Status
Not open for further replies.

greliu

Distinguished
Apr 7, 2009
117
0
18,690
Alright, I have looked for this answer everywhere when I should have came here in the first place. I was wondering if there was a performance differences between Hypertranport 1.0, 2.0, and 3.0. I mean, I know there is a bandwidth difference, but is it noticeable in the "real" world, particularly with gaming? Thanks for any information! :bounce:
 
Solution
Strictly speaking Hyper Transport isn't related to any DDR2/DDR3 support or the ability to overclock an AMD CPU.
Think of Hyper Transport as a data pipeline. It only has to be wide enough to not dam up the rapid flow of data communications between the CPU and all the other devices in the system.
m4a78te3.jpg


As Hyper Transport versions have changed the 'pipeline' has gotten fatter with the increased bandwidth (speed) keeping it from being a bottle neck in the system. Increasing the size of the pipeline won't change the amount of data being sent or received. As the MadShrimps article showed only a video task (3DMark06) showing an improvement when the HT was raised from 200Mhz to 1800Mhz and...

greliu

Distinguished
Apr 7, 2009
117
0
18,690
So, other than HT allowing you to support newer RAM, like DDR3, and the ability to overclock your CPU higher, there's not that much of a preformace increase? I was reading that the only why HT is used is in severs, where a ton of multi-tasking is required. Is all this true?
 
Strictly speaking Hyper Transport isn't related to any DDR2/DDR3 support or the ability to overclock an AMD CPU.
Think of Hyper Transport as a data pipeline. It only has to be wide enough to not dam up the rapid flow of data communications between the CPU and all the other devices in the system.
m4a78te3.jpg


As Hyper Transport versions have changed the 'pipeline' has gotten fatter with the increased bandwidth (speed) keeping it from being a bottle neck in the system. Increasing the size of the pipeline won't change the amount of data being sent or received. As the MadShrimps article showed only a video task (3DMark06) showing an improvement when the HT was raised from 200Mhz to 1800Mhz and even that was a very small percentage increase. The increase from 800Mhz (HT 1.0) to 1800Mhz (HT 2.0 plus 400Mhz) was an even smaller increase.
spec_table.gif


In servers HT gets a real workout since it's used to connect multiple CPUs on a server board and servers can be heavily I/O dependent on storage devices which run data over the HT links.

In PCs it's more likely the graphics card (or cards in CF systems) will put the largest load on the HT followed by disk IO.
 
Solution
Status
Not open for further replies.