Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Performance Evaluation of Compute Unified Device Architecture (CUDA) compared to Traditional Central Processing Unit (CPU) Architecture
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
Prestandautvärdering ut av Compute Unified Device Architecture (CUDA) i jämförelse med traditionell Centralprocessor (CPU) Arkitektur (Swedish)
Abstract [en]

Nvidia’s toolkit Compute Unified Device Architecture (CUDA) have opened the door for anyone with a Nvidia Graphics Processing Unit (GPU) to harness its power and let them use the GPU’s strength in parallelism to solve computational heavy tasks. This thesis will discuss if a CUDA powered GPU is worth investing into compared to a traditional CPU. The thesis compares performance (execution time, power usage and memory usage), cost and the complexity of the code. The performance of the CPU and GPU is compared by benchmarking a computational heavy model problem, which consists of solving a one- and three-dimensional wave-equation with the finite-difference method. The results show that a CUDA powered GPU have a great performance advantage for larger data sets, while a CPU show a better result on smaller data sets. The GPU seems promising for larger data sets, however there is a drawback with the higher cost for investing in a GPU, both in retail price and the complexity of developing in CUDA. Due to limited access to hardware the research has been concentrated on one set of hardware and for future work it is recommended it is extended to multiple machines to diversify the result.

Abstract [sv]

Nvidias Compute Unified Device Architecture (CUDA) har öppnat dörrarna så att alla med en Nvidia grafikprocessor (GPU) kan använda dess prestanda och utnyttja GPU:ens styrka i parallellisering till att lösa beräkningstunga uppgifter. Detta arbete kommer att diskutera om en CUDA driven GPU är värd att investera i, jämfört med en traditionell CPU. Detta arbete jämför prestanda (beräkningstid, strömförbrukning och minnesanvändning), kostnad och komplexiteten av koden. Prestandan för CPU:n och GPU:n är jämförda med ett beräkningstungt modellproblem vilket löser en- och tre dimensionella vågekvationer med den finita differensmetoden. Resultaten visar att en CUDA driven GPU har ett bättre resultat för större datamängder, medan en CPU har ett bättre resultat på mindre datamängder. Trots att GPU:n visar ett lovande resultat för större datamängder finns en nackdel med den högre kostnaden med att investera i en GPU, både i detaljhandelspriset och komplexiteten med att utveckla program i CUDA. På grund av begränsad tillgång till hårdvara har arbetet koncentrerats till en uppsättning av hårdvara och för framtida arbete rekommenderas det att den utökas till flera maskiner för att diversifiera resultatet.

Place, publisher, year, edition, pages
2024. , p. 40
Series
TRITA-EECS-EX ; 2024:926
Keywords [en]
CUDA, CPU, GPU, Finite-Difference Method, Wave-equation
Keywords [sv]
CUDA, CPU, GPU, Finita differensmetoden, Vågekvation
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-361053OAI: oai:DiVA.org:kth-361053DiVA, id: diva2:1943546
Supervisors
Examiners
Available from: 2025-03-17 Created: 2025-03-11 Last updated: 2025-03-17Bibliographically approved

Open Access in DiVA

fulltext(578 kB)45 downloads
File information
File name FULLTEXT02.pdfFile size 578 kBChecksum SHA-512
dbb53041bb43346ac48a7fe50f2bf7a540ecf664f86a7efa853756403474b3dcee6a7a644954175c8b1155b3bf170abc396bc5e6ee47e508ed462e9de59f5573
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 45 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 344 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf