What kind of ASIC designer is better off, frontend or backend?

[steven852]

I bring up this question just for curiosity when browsing the post "does ASIC designers require knowledge of HDL?". So my question is: if you have option to start from frontend, which means algorithm development, RTL coding, and handing out netlist versus backend, which equals to physical layout and landing out GDSII, etc, which way is the best way to become a skilled ASIC designer? I assume that both frontend and backend are available options for a new comer with reasonable background on both sides.

In my own opinion, frontend requies more knowledge and should be a better starting point. This is especially true now since the EDA tools become very powerful. However, the algorithm hardware implementation will (or have) become major barrier in ASIC designs. On the other hand, backend has become more routine (no offensive meanings here). I think backend takes less time to learn in this regards. There is little doubt that have strong background on both of them would assist others. But if any personal suggestions on this development path can be provided, that will be a great help.

I also wonder if there is any clear difference in terms of job security and pay between these two.

Thanks

 

[stevepre]

frontend vs backend semiconductor

i'd recommend you go with frontend first. the reason is, once you get into the backend world, it's hard to get out. No company is going to hire a backend guy to do the frontend design. But uh...if you are a talented frontend designer and you always have the chance to learn backend stuff when you are free.

 

[omara007]

90nm backend

Good words guys .. I completely do agree with you .. I do see the front-end a better starting point .. though that backend is highly demanded in service companies .. yet, front-end is more into science , while back-end is more into tools.

I need just to give a small comment .. ASIC designer is vague title .. so, don't think that the ASIC designer should start as frontend .. or as a backend .. the ASIC designer is either a front-end or a back-end engineer .. if you can be both, it's ok for u .. but usually they are TWO.

 

[neo_chip]

front-end vs back-end

This discussion really gives me confidence......
I always wanted to be in backend........... but now Iam working as a verification engineer.
Whenever I come across openings in backend, they ask for tapeout experince..........
If this is the case how a guy can move frm frontend to backend

 

[wangkl]

learn backend

you always have the chance to learn backend stuff when you are free.

I like these words : )

 

[cdic]

difference between frontend and backend languages

That is a very good topic. My opinion is that, once you choose one field, it's hard for you to change again, each field contains huge stuff.
in general, up to 0.13um, the back-end is more routine job, but under 90nm, so many stuff pop up, the requirement for back-end designer is much more chanllenging, and usurally it's crossing different domains, such as timing(front-end), test(dft), yield(operation,foundry), power(application), packaging(application), it's not only about tool any more, the enormous knowledge required. for some of the very large, complex 90nm design, 80% yield is very good number, compare to 95% and above yiled in 0.15,0.25um process, you can image what is going on behind the numbers.
As to front-end, it's different, in terms of design methodology, nothing change from 0.35u to 0.09u, it all depends on how good the designer's knowledge to the design, how good the designers at HDL, actually front-end contains two categories, 1. high level: architecture, 2. low level: RTL designer. The first category is more valuable, much higher requirement compare to second category, as to 2nd category, it's implementation job like back-end, just implement to the RTL, maybe one engineer can make it with 1k gate, the other one make it with 2kgate and extra 10days, that is the difference. no matter how good you are, how strong you are, you'll alway face chanllenges from young people, HDL is just another language, like C, just different way to think, tools are so cheap, any new graduate can chanllege you in very short time, so you need to move to the first category.
IMHO, it doesn't matter which way you want to go, if you go for back-end, go for 90nm and below, if you go for front-end, go for 1st category, if you can be either one of them, you don't need to worry about your engineering job any more(at least 3-5years), other-wise, just another poor labour, no much value. As to verification job, it's even worse, I don't see any difference between verification engineer and software alpha test engineer, game tester. pure labour except for simple simulator/hdl/design knowledge.

Hopfully I didn't offense so many people, any discussion will be welcome.

Actually, talking about engineering job, except for what I mentioned above, you can also go for yield field, job will be much more secure. Of course, if you can be hands-on manager or higher, that will be the final goal for the engineers if you just want to make more money, earies life, instead of really enjoying the engineering hard working.

cdic

 

[salma ali bakr]

back end vs physical design

hi guys,

really interesting topic indeed
well, I just want to ask about the skills and qualifications a verification engineer should have, and the chances of getting such a job in the mean time? and in which applications mostly

thanks,
Salma

 

[omara007]

what does backend verification mean

This discussion really gives me confidence......
I always wanted to be in backend........... but now Iam working as a verification engineer.
Whenever I come across openings in backend, they ask for tapeout experince..........
If this is the case how a guy can move frm frontend to backend

If you work as a front-end engineer, in a company that continues to do the back-end for the same design, you can always sit with the physical design guys .. at least when they feed you back with the RTL idiot problems generated by yourself

In general, you can add the tape out experience to your history in case your company tapes out .. but if you work for a service or IP company, this won't be feasible unless they silicon verify their designs.

