[Anandtech]: GlobalFoundries Stops All 7nm Development !!

[moonbogg]

So if I live an average lifespan I'll get to see them hit physical limits and have to innovate in some interesting new way. Cool.

 

[dark zero]

I wonder if GloFo will catch TSMC, Samsung and Intel up with that desicion... since AMD would simply switch to Samsung or TSMC and they won't have any client to attend.

 

[NostaSeronx]

GloFo will probably catch up once they shift to FDSOI. Basically, majority of the 300mm foundries from GloFo can integrate 22FDX today.

There is already RF design wins for 22FDX.
https://www.imgtec.com/communications/connectivity/gf22-bluetooth/
https://www.imgtec.com/communications/connectivity/ensigma-gf22fdx/

From TSMC's 40nm:
https://www.imgtec.com/communications/connectivity/bluetooth-hard-macro/
https://www.imgtec.com/communications/connectivity/ensigma-tsmc-40nm/

STMicroelectronics also detailed that they aren't going GlobalFoundries 7nm:

TSMC 16FF => Q1 2018
GF 22FD => Q4 2018
TSMC 7FF => Q1 2019
This roadmap doesn't detail anything that goes through 2020. So, the boxes that do go beyond 2020 have been removed. So, 12FDX can be there but not in this roadmap. Since, it is in another STM slide.

Especially, with the whole OCEAN12 project: https://cordis.europa.eu/project/rcn/216110_en.html
https://www.ecsel.eu/projects/ocean12

 

[amd6502]

22FDX looks the most promising. Ultra simple, cheap, and perfect for small die products, and pretty much available now.

It would be a small bet. 12FDX seems to have potential but would be a bigger and riskier bet. Better to test the waters with a small bet and learner project on what's available now.

The size limitation would seem to be a big limitation for GPUs and also high core counts in CPUs.

Let's say what the CTO of GlobalFoundries had to say about the subject:

Garry Patton said:
You can get performance when you need it, so on our 22FDX you can easily get 14nm FinFET performance. You can’t get it all the time, but when you need it. With our 12FDX, you can get close to 7nm performance when you need it.

When I say that, I should qualify: you are not going to get that from a big chip. If you are making a big chip, you got tons of wire capacitance, so you are not going to beat a FinFET design. FinFET has huge drive current, so if you're dealing with smaller chips, wiring capacitance isn't as a big, but gate capacitance of the transistor is more important. FD-SOI has less gate capacitance than FinFET because you don't have the gate wrapping around the whole device. So it is ideally suited for smaller chips, I say less than 150 square millimeters.​

So this means low to mid mainstream APUs. It looks like the higher end stuff and large GPUs will remain fin fet.

 

[DrMrLordX]

As an end-user, this news means I do not have to watch AMD tie their launch schedule for 7nm products to GF's success or failure on future nodes.

FWIW, I think GF has executed fairly well since taking on the former IBM staff: they improved 32nm significantly (making the 8320e and 8370e possible, at a minimum), and they did manage to adapt Samsung's 14LPP to their needs without major hiccups. 12nm also went without a hitch. They kept AMD mostly on-schedule for Ryzen and completely on-schedule for Ryzen 2. I honestly was not THAT worried about GF flubbing 7LP so badly that we would see launch delays for Matisse/Ryzen 3. Now I don't even have to think about it. Since Rome was already taping out at TSMC anyway, it should make it fairly easy for AMD to let TSMC handle Matisse as well.

If GF wants out of the leading-edge game, then so be it. No skin off my nose.

And the likelihood of desktop/laptop CPUs or APUs using FDX nodes seems very, very low.

 

[AkulaMD]

Since Rome was already taping out at TSMC anyway, it should make it fairly easy for AMD to let TSMC handle Matisse as well.

Heard somewhere that Rome was already sampling, I maybe wrong though. At least as an end user I don't have to worry that much about AMD being able to keep their plan pretty much on time.

 

[iBoMbY]

Just a reminder regarding GloFo not having enough 7nm demand:

The company’s chief technologist, Gary Patton, told EET’s Rick Merritt that AMD “will have more demand than we have capacity

https://www.extremetech.com/computi...7nm-capacity-than-globalfoundries-can-provide

...

Edit: Well, technically this would of course also still be true if the capacity is zero.

