The OC War....Multi or Ref. Clock?? - Overclock.net

Altimax98 · 05-01-09

OK well, ive asked this question many many times. So I figured I would do a test to see what was the better OC - Multi Vs. FSB war had started.
Heres the rig it was tested on:

Asus M4A79T Deluxe AM3 motherboard \ 2x2g Corsair 1600mhz ram \ HIS Radeon 4870 OC'd to 805/1010 \ AMD 720BE x3 Phenom II \ Corsair 850TX PSU
I used the trial version of 3Dmark06 because I dont have 3DmarkVantage and I dont own 06, so I used the trial at trial settings and the resulution was at
1280x1024 for the test. My monitor used for testing was a Acer x233H 23in LCD plugged in with a DVI input. Windows 7 x64 was the OS on a SATA drive.

I first started by resetting my Bios to stock and leaving all settings stock except for my NB which I had at 1.4v, my HT @ 1.3v and Ram @ 1.8v. ON all OC tests my CPU V was < 1.5v
I had my ram setting set to the setting that would give me 7 latency for my ram.
I had ACC set to AUTO because I cannot set it to Per-Core due to my 4th core being a real dud and crashing my system.
I did the different tests with about 10mins cool down between them. I monitored my temps and I would let them sit at idle temps for 2mins before OC to the next level.

Heres the Stock results:

Bare Bones Run, stock settings 2.8ghz

Now for the OC. First the multiplier OC:

Multiplier OC to 3.4GHZ

Multiplier OC to 3.6GHZ

And thats the tests for the Multiplier OC's.

Now heres OC by FSB (ref clock). I left the multi at 14 just like stock and raised the ref clock till it achieved desired clocking

Heres the Results:

FSB OC to 3.4GHZ

FSB OC to 3.6GHZ

Now I decided to run my normal OC of 270mhz Ref Clock, 3.5ghz w/1.475v and a multi of 13:

Now I decided to run 2 more tests, 3.6ghz Multi OC with 2700mhz NB speed:

And 3.7GHZ Multi OC with 2700mhz NB Speed:

Now obviously the HB Speed helps, alot... In fact it added close to 800 points to the scores.

So a recap:

OC Style Core Clock Multi.. FSB NB ... Points
Stock ........ 2.8ghz .. 14 .. 200 ..2000 .. 13605
Multi ........ 3.4ghz .. 17 .. 200 .. 2000 .. 15116
FSB ........3.4ghz .. 14 .. 245 .. 2450 .. 15082
Multi ........3.6ghz .. 18 .. 200 .. 2000 .. 15559
FSB ........ 3.6ghz .. 14 .. 257 .. 2575 .. 15283

EXTRA TESTS
FSB ......... 3.5ghz .. 13 ..270 .. 2700 .. 15798
Multi/NB .....3.6ghz .. 18 .. 200 ..2700 .. 15999
Multi/NB ....3.7ghz .. 18.5 .. 200 ..2700 .. 16033

In each test the Multiplier OC performed better then the FSB overclocked counterpart, with the exception of rasing the NB speed to 2700mhz.
Example: the FSB overclock to 3.5ghz w/2700mhz NB speed was over 600 points BETTER then 3.4ghz Multi OC and 200points over the 3.6ghz Multi w/2000mhz NB speed.
I think this can be remidied by lowering the multi to the desired level while raising the FSB to 2700mhz.

One little thing I noticed is that when I used AMD overdrive and ran 3dmark06 right after 3Dmark would pick up a slower CPU speed then actual and verified by CPUZ.

So in all I added 2428 points to my system with a smarter CPU OC. The best results came from a multiplier + NB speed bump, with my top score of 16033... Not bad for a triple core, single 4870 system.

SubstancenUnknown · 05-01-09

cool man good info ill have to mess around with mine now...lol

0m3g4 · 05-02-09

Good work, I find that very interesting.
+1

Nenkitsune · 05-02-09

according to your tests, the reference clock overclocks actually did BETTER (if you go based off the CPU scores)

however, that is because you also ramped up your memory frequency, your HT frequency, and your NB frequency.

I've found that for best stability though, it takes a combination of a reference overclock and a multi overclock.

Altimax98 · 05-02-09

Quote:

Originally Posted by Nenkitsune

according to your tests, the reference clock overclocks actually did BETTER (if you go based off the CPU scores)

however, that is because you also ramped up your memory frequency, your HT frequency, and your NB frequency.

I've found that for best stability though, it takes a combination of a reference overclock and a multi overclock.

You are correct in that. But I was only counting the total number for the OC. But you have a good point there.

Nenkitsune · 05-03-09

by including the entire test scores though, you're also counting in variables in how the GPU runs, which is why when overclocking is concerned, I focus more on the CPU scores, then tweak the GPU to get better results.

Altimax98 · 05-03-09

Quote:

Originally Posted by Nenkitsune

by including the entire test scores though, you're also counting in variables in how the GPU runs, which is why when overclocking is concerned, I focus more on the CPU scores, then tweak the GPU to get better results.

The reason I counted the entire score was because i was not sure a total system OC by FSB would be reflected in the total score.

The GPU settings never changed, They were always the same, so whatever fluctuations occured would also occur between tests of just the cpu.

pewpewlazer · 05-03-09

You really need to run more than one trial per clock speed and reboot between every single run for there to be any validity. Running back to back results in varying SM2 and SM3 scores as your results show.

Altimax98 · 05-03-09

Quote:

Originally Posted by pewpewlazer

You really need to run more than one trial per clock speed and reboot between every single run for there to be any validity. Running back to back results in varying SM2 and SM3 scores as your results show.

I thought the same thing. So I did another test on my same settings that I got 16033 points on, and got 15957. So even restarting it doesnt seem to matter as to helping the fluctuations. Those numbers seem to fluctuate anyways.

I did the test 1 day later. So I would guarantee it wasnt too hot or anything.

		Overclock.net - Overclocking.net > AMD > AMD CPUs
The OC War....Multi or Ref. Clock??

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