Both front-end and back-end can be done in leisure time .. yet, front-end is more into research , while the back-end is more into tools and scripting skills .. the problem about back-end is that its tool is not pretty much availble at home .. unlike front-end .. you can always find educational versions of them ..

 

[raffusc]

front end and backend eda tools

Hello,

I worked in both "arenas" (FE first then BE). Nowdays it is required that BE has good design background since design requirements are thougher ; BE needs to understand clocking requirement, DDR interfaces, Static Timing Analysis; needs to help FE in timing closure. When I started my career If something worked in simulation without delays it also worked in Silicon; the BE guys were called "polygon pushers"; then when the design had few thousand gates they were called "tool guys". BUT NOW. I have no doubt, BE is much thougher for those reasons:

-1- BE comes later in the design cycle, the RTL guys end up always stealing weeks out of the schedule to the BE guys (RTL freeze slips, but PG tape does not !)
-2- No margin of errors for BE during final timing closure; FE they have ECOs (here are other days "stolen" to BE).....
-3- BE needs to have logic design knowledge to understand FE requirements and "drive" the BE tool to achieve them.
-4- In SOC BE needs to take care of Mixed signal and interface requirements;
FE guys are mostly blissfully unaware of those issues.
-5- BE helps FE in "refining" the constraints; I have never seen FE helping BE during an LVS ...............

In my group FE guys go home at 6 pm, BE guys at 8 pm ........

 

[richardleo]

front end and back end means in hdl

actually front-end contains two categories, 1. high level: architecture, 2. low level: RTL designer. ....As to verification job, it's even worse, I don't see any difference between verification engineer and software alpha test engineer, game tester. pure labour except for simple simulator/hdl/design knowledge.
cdic

Interesting topic.
1. Yes, I saw some very experienced architects, but I don't clearly know how they grow up starting from a common engineer. Maybe some one can give some tips, and do they need to know the algorithm very weill or algorithm engineer in hardware should be another category, in another words just the same as the software guys. I don't think RTL engineer will work on that, who just implement the algorithm.
2. Yes, the verification engineer is similar to software enginner, but are they really so worse? I don't see difference between verif and rtl engineer, for they just do the same thing in different ways some times. Moreover, rtl engineer only need to focus on the rtl code and some sythesis work, but the verif engineer should take care of all other jobs like create/maintain the environment, write scripts in my company. Consequently, the verif engineer always work on the critical path of the schedule...

 

[omara007]

front end guy in backend world

hi guys,

really interesting topic indeed
well, I just want to ask about the skills and qualifications a verification engineer should have, and the chances of getting such a job in the mean time? and in which applications mostly

thanks,
Salma

Verification job become one of the most important jobs, namly when the designs started to be huge.

The verification engineer should at least know one HDL language (like vhdl or verilog) and one HVL language (like C/C++, SystemC, Vera, E).

He should have the testbench skill, and be able to integrate large number of blocks together. Integration is not an easy job.

He also needs to understand the design pretty well, same as the RTL guy. If he doesn't know what's going on inside the design, he won't be able to generate test cases.

One of his jobs, as well, is to run code coverage. Some verification engineers believe that Synthesis/STA/DFT/Formal Verification are part of their job and they have to posses these skills.

Years ago, there wasn't an engineer solely called a verification engineer. RTL engineer used to do the job. Nowadays, the amount of verification work mandated to have a seperate engineer for this job with the evolving of the HVLs.

 

[laglead]

front end vs back end jobs

Hi,
I am a BE. I think the best career development is from BE to FE.
As everyone knows, IC is basically silicon. Every designer should have a profound understanding of semiconductor physic. As the clock is speeded up signal and power integrity becomes critical problem. Of course successful avoidance should require knowledge of electromagnetics. SOC is constructed on the complex network of silicon and metal. Familiar with these backend knowledge is obsolute necessary for a robust SOC artchitecture. But understanding design from FE perspective is a final goal. Moving from BE to FE with deep understanding of BE can only make final success.