 

[LightningZ71]

So, in the above interview, Garry gives a rough cap on SOI chips at 150mm^2. Does that statement just apply to 22FDX, 22FDX and 12FDX, or all SOI chips in general? The utility of a design at their conjectural 7FDX node at 150mm^2 is MUCH higher than at 22FDX as they can easily stick 8X as many circuits in the same space. I suspect that, if GloFo did manage to get a 7FDX out the door, it could be quite a product for its targeted markets and still produce competent mobile APUs for AMD, that is assuming that GloFo actually does that node.

 

[NostaSeronx]

So, in the above interview, Garry gives a rough cap on SOI chips at 150mm^2. Does that statement just apply to 22FDX, 22FDX and 12FDX, or all SOI chips in general?

Drive current is about high frequency design.
Wire capacitance is about wire length.

>150 mm squared is affected by "longer" wires which cause "more" wire capacitance on the FDSOI nodes.
High frequency design is affected by the "weaker" drive current on the FDSOI nodes.

Gary Patton needs to have a refresher course done at Dresden. He has been at Malta and Albany too long. He is one of the people sandbagging the FDX nodes.

Is he somehow shocked that the 5nm nanosheet node also has lower drive current. With the 5nm node you have more wires closer together and more capacitance! Has he ignored the novel MOL/BEOL structures that 22FDX/12FDX use which lower wire capacitance. While, in the macro device NoCs are widely used to reduce wire length between micro-devices(CPU, GPU, DSP, etc). Drive current can be solved with FIVR and mVRM. All of it points to bad marketing rather than being fact.

1. 22FDX can have higher frequency than FinFETs in low power and high performance.
2. 22FDX can have larger die sizes while consuming less power in high Vdd through low Vdd scenarios.

TSMC's 20nm being planar also has some of the FDSOI issues and more.
Oracle's 20nm processors M7/M8 are going strong!

22FDX design cost => $50 mil peak
12FDX design cost => $100 mil peak
7LP design cost => >$200 mil peak.

Double the 22FDX or 12FDX cost, I am pretty sure wire capacitance and drive current won't be an issue. These are literally issues that throwing money at can solve or design around.

22FDX beyond 1.2 Vdd and below 0.8Vdd is higher everything than 14LPP/12LP and 7LP.

 

[PeterScott]

So, in the above interview, Garry gives a rough cap on SOI chips at 150mm^2. Does that statement just apply to 22FDX, 22FDX and 12FDX, or all SOI chips in general? The utility of a design at their conjectural 7FDX node at 150mm^2 is MUCH higher than at 22FDX as they can easily stick 8X as many circuits in the same space. I suspect that, if GloFo did manage to get a 7FDX out the door, it could be quite a product for its targeted markets and still produce competent mobile APUs for AMD, that is assuming that GloFo actually does that node.

7FDX at GF is nothing but fantasy. It has similar features sizes to 7nm FinFet. If GF can't afford equipment investment in 7nm FinFet, then they can't afford the investment in 7FDX.

Think about what GF-CTO said in this interview:
https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries

We have been doing work on 12FDX here in NY for over a year. We have working devices and they’re not far off the performance target so, and we know how to do 12nm technology obviously since we're doing 12LPP here. We expect to be taking risk production on the parts early next year (2019), so we are pretty far along with the technology.

12FDX is based off having 12nm expertise in the first place. There is a reason FDX lags FinFet by 2 years.

If you don't have 7nm FinFet, you aren't going to have 7FDX.

GF has thrown in the towel on leading edge Semiconductor fabrication. The will soon be as irrelevant to PC Enthusiasts as UMC, SMIC, or Chartered Semi (before GF bought them).

They may have a strong business supplying a multitude of industry semiconductors in the future, but they won't have any more new CPU and/or GPU designs that PC enthusiasts pay attention to.

 

[french toast]

Globalfoundries has clearly stated that they are not producing or researching leading edge nodes, that means no 7nm/5nm/3nm finfet obviously, but also means by extension no 7FDX/FDSOI.

This is a disaster for them, they had arguably the best process of that node 'era', experience that could also be used to push 7nmFDX in future, giving them a differentiator and a competitive advantage, to pull out funding at this point seems rather dumb if you ask me, even if they skipped 5nm EUV+ nanosheet generation cut costs and instead heavily developed EUV 7nm ++ process for cost effective design, (ala intel 14++++) alongside 7nm FDX...this would be a good strategy until 2022, when they then could introduce 3nm...once EUV has matured and they have increased capacity.

