Ok, logic is now out the window..

[magnux]

Ok, my system:

EPOX 8RDA+ (C10 BIOS)
XP 2400+ (2300@200x11.5)
XMS3200 512MBx1
Radeon 9700
Ram @ 4-2-2-2 / CPU Aggressive

My friend's system:

EPOX 8RDA+ (C10 BIOS)
XP 2400+ (2172@181x12)
XMS3200 512MB (256MBx2)
Radeon 9700
Ram @ 5-2-2-2 / CPU Aggressive

My 3DMark: 15448
His 3DMark: 15537

Ok, there is a +128MHz and +19FSB difference.. yet, I still can't MATCH his score, he has +89Mark's higher than me!! Can someone explain this? Granted, he's utilizing dual-ddr.. but COME ON!

Our systems are IDENTICAL in software and settings. We are both proven to be rock stable.. I've let prime run for OVER 24 hours, and so has he. This makes -no- sense whatsoever.

[datamage]

I happen to be that friend. :devil:

Seriously though, it is odd that I score slightly higher than him even though his system is running everything higher than mine. Then again, it's only 3Dmark right?

- D.

[RandyB]

Could it be the difference in memory bandwidth? Dual DDR does make a difference does it not?

[magnux]

Originally posted by RandyB
Could it be the difference in memory bandwidth? Dual DDR does make a difference does it not?

Of course it does, but did you not realize that I am at 200FSB while he is at only 181FSB?? I'm at 2300MHz and he's at 2172MHz!! Sure, dual-DDR is great, but don't you think I should at the very least match his score???

[RandyB]

Not necessarily, I am no computer expert but I do know that just because you have equal systems does not mean you will have equal performance. I build 2 race motors every winter for the upcoming race season, identical components both assembled side by side. One always makes better numbers on the Dyno
than the other with the same fuel system used on both motors.
It seems IMHO that not all PC components are created equal and all it takes is for one piece of the entire system to be performing below it's potential to make the difference. Try matching his settings exactly MHz to MHz FSB to FSB driver to driver etc, etc,
then see how much of a difference there is between systems.

[magnux]

Originally posted by RandyB
Not necessarily, I am no computer expert but I do know that just because you have equal systems does not mean you will have equal performance. I build 2 race motors every winter for the upcoming race season, identical components both assembled side by side. One always makes better numbers on the Dyno
than the other with the same fuel system used on both motors.
It seems IMHO that not all PC components are created equal and all it takes is for one piece of the entire system to be performing below it's potential to make the difference.

Oh, I agree. We benchmarked our systems at default speeds, and they were pretty much on par with each other.

Since we're both gearheads, this may work:

I installed a T-60 turbo on my LS1 (SBC 350), and @ 3psi I dynoed 437RWHP.

Various other people installed the same turbo on their LS1's, and can't break 400RWHP @ 3psi. So, they need to resort to ~ 6psi just to match/beat me.

Why? Many factors.. tuning, fuel, etc.. but this isn't suprising, because my motor dynoed +25RWHP _more_ than theirs did stock.

That's not the case with the PC's... we took two systems which benchmarked nearly the same. Now, it's not even close. I've done _more_ to my system and I can't beat him.. *shrug*

I'm not really "pissed" because my PCMark scores are tremendous, but it's just the little things that irritate me I guess.

[mookydooky]

Are you sure your friend isn't accidentally OCing his AGP bus or using a software overclocking utility on his 9700?

[RacerX1]

I think dual channel at 181fsb beats a single stick of 512 at 200fsb for bandwidth. Might not show in a benchmark like Sandra but I bet in does in 3dmark.

If your friend lives close enough why not exchange ram modules and see what happens.

[datamage]

Are you sure your friend isn't accidentally OCing his AGP bus or using a software overclocking utility on his 9700?

Nope. My AGP bus is @ 66mhz, and I don't even think about o/c'ing my Radeon.

If your friend lives close enough why not exchange ram modules and see what happens.

We're actually going to do that one of these days to see what happens.

- D.

[EluSiOn]

I think that just proved the benefits of Dual Channel DDR...

instead of cpu fight bandwidth with gpu..... both now are fed with data at the same time.

[avatar82]

very interesting to see if dual channel makes a diffrence, you guys should swap ram and rebench

[RacerX1]

I've been doing some testing with my ti4200. With dual channel I gain over 400pts more in 3dmark compare to single. Eveything else being the same of course.

So I really do believe dual at 181 could easily beat 200 single in 3dmark.

[Aedan]

As many people have noticed, the extra bandwidth from DualDDR doesn't appear to show up in Sandra benchmarkings. Obviously this is due to the fact that the weak link is the FSB!

It does rather suggest that the AGP card makes use of that second DDR controller, whilst the CPU is hammering the first DDR controller.

What would be more interesting would be to drop back into standard DDR mode, and see if the results drop that way. The other thing to try is to make everything identical between the two machines (ie, ram timings, fsb, multiplier etc) and see what two equally spec'd machines will do. That way, you can rule out any problems with the resyncing of data.

Áedán

[Galifrey]

LOL 181mhz x128bit is MUCH faster than 200mhz x64bit

It doesnt show on sandra because sandra only supports 64bit wide DDR benchies... max bw of a single ddr system is 3200mb/s max bw of dual ddr system is 6400mb/s

Of course his system will bench faster as lobby high detail and car chase high detail are EXTREMELY bw dependant....

Its no good having a fast CPU and GPU if the memory cant feed it data fast enuff...

[Corn]

I already posted this in an other thread but no 1 responded...
It would be interesting if some 1 could try to run FSB @ 400
and the DUAL memory @ 333 (eg. 83%) and see what happens.

[EluSiOn]

i doubt that will work corn.... I tried 83% ratio.... it the ram cant take it 200mhz intially... it will not able to drop the 83%

[Corn]

hmm... so you meen that the ram runs @ 200 and then get dropped to 333 when useing the 83% option?
That didn´t sound to clever to have that option in the BIOS :/
on the other hand it would explain why my comp didn´t post
when I had it set like that.
but I have some not so speedy DDR sticks so I thougth it was something else...
The idea was to increase the CPU bandwidth since the DDR already has a lot to give when run in dual mode.
Well thanks anyway EluSiOn.

[Galifrey]

the increased latency between the memory and the cpu means you are better off running sync most of the time...

Although, your arguement holds a lot of water in theory...