AMD_UCF_JAN_09 - AMD AMD AMD AMD Entering the Golden Age of...

Info iconThis preview shows pages 1–8. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: AMD AMD AMD AMD Entering the Golden Age of Heterogeneous Computing Michael Mantor Senior GPU Compute Architect / Fellow AMD Graphics Product Group 1 michael.mantor@amd.com The 4 P illars of massively parallel compute offload Performance Power Moores Law ! 2x < 18 Months Frequency\Power\Complexity Wall Parallel ! Opportunity for growth Price Programming Models Programming Models GPU is the first successful massively parallel COMMODITY architecture with a programming model that managed to tame 1000s of parallel threads in GPU is the first successful massively parallel COMMODITY architecture with a programming model that managed to tame 1000s of parallel threads in hardware to perform useful work efficiently hardware to perform useful work efficiently 2 The 4 P illars of massively parallel compute offload Performance Power Moores Law ! 2x < 18 Months Frequency\Power\Complexity Wall Parallel ! Opportunity for growth Price Programming Models Programming Models GPU is the first successful massively parallel COMMODITY architecture with a programming model that managed to tame 1000s of parallel threads in GPU is the first successful massively parallel COMMODITY architecture with a programming model that managed to tame 1000s of parallel threads in hardware to perform useful work efficiently hardware to perform useful work efficiently 2 Quick recap of where we are Perf, Power, Price ATI Radeon HD 4850 4x Performance/w and Performance/mm in a year ATI Radeon X1800 XT ATI Radeon HD 2900 XT ATI Radeon HD 3850 ATI Radeon X1900 XTX ATI Radeon X1950 PRO ATI Radeon HD 2900 XT 3 Source of GigaFLOPS per watt: maximum theoretical performance divided by maximum board power. Source of GigaFLOPS per $: maximum theoretical performance divided by price as reported on www.buy.com as of 9/24/08 ATI RadeonHD 4850 Designed to Perform in Single Slo Designed to Perform in Single Slo Designed to Perform in Single Slot Designed to Perform in Single Slot SP Compute Power 1.0 T-FLOPS DP Compute Power Core Clock Speed 200 G-FLOPS 625 Mhz Stream Processors Memory Type 800 GDDR3 Memory Capacity Max Board Powe 512 MB 110W Max Board Power Memory Bandwidth 110W 64 GB/Sec 4 ATI RadeonHD 4870 First Graphics with GDDR5 First Graphics with GDDR5 SP Compute Power 1.2 T-FLOPS DP Compute Power Core Clock Speed 240 G-FLOPS 750 Mhz Stream Processors Memory Type 800 GDDR5 3.6Gbps Memory Capacity Max Board Powe 512 MB 160 W Max Board Power Memory Bandwidth 160 W 115.2 GB/Sec 5 ATI RadeonHD 4870 X2 X2 Incredible Balance of Performance, Power, Price Incredible Balance of Performance, Power, Price Compute Power 2.4 TFLOPS Core Clock Speed 750 Mhz DP Compute Power 480 G-FLOPS Stream Processors Memory Type 1600 GDDR5 3.6Gbps Memory Capacity Max Board Powe 2 GB 285W Max Board Power Memory Bandwidth 285W 230 GB/Sec 6 AMD FireStream 9250 AMDs Second Generation Stream Computing Product Single PCI Slot...
View Full Document

This note was uploaded on 10/03/2011 for the course CDA 6938 taught by Professor Zou,c during the Spring '08 term at University of Central Florida.

Page1 / 75

AMD_UCF_JAN_09 - AMD AMD AMD AMD Entering the Golden Age of...

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online