PDA

View Full Version : Lower multiplier gives a lower 3dmark score.!


demonized
01-14-03, 06:14 AM
http://forums.overclockers.co.uk/showthread.php?s=&threadid=20513

I found this review of an Epox 8rda+ mobo. If you look at the last 3dmark benchmarks youll see that he got a lower score when he overclocked the system.


stock speed score:
(13x166) 2160= 14911
http://service.madonion.com/compare?2k1=5232359

overclocked speed score:
(12.5x200) 2500 = 11299
http://service.futuremark.com/compare?2k1=5232374

Why did he get a lower score? The same thing is happening to me.

slightly overclocked: default multiplier used.
(12x142) 1704 = 12000+scores

(10x170) 1700 = 10000+scores


Can you guys give me an explanation for this..

pelly
01-14-03, 06:39 AM
Try running the benchmarks at a bit lower setting...( in your case, try a 160FSB ). This "answer" may not apply to every case...but it certainly provides some light for most...

As the systems nears its limitations...things can get a bit hectic as tolerances begin to get thrown out he window. Often, when a system is on the verge of being unstable...errors can develop as the data is being written to memory and communicated...The lower scores are a result of this as the system cannot communicate as efficiently as before...causing the additional cycles to get the job done and resulting in a lower score...

:D

demonized
01-14-03, 09:03 AM
So is it my motherboard? I just got an OCZ pc3500 just to find out i would have the same problem. lower score but high memory bandwith and still cant go over 170fsb. now i have to wait for my next paycheck to get an nforce2:( could it be that im using a palomino?

pelly
01-14-03, 10:22 AM
Depending on what chipset your mobo is using...you'll have limited overclocking success...I know that I had a KT4 Ultra ( VIA KT400 ) that didn't like any FSB over 165...

9 times out of 10...the limiting factor ( when using good RAM like you are ) for overclocking is the motherboard...The CPU is taken out of the picture once you start lowering multipliers and raising FSB ( not really overclocking as the final MHz remains the same for CPU ).

Let us know what mobo you're using...Jump around and dig up a few reviews of the mobo and find out how their board overclocked using their equipment...That will give you some good insight...

Were you raising overall MHz for the CPU when you overclocked ( ie: trying to hit 1800MHz...etc... ) I would say that the power supply might be the culprit...However, since you're not aiming to draw any additional juice here...that should be ok...

One thing you might want to try is bumping the DIMM voltage up to 2.8V....I know that Geil's high-end DDR is meant to run at its advertised spec using that high a voltage...Give it a shot and see what sort of results you get...

Good luck...I hope something in all this helps!

:D

demonized
01-14-03, 10:48 AM
Heh, ive read like all the reviews for my mobo(MSI KT3 ultra). Most reviews didint do any overclocking. But there was one review that was able to get upto 182fsb. the other reviews that werent able to get pass 142 were using kingston ddr modules which is the one i was first using so i though it was the problem.
I have my cpu voltage at 1.85 already which only gave more heat.
Are there any bios setting that i may have enabled thats limiting me? thx. for helping on this pain in the ass non stop upgrades.


:rolleyes:

pelly
01-14-03, 11:09 AM
You shouldn't need to touch your CPU voltage ( leave that at stock or up to .15V above )....change the DIMM voltage ( voltage to memory ) up to 2.8V. That should make a big difference...

:D

demonized
01-14-03, 08:39 PM
Originally posted by pelly
As the systems nears its limitations...things can get a bit hectic as tolerances begin to get thrown out he window. Often, when a system is on the verge of being unstable...errors can develop as the data is being written to memory and communicated...The lower scores are a result of this as the system cannot communicate as efficiently as before...causing the additional cycles to get the job done and resulting in a lower score...

:D

So whats the point of overclocking upto 200fsb when your just gonna get lower performance? So its not advisable for people to go that far even though their system can run stable at those speeds?

pelly
01-14-03, 10:24 PM
So whats the point of overclocking upto 200fsb when your just gonna get lower performance? So its not advisable for people to go that far even though their system can run stable at those speeds?

You misinterpreted what I'm saying...

