How do hardware review sites get those shots of CPU cores?

[desura]

Like this one from Toms Hardware?

 

[2timer]

How do hardware review sites get these shots of CPU cores?

Well, first a paparazzi hides in the bushes outside of the CPU core's house, like this:

Then when the CPU core comes outside to drive to the store, the paparazzi jumps out of the bushes and snaps a photo before the core can get in it's car!

Just kidding. Actually I have no idea, that's a good question!

 

[ShadowVVL]

Well, first a paparazzi hides in the bushes outside of the CPU core's house, like this:

Then when the CPU core comes outside to drive to the store, the paparazzi jumps out of the bushes and snaps a photo before the core can get in it's car!

Just kidding. Actually I have no idea, that's a good question!

I believe it.My guess is they get them from from the manufacturer but I could be wrong.

 

[dma0991]

I'm guessing its reverse engineering. Example of a teardown.

 

[GreenChile]

I'm guessing its reverse engineering. Example of a teardown.

Interesting theory. Or they could just go to Intel or AMDs website and download the promotional images. http://newsroom.intel.com/docs/DOC-2735

 

[Haserath]

Pretty sure Intel/Amd have to sacrifice a wafer. They shave a few layers off then take a photo of it and use photoshop.

I might be remembering wrong.

 

[ViRGE]

Pretty sure Intel/Amd have to sacrifice a wafer. They shave a few layers off then take a photo of it and use photoshop.

I might be remembering wrong.

No, that's basically it. There's no practical way to get a shot like this from a finished product. All of these good shots are coming from the manufacturer.

 

[ghost03]

Pretty sure Intel/Amd have to sacrifice a wafer. They shave a few layers off then take a photo of it and use photoshop.

I might be remembering wrong.

That sounds about right.

Alternatively, at a guess, they might just use infrared cameras and shift the color spectrum. Silicon is transparent to infrared.

 

[Idontcare]

They don't have to sacrifice the wafer or the chip. They (the manfucturer or foundry) just take a wafer to a microscope tool mid-production and take photos of it. Then they put the wafer back into production.

Most micrographs of silicon dies that we see in the public are not of a chip that has been fully processed, they might still have 6 metal levels to go before they even leave the fab.

We take pictures of wafers all the time as routine analysis for yield enhancement. If it was destructive then that would be counter-productive.

Now it is true you can get the images through destructive means such as teardowns on Chipworks. But that is expensive (~$2000) and I guarantee you no one does that just to write an internet article on a review site like Toms.

 

[desura]

So what's with the colors? Is it like added for effect or is it really needed as part of the mfg process?

 

[2timer]

So what's with the colors? Is it like added for effect or is it really needed as part of the mfg process?

I think it's s added to highlight the specific components. Otherwise everything would be as monocolor and confusing as looking at a tiny circuit board.

Or maybe it's part of the lithography process? :hmm:

 

[Idontcare]

So what's with the colors? Is it like added for effect or is it really needed as part of the mfg process?

The colors are usually real, not added, but they are not "real" in the sense of the colors you see on a painted wall. They are more akin to the color you see when looking at a peacock's tail feathers or from soap bubbles.

Iridescence in soap bubbles

Also the same thing happens with butterfly wings.

It is called thin-film interference. The different colors in the silicon die image are due to the various structures you are looking at having different dimensions and materials composition, changes the physics of the thin-film interference which then changes the color.

Figure 2. White light incident on an oil film. A colorful interference pattern is observed when light from the sun is reflected from the top and bottom boundaries of a thin oil film.

 

[StrangerGuy]

Why would manufacturers not release die shots? They don't divulge any critical information and corporate spies would have much more indepth information already.

 

[Phynaz]

They don't divulge any critical information

Apparently they do. Remember when AMD was releasing obfuscated BD die photo's? They specifically said they didn't want the competition to glean any intelligence from them.

 

[Idontcare]

Apparently they do. Remember when AMD was releasing obfuscated BD die photo's? They specifically said they didn't want the competition to glean any intelligence from them.

They sure showed them! :|

 

[Tsavo]

I wonder how many Libraries of Congress they can fit on a 3770k die.

Relative scale is important in order to comprehend sizes of things.

 

[Idontcare]

I wonder how many Libraries of Congress they can fit on a 3770k die.

Not many...the die is ~160mm^2 and the library of congress is on the order of a whole city block in size.

 

[erunion]

Not many...the die is ~160mm^2 and the library of congress is on the order of a whole city block in size.

lol.

Supposedly all 24 volumes of the Encyclopedia Britannica would fit on the head of a pin if printed at 8nm.



http://www.chipworks.com/blog/techn...plenty-of-room-at-the-bottom-intel-thinks-so/

 

[djshortsleeve]

Just a wafer. I can take a pic for you right now of one...

 

[pm]

Back when I worked on Itanium 2 9000-series (codenamed Montecito) and was giving a presentation at ISSCC, we wanted some good die photos, so we took a few dies that we stopped fabrication at something like metal 4 or metal 5 and took them over to Colorado State University and they took the photos using a conventional optical microscope and then they email'd us some really massive .TIF files. As Idontcare said, the colors are not false-color - they are thin-film interference patterns. You can see them with the naked eye.