====== Differences ====== This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
soc:2010:cooldavid:notes:benchmark [2010/05/11 18:54]
cooldavid created
soc:2010:cooldavid:notes:benchmark [2010/05/12 01:32] (current)
cooldavid
Line 4: Line 4:
  
 I've done some benchmark with imgfetch a 500MByte image via HTTP. I've done some benchmark with imgfetch a 500MByte image via HTTP.
 +The test was done by directly connect two computer with a CAT.6 cable:
 +  * Server:
 +    * Processor: Intel(r) Core(tm)2 Quad CPU    Q8400  @ 2.66GHz
 +    * RAM: 8G DDRII-800
 +    * MB Chip: G965 (ICH8)
 +    * NIC: Marvell 88E8056 PCI-E Gigabit Ethernet Controller
 +    * OS: Gentoo (Linux 2.6.33.3)
 +    * HTTP Server: Lighttpd 1.4.26
 +  * Client:
 +    * Processor: AMD Athlon(tm) 64 Processor 3000+ (1.8HGz)
 +    * RAM: 2G DDR400
 +    * MB Chip: nVidia CK804
 +    * NIC: JMicron JMC250 PCI Express Gigabit Ethernet Controller
 +    * OS: gPXE latest ​   ​
 +
 +[[http://​git.etherboot.org/?​p=people/​cooldavid/​gpxe.git;​a=shortlog;​h=refs/​heads/​jme_driver_cso|The RX checksum offload source tree]]
  
 The following table shows the result of different TCP_MAX_WINDOW_SIZE The following table shows the result of different TCP_MAX_WINDOW_SIZE
 and with or without RX checksum offloading. and with or without RX checksum offloading.
- 
-[[http://​git.etherboot.org/?​p=people/​cooldavid/​gpxe.git;​a=shortlog;​h=refs/​heads/​jme_driver_cso|The RX checksum offload source tree]] 
- 
  
 ^             HTTP imgfetch benchmark ​         ^^^ ^             HTTP imgfetch benchmark ​         ^^^
-|         ^ Without RX Offload ^ With RX Offload ^ +^ WinSize ​^ Without RX Offload ^ With RX Offload ^ 
- 4k win |      136Mbit/​s ​    ​| ​    ​136Mbit/​s ​  | +|           ​Window Size 1024 byte step         ||| 
-64k win |      528Mbit/​s ​    ​| ​    ​896Mbit/​s ​  | +   ​4096 ​|      136Mbit/​s ​    ​| ​    ​136Mbit/​s ​  | 
 +   5120 |      200Mbit/​s ​    ​| ​    ​208Mbit/​s ​  | 
 +^    6114 |      224Mbit/​s ​    ​| ​    ​264Mbit/​s ​  | 
 +^    7168 |      264Mbit/​s ​    ​| ​    ​288Mbit/​s ​  | 
 +^    8192 |      256Mbit/​s ​    ​| ​    ​272Mbit/​s ​  | 
 +^    9216 |      368Mbit/​s ​    ​| ​    ​416Mbit/​s ​  | 
 +^   10240 |      368Mbit/​s ​    ​| ​    ​400Mbit/​s ​  | 
 +^   11264 |      424Mbit/​s ​    ​| ​    ​496Mbit/​s ​  | 
 +^   12288 |      448Mbit/​s ​    ​| ​    ​536Mbit/​s ​  | 
 +^   13312 |      496Mbit/​s ​    ​| ​    ​616Mbit/​s ​  | 
 +^   14436 |      496Mbit/​s ​    ​| ​    ​592Mbit/​s ​  | 
 +^   ​15360 ​|      528Mbit/​s ​    |     ​704Mbit/​s ​  | 
 +|           ​Window Size 2048 byte step         ||| 
 +^   16384 |      528Mbit/​s ​    ​| ​    ​696Mbit/​s ​  | 
 +^   18432 |                    |     ​768Mbit/​s ​  | 
 +^   20480 |                    |     ​824Mbit/​s ​  | 
 +^   22528 |                    |     ​880Mbit/​s ​  | 
 +^   24576 |                    ​|     ​896Mbit/​s ​  | 
 +|                    Full Speed                ||| 
 +^   32768 |      536Mbit/​s ​    ​| ​    ​896Mbit/​s ​  | 
 +^   65532 |      536Mbit/​s ​    ​| ​    ​896Mbit/​s ​  |
  

QR Code
QR Code soc:2010:cooldavid:notes:benchmark (generated for current page)