Homework and Quiz Discussion

[Guofeng]

This is a general discussion space. Do NOT post solutions to any
homework or quiz questions here.

[Guofeng]

Quiz 1 question 6(c)

I was asked if (1/8) * (1-Yield) is the probability that only 1/8th
area of the die (which is 1 out of 8 cores) has a defect. The answer
is no.

(1-Yield) gives the probability that there is defect on a die.
However, there might be multiple defects and defects may be on
multiple cores. If you divide (1-Yield) by 8 to calculate the
probability of defect occur on a single core, you are actually
assuming all defects occur on a single core. So, it is incorrect. In
fact, this calculation obtains a value smaller than the actual value.
Do you know why?

Guofeng

[aarul...@asu.edu]

Is it because cases where multiple defects occur on same core are also
excluded from this calculation? And to account for this, one should
divide the die area by 8 and then calculate yield on that area
(similar to what you said).

Thanks
-Aarul

> > homework or quiz questions here.- Hide quoted text -
>
> - Show quoted text -

[Ayan]

HW 1 doubt

In question number 1.2 c it says that the demand for Woods chips is
50000 per month and that for Markon chips is 25000 per month. But the
fabrication facility allows only 150 wafers. But I think we cannot
manufacture 50000 Woods chips and 25000 Markon chips from only 150
wafers. In that case the demand can never be satisfied. Please
suggest.

Ayan

[Aarul Jain]

If the question is correct (and if my calculation is correct), I guess we have two alternatives

i) Look to maximise the profit even if it does not meet the demand. I will assume that there are other chip manufacturers for meeting the demand :)-

ii) Open a new fabrication facility, for long term profit and meeting demand.

 

The question does not ask us to consider (ii) option, so option (i) is better option in that sense.

Any comments???

 

Regards

-Aarul

 

Arizona State University, Tempe, US
Ph: 480-278-9230

[Guofeng]

Question 1.2d, which is similar to 1.2c, asks how to make the most
profit.
In this case, specifically, if you make more than 50000 Woods chips
per month, you won't be able to sell it; similar to the Markon chips.

Regards,
Guofeng

[YI-HSI...@asu.edu]

Questions on original Case Study

I have some questions on original Case Study
In Case Study 1.3(b)
Prob of one defect = 0.29×0.71^7×8 = 0.21
Prob of two defects = 0.29^2×0.71^6×28 = 0.30

Why does it multiply 8 in Prob of one defect and 28 in Prob of two
defects ?
I cannot figure out with it.

Thanks,
Yihsin

> > > > - Show quoted text -- 隱藏被引用文字 -
>
> - 顯示被引用文字 -

[Sugan]

Question 1.8(a)

This question is about the downtime cost. Is the downtime cost same as
the sales for the day? & Can we assume that the downtime cost asked is
for Dec 12, 2005 or is it 2004?

Thanks,
Sugan & Ashay.

On Sep 7, 11:03 am, Guofeng <guofeng.d...@gmail.com> wrote:

[Guofeng]

0.29×0.71^7 tells you the probability of one core has defect and the
other 7 cores are defect free. Since the one defect can happen to any
one of the 8 cores, the chances are 8 times of this.

0.29^2×0.71^6 tells you the probability of 2 core have defects and the
other 6 cores are defect free. Since the two defect ones can happen to
any two of the 8 cores, the chances are 8 * 7 / 2 times, which is 28.

These involve some basic stochastic computation.

Regards,
Guofeng

[Guofeng]

In this question, the downtime cost is same as the online sales for
the day.

>From the last two rows in the table, you can compute the percentage of
items sold on that date. From the corresponding rows associated with
Gap, you can compute the percentage of online sales (kind of
confusing). Then you can get the cost.

Regards,
Guofeng

[Guofeng]

In question 1.3(b), assume the number of dies per wafer is reduced by
half.

You are not provided enough information to compute exact cost of
wafter or dies per wafer. But you don't need exact values for either.

Regards,
Guofeng

> My question is on 1.3 (b)
>
> I am using the following equation for cost of chip:
>
> cost of die = cost of wafer / (dies per wafer * die yield)
>
> We don't have cost of wafer but we do have cost of die for the old chip.
> Using that, we can find the cost of wafer to calculate the cost of new chip.
> We are also missing the wafer diameter for AMD which is used to find dies per wafer.
> The new die area with two cores per die will be double the old area. So can we assume
> that number of dies per wafer for the new two core dies will be roughly half the
> number of old dies as we don't have enough parameters to calculate dies per wafer?

On Sep 7, 11:03 am, Guofeng <guofeng.d...@gmail.com> wrote:

[Guofeng]

Homework 1 questions and answers:

I received a couple of questions on Q1.8. The question is to estimate
the economic cost to the showdown based on possibly incomplete
statistics. So, don't worry too much about the accuracy, but provide
reasonable estimations.

> Q 1.8 (c) => The last line says, "how much does it cost to support online traffic
> at the two websites?" which two websites are we talking about here? Or is it just
> a typo and do we have to consider OldNavy.com only?

Consider OldNavy.com only.

> Q 1.11 (a) => My question is for original(book) case study solution.
> Why are we multiplying 3334 (which comes from 10000 * 1/3) with 10^7 ? shouldn't it be 10^9 instead?

This question specifically follows question 1.9, which says assuming
an FIT of 100.

Regards,
Guofeng

On Sep 7, 11:03 am, Guofeng <guofeng.d...@gmail.com> wrote:

[Guofeng]

Question 1.11a was from the book case study. Thanks Justin for
pointing that out.

Regards,
Guofeng

[Guofeng]

The programming assignment has been assigned today. You will implement
game of life using HPC resources. Here are some additional resources
you may want to look at if you are not familiar with the game of life.
Wikipedia has the complete story and examples of the game of life
(http://en.wikipedia.org/wiki/Conway's_Game_of_Life). Youtube has an
interview on it (http://youtube.com/watch?v=HqR0Ao1JiuI).

The assignment is due in about 3 weeks. The specification of
submission and grading criteria are coming soon. You are encouraged to
start it as early as possible. According to Doug, it is better to
implement a sequential version.

Let us know if you have any questions.

Regards,
Guofeng

On Sep 7, 11:03 am, Guofeng <guofeng.d...@gmail.com> wrote: