Overclock.net › Forums › Intel › Intel CPUs › FSB and Dividers: My Findings
New Posts  All Forums:Forum Nav:

FSB and Dividers: My Findings - Page 9

post #81 of 85
Given the complaints from some members about Choggs396's testing method, I'm undertaking my own investigation into the old speed vs timings debate. I'd appreciate it if people could give some ideas on applications to use for benchmarking and a few opinions regarding the questions in my opening post here.

Reps will be given for all useful input before I begin my analysis.
post #82 of 85
Quote:
Originally Posted by Witchfire View Post
And for the last time... It was not just a memory test, it was a quick way to test overall system performance. 3DM is a standardized test that can and does show changes in overall system performance very well. Yes, it's very biased towards a GPU, but if the GPU doesn't change from test to test, it will be a constant, and changes in the score will reflect changes made to other parts of the configuration.

Just because his choice of testing doesn't fall into your narrow view of what is or isn't an acceptable method does not invalidate it. Sorry, but there ARE people who live outside your little personal bubble of reality.
I'm not trying to start another arguement here but just want to bring a few things back up again. witchfire 3dmark almost seems to ignore memory performance all together. I have ran it going from arm at 466mhz (233mhz fsb) and then up to 700mhz (350mhz fsb). which is aruond 50% increase in both fsb and memory speed. plus the timings stayed the same which would mean the faster ram has better latency. it showed a .3% difference. I would say in that it just about ignores memory performance all together. only other thing I could try on that to prove it further is dual channel but I can't yet untill I get two sticks of ram back in again, but with what I have seen with just frequency changes not affecting it I could almost promise you what the results would be. I believe that is part of why mr bungles doesn't not agree with 3dmark06. I know it was along those lines as to why I didn't like that test.



t4ct1c47
your if your trying to further what choggs was doing your going into the wrong thing if you diong the debate of speed vs timings. this was more of a divider vs performance issue.
compy jr
(13 items)
 
  
CPUMotherboardGraphicsHard Drive
xeon 3350 at 3.6ghz Asus P5k-dlx HIS 7970 @ 1150/1900 Lots of them.  
OSMonitorKeyboardPower
vista ultimate x64/xp sp3 compaq 21" junk dell keyboard corsair tx850m 
CaseMouseMouse Pad
upgraded metal box You don't need a mouse Desk 
  hide details  
Reply
compy jr
(13 items)
 
  
CPUMotherboardGraphicsHard Drive
xeon 3350 at 3.6ghz Asus P5k-dlx HIS 7970 @ 1150/1900 Lots of them.  
OSMonitorKeyboardPower
vista ultimate x64/xp sp3 compaq 21" junk dell keyboard corsair tx850m 
CaseMouseMouse Pad
upgraded metal box You don't need a mouse Desk 
  hide details  
Reply
post #83 of 85
Quote:
Originally Posted by rx7speed View Post
I'm not trying to start another arguement here but just want to bring a few things back up again. witchfire 3dmark almost seems to ignore memory performance all together. I have ran it going from arm at 466mhz (233mhz fsb) and then up to 700mhz (350mhz fsb). which is aruond 50% increase in both fsb and memory speed. plus the timings stayed the same which would mean the faster ram has better latency. it showed a .3% difference. I would say in that it just about ignores memory performance all together. only other thing I could try on that to prove it further is dual channel but I can't yet untill I get two sticks of ram back in again, but with what I have seen with just frequency changes not affecting it I could almost promise you what the results would be. I believe that is part of why mr bungles doesn't not agree with 3dmark06. I know it was along those lines as to why I didn't like that test.



t4ct1c47
your if your trying to further what choggs was doing your going into the wrong thing if you diong the debate of speed vs timings. this was more of a divider vs performance issue.
rx7speed, While I may not agree with you on whether or not 3DM06 is affected much by memory speeds, you at least put up an intelligent and respectful discussion about it. You can state and back your position with discussion, rather than by attacking others and questioning their intelligence or other faculties.
    
