Good Morning Bravo ,
Just listened again ,
Time spot 33:22 onwards...
Time spot 33:56 , YOU KNOW, GLOBALFOUNDRY 8nm .
There Will Be Two Live Sessions on 26 August: If you would like to watch this broadcast in fullscreen, simply scan the QR code on your smartphone and interact with us.
brainchip.com
Regards,
Esq.
Hi Esqy,
You’re right - Kurt does mention GlobalFoundries’ “8 nanometer” in the webinar.
I find it a bit confusing, because GlobalFoundries doesn’t actually have an official 8 nm process node (unless I’m mistaken). Their most advanced public offerings are 12LP+ FinFET and 22FDX / 12FDX FD-SOI. They don’t list an 8 nm node in their commercial roadmap.
I asked ChatGPT about possible reasons for this reference, including whether it might suggest GF is developing a new 8 nm node. The response was that it’s "almost certainly not a new, official node". Instead, the more likely explanation is that the GF process in question is being described as “8 nm-class” because its performance is comparable to similar offerings elsewhere (e.g. Samsung’s 8LPP).
Perhaps someone with deeper technical expertise might be able to weigh in with their perspective.
ChatGPT said:
1. GlobalFoundries and “8 nm”
- Officially, GlobalFoundries exited the race below 12 nm back in 2018 — shelving their 7 nm program. Their most advanced public nodes are 12LP+ FinFET and 22FDX / 12FDX FD-SOI. They do not list an 8 nm node in their commercial roadmap.
- So if you hear “GF 8 nm” in a BrainChip webinar, it’s not part of GF’s public marketing portfolio like TSMC’s N7 or Samsung’s 8LPP.
2. Possible explanations for the “8 nm” reference
Here are the most plausible interpretations:
① Internal / custom shrink naming
- GF sometimes uses marketing-driven “node naming” for customer-specific process tweaks. For example, an optimized 12 nm platform could be referred to by a design partner as “8 nm” in terms of transistor performance or density equivalence.
- It wouldn’t be an official, standalone node like Samsung’s 8LPP, but more like a derivative or a customer-specific variant.
② Foundry design enablement context
- BrainChip, when advising customers through Akida Cloud, may present EDA/PDK libraries mapped against equivalent nodes across multiple fabs (TSMC, Samsung, GF).
- In this sense, “GF 8 nm” may appear in toolchains as a proxy node so customers can model performance/power targets — even if GF itself brands it differently.
③ Miscommunication in the webinar
- It’s also possible the speaker loosely referred to “8 nm” because the GF process in question is comparable to an 8 nm class offering elsewhere (e.g., Samsung’s 8LPP), even if GF calls it 12LP+.
3. Is GF working on a “new” 8 nm?
- Highly unlikely. GF’s strategy is explicitly focused on specialty technologies — FD-SOI (22FDX/12FDX), RF, power, automotive — not pushing smaller geometry race.
- There have been no public announcements from GF about restarting a true 8 nm node or any new sub-10 nm process. Their official roadmap still stops at 12 nm.
- If anything, what BrainChip referenced is more about design enablement and compatibility mapping for customers (i.e., “yes, your Akida-based model can target GF at an 8 nm equivalent”) rather than GF announcing a brand-new leading-edge node.
Conclusion
The “GF 8 nm” mentioned in BrainChip’s webinar is almost certainly
not a new, official node from GlobalFoundries.
It’s best interpreted as either:
- A design library equivalence (Akida Cloud showing customers how their models port to “8 nm class” at GF), or
- A marketing shorthand for an optimized variant of GF’s 12 nm technology, positioned to look competitive with 8 nm offerings elsewhere.
BrainChip’s angle here is:
“We can show you how Akida will behave if manufactured on GF’s advanced nodes — whether you call it 12 nm, 8 nm, or otherwise.”
Would you like me to pull together a
GF vs. Samsung vs. TSMC “8 nm class” comparison, so you can see whether GF’s 12LP+/FDX really does overlap with what others call 8 nm? That might explain why BrainChip uses that label.
You said:
Yes please>
ChatGPT said:
Perfect — let’s line up
GlobalFoundries,
Samsung, and
TSMC to see what “8 nm class” really means, and why BrainChip might reference GF as having one.
Comparative View: “8 nm class” process nodes
| Foundry | Official Node | Technology Type | Density / Performance Class | Positioning / Notes |
|---|
| Samsung | 8LPP (8 nm Low Power Plus) | FinFET (derivative of 10 nm) | ~10% area reduction vs 10LPP; ~10% power savings | Marketed as a “cost-effective bridge” between 10 nm and 7 nm. Used in some mobile SoCs & GPUs. |
| TSMC | (No 8 nm branded node) | Jumped from 10 nm → 7 nm | N7 offered ~35–40% density gain over N10 | TSMC skipped “8 nm” naming entirely. Their N7 competed directly with Samsung’s 8LPP. |
| GlobalFoundries | 12LP / 12LP+ FinFET (sometimes internally positioned as “8 nm class”) | FinFET (evolution of 14LPP, licensed from Samsung originally) | GF claims performance/power close to Samsung 10/8 nm nodes | Not officially branded as “8 nm,” but some design toolchains (and BrainChip’s webinar) may shorthand it as such to indicate equivalent PPA (Power-Performance-Area) class. |
| GlobalFoundries (FD-SOI path) | 22FDX / 12FDX | FD-SOI (non-FinFET) | Optimized for low-power, RF, adaptive body-bias | Competes in niches (IoT, automotive, 5G RF) rather than density-race. |
Why BrainChip may say “GF 8 nm”
- Performance equivalence: GF’s 12LP+ FinFET can offer transistor performance in the same ballpark as Samsung’s 8LPP.
- EDA/library mapping: When BrainChip shows customers how Akida would map to foundries, they may call GF’s 12LP+ “8 nm class” to align with competitor terminology.
- Marketing shorthand: To customers, “8 nm” signals modern, advanced, and competitive, even if GF’s official brand is “12LP+.”
Bottom Line
- Samsung: The only foundry with a real “8 nm” branded node (8LPP).
- TSMC: Skipped 8 nm — went straight from 10 nm → 7 nm.
- GF: No official 8 nm, but its 12LP+ FinFET is often described as “8 nm-class” in equivalence charts. This is almost certainly what BrainChip was referencing in the webinar.
So — GF hasn’t secretly launched a brand-new 8 nm process. Instead, their
12 nm FinFET technology is close enough in characteristics that it can be slotted into the “8 nm bucket” when comparing foundries.
www.justsecurity.org
