數字ASIC設計的職位要比ASIC驗證的職位好么?


來源:Is a career in Digital ASIC design better than asic verification?

Is a career in Digital ASIC design better than asic verification?

數字ASIC設計的職位要比ASIC驗證的職位好么?

As a new grad without any industry experience, which would be a better career choice in terms of growth, learning, compensation, work-pressure and future opportunities? I want to know the pros and cons of both fields.

作為沒有任何行業經驗的的新畢業大學生,從發展,學習,薪酬,工作壓力和未來機會等方面來看,哪個職位算是更好的選擇呢?我想知道這兩個領域的優點和缺點。

I have heard that a verification engineer is a generalist and hence in more demand with faster growth, whereas, a design engineer"s scope gets narrowed and growth is slow. Please give your perspective on this.

我聽說,驗證工程師需要的是通才,因此會有更快的發展。然而,設計工程師的視野則被縮小,發展緩慢。請提供你對此的觀點。

Akshay Gopalan, Always the hardware guy.

30 upvotes by Quora User, Rishabh Gupta, Anonymous, (more)

Good answers so far. I want to answer this from the perspective of a new grad who juggled with the same decision 6 months ago.

到現在為止的回答都非常好。我也是剛畢業的學生,6個月前被同樣的選擇做折磨。希望來自這個角度的觀點能幫助你。

In Silicon Valley it appears that on the surface ASIC verification is more in demand as well. I applied for ASIC design jobs and was offered an incredible opportunity to "try out" verification in a job environment before deciding for sure.

在矽谷,表面上看來,ASIC驗證的需求依然很大。我申請了ASIC設計的職位,並被offer了一個令人難以置信的機會,「try out」驗證職位的工作環境,在我作出決定之前。(你好牛~居然同時拿到了兩個職位的offer

I ended up picking verification. But that"s not the point of this question. The point of mentioning it was to provide perspective of how I judged the two. I believe that there are no universal pros and cons of either only pros and cons for each individual, and so I will just list the characteristics that helped me decide:

我最終選擇了驗證的職位。但這是不是問題的關鍵。我提及這段經歷,是想說明我從何角度來看待這兩種職位。我不相信普適的優點和缺點,對每一個人而言,優缺點都是不同的,我僅僅列出幫助我作出選擇的原因:

