Intel Alder lake source code may have been leaked

winjer

Gold Member

VX-Underground has tweeted that the source code to the Intel 12th Gen Alder Lake CPUs has been leaked online. Intel's Alder Lake CPUs were released last year on the 4th of November, 2021 & the data include 2.8 GB worth of compressed source code (5.86 GB in full), and the leak allegedly came from 4chan. The code base is said to be massive but the contents have yet to be verified.

In a second tweet, glowingfreak posted a GitHub link that shows the BIOS files & was shared 8 days ago. Based on the files, it looks like the data and files are mostly related to BIOS and chipsets & it's not officially known if this hack happened at Intel or some other vendor such as a major OEM since there are documents that refer to Lenovo's "Feature Tag Test Information". Whether or not the files were obtained by hacking Intel or their partner, they are still very important and can cause big and major trouble for the blue team.

This won't be the first hack of its kind. NVIDIA, AMD, and Gigabyte have all been hacked in recent times. NVIDIA was hacked back in early 2022 with several documents and source codes for future technologies leaking out. AMD was hit by a Ransomware attack and 450 GB worth of data was all but stolen. Even Gigabyte also had 112 GB of data which included confidential documents for Intel and AMD products hacked. The information gave away details for next-gen GPUs and CPUs that all three vendors were working on.



 
Last edited:

winjer

Gold Member
What does one do with this information?
Security holes?
Reverse engineer to make Sugar Free Alder Lake?

Yes, security on Intel CPUs might get seriously compromised.
Some companies might try to use it to develop their own products. Not so much in the west, as there are strong laws in place, but in China and Russia, it's a free for all.
 

Bernkastel

Ask me about my fanboy energy!
Some companies might try to use it to develop their own products. Not so much in the west, as there are strong laws in place, but in China and Russia, it's a free for all.
Chinese companies can't buy EUVs from ASML due to sanctions. They can't really manufacture advanced semiconductors even if they want to.
 

winjer

Gold Member
Chinese companies can't buy EUVs from ASML due to sanctions. They can't really manufacture advanced semiconductors even if they want to.

SMIC, which is Chinese, is already at 7nm process node.
EUVs are impossible for SMIC for now, and for several years.
But to make a 7nm CPU, they can already do it, and that is sufficient for something on the tech of Alder lake.
 

Bernkastel

Ask me about my fanboy energy!
SMIC, which is Chinese, is already at 7nm process node.
EUVs are impossible for SMIC for now, and for several years.
But to make a 7nm CPU, they can already do it, and that is sufficient for something on the tech of Alder lake.
I doubt there 7 nm is anywhere near Intel 10 nm. nm mean a different thing for each fab manufacturers, it's like comparing AMD and Nvidia TFs. Going forward PSMC can't compete with TSMC, Samsung or Intel since they are not permitted to buy EUVs from ASML. Even if they have Alder Lake source, it will take them years to reverse engineer that tech and manufacture it with DUVs.
 

winjer

Gold Member
I doubt there 7 nm is anywhere near Intel 10 nm. nm mean a different thing for each fab manufacturers, it's like comparing AMD and Nvidia TFs. Going forward PSMC can't compete with TSMC, Samsung or Intel since they are not permitted to buy EUVs from ASML. Even if they have Alder Lake source, it will take them years to reverse engineer that tech and manufacture it with DUVs.

SMIC 7nmprocess was made by Liang Mong Song, so it's probably very similar to TSMC N7. That means it's competitive with Intel 10nm.

 
SMIC 7nmprocess was made by Liang Mong Song, so it's probably very similar to TSMC N7. That means it's competitive with Intel 10nm.
Well Raja Koduri made Intel Arc A770, so I guess the GPU he made is very comparable to AMD's.

It might be, but it also might not be. Depending on the team and tech one has available, or doesn't. What you said is still good information for the thread mind you, but I'm not so sure about comparability unless there was data to corroborate it. :)

Even if the transistor size is comparable, how much density are they getting away with? I don't know, and the fact they don't say it might be a red flag.
 
Last edited:

winjer

Gold Member
Well Raja Koduri made Intel Arc A770, so I guess the GPU he made is very comparable to AMD's.

It might be, but it also might not be. Depending on the team and tech one has available, or doesn't. What you said is still good information for the thread mind you, but I'm not so sure about comparability unless there's data to corroborate it. :)

You clearly don't know who Liang Mong Song is, otherwise you woudn't be comparing him to Raja Koduri.
 
You clearly don't know who Liang Mong Song is, otherwise you woudn't be comparing him to Raja Koduri.
It was an example.

Fact is, the little we know about said node, doesn't give us those values, and a big part of the result of manufacturing relies on machinery and how clean everything is.

