why is XBoxLPU's OCed GF4 4400 faster then my 4600?

[THUGSROOK]

ive seen some really stupid computer oddities in my time ~ but this one takes the freakin cake....
why is XBoxLPU's OCed GF4 4400 faster then my 4600? clock for clock

retested results....

3dm2k 32bit mode
GF4 4400 295x645 = 15014
GF4 4600 295x645 = 14894
GF4 4600 300x650 = 14996

3dm2k1se
GF4 4400 295x645 = 12219
GF4 4600 295x645 = 12106
GF4 4600 300x650 = 12174

so i ask again, why is the 4400 faster at the same clock speed?

:disgust: :| :Q :frown:

 

[Dug]

Because with the 4400 being o/c'd your raising its fsb. j/k

Actually I couldn't answer that one. I do know that I see 100 point differences in my 3dmark scores all the time, at the same speed.

 

[whiteboyatcal]

Why did the chicken cross the road?


I would say that if they are both running at the same speed the results are not that surprising, since variances in the microfabrication process alone could create much bigger differences. Factors such as mask alignment and material parameters will cause variances of performance in GPU IC's which vary from batch to batch. As process sizes become smaller, these issues become even more important.

 

[THUGSROOK]

oh i forgot ~ both cards are VisionTeks.

thx for the replies so far... still looking for a better excuse tho
(card latency?)

 

[DAPUNISHER]

Well, since I know your testing methodology is excellent I'm I know you would have accounted for actual memory speeds and amount of memory of each card, system config, used the same drivers, and run the benchies after fresh boots so I got nothin'! EDITED: because cards are from same company

 

[THUGSROOK]

hiya DAPUNISHER

yes all i did was swap out cards, clean out the drivers and reinstalled them. (29.42s)
i dont think there were any variables besides swapping out the cards.

of course this 4600 OCs higher then my 4400 was able too ~ so its not like im upset ~ just baffled?

 

[SupermanCK]

i think you system feels for your ti4400...they don't like big bro 4600...
btw...did you do several test for each or just one tests???

 

[THUGSROOK]

im still crankin up the 4600 ~ im @ 325x725 benchtesting still w/o problems.

ill go back and retest it @ stock speed when im done.
maybe it just needs to get aquainted with the rest of my system
i know the new drivers need to be burned in for a few hours ~

 

[CrazySaint]

What, you still have a lowly Ti4600?? Get with the times, man - Ti4600s are just so first-half 2002!

 

[LastRide]

Well knowing thugs he's probably got AS3 and one of those socket A Silver Orbs mounted to his graphics chip on the ti4400.Its not so much the 4600 cause that seems about right for the score with no O\C.The 4400 is scoring higher than it should for some reason unknown.I wish I had a analogy on the reason for it.To me it just looks you have a killer ti4400 that has some extra boost to it:Q.I wonder if the 4400 could be running at higher clock speeds than it shows at default.I am not so sure that would be possible.Back to analogy 1,you have a killer ti4400

 

[kylebisme]

Originally posted by: THUGSROOK
hiya DAPUNISHER
clean out the drivers and reinstalled them. (29.42s)

i dont think cleaning out the drivers was nessacary at all as they are both the same technology just diffrent clock speeds. speaking of that, did you check the clock speeds on both cards? ohh and lastly, you realy cant say one is faster with those scores just that they are comeing up about even, 3dmark has a prety high marjon for error, ~300 points.

 

[THUGSROOK]

well my margin for error was only about +/- 10 on the 4400.
my system is really solid

this is purely a "card" thing. there were no variables.
could it be a latency or buffer thing?

 

[LastRide]

Their should be about a 1000 point difference in cards.I never had 3D Mark2001 give me a 300 point difference.I have seen 12-50 point difference rerunning 3D in that area never more than that as far as I remember.As far as the latency and buffer thing I have no clue in that area.I was expecting a little higher score with your 4600 with your trophy chip at over 2.6ghz.

 

[THUGSROOK]

well i kind of completly dismissed this at first because i bought a used card ~ but it looks as tho it isnt even broken in yet :Q
scores are varying wildly kinda making verifying ocing difficult. ill have to "creap up" on its highest settings slowly til its gains more stability. pretty much the same thing i had to do with the 4400.

i still havnt retested stock speed but ill get around to it

 

[Mikki]

I'm at a loss, too. Nice scores tho Thugs! Jeez!

 

[THUGSROOK]

i think a few of you guys missed the point that the 4400 was overclocked ~ but still clocked less then a 4600 ~ and the 4400 was faster.

ill test stock speeds again on the 4600 to see if it improved after some break in.
overclocking results are really nice so far tho

<EDIT> actually now that i think about it ~ im gonna run tests on the 4600 underclocked to match my 4400 speed too. that could reveal some interesting results

 

[Mikki]

Yeah, I missed that. What was your default speed score?

 

[THUGSROOK]

a 4400 default speed is 275:550.

obviously what happened was i upgraded my overclocked 4400 to a 4600 ~ and it was slower out of the box :Q
ill get to the bottom of this yet!

 

[Mikki]

Sorry Thugs, I meant what were your 3DMark scores with your Ti4400 at default speed?

 

[LastRide]

I misread your post.I didn't know what the default settings were on those cards so I assumed you were running default.With an overclocked 4400 you should reach 4600 speeds at default.My visiontek 4200 overclocked 310/530 score was 12,684 1024x768x32.I am using the 40.41 driver that give me an extra 600-700 points on 3D.I don't see anything fishy since a 4200 can reach 4400-4600 speeds.A 4400 can reach 4600 speeds.The 4600 O\C'd will seperate itself from the other cards.

 

[SupermanCK]

how the testing go?

 

[ndee]

Well the diffrence is about 50 points, no biggie in 3dmark 2001.

 

[THUGSROOK]

i dont understand why so many of you are missing the point of this thread

Mikki ~ i have no idea

SCK ~ its "sweet spot" seems to be 320x714 so far.
(it has 2.8ns mem so i just cranked it to 714DDR ~ 1000/2.8x2=714 ~ i gotta retest higher mem speeds)
i still think it needs more breakin and ramsinks.

 

[THUGSROOK]

posted retested results....

 

[SupermanCK]

any chance you will retest the ti4400???