  • Design/product life cycle: Design roles are typically earlier in the cycle of the product than verification. Of course as the project continues these overlap quite a bit, especially when verif finds bugs or reasons why features need to change. Typically by the middle of a project Design and Verif end up being a back and forth see saw game. Which end of the product cycle do you prefer? Do you want to be closer to product launch, or product conception?
  • 設計/產品生命周期:設計職位通常比驗證職位處於產品更早的開發階段。當然隨著項目的進行,這兩種職位的重疊會更多,特別是當驗證端發現bugs或者功能需要改進的原因。通常在項目中期,設計和驗證就像終成為一個來回蹺蹺板遊戲。到底哪個產品周期你更喜歡?你想更接近產品的launch階段,或產品conception階段?(譯者:我的理解是,conception是初期的概念設計,launch是最終的產品定型,設計和驗證都會經歷整個產品周期,但設計偏早,驗證偏晚
  • Pure Hardware vs Hardware/Software: ASIC Design ends up being 95% hardware related. You work in VHDL/Verilog and the closest you get to software is EDA tools or Perl scripting. Maybe if you"re really ambitious you will write your own software drivers for something but in a larger company setting there will be someone whose job it is to do that for you. Power/area matters. Timing matters. It"s easy to come up with a functional design but hard to come up with a design that beats timing, meets area constraints and is scaleable for the next project you work on. ASIC Verification is, at the core of it, solving programming problems. You work in SystemVerilog/ C++/ something similar. (Some designs still use Verilog based testbenches but that is decreasing day by day). The challenge becomes building environments that are reusable, scalable and high coverage achieving. You might have to make use of multi-threading, networking concepts, databases and systems - but all while keeping the hardware aspect in mind.
  • 純硬體vs硬體/軟體:ASIC設計95%最終都和硬體相關。你工作中主要使用VHDL/Verilog,EDA工具或Perl腳本算是你能最接近的軟體。也許,你真的雄心勃勃,會寫自己的軟體驅動程序。但在一個大公司,會有專人的工作就是為你做這些事。Power/area很重要。Timing也很重要。設計一個功能很簡單,但想設計出考慮了timing,滿足area約束,並能夠scaleable到下一代項目的電路,那就非常之難了。ASIC驗證的核心在於,解決編程問題。你工作中主要使用SystemVerilog/C++/類似的東西。(某些設計師仍採用基於Verilog的測試平台,但已經隨時間減少)。我們面臨的挑戰是,構建可重用,可擴展,高coverage的測試環境。你可能需要使用多線程,網路,資料庫和系統的概念,但都同時需要考慮硬體方面。(做驗證,面對越來越複雜的系統,可復用性非常重要,UVM已經快成為行業標準了
  • Specialization: I wanted to specifically addressing your concern about specialization within these fields. Sure, while verif folks are more generalist than design folks, in larger designs this changes a bit. Whenever a project team sits down to divide up tasks, the most productive way to do this is to have people pick up some tasks that they are comfortable and familiar with. I like to take on a couple things I know I can knock out of the park and then a couple things that are new and more challenging that will help me grow. But what always ends up happening is that some tasks are best left to certain people. In general, this shouldn"t be a driving force to make your decision. No matter what industry you work in, your work will be a mix of something new and something you have done before.
  • 專精:我想特別討論下你關心的行業專精問題。當然,儘管verif的工程師比設計工程師更多面手,但從大方面來說,差別很小。每當一個項目團隊坐下來劃分任務,最有效的方式就是讓人們做他們熟悉舒適的任務,我喜歡做一些我知道我能搞定的事情,同時做一些新的,更具挑戰性的任務,這將幫助我成長。但最終總發生的是,有些任務最好留給某些人。通常來說,這不會成為影響你決定的原因。因為無論你工作在什麼行業,你的工作總將是一些新東西和一些你以前做過東西的組合。(譯者:設計比驗證要更專精,也許驗證工作會接觸到更多新的東西,但總體來說,兩種工作都會遇到新的知識

  • Industry growth: This is the only bullet point in which there is a clear winner in favor of verification. As designs become more and more complicated, the demand for verification has increased. Just 10 years ago, the split between ASIC Design and Verification on any given team was 70-30 in favor of designers, whereas now it"s approaching 50-50 if not 45-55 in favor of verification. Think of the Intel bug (Intel"s Billion-Dollar Mistake) that cost them a billion dollars. You don"t need that on your hands as a design team. Especially when you can"t cover losses like Intel can.
  • 行業發展:這是唯一一點,驗證工作成為明顯的贏家。隨著設計變得越來越複雜,對驗證的需求也逐漸增加。就在10年前,ASIC設計和驗證組之間的分工還是有利於設計師的70-30,而現在則是接近到50-50或45-55的有利於驗證工作。想想英特爾的bug(Intel"s Billion-Dollar Mistake),這花了他們十億美元。你不會想讓你的設計團隊發生這樣的情況,尤其是當你不能承擔像英特爾那麼大的損失。

Overall, pick what you"re good at. You"ll grow just fine.

總之,選你擅長的(喜歡的)。你一定沒問題的。

Written 3 Sep, 2013.

Upvote30


if you have passion, either one can be your best career path.

Usually verification resource allocation is about 3 times of design resource allocation in a typical project.

Companies outsource verification job much more than design job.


這個問題猶如小馬過河,不同情況的人看法也不一樣。其實無論設計和驗證都各有好處。驗證更好一點吧,個人理解。


我就來默默地說一句,你圈了Asics幹嘛=。=


推薦閱讀:

如何才能看到verilog,就知道綜合出來的電路呢?
初學Cadence IC的小白應該在哪些論壇或者社交平台提問?
提高核心數和改善架構,這兩種方案哪種更能提升 CPU 的性能?
多核心處理器設計有哪些特別難的技術門檻?
985碩士一枚,專業方向射頻集成電路、通信數字接收機集成電路、超大規模集成電路。今後發展方向有哪些?

TAG:晶元集成電路 | 愛世克斯Asics |