CPUMotherboardGraphicsRAM
E4300 @ 3150MHz Gigabyte P35-DS3R Diamond HD2900XT 1GB 4GB DDR2-1000 G.SKILL 
Hard DriveOptical DriveOSMonitor
Samsung 750GB 32mb + 3x 250GB Samsung SATA Windows XP SP2 Samsung 941BW (HANNS.G 28" on the way) 
KeyboardPowerCaseMouse
Old Dell Clicky-Type w/ Cherry Sliders PCP&C 510 SLI LIAN LI PC-7B Razer Lachesis 
Mouse Pad
Func F10.S 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
E4300 @ 3150MHz Gigabyte P35-DS3R Diamond HD2900XT 1GB 4GB DDR2-1000 G.SKILL 
Hard DriveOptical DriveOSMonitor
Samsung 750GB 32mb + 3x 250GB Samsung SATA Windows XP SP2 Samsung 941BW (HANNS.G 28" on the way) 
KeyboardPowerCaseMouse
Old Dell Clicky-Type w/ Cherry Sliders PCP&C 510 SLI LIAN LI PC-7B Razer Lachesis 
Mouse Pad
Func F10.S 
  hide details  
Reply
post #84 of 85
mind me asking why you don't agree on that?
and actually I just looked again at my 3dmark scores (which I could link but not pro member and can't screenshot due to work problems) but I ran one test at 350*9 the other at 233*9 it should of been 350*6 and 233*9. my mistake and this is what I was refering to in the previous post and another one ealier in the thread.
so that was kind of an uh oh I screwed up. but even then the first system should of kicked arse.
Main Test Results
3DMark Score 6188 3DMarks 3DMark Score 6167 3DMarks
SM 2.0 Score 2437 Marks SM 2.0 Score 2428 Marks
SM 3.0 Score 2807 Marks SM 3.0 Score 2802 Marks
CPU Score 1879 Marks CPU Score 1867 Marks
350*9 233*9
both clocked gpu at 688*1035
nothing was running in background. b ut since I screwed up on the test these are kinda voided out. but it does give something interesting there. you would think the first system at 3150mhz cpu speed higher fsb and higher ram speed would easily waste in the cpu scores. it doesn't

only other thing I can think of is the program even when doing cpu test is still limited by gpu. I noticed aquamark is like that. before I had my 1950xt it would score my cpu as something like 2k points, put the 1950xt in and my cpu score shoots up to like 20k+.

other then that though witchmark do you have any ideas why two setups both the same gpu but one running highe fsb, ram, and cpu speeds would score the same as system two running slower on everying other then gpu?
compy jr
(13 items)
 
  
CPUMotherboardGraphicsHard Drive
xeon 3350 at 3.6ghz Asus P5k-dlx HIS 7970 @ 1150/1900 Lots of them.  
OSMonitorKeyboardPower
vista ultimate x64/xp sp3 compaq 21" junk dell keyboard corsair tx850m 
CaseMouseMouse Pad
upgraded metal box You don't need a mouse Desk 
  hide details  
Reply
compy jr
(13 items)
 
  
CPUMotherboardGraphicsHard Drive
xeon 3350 at 3.6ghz Asus P5k-dlx HIS 7970 @ 1150/1900 Lots of them.  
OSMonitorKeyboardPower
vista ultimate x64/xp sp3 compaq 21" junk dell keyboard corsair tx850m 
CaseMouseMouse Pad
upgraded metal box You don't need a mouse Desk 
  hide details  
Reply
post #85 of 85
Here's my findings so far with regards to a SuperPi 1MB run. I ran SuperPi 30 times and saved the highest score attained in two different configurations. Both times I ran with my E6600 at 333FSB x9 (3.0GHz), though one run was with the RAM at 667Mhz 4-4-4-8 1T, and the other had the RAM at 800Mhz 5-5-5-12 1T. Seeing as people usually have to sacrifice lower timeings in order to run their RAM at higher speed, I'm trying to find out if it's more optimal to simply keep RAM running the same speed as the CPU (1:1), and use lower timings.

Low timings and low speed;


High timings and high speed;


I didn't expect the higher speed test to come out better than the lower timings test, then I understood why. There's not really any accurate way that any typical member on OCN could answer this question fully. Communication between the CPU and RAM is handled by the MCH, which performs drastically different from chipset to chipset. There's also the fact then when running my tests I would have to use the Unlinked mode of the 680i which would further throw overall scores off. To be fair, in order to run a thourough set of tests I would have to have accss to a number of diffrent systems with different chipsets.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › FSB and Dividers: My Findings