Page 1 of 5 - Cover Page Submission ID trn:oid:::3618:110116755
340 [Link]
Hajee Mohammad Danesh Science And Technology University,Bangladesh
Document Details
Submission ID
trn:oid:::3618:110116755 3 Pages
Submission Date 1,398 Words
Aug 29, 2025, 10:26 PM GMT+6
7,953 Characters
Download Date
Aug 29, 2025, 10:27 PM GMT+6
File Name
340 [Link]
File Size
159.9 KB
Page 1 of 5 - Cover Page Submission ID trn:oid:::3618:110116755
Page 2 of 5 - AI Writing Overview Submission ID trn:oid:::3618:110116755
0% detected as AI Caution: Review required.
The percentage indicates the combined amount of likely AI-generated text as It is essential to understand the limitations of AI detection before making decisions
well as likely AI-generated text that was also likely AI-paraphrased. about a student’s work. We encourage you to learn more about Turnitin’s AI detection
capabilities before using the tool.
Detection Groups
0 AI-generated only 0%
Likely AI-generated text from a large-language model.
0 AI-generated text that was AI-paraphrased 0%
Likely AI-generated text that was likely revised using an AI-paraphrase tool
or word spinner.
Disclaimer
Our AI writing assessment is designed to help educators identify text that might be prepared by a generative AI tool. Our AI writing assessment may not always be accurate (i.e., our AI models
may produce either false positive results or false negative results), so it should not be used as the sole basis for adverse actions against a student. It takes further scrutiny and human
judgment in conjunction with an organization's application of its specific academic policies to determine whether any academic misconduct has occurred.
Frequently Asked Questions
How should I interpret Turnitin's AI writing percentage and false positives?
The percentage shown in the AI writing report is the amount of qualifying text within the submission that Turnitin’s AI writing
detection model determines was either likely AI-generated text from a large-language model or likely AI-generated text that was
likely revised using an AI paraphrase tool or word spinner.
False positives (incorrectly flagging human-written text as AI-generated) are a possibility in AI models.
AI detection scores under 20%, which we do not surface in new reports, have a higher likelihood of false positives. To reduce the
likelihood of misinterpretation, no score or highlights are attributed and are indicated with an asterisk in the report (*%).
The AI writing percentage should not be the sole basis to determine whether misconduct has occurred. The reviewer/instructor
should use the percentage as a means to start a formative conversation with their student and/or use it to examine the submitted
assignment in accordance with their school's policies.
What does 'qualifying text' mean?
Our model only processes qualifying text in the form of long-form writing. Long-form writing means individual sentences contained in paragraphs that make up a
longer piece of written work, such as an essay, a dissertation, or an article, etc. Qualifying text that has been determined to be likely AI-generated will be
highlighted in cyan in the submission, and likely AI-generated and then likely AI-paraphrased will be highlighted purple.
Non-qualifying text, such as bullet points, annotated bibliographies, etc., will not be processed and can create disparity between the submission highlights and the
percentage shown.
Page 2 of 5 - AI Writing Overview Submission ID trn:oid:::3618:110116755
Page 3 of 5 - AI Writing Submission Submission ID trn:oid:::3618:110116755
Introduction:
In this era of artificial intelligence, AI-powered mirrors are becoming rapidly relevant because of their
effectiveness on the “try on” clothing system. According to FFAC, the key features are interactiveness,
personalization, customizability, and multi-industrial integration. Because of the systems processor
design, memory hierarchy, and real-time AI tasks, it is indeed relevant to computer architecture. The main
objective of this paper is to analyze the issues which are: real-time processing, memory overloading,
thermal throttling and SSD wear.
Background:
From chapter 5 of Patterson & Hennessy, we can understand cache partitioning. It is used to
isolate streams. We can also know about reducing conflict misses. Firstly, the concept of cache
partitioning is to avoid overlapping different data streams which might compete for the same cache. For
instance, if two streams are to access the same cache, one might replace the other stream. Later when the
replaced stream will be needed, the program will not be able to access it. It happens because of how the
cache is handled, not depending on the size of the cache. Secondly, the generated clothing flickering
problem. Generations are dependent on memories. When under load, the real time pipelining uploads
partially rendered or partially generated images. Thirdly, overheating also causes performance throttling.
According to the formula, P = C * V^2 * f, the internal thermal management system of chips (cpu,
memory) lowers the frequency and voltage to stop the chips from damaging itself under load. Lastly, the
SSD wear happens because of rapid read/write, deletions. From chapter 6, we can learn about the parallel
processing, which is useful for faster processing.
The actual hardware used in this system is:a quad-core ARM Cortex @ 2 Ghz, 1.5 mb l2 cache, 6gb
lpddr4 ram, rgb camera, depth sensor, generative model, passive cooling, local 64gb SSD with 100 TBW
endurance. This system allows the user to generate virtual clothes according to their shape in real time. It
is done by segmentation and posture detection. The user can also save the rendered image to the local
SSD.
Methodology:
The problem we are facing is that, when the user rotates too quickly, it lags. Secondly, when the system is
overloaded, the image flickers. This happens because of memory contention and task concurrency. It
means that multiple tasks are trying to access the same memory. It causes higher cpu stalls, increases
latency. Task concurrency means multiple tasks are running at the same time. When it is high, all the tasks
continuously request to read/write from the memory. This leads the operating system to slow the tasks
down. Because multiple tasks are running at the same time and continuously requesting for data, the SSD
shuffles through all the requests, causing overheating and wear. This decreases the life span of the SSD.
We can propose an architecture which can solve this problem:
1. Cache partitioning: We can divide the cache and cores to perform different tasks. Because it is a 4
core cpu, we can divide the rgb camera and depth sensor to Core 0. Than, we can use the Core 1
to segment and detect the posture. Basically, the first two cores (0,1) will work on the visionary
part of the system. Then we will use the Core 3 to render the images through artificial
intelligence. And the last core 4 will be used to save the images to local storage.
2. Frame Buffering: There is a dedicated block in the memory which stores the image as pixel
[Link] is necessary because the display continuously updates the screen. For smooth transition,
frame buffering is important.
3. Deferred Writes: It means that the system holds the data in cache temporarily to use it when it is
efficient. It increases the performance of the system. It also helps multiple data to be used at the
same time.
Page 3 of 5 - AI Writing Submission Submission ID trn:oid:::3618:110116755
Page 4 of 5 - AI Writing Submission Submission ID trn:oid:::3618:110116755
Here is a proposed diagram of how the system will work:
Figure : Flow-Chart
The assumptions we made is, each rendered frame/image is more or less 3 mb. When the used interacts,
it costs more or less 10 mb. The estimated lifespan is 100 terabytes written.
This system improves performance via cache partitioning, frame buffering and deferred writes. The
cache partitioning makes sure that the vision part and the ai part do not collide in the same cache/core.
The frame buffer stops partial updates so that it doesn’t flicker. Deferred writes trun small memory
requests into large sequential requests, which reduces stalling. It also uses energy efficiently by smoot
loading, which happens through frame buffering. By reducing missing caches, it sustains energy per
instruction. And batching small memory requests also lower energy consumption. Finally, the cost is
improved by increasing the SSD lifespan.
Discussion:
The proposed architecture improves real-time interaction between ai and human. The cache partitioning
helps reducing task overlapping by allocation different streams to different cores/caches. The frame buffer
ensures a smooth update which prevents flickers when user move too quick. Deferred write also improves
the delay. But it all comes with trade offs. Firstly, cache partitioning works well when it is fully utilized
by a single stream. If a single stream is not large enough to use the cache properly, it wastes memory,
which could be used by another stream. Secondly, frame buffer prevented partially generated images from
showing up. But in this process, it might cause delay and use up more space and resources. During
designing the solution this problem, I was able to find some novel ideas like, adaptive cache slicing and
thermal aware write scheduling.
The impact of fast-changing computer architecture and technology is both positive and negative. For
much faster chip architecture, we can now solve more complex problems. We can make better health
tools, serve people much faster, work remotly and many more. But, the rapid growth of new architectures
leads to e-wastes. Because people likes to be up to date, so whenever a new device/chip comes, people
replaces it with the older one. And making new chips leades to large amount of wastage of water.
Additionally, large manufacturing plants uses a lot of electricity, which comes from nuclear power plants
or coal/oil based power plants. This leads to high carbon emission. Lare data centers causes huge
pollution to nearby water, air and soil. But, moder technology improves the energy usage and sustains
resources. Old chips are recycled so waste is reduced. Modern architecture ensures efficient usage of
chips, so the lifespan increases, thus reducing waste.
Page 4 of 5 - AI Writing Submission Submission ID trn:oid:::3618:110116755
Page 5 of 5 - AI Writing Submission Submission ID trn:oid:::3618:110116755
Conclusion:
From this problem, we got insight about cache partitioning system, frame buffer and deferred write. We
got to understand how to allocate cache to different stream and how to use memory to reduce partial
rendering. My proposed design uses efficient cache memory, processes memory requests sequentially,
which lead to less energy usage and improved SSD lifespan.
References:
● Patterson & Hennessy, Computer Organization and Design: RISC‑V Edition —
Chapter 5 and 6. Publisher page or library copy.
[Link]
● Evaluation of Cache Partitioning for Hard Real‑Time Systems.
[Link]
.pdf
● Compiler support for software‑based cache partitioning.
[Link]
● A Hardware Evaluation of Cache Partitioning to Improve Utilization.
[Link]
ccd53212c509fb5548226b774c06b
● Types of cache misses (conflict/capacity/compulsory).
[Link]
he-misses/ and
[Link]
onflict-miss-and-capacity-miss
● Double buffering concept and practice.
[Link] and
[Link]
educe-graphics-flicker-with-double-buffering-for-forms-and-controls
● Thermal throttling overview and frame‑rate impact.
[Link]
rame-rates/ and [Link]
● Mobile thermal management trade‑offs.
[Link]
● Write amplification definition and mitigation.
[Link] and
[Link]
and
[Link]
[Link] (mirror of article) or original:
[Link]
/ and write coalescing explainer:
[Link]
● SSD endurance metrics (TBW/DWPD).
[Link]
-endurance-tbw-dwpd and [Link]
● AR/VR latency and low‑latency display context.
[Link]
Page 5 of 5 - AI Writing Submission Submission ID trn:oid:::3618:110116755