Really disappointed in this, I could see them packing it in after 14XM went bust, that would have made sense at least, but now? When they have good technology and have been meeting production targets and executing well?....seems dumb to me.
Wouldn't be surprised if Apple bought this up on the cheap, they have the resources to make this work for themselves.

On the other hand, Apple might pull an Imagination tech 'homebrew' move and ...out of the blue ...an identicle Global foundries Dresden fab clone pops up in California one infinate loop.

 

[NostaSeronx]

@PeterScott
L1: GlobalFoundries afforded equipment for 7nm FinFET, they didn't have customers or demand to immediately get profit.
-> 22FDX $2billion in revenue. Where is the 7LP $xbillion in revenue.
-> GLOBALFOUNDRIES (GF) announced FDSOI design wins worth $2 billion dollars in revenue with $1 billion dollars booked in 2017 and another $1 billion dollars in revenue booked in the first half of 2018.
1. https://www.semiwiki.com/forum/content/7604-fdsoi-status-roadmap.html
(Based on the above, 22FDX is speeding up! So, $1 bil in Q3, $1 bil in October and November, $1 bil in December, $1 bil in the first two weeks of January, $1 bil in the third week of January, $1 bil in the first half of fourth week of January. I'm joking.)

L2/L3: It went 14LPP to 12FDX to 12LP.
2. https://www.globalfoundries.com/new...-extends-fdxtm-roadmap-12nm-fd-soi-technology
Sep 08, 2016
3. https://www.globalfoundries.com/new...-technology-for-high-performance-applications
Sep 20, 2017

12LP utilizes things learned from the 22FDX node. CNRX, T/L-structures, etc.

L4: GlobalFoundries had a 7LP node and cancelled production of said 7nm FinFET, PDK was final. They didn't say they are decreasing FDX investment, GloFo stated they are increasing FDX investment.

https://www.youtube.com/watch?v=Zbz0R_yFFrQ
Only 4 customers developing 7nm. 22FDX has 75 ecosystem partners. 22FDX manufacturing for a decade. You can tell the only thing going for the 7nm FinFET is the Logic on SRAM or SRAM on Logic, 3D device. Doesn't list 22FDX-UHP, hmm I wonder what he might be hiding.

@french toast
GlobalFoundries is leaving the SUNY FinFET/Nanosheet research. They said nothing about leaving the CEA Tech/Dresden/STM/SOITEC FDSOI research. Until stated otherwise by GlobalFoundries the roadmap for FDX continues beyond FinFETs. Research of FDSOI is done in France to Germany, not at Albany/Malta. The only research on that side is post-development and validation.

https://www.globalfoundries.com/new...nter-long-term-supply-agreement-fd-soi-wafers
GlobalFoundries is required to have FDSOI product for the 5 years after that WSA. While, also participating in next-gen FDSOI research for the same amount of years. Being SOITEC's #1 FD-SOI customer comes with advantages. Like significantly lower SOI wafer costs, which can go directly to the consumer and more.
Sep 20, 2017 + 5 => 2022

 

[PeterScott]

This is a disaster for them, they had arguably the best process of that node 'era', experience that could also be used to push 7nmFDX in future, giving them a differentiator and a competitive advantage, to pull out funding at this point seems rather dumb if you ask me, even if they skipped 5nm EUV+ nanosheet generation cut costs and instead heavily developed EUV 7nm ++ process for cost effective design, (ala intel 14++++) alongside 7nm FDX...this would be a good strategy until 2022, when they then could introduce 3nm...once EUV has matured and they have increased capacity.

A disaster in the eyes of PC enthusiasts, since they are out of the picture. But for GF it may finally be a chance to actually make money.

Given the nearly exponential rise in the cost of developing new process nodes, it was pretty much inevitable that they would hit this wall at some point before the bigger fabs ahead of it (Intel,Samsung, TSMC). It's too bad they didn't decide this much earlier, they could have saved a lot of money spent on 7nm development.

