AOA Forums AOA Forums AOA Forums Folding For Team 45 AOA Files Home Front Page Become an AOA Subscriber! UserCP Calendar Memberlist FAQ Search Forum Home


Go Back   AOA Forums > Hardware > AMD Motherboards & CPUs

AMD Motherboards & CPUs Questions or comments on AMD products?


Reply
 
LinkBack Thread Tools Rate Thread
  #1 (permalink)  
Old 31st January, 2006, 06:52 PM
Super Nade's Avatar
Member
 
Join Date: December 2005
Location: Indianapolis, USA
Posts: 157
Send a message via Yahoo to Super Nade

Question Strange occurances with dual core CPU's ?

This morning, I was messing around with a set of Patriot XBLK TCCD. Previously, I could not hit DDR600 with the6-23-3BIOS, no matter what. So, I decided to use the BigToe modded 704-2BTA BIOS,which was supposedly optimized to support Dual core CPU's.

Well, I hit DDR608 2.5-4-3-8-1T,2.9V, but my memory bandwidth (as shown by memtest) was shot to hell (2.8k). With the same TCCD, using my Venice, I achieved a BW in excess of 3.2k!

This is not only confined to TCCD but to BH5 also. Running the Corsair BH5 in my sig at 270MHz (2-2-2-5-1T) yeilds only 2.7k, whereas with my Venice, I hit 3.1k

What the heck is going on? My CPU is recognized as an X2, so memtest should be tuned to load both cores, right? Is this a manifestation of a load-sharing problem or is it because the cache latency of my Opteron being less than the Venice is somehow bottlenecking the RAM (logically, it should not happen, but this is something I really have not analyzed)?

Strange..
__________________
R.P Feynman : "If I can't create it...it doesn't exist "
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #2 (permalink)  
Old 31st January, 2006, 06:56 PM
Favu's Avatar
AOA's resident barman
 
Join Date: October 2005
Location: /Wales/Abergavenny
Posts: 4,004
Send a message via ICQ to Favu Send a message via AIM to Favu Send a message via MSN to Favu

Could it be that the BIOS is modded to allow higher Mhz on the RAM, but at a higher latiency?
__________________
AOA Team fah
 

Custom 8-bit Sharp Z80 @ 4.194304 MHz
Reflective LCD @ 160 × 144
8 kByte S-RAM






Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #3 (permalink)  
Old 31st January, 2006, 08:24 PM
Chief Systems Administrator
 
Join Date: September 2001
Location: Europe
Posts: 13,075

Both CPUs were dual core or not?
__________________
Any views, thoughts and opinions are entirely my own. They don't necessarily represent those of my employer (BlackBerry).
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #4 (permalink)  
Old 31st January, 2006, 09:45 PM
Super Nade's Avatar
Member
 
Join Date: December 2005
Location: Indianapolis, USA
Posts: 157
Send a message via Yahoo to Super Nade

All high BW scores were obtained with a single core (Venice) 3200+. I should have mentioned that..silly me. Low BW scores were obtained with a Dual core Opteron 165.

Favu, I don't think one can change CPU cache latency. AFAIK,it is programmed into an on-die ROM. Maybe I misunderstood what you were trying to say.

The thing is, NO S939 main-stream enthusiast Mobo maker will support the Opteron (officially). The nice thing about DFI is that, like the open source community, some good soul would write up a BIOS. In this case it was Tony a.k.a BigToe from OCZ. Maybe, its all BIOS dependent.
__________________
R.P Feynman : "If I can't create it...it doesn't exist "

Last edited by Super Nade; 31st January, 2006 at 09:51 PM.
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #5 (permalink)  
Old 31st January, 2006, 09:54 PM
Chief Systems Administrator
 
Join Date: September 2001
Location: Europe
Posts: 13,075

Stupid question #2, which memtest are you using?
__________________
Any views, thoughts and opinions are entirely my own. They don't necessarily represent those of my employer (BlackBerry).
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #6 (permalink)  
Old 31st January, 2006, 10:02 PM
Super Nade's Avatar
Member
 
Join Date: December 2005
Location: Indianapolis, USA
Posts: 157
Send a message via Yahoo to Super Nade

I'm not really sure mate. I'm using the one which came with the BIOS. It recognizes my CPU as an X2, so it must be the right one. However, even my everest/Sciencemark BW scores are low. I'll post a few comparision SS's when I get home.
__________________
R.P Feynman : "If I can't create it...it doesn't exist "
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #7 (permalink)  
Old 31st January, 2006, 10:06 PM
Chief Systems Administrator
 
Join Date: September 2001
Location: Europe
Posts: 13,075

If a program uses both cores, and the memory accesses are not localised, I'd expect the memory performance to be seriously impaired. Non localised memory access would result in large amounts of page closure/openings, with their associated overheads...

Would be interesting to find out if the program is multithreaded too...
__________________
Any views, thoughts and opinions are entirely my own. They don't necessarily represent those of my employer (BlackBerry).
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #8 (permalink)  
Old 1st February, 2006, 12:38 AM
Super Nade's Avatar
Member
 
Join Date: December 2005
Location: Indianapolis, USA
Posts: 157
Send a message via Yahoo to Super Nade

Mate, what do you mean by "nonlocalised memory access"? Could you please explain this as I don't know anything about this. What overhead are you referring to?

Well, I can run memory bandwidth benchmarks loading one core heavily (CD ripping) and in the next case not loading a particular core but setting the benchmark affinity to one core. Let us see what happens.
__________________
R.P Feynman : "If I can't create it...it doesn't exist "
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #9 (permalink)  
Old 8th February, 2006, 05:14 PM
Member
 
Join Date: August 2003
Location: Belgium
Posts: 211

Exclamation

Quote:
Originally Posted by Super Nade
Mate, what do you mean by "nonlocalised memory access"? Could you please explain this as I don't know anything about this. What overhead are you referring to?

Well, I can run memory bandwidth benchmarks loading one core heavily (CD ripping) and in the next case not loading a particular core but setting the benchmark affinity to one core. Let us see what happens.



Try, by going into Task Manager and processes, to change the affinity of the test program to 1 of the 2 CPU (cores) ...
Right-click on the process (.exe) and choose affinity; in the new Window, choose CPU0 or CPU1.

See what that yields !





__________________
F_O_L_D_E_R_S

1] ... Athlon XP2500+ Barton @ 3000+ & 512 MB RAM WinXP
2] ... Athlon XP2500+ Barton @ 3100+ & 1024 MB RAMWin2k
3] ... Athlon XP2500+ Mobile Barton @ 3200+ & 512 MB RAM WinXP
4&5] Athlon A64X2 3800+ @ 2.6 GHz (~ 5000+) & 2x 1024 MB RAM running @ 237 MHz WinXP

All folding 24/7
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
  #10 (permalink)  
Old 8th February, 2006, 05:51 PM
Member/Contributor/Resident Crystal Ball
 
Join Date: March 2004
Posts: 7,451

Each cpu core revision from AMD has different bandwidth numbers, although the problem you surmise is most likely due to alpha timings set @ default, rather than the core itself, although i myself have noticed a 300mb/sec in memtest+x86 between sandiego and venice, as well as winchester, toledo, and manchester.

plain and simple, the reason why is easily explained. because the dualcore CPU's and sandiego have a larger cache, it takes a wee bit longer for data to pass through it(all calls from cpu must pass through the L2 cache). the other day andy and i were discussing a certain bios setting that can help overcome this slight performance drop, however it can also bring instability and lower numbers as well.
__________________
Digg this Post!Add Post to del.icio.usBookmark Post in TechnoratiFurl this Post!
Reply With Quote
Reply



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Single core to dual core upgrade problem. Daston AMD Motherboards & CPUs 11 19th September, 2006 12:18 AM
dual core or dual xeon? acgraham Intel Motherboards & CPUs 4 16th September, 2005 12:37 PM
Users of new 333 FSB CPU's...what RAM for dual channel really works? Oofda EPoX MotherBoards 2 1st January, 2003 01:49 AM
Dual CPU's Spode EPoX MotherBoards 8 13th September, 2002 11:27 AM
Safe temps for dual CPU's on an Asus A7M266-D motherboard? Nate-X Cooling & Temperature Monitoring 3 22nd April, 2002 03:47 PM


All times are GMT +1. The time now is 03:18 PM.


Copyright ©2001 - 2010, AOA Forums
Don't Click Here Don't Click Here Either

Search Engine Friendly URLs by vBSEO 3.3.0