Systems that CAN handle high FSB's ( ie: 200MHz for example ) while remaining stable see an ENORMOUS performance gain. My point is that systems that are borderline stable at a given overclocked speed can actually score lower than a stable default-clocked system b/c of write errors.

Here's an experiment to illustrate my point:

Try to find the maximum FSB that your system can run and still get by ( albeit, barely ) a run of 3dMark or some other stressful benchmark...

Back the FSB down 5MHz and do the same...

Back another 5 MHz and do the same...

Looking at the results, you will almost always find the system 5-10MHz below the supposed "max" FSB will yield the fastest results...Why? B/c of errors...

:D

demonized
01-15-03, 12:49 AM
So the guy who did a review on the epox nforce2 mobo with an athlon xp 2700 cant handle it? So thre's no point for me to get and epox nf2 since the result would be the same? would you suggest getting a Asus nf2 instead? Im so confused.:confused:
epox,asus and msi are the only nf2's available here in toronto.
Thx. for your help:beer:

demonized
01-15-03, 02:01 AM
i checked the thing on cpu which people seem to talk about alot.
here's what it says.

XC-AX2000DMT36
AGOIA0217MPM
41053802057

Whats that :rolleyes: mean?
Also, will mixing different brands of ram damge anything.
I wanna put my kingston pc 3200 256mb stick back. IS this ok?

-=DVS=-
01-15-03, 02:50 AM
Lol why not try lowering bus speed and riseing multiplier :rolleyes: , and see what we gonna happen :p

StealthHawk
01-15-03, 04:09 AM
not all hardware is created equal. things pumped out at the factory are each slightly different. so in other words, YMMV. trading one "exact" mobo for another can change your results, same with CPU, same with RAM, etc, etc.

it's why there are never guarantees when overclocking anything. some parts got it, some don't.

pelly
01-15-03, 09:35 AM
In my experience, mixing RAM = bad idea when overclocking...

demonized
01-15-03, 09:44 AM
I tried to overclock again was able to do this with a 10x multiplier.
Only problem is its not steady.

demonized
01-15-03, 09:50 AM
right now im at 11x166. And its stable.
I also fixed the low 3dmark score when using high fsb's.
I just had to put this setting FAST COMMAND: to normal.:D
with a 9x multiplier i was able to get as high as 192 but then again it wasnt stable. I cant wait to get an nforce2. :p
i need more voltage.:clap:

demonized
01-15-03, 09:51 AM
i forgot to attach the pic.

demonized
01-15-03, 10:08 AM
One last question? which would cool mu CPU better?
Vantec VT-VA4 w/ TMD Fan or
SLK-800 with an 80mm Panaflo high fan


btw. pelly. your work on hardocp now right? Have you ever tried or mistakenly typed www.hardcop.com? It's a freakin porn site :p

pelly
01-15-03, 02:27 PM
LoL...With "hard" in there...it's no surprise...

Without question...the best air coolers out there are the Thermalright SLK-800 and the Swiftech MCX462-U...

:D

|JuiceZ|
01-15-03, 03:16 PM
Originally posted by demonized
btw. pelly. your work on hardocp now right? Have you ever tried or mistakenly typed www.hardcop.com? It's a freakin porn site :p

lol, I'm suprised I haven't done that yet :D

Anyways, for you guys running stable w/ higher fsb (150+) what kind of limitions have you discovered in terms of vid card o'cing, since the agp bus is obviously already stressed?

lothodon
01-15-03, 06:13 PM
ok, not here often, but i have to tell you all this.

currently running an epox rda+, with 512mb (2x256, dual channel) of corsair pc3200 and an athlon 2400+...all at 200mhz fsb and stable.

AND FAST, my 3dmark2001 score is 15,5xx when running like this (200x10) and was only 13,8xx when running at spec (133x15).

so i'm leaning toward user error or memory error in the review. it's also proven that running the cpu and memory at different speeds really hoses your performance. perhaps this was the problem in the review.

added - the agp and pci bus speeds are locked on the nforce2 boards at 66 and 33 respectively. the pci is always at 33 and the agp bus is fully adjustable in the bios.