 

[omara007]

Re: Frontend vs. Backend

Hi,
I am a BE. I think the best career development is from BE to FE.
As everyone knows, IC is basically silicon. Every designer should have a profound understanding of semiconductor physic. As the clock is speeded up signal and power integrity becomes critical problem. Of course successful avoidance should require knowledge of electromagnetics. SOC is constructed on the complex network of silicon and metal. Familiar with these backend knowledge is obsolute necessary for a robust SOC artchitecture. But understanding design from FE perspective is a final goal. Moving from BE to FE with deep understanding of BE can only make final success.

But don't you think that this is against the design flow nature ??
I mean te design flow itseld starts from FE down to BE .. this means that you can't learn backend unless u have a FE stuff to work on .. so, logic says that u need to know what is in the FE and what does it pass to the BE in order to do the BE .. don't you agree ?:idea:

 

[vijay_nag]

Frontend vs. Backend

i think frontend job is more better since it requires us to know how the hardware works and so we have the liberty of designing it to our needs. whereas backend requires us to know about the tools

 

[steven852]

Re: Frontend vs. Backend

Hi,

I didn't know there were replies on this topic since I posted it. Also, since the period of time, I gained more experience both from BE and FE. From there, opinions from Omara007 (the first post is particularly useful), cdic. Ruffusc pointed out exactly what is going on now in the industry. I thanks these 3 posts have shed light on this topic and very valuable for young ASIC engineers.

Thanks all.

 

[cdic]

Re: Frontend vs. Backend

Talking about High level architector, there are only 3 types:
1. coming from pure high level, building models, C/system C/mathlab level, strong math background, expert in particular field such as CPU, CDMA, vedio compression, this kind of big systems.
2. coming from low level, start from RTL design, with many years experience, understand/expert in many majoy hardware component, such as USB, PCI, MEMORY ....
3. Genis. ( I saw two so far, ee/cs related background).

No matter what kind of job you are working on, always trying to work on higher level, bigger picture, other-wise, labour. You will be easied beat when you are 40s no matter how good you are.

 

[albred]

Re: Frontend vs. Backend

I'm a BE engineer, in what way the knowlege of FE help my job?
If I'm a FE, what the knowlege of BE should I know? Waht's the useful things it offer me?
Maybe I should make a division between Architecter/RTL coder/verificer from FE job. Everyone would say the knowlege in both FE and BE side would helpful for your job, but this conclusion is too simple and not enough.
Now, I think maybe Archit view should be useful for a BE engineer, for example, I can divide power domain according to the system architecture, I may use clock gating or power domain island to implement the design. On the other hand, if I'm a
architecture designer, I know our BE team has the clock gate or power island techonology, so I will design a effective power consumption structure for the design. If I'm a RTL corder, I may take a tension to the clock structure for clock gating.
What I need is the material examples of relationship between FE and BE, the more the better.

 

[laglead]

Re: Frontend vs. Backend

Hi,
I am a BE. I think the best career development is from BE to FE.
As everyone knows, IC is basically silicon. Every designer should have a profound understanding of semiconductor physic. As the clock is speeded up signal and power integrity becomes critical problem. Of course successful avoidance should require knowledge of electromagnetics. SOC is constructed on the complex network of silicon and metal. Familiar with these backend knowledge is obsolute necessary for a robust SOC artchitecture. But understanding design from FE perspective is a final goal. Moving from BE to FE with deep understanding of BE can only make final success.

But don't you think that this is against the design flow nature ??
I mean te design flow itseld starts from FE down to BE .. this means that you can't learn backend unless u have a FE stuff to work on .. so, logic says that u need to know what is in the FE and what does it pass to the BE in order to do the BE .. don't you agree ?:idea:

I am not specialized in FE area, so I cannot provide specific examples.
But my opinion is that if u find problem at the last stage, e.g. power integrity or signal integrity, then go back RTL or even system level to improve the design ,then much time and resouces are wasted. Why not keep those critical points in mind at the very beginning of the flow? For example, some noise coupling knowledge will help u to realize it is advatageous to make block partitions---many high speed subblocks should not be grouped togather. These back-end insight can help one to optimized design to some extent and reduce the probability of re-work.

 

[shiv_emf]

Frontend vs. Backend

excellent discussion here!!!

what are prospectives for BE and FE engineers in future for research work!!

 

[rajkishore12345]

Re: Frontend vs. Backend

Front end and back end have their own difficulties.There is no problem in starting a study from back end