Abstract
Due to the off-chip I/O pin and power constraints of GDDR5, HBM has been proposed to provide higher bandwidth and lower power consumption for GPUs. In this paper, we first provide detailed comparison between HBM and GDDR5 and expose two unique features of HBM: Dualcommand and pseudo channel mode. Second, we analyze the effectiveness of these two features and show that neither notably contributes to performance. However, by combining pseudo channel mode with cache architecture supporting fine-grained cache-line management such as Amoeba cache, we achieve high effciency for applications with irregular memory requests. Our experiment demonstrates that compared with Amoeba caches with legacy mode, Amoeba cache with pseudo channel mode improves GPU performance by 25% and reduces HBM energy consumption by 15%.
| Original language | English (US) |
|---|---|
| Journal | IEICE Electronics Express |
| Volume | 13 |
| Issue number | 14 |
| DOIs | |
| State | Published - Jan 1 2016 |
Fingerprint
Keywords
- Amoeba cache
- DRAM
- Energy
- GPU
- HBM
ASJC Scopus subject areas
- Electronic, Optical and Magnetic Materials
- Condensed Matter Physics
- Electrical and Electronic Engineering
Cite this
Exploring new features of high-bandwidth memory for GPUs. / Li, Bingchao; Song, Choungki; Wei, Jizeng; Ahn, Jung Ho; Kim, Nam Sung.
In: IEICE Electronics Express, Vol. 13, No. 14, 01.01.2016.Research output: Contribution to journal › Letter
}
TY - JOUR
T1 - Exploring new features of high-bandwidth memory for GPUs
AU - Li, Bingchao
AU - Song, Choungki
AU - Wei, Jizeng
AU - Ahn, Jung Ho
AU - Kim, Nam Sung
PY - 2016/1/1
Y1 - 2016/1/1
N2 - Due to the off-chip I/O pin and power constraints of GDDR5, HBM has been proposed to provide higher bandwidth and lower power consumption for GPUs. In this paper, we first provide detailed comparison between HBM and GDDR5 and expose two unique features of HBM: Dualcommand and pseudo channel mode. Second, we analyze the effectiveness of these two features and show that neither notably contributes to performance. However, by combining pseudo channel mode with cache architecture supporting fine-grained cache-line management such as Amoeba cache, we achieve high effciency for applications with irregular memory requests. Our experiment demonstrates that compared with Amoeba caches with legacy mode, Amoeba cache with pseudo channel mode improves GPU performance by 25% and reduces HBM energy consumption by 15%.
AB - Due to the off-chip I/O pin and power constraints of GDDR5, HBM has been proposed to provide higher bandwidth and lower power consumption for GPUs. In this paper, we first provide detailed comparison between HBM and GDDR5 and expose two unique features of HBM: Dualcommand and pseudo channel mode. Second, we analyze the effectiveness of these two features and show that neither notably contributes to performance. However, by combining pseudo channel mode with cache architecture supporting fine-grained cache-line management such as Amoeba cache, we achieve high effciency for applications with irregular memory requests. Our experiment demonstrates that compared with Amoeba caches with legacy mode, Amoeba cache with pseudo channel mode improves GPU performance by 25% and reduces HBM energy consumption by 15%.
KW - Amoeba cache
KW - DRAM
KW - Energy
KW - GPU
KW - HBM
UR - http://www.scopus.com/inward/record.url?scp=84979555724&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84979555724&partnerID=8YFLogxK
U2 - 10.1587/elex.13.20160527
DO - 10.1587/elex.13.20160527
M3 - Letter
AN - SCOPUS:84979555724
VL - 13
JO - IEICE Electronics Express
JF - IEICE Electronics Express
SN - 1349-2543
IS - 14
ER -