argument rather ridiculous.
Larger surface = better heat dissipation. Even disspation = better. A triangle would not have an 'even' dissipation considering the shape of the heat spreader. A triangle needs to have larger sides for the same area, so what's the point? It doesn't use the heat dissipation properly, I reckon. Since a triangle will have certain parts in the rectangular heat spreader's center-wise area (where it dissipates from!) where there is much less heat. (because there's empty space) so at full load it will be much less efficiently to spread it, as heat will be spread more towards one end (where the triangle is 'filled' compared to a square/rectangle).

Imagine having a CPU square sliced into a triangle (diagonal)... the other half of the heat spreader will receive less heat compared to the side where the triangle is, even at full load -- inefficient.

I mean do you have a specific reason why triangles should be used? Because clearly the possible problems exist with it. So what's compelling you to think about them?

_________________
Previously known as The_Grey_Beast
Post 02 Jan 2010, 21:27
View user's profile Send private message Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17279
Location: In your JS exploiting you and your system
revolution
Borsuc: If your reasoning were correct then chips should be hexagonal.
Post 02 Jan 2010, 21:35
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
Actually the heat spreaders should have a square/rectangular shape but with rounded corners (to preserve distance from corner) -- which they actually do, in fact. Wink

_________________
Previously known as The_Grey_Beast
Post 02 Jan 2010, 23:49
View user's profile Send private message Reply with quote
kalambong



Joined: 08 Nov 2008
Posts: 165
kalambong
vid wrote:
Is the CPU somehow fucked up, or did you mean "dye" instead of "die"?
Dye is about coloring fabrics, Die on the other hand, has several meanings.

Other than the meaning of "Kaput", it also means to "permanently cast". Hence the term "Die cast".

English ---> A very confuse language.

More example of how confuse English is:

Tall = Tall
Taller = More tall.

The above is correct, however,

Corn = Corn
Corner != More corn.

Very Happy
Post 05 Jan 2010, 08:19
View user's profile Send private message Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 2915
Location: [RSP+8*5]
bitRAKE
Modern ICs have arrays of functionality (i.e. buses, cache memory, etc.) which are laided out in a rectangular grid. If a more efficient (in terms of power or time) arrangement were possible in other shapes it would be quite valuable. Although I don't design ICs, everything I've read seems to indicate the majority of tools and research are based on rectangular arrangements.

Presently, doping creates an irregular film, but there is some work based on surface atomic structure -- self-organization at the atomic scale is likely to play a role if it can be integrated into the manufacturing process. Which might lead to changes in the shape of these larger scale functional units.

It might be the cutting process which favor rectangles?

Triangles seem more fragile to me.
Post 05 Jan 2010, 17:44
View user's profile Send private message Visit poster's website Reply with quote
El Tangas



Joined: 11 Oct 2003
Posts: 120
Location: Sunset Empire
El Tangas
revolution wrote:
Borsuc: If your reasoning were correct then chips should be hexagonal.


That's right, most physical arguments presented here favoring squares over triangles would then favour hexagons over squares...

Check this patent: US 6,030,885
Post 05 Jan 2010, 19:10
View user's profile Send private message Reply with quote
DustWolf



Joined: 26 Jan 2006
Posts: 373
Location: Ljubljana, Slovenia
DustWolf
El Tangas wrote:
That's right, most physical arguments presented here favoring squares over triangles would then favour hexagons over squares...


And hexagons can be made out of triangles. Smile

bitRAKE wrote:
It might be the cutting process which favor rectangles?


AFAIK in printed circuits this is not an issue (I've seen triangular boards with diagonal wires before, in very cheap PCI cards).

For IC it doesn't seem to be a problem either. Since light is used for the "cutting" the mask can be much bigger than the chip, allowing one to cut whatever shape into them. I suppose triangles too?

LP,
Jure
Post 05 Jan 2010, 20:53
View user's profile Send private message AIM Address Yahoo Messenger MSN Messenger Reply with quote
Display posts from previous:
Post new topic Reply to topic

Jump to:  
Goto page Previous  1, 2, 3  Next

< Last Thread | Next Thread >
Forum Rules:
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum


Copyright © 1999-2020, Tomasz Grysztar.

Powered by rwasa.