Regardless, of all that, from the little we know, it's not the exact same method of manufacturing:



It's very similar, but... It has at least one difference and fact is even if it was the same on paper the difference could still be big in regards to density/yields.

I don't know enough to lean to one way or the other. All I'm saying.
 
Last edited:

winjer

Gold Member
It was an example.

Fact is, the little we know about said node, doesn't give us those values, and a big part of the result of manufacturing relies on machinery and how clean everything is.

Regardless, of all that, from the little we know, it's not the exact same method of manufacturing:



It's very similar, but... It has at least one difference and fact is even if it was the same on paper the difference could still be big in regards to density/yields.

I don't know enough to lean to one way or the other. All I'm saying.

Liang Mong Song worked at TSMC, as one of the major leaders in development in new process nodes. A real genius.
Some years back he left TSMC and started working for Samsung, covertly, because he had a non-compete clause.
After a while Samsung process nodes caught up to TSMC. This year, Samsung and TSMC announced production for their 3nm process at the same time.
SMIC for years was struggling to make advances in old process nodes, and all of a sudden they managed to get a 7nm process node working. And Liang Mong Song was there to make it happen.
At this point we don't know yet about yields, and there aren't yet any major products released by SMIC on 7nm. But this is just a matter of time.

The big obstacle SMIC has now, is EUV. The only company that makes machines with EUV is ASML on the Netherlands. And they are not selling them to China.
This means China has to develop their own EUV machines. And this is going to take a lot of time, money and effort.
 
The big obstacle SMIC has now, is EUV. The only company that makes machines with EUV is ASML on the Netherlands. And they are not selling them to China.
This means China has to develop their own EUV machines. And this is going to take a lot of time, money and effort.
Yup.

Anyway, in regards of their node, it's technically 7nm, but that's not the whole story even taking into account LMS involvement, it's probably the possible result given the problems faced.

Tech Insights has classed the N+1 process from Chinese foundry Semiconductor Manufacturing International Corp. (SMIC) as a 7nm process, albeit on a relaxed pitch for polysilicon, fin and metal interconnect.

(...) it falls into the 7nm group with respect to standard cell height. From the height and width of the SMIC standard cells, SMIC scaling puts their technology in the same 7nm group as Intel, TSMC and Samsung.

However, in the documents shown under open access Tech Insights makes no reference to M1 metal pitch, which is one of the key defining parameters of a process.

SMIC has had to develop its process using only deep-UV immersion lithography equipment. This is likely to make production of 7nm, while technically possible, more complex, lower yielding and considerably more expensive.

SMIC 7nm process shows more relaxed scaling for contacted poly, metal and fin pitches, setting it apart from TSMC N7+ and Intel 10nm – all of which use single diffusion break (SDB) and design technology co-optimization (DTCO). However, a more relaxed scaling means fewer transistors and gates per unit area.

MinerVa Bitcoin mining chip is based on this N+1 7nm process (...) the application is less sensitive to process variation and makes use of a higher threshold voltage for transistors in the more typical logic circuits.
Source: https://www.eenewsanalog.com/en/smic-process-is-7nm-says-tech-insights/

Less density, worse yields and more energy needed (therefore, more heat generated as well).

At the bare least, I believe it behaves like TSMC 10nm. At worst, I won't even make a suggestion.
 

winjer

Gold Member
Yup.

Anyway, in regards of their node, it's technically 7nm, but that's not the whole story even taking into account LMS involvement, it's probably the possible result given the problems faced.


Source: https://www.eenewsanalog.com/en/smic-process-is-7nm-says-tech-insights/

Less density, worse yields and more energy needed (therefore, more heat generated as well).

At the bare least, I believe it behaves like TSMC 10nm. At worst, I won't even make a suggestion.

It's probably at the stage where Intel's 10nm was, a few years ago.
Eventually, Intel solved their issues. SMIC will probably manage to do it as well.
But the thing to take into consideration is that SMIC manage to have a huge jump in tech, thanks to Liang Mong Song. Just like Samsung did.
 
It's probably at the stage where Intel's 10nm was, a few years ago.
What I read reminded me of that, yes. Although iirc intel issue had to do with EUV passages. Point stands though as it'll always improve but they'll need EUV to go forward, 7nm with good density included, probably.

I'm sure they'll solve it, but right not it's not performant enough for something like a high performance CPU, for that they might as well fall back on 14-16nm for now.
 
Last edited:
Wow it seams that Russians are taking chips from washing machines to build state of the art aircrafts and there are idiots who believe this kind of things.

China is under sanctions and can't make advanced semiconductors. How can you function believing all this crap. They are under sanctions and now they are trying to build their own machines and in couple years they will able to create advanced semiconductor like they make other millions of items.
 
Top Bottom