Strange Score In 3dMark06... Help!

[trog100]

i dont notice any of those things ket.. hard to compare what we notice.. there are no figures to back it up..

all i read and every test i run plus what i notice tells me the amd 64 chip with its onboard memory controler has changed the rulles about the benefits of expensive fast memory..

i can mess with my memory speeds and timings and get big changes in a sandra bandwidth benchmark but in real life tests or observations i see no benifit or loss..

for example the fear in game test thingy drops or goes up depending which way u are going by one frame per second (average) difference by altering memory timings from 1T to 2T..

quake 4 plays pretty much the same as well whatever.. so do all my games..

running cas 3 at a super low speed of 320 as opposed to cas 2.5 at a memory speed of 420.. we are taliking big differences here.. produces a difference of about 25 points in 2005 around the mid 6000s mark.. i assume it would produce double that around the 12000 mark..

cpu speed makes a big difference to 2001 scores.. but we are talkng 600 frames per second here and scores of over 30,000.. i dont play real games at those kinda fps..

what else can i believe other than memory speeds or timings dont make deal of difference with the amd chip and its onboard memory controler..

believe u me i have tried hard to spot these real life (not sandra) differences.. i cant find any..

and we were originally talking about improving a 2006 score.. i recon i have proven that memory tiimings or speeds dont alter that much.. which was my main point..

sooo (at least where gamng is concerned) cpu speeds make far less difference than folks think they do.. and neither do memory speeds or timings..

i also cant notice a difference while gaming of whether my cpu is at 2.2 or 2.9 gig.. that i suppose would be another myth destroyed.. overclocking your cpu by 700mhz dosnt produce any noticable benifits.. he he he

if i didnt have it i would dream about the benfits of it.. now i have it i know just how little its worth.. now a faster grfx card.. thats a different story.. if u game thats where all your money should go.. i can notice differences with one of those..

praps i am just that stupid i cant notice all these other expensive performance differences.. praps they really are there and its just me that cant see em.. he he he.. either that or the emperor really dont have any clothes.. he he he

trog

 

[Ketxxx]

If you want figures just look at memtest, or run everest, or simply just time how long it takes a level to load. Thats just 3 of many ways. in doom especially you should of noticed level loading time differences, they hit you in the face like a freight train.

 

[trog100]

memtest is just another memory bandwidth test.. they dont relate to real life system performance..

for example if i do something to my system that doubles my score in 3d mark 2005 or 2006.. in real life i see a similar gaming frame rate increase.. or at least a pretty noticable one..

if i do something that doubles my memory bandwidth in any memory speed/bandwidth benchmark in real life i see no frame rate increase or any other difference in any game thats user noticable..

if for example i take the cpu speed benchmark in sandra that tells me a dual core cpu will exactly double my cpu performance to be true.. i go out and buy one and run a game.. i will see bugger all difference in frame rates..

tell u one thing that will hit u in the eye as regards game loading times.. i do have all these heavy games by the way.. that is having 2 gigs of system memory as opposed to just the one..

whilst argueing that memory speeds and timing make little difference to real life performance.. i would argue the opposite about the amount of memory.. having two gigs hits u in the eye as regards loading times.. so does the size of the textures u use..

level loading comes of your hardrive anyways.. cant see how memory speeds are gonna speed up that slow beast much.. with more memory less has to come of the hardive.. that speeds things up significantly..

its all these memory and cpu benchmarks we run that are going around proclaiming the wonders of the emperors new clothes.. they have a lot to answer for.. except 3dmark 2005.. where games are concerned that compares well with real life.. memtest.. sandra and everest or rightmark dont..

a good overall system test would be 3dmark 2001.. its not just grfx card limitted..

but the only thing that i need PC power for is for playing games.. and even then it pretty much all down to the amount of system memory and the grfx card..

all my games play very smoothly now i have a card with enough grunt and 2 gigs of system memory.. they also load quickly.. even with cas 3 memory.. honest..

my system is software clocked.. i can change things with the click of a button.. i really can mess about and look for differences.. i know no one is gonna believe a change in cpu speed of 500mhz and a relative change in memory timings or speeds aint gonna show up in a game or general system usage.. but it dosnt..

he he

trog

 

[Ketxxx]

It does, I can physically see it with my system, you just need to know where and when to look and know exactly how a system would react according to the parameters changed, and as any pro overclocker will tell you, software is NOT the way to overclock, for stability and performance reasons, it needs to be done through the BIOS, and even with VCs hardware fixing anOC will neta few more points still over software.

 

[MZ3692]

I agree completely, Ketxxx.

 

[Steevo]

I scored 4800 in 3D06 but........

1280X1024

6X-16X All high quality HDR is available reflect all.

 

[trog100]

its rather handy having software that controls cpu speeds automatically u kinda get used to it and miss it if it isnt there.. i used atitool to auto control my x850 grfx card speed and abit guru to control cpu speeds.. it did all work beautifully..

but a couple of days back i bought a x1900 grfx card.. thats buggered up my nice reliable grfx card control..

tonight i downloaded and installed the latest via hyperion drivers.. my abit guru seems to be doing its job as it always has and recording the cpu speeds its supposed to.. but.. unlike it used to do.. cpu-z is showing me to be at a fixed cpu speed of 2.75 gig... he he he

i always used to check the abit guru stuff with cpu-z to make sure it was doing its job.. its always checked out 100%.. but tonight cpu-z tells me one thing.. abit guru another.. he he

soo ket u might be right about software clocking not being as reliable as bios clocking.. he he.. best to ignore my cpu slow down figures i posted earlier i aint sure it was slowing down.. he he

the two different memory timings cas 3 and cas 2.5 would be correct they were altered in the bios.. but as of now i aint sure what the hell my cpu was running at for the cpu slowdown down figures..

trog

ps.. my bios is set at 11 x 220.. which would be 2.4 gig.. guru tells me what it tells me depending where its set.. everest tells me 11 x 250 which would be 2.75 gig.. cpu-z backs up everest.. he he he.. so i have three different things reporting three different speeds.. something has gone apeshit.. he he he.. could be the hyperions i installed tonight or it could have been like this for a couple of weeks.. dunno.. some iinvestigating needs to be done.. he he

 

[Tatty_Two]

If you want figures just look at memtest, or run everest, or simply just time how long it takes a level to load. Thats just 3 of many ways. in doom especially you should of noticed level loading time differences, they hit you in the face like a freight train.

But I think that in part that is down to quantity of memory and not so much quality/speed/timings but cant argue with your point however I think you would find in most cases, once you are into the game (loading completed) you would see little or no difference.

My thoughts for some time have been identical to Trog's and believe him to be right, I just have not had the time (or the patience) to back it up with such eloquent testing and facts!

 

[Steevo]

I have my system set at 230FSB 1:1 2.8Ghz 2.5,3,3,7 1.45v boot. Then I can turn it up to 234MhzFSB 2.5,3,2,6 1.55v for gaming as it is still a tiny bit unstable at those speeds using ntune and CBI.

 

[trog100]

its not just me that says memory timing and speeds dont make a deal of difference in real life performance.. there is a recent article on toms hardware that takes a whole bunch of different memory types and