 

[Topweasel]

Personally I see 7nm anything as being useless to them. They aren't making anyone's ARM chips. They lost competitive CPU and GPU manufacturing. AMD isn't going to spend the money on new work to be done at GF any time soon. Their best bet is going to be making smallish chips that don't really need to be smaller at competitive prices and just stagnate at 12nm till they figure out how they want to course correct.

 

[NostaSeronx]

2015~2016: GlobalFoundries announces 7nm FinFETs + rumors that 7nm wasn't from Samsung.
Of the customers going to 7LP: AMD, IBM(Extended dev to 10HP(7LP on SOI)), HiSilicon, MediaTek, and Qualcomm.
Of the above, only IBM was going to be exclusive to GlobalFoundries. AMD, HS, MTK, QC were all second sourcing to TSMC. When I say second sourcing I mean TSMC first, GlobalFoundries last.

Meanwhile, GlobalFoundries customers on 40nm and 28nm. Didn't have the wealth to jump on 14LPP/12LP. While, also 14LPP/12LP weren't successors to GF's 28 enhanced SLP:
28SLP – Low Complexity GHz Performance
- ULP option in SLP with Vdd of 0.8V results in additional 45% power reduction
- Additional devices for ultra-low static leakage power
- Integrated RF and analog to reduce system cost, power and time to market
- Wide selection of foundation IP optimized for performance or power

These customers that didn't go 14LPE/14LPP at GlobalFoundries in 2015 looked at 22FDX. Which launched with the successors to above 28SLP. Also 28HPP/28SHP(with 22FDX-UHP).
- 22FDX-ULP
- 22FDX-ULL
- 22FDX-RFA
- FDXcelerator

They also wanted a roadmap which they got with the 12FDX node.

2017~2018: GlobalFoundries started getting revenue from 22FDX faster than 14LPP in the same timeframe.
More differentiated options:
- MRAM
- FeRAM
- PCM, MeRAM (STMicroelectronics)
- CTT (Charge Trap Transistor)

22FDX risk production began Q1 2017 with the 2016 options.
22FDX risk production began Q1 2018 with the 2017 options.
If the assumption holds: 22FDX+ with the option UHP and the improved drive current transistor(22FDX+ not UHP) will begin Q1 2019.

//
Back to the roadmap:
14LPP -> 12LP -> 7LP
Would not have made revenue to produce 5LP or 3LP.

40/28-nm -> 22FDX -> 12FDX
Surprising amount of support. Initially, half of GlobalFoundries customers with more customers hoping on the FDSOI train. With a clear preference over Samsung's FDSOI.

To support the FinFET roadmap, GloFo would need to leach from the FDSOI revenues.

Before the decision: R&D support(most to least)
1. FinFET
2. FDSOI
3. RF-SOI
4. Other technologies

After the decision: R&D support(most to least)
1. FDSOI
2. RF-SOI
3. Other
4. FinFETs

What I have researched is that after 14LPP/12LP gets: -MRAM -ULL -ULP -RFA. Those workers are going to be laid off or switched to the FDSOI roadmap. So, effectively FinFETs are done after those options are provided for.

Other is photonics, enhanced BEOL(Doped CNT/Composite Cu-CNT), Quantum design. Most of the other technologies is for down the road or enhancing/differentiating what is currently available.

 

[dark zero]

7FDX at GF is nothing but fantasy. It has similar features sizes to 7nm FinFet. If GF can't afford equipment investment in 7nm FinFet, then they can't afford the investment in 7FDX.

Think about what GF-CTO said in this interview:
https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries


12FDX is based off having 12nm expertise in the first place. There is a reason FDX lags FinFet by 2 years.

If you don't have 7nm FinFet, you aren't going to have 7FDX.

GF has thrown in the towel on leading edge Semiconductor fabrication. The will soon be as irrelevant to PC Enthusiasts as UMC, SMIC, or Chartered Semi (before GF bought them).

They may have a strong business supplying a multitude of industry semiconductors in the future, but they won't have any more new CPU and/or GPU designs that PC enthusiasts pay attention to.

They are on bad shape... they even lost that strong bussiness... all AMD chips are shifting to Samsung or TSMC.
And after that, they won't have any products to show to us.
GloFo is turning to be a stinking dead company that started to do bad moves all over again.

 

[piesquared]

They are on bad shape... they even lost that strong bussiness... all AMD chips are shifting to Samsung or TSMC.
And after that, they won't have any products to show to us.
GloFo is turning to be a stinking dead company that started to do bad moves all over again.

Why do you think that the sign of success can only be determined by if they have a product to show you? You don't see most of the products coming from other fabs yet they can survive. If they are investing heavily in FDSOI and have the best RF and have the number of interested customers that they appear to have I bet they'll do quite well.

 

[AtenRa]

7FDX at GF is nothing but fantasy. It has similar features sizes to 7nm FinFet. If GF can't afford equipment investment in 7nm FinFet, then they can't afford the investment in 7FDX.

Think about what GF-CTO said in this interview:
https://www.anandtech.com/show/1243...ew-with-dr-gary-patton-cto-of-globalfoundries


12FDX is based off having 12nm expertise in the first place. There is a reason FDX lags FinFet by 2 years.

If you don't have 7nm FinFet, you aren't going to have 7FDX.

Im not saying here that GF will create a 7FDX but since FDX is planar, what makes you think they need 7nm FinFet to create 7nm FDX ???

Also, it seems that FD-SOI is more suited for IoT, Automotive and RF. GF has a long time expertise with SOI and they seems to think that FD-SOI will have a big TAM the coming years. Development cost seems to be smaller than FinFets and number of customers to become larger for FDX than FinFets.
I wouldnt count out Mobile Companies to produce Mobile SoCs at 12FDX in 2020 perhaps.

 

[NostaSeronx]

Im not saying here that GF will create a 7FDX but since FDX is planar, what makes you think they need 7nm FinFet to create 7nm FDX ???

Samsung explains it a bit better:

Essentially, the FinFET back-end will be reused for the next FDSOI node. Basically, Samsung could do (mature back-end from):
28FDS -> 18FDS(14LPP) -> 15FDS(11LPP) -> 11FDS(7LPP) -> 8FDS(4GAA)

This is to make the FDSOI node cheaper than the FinFET node at initial consumption. As FDSOI will always be cheaper than FinFET @ same node.

 

[iBoMbY]

all AMD chips are shifting to Samsung or TSMC.

Yeah, no:

We also continue to have a broad partnership with GLOBALFOUNDRIES spanning multiple process nodes and technologies. We will leverage the additional investments GLOBALFOUNDRIES is making in their robust 14nm and 12nm technologies at their New York fab to support the ongoing ramp of our AMD Ryzen, AMD Radeon and AMD EPYC processors.

https://community.amd.com/community...mance-leadership-with-focused-7nm-development

 

[NTMBK]

Yeah, no:



https://community.amd.com/community...mance-leadership-with-focused-7nm-development

I think he meant all post-12nm chips.

 

[iBoMbY]

Yes, new dies using more advanced nodes, although I think it is not unlikely they will have to keep doing stuff in 12nm, or whatever, at GloFo until the WSA runs out in 2024, or they have to make a huge one time payment to get free early.

 

[PeterScott]

Yes, new dies using more advanced nodes, although I think it is not unlikely they will have to keep doing stuff in 12nm, or whatever, at GloFo until the WSA runs out in 2024, or they have to make a huge one time payment to get free early.

They aren't going to forced to hold back anything, stuck on 12nm. AMD was clear, they are all in on 7nm. All their new designs are 7nm.

They will keep producing >12nm stuff with GF as long as they have a market need for that stuff, but they aren't going to produce to a quota with no market.

By now all those detail have likely already been worked out behind the scenes between GF/AMD. This is GF's failure to deliver after all.

 

[moinmoin]

AMD promises long-term availability of all their embedded products for up to 10 years. These existing products will all stay at GloFo.

 

[darkswordsman17]

Why do you think that the sign of success can only be determined by if they have a product to show you? You don't see most of the products coming from other fabs yet they can survive. If they are investing heavily in FDSOI and have the best RF and have the number of interested customers that they appear to have I bet they'll do quite well.

Its possible, but its not like no one else is doing that or couldn't do that if demand does in fact grow substantially. Plus those companies will have advanced Finfet processes they can adapt/utilize aspects of to get there, GF no longer will (which means they can pretty quickly be put at a disadvantage in FD-SOI).

This is obviously coming from PC enthusiast perspective (so yes, GF can make money and even do well, it just won't be with chips that leading edge computing people get terribly hyped about - that doesn't mean they won't benefit from them, as more efficient wireless chips will help - for VR to take off I think it'll need to be wireless and that means smaller and more efficient wireless chips). I have my doubts that we'll be seeing high performance (by that I mean at the time, not previous high end) ARM designs at GF either. I would actually like to be proven wrong (and also for GF to prove they can make a competitive process).

And I'm full on not buying Nosta's hype because he's been spouting crazy nonsense like this for years and it has never materialized like he said (at this point its almost total opposite of what he says, or its years behind schedule). He just doesn't want to accept that. He refuses to accept that GF's FDX has limitations that means only certain chips will perform really well on them (which does not include his wonder CPU that he's been posting for years about as well), and that he thinks GF is actually becoming true leading edge despite them outright saying they're intentionally not doing that any more. My issue there is him intentionally trying to distort facts to support his beliefs.

Im not saying here that GF will create a 7FDX but since FDX is planar, what makes you think they need 7nm FinFet to create 7nm FDX ???

Also, it seems that FD-SOI is more suited for IoT, Automotive and RF. GF has a long time expertise with SOI and they seems to think that FD-SOI will have a big TAM the coming years. Development cost seems to be smaller than FinFets and number of customers to become larger for FDX than FinFets.
I wouldnt count out Mobile Companies to produce Mobile SoCs at 12FDX in 2020 perhaps.

Yeah, have fun with that when its supposed to be comparable to 7nm FF, which companies are making those types of SoCs on that process this year. If that doesn't key you off to the fact that its not going to be leading edge chips using GF, I don't know what will. And that's if GF actually delivers on 12FDX in 2020, which I have my doubts on to be frank. GF has to first prove their process, and then woo companies. The thing is, those companies aren't going to just wait around, so that means they have to decide to design around finfet or planar. I don't know, maybe its possible for them to design it for finfet and then alter it fairly easily for planar should that prove worthwhile. But if they do wait around, their competition is going to already be producing chips (selling actual products), and working on their next gen stuff (on potentially better process, its not like FF development will be standing still).

And let's not forget, GF has had major problems with execution. So, great, their 22FDX is finally decent. Yay, I guess. In the meantime we've had Nosta telling us that it was going to come raging out and squash 16/14FF for years, and the industry is moving on to 7FF now. Its here now, not "on the roadmap".

Samsung explains it a bit better:

Essentially, the FinFET back-end will be reused for the next FDSOI node. Basically, Samsung could do (mature back-end from):
28FDS -> 18FDS(14LPP) -> 15FDS(11LPP) -> 11FDS(7LPP) -> 8FDS(4GAA)

This is to make the FDSOI node cheaper than the FinFET node at initial consumption. As FDSOI will always be cheaper than FinFET @ same node.

Er, I'm not sure what you're saying. Are you saying FDSOI, when it initially starts, to be cheaper (in which case your claim about it always being cheaper wouldn't be true if they weren't utilizing the same node FinFET back end), or are you saying for FDX to be cheaper when both are starting out it needs the FF backend development (which, well no duh that'd be cheaper, and the reason it'd "always be cheaper" is because its behind and using the FF backend to get there not because its superior absolutely). Plus since GF is giving up having that matured development post 12nm, that means their next FDX will be going up against a mature FinFET from a company doing much higher volume production. Sounds to me like there's a very good chance that 7nm FDX at GF might not be cheaper than 7FF for 5+ years. Meanwhile companies can be producing at TSMC for years before they'd even have the option to consider GF FDX at similar node. That would be a losing proposition for GF I'd think. Heck, even for companies where FDX at the previous node can be equal to FF at the new node, that still means years of delay between when they can produce on the FDX previous node (and that's at "FD-SOI leading edge GF") for 7 FF vs 12FDX.

The thing is, it won't be going up against initial FinFET, it'll be going up against FinFET that is already up and running and tweaked even, for probably a good 2-3 years.

Also remember, this is a situation where other factors come in (so sure if its the same company offering both, FDSOI should be cheaper, but when competing against another company that has their process up and running for years while you're just now trying to get into volume production without the benefit of the FF developed back end, that very possibly means your FDSOI is not cheaper than their FinFET, and performance benefits could very quickly go out the window depending on your chip).