各位使用N卡的兄弟，发一下CUDA-Z（真。粗大）的数据

xiaoren6412 · 发表于 2011-11-20 22:20:31

显卡差，随便发来玩的，各位大神速度发出来瞧瞧

下载地址：http://www.onlinedown.net/soft/71512.htm

金鹏 · 发表于 2011-11-20 22:40:46

测试时GPU负载大概才13%，这个靠谱么？

捕获.PNG

xiaoren6412 · 发表于 2011-11-20 22:50:31

回复 2# 金鹏

应该没有影响，主要看单浮点和双浮点，我刚才试了下，99%的GPU运行率，数值基本一样的

xiaoren6412 · 发表于 2011-11-20 23:02:25

本帖最后由 xiaoren6412 于 2011-11-20 23:08 编辑

回复 2# 金鹏

看错了，失误

panda7456 · 发表于 2011-11-20 23:35:23

cuda · 发表于 2011-11-20 23:43:48

这个CUDA测试程序是靠谱的，它给出的是实测的峰值性能。
GPU占用率不高的原因是它只用1个MP进行测试(1个GPU有N个MP)，最后把结果乘以N，得到整个GPU的峰值性能。

回复 2# 金鹏

nekoko · 发表于 2011-11-20 23:48:38

本帖最后由 nekoko 于 2011-11-20 23:51 编辑

算着gpugrid测的，懒的停

xiaoren6412 · 发表于 2011-11-21 04:43:49

都很迅猛呀

ocw · 发表于 2011-11-21 05:01:29

CUDA-Z Report
=============
Version: 0.5.95
http://cuda-z.sourceforge.net/
OS Version: Windows AMD64 6.1.7601 Service Pack 1

Core Information
----------------
Name: GeForce GTS 450
Compute Capability: 2.1
Clock Rate: 1770 MHz
Multiprocessors: 4
Warp Size: 32
Regs Per Block: 32768
Threads Per Block: 1024
Watchdog Enabled: Yes
Threads Dimentions: 1024 x 1024 x 64
Grid Dimentions: 65535 x 65535 x 65535

Memory Information
------------------
Total Global: 961.688 MB
Shared Per Block: 48 KB
Pitch: 2.09715e+06 KB
Total Constant: 64 KB
Texture Alignment: 512
GPU Overlap: Yes

Performance Information
-----------------------
Memory Copy
Host Pinned to Device: 1376.34 MB/s
Host Pageable to Device: 655.843 MB/s
Device to Host Pinned: 1307.77 MB/s
Device to Host Pageable: 891.587 MB/s
Device to Device: 19710.9 MB/s
GPU Core Performance
Single-precision Float: 394582 Mflop/s
Double-precision Float: 56673.4 Mflop/s
32-bit Integer: 226368 Miop/s
24-bit Integer: 226227 Miop/s

Generated: Mon Nov 21 05:01:01 2011

xiaoren6412 · 发表于 2011-11-21 05:17:59

回复 9# ocw

5点了，是没睡，还是睡醒了？

ocw · 发表于 2011-11-21 05:26:07

回复 ocw

5点了，是没睡，还是睡醒了？
xiaoren6412 发表于 2011-11-21 05:17

睡醒了

wpf999 · 发表于 2011-11-21 10:00:15

回复 6# cuda

cuda兄弟研究cuda编程的？

青島人 · 发表于 2011-11-21 10:01:11

CUDA-Z Report
=============
Version: 0.5.95
http://cuda-z.sourceforge.net/
OS Version: Windows AMD64 6.1.7600

Core Information
----------------
Name: GeForce GT 520
Compute Capability: 2.1
Clock Rate: 1660 MHz
Multiprocessors: 1
Warp Size: 32
Regs Per Block: 32768
Threads Per Block: 1024
Watchdog Enabled: Yes
Threads Dimentions: 1024 x 1024 x 64
Grid Dimentions: 65535 x 65535 x 1

Memory Information
------------------
Total Global: 482.438 MB
Shared Per Block: 48 KB
Pitch: 2.09715e+06 KB
Total Constant: 64 KB
Texture Alignment: 512
GPU Overlap: Yes

Performance Information
-----------------------
Memory Copy
Host Pinned to Device: 3110.56 MB/s
Host Pageable to Device: 1558.1 MB/s
Device to Host Pinned: 3111.97 MB/s
Device to Host Pageable: 1627.72 MB/s
Device to Device: 4971.42 MB/s
GPU Core Performance
Single-precision Float: 92727.7 Mflop/s
Double-precision Float: 13279 Mflop/s
32-bit Integer: 53050.9 Miop/s
24-bit Integer: 52952.9 Miop/s

Generated: Mon Nov 21 10:00:54 2011

eofliyws · 发表于 2011-11-21 10:42:22

为何我的470如此之低？

GTX470

CUDA-Z Report
=============
Version: 0.5.95
http://cuda-z.sourceforge.net/
OS Version: Windows AMD64 6.1.7601 Service Pack 1

Core Information
----------------
Name: GeForce GTX 470
Compute Capability: 2.0
Clock Rate: 1320 MHz
Multiprocessors: 14
Warp Size: 32
Regs Per Block: 32768
Threads Per Block: 1024
Watchdog Enabled: Yes
Threads Dimentions: 1024 x 1024 x 64
Grid Dimentions: 65535 x 65535 x 65535

Memory Information
------------------
Total Global: 1280 MB
Shared Per Block: 48 KB
Pitch: 2.09715e+06 KB
Total Constant: 64 KB
Texture Alignment: 512
GPU Overlap: Yes

Performance Information
-----------------------
Memory Copy
Host Pinned to Device: 736.897 MB/s
Host Pageable to Device: 690.035 MB/s
Device to Host Pinned: 736.66 MB/s
Device to Host Pageable: 688.496 MB/s
Device to Device: 55276.1 MB/s
GPU Core Performance
Single-precision Float: 1.17656e+06 Mflop/s
Double-precision Float: 148087 Mflop/s
32-bit Integer: 591011 Miop/s
24-bit Integer: 590390 Miop/s

Generated: Mon Nov 21 10:20:27 2011

金鹏 · 发表于 2011-11-21 10:51:55

回复 6# cuda

原来如此，学习了，感谢兄弟赐教

		自动登录	找回密码
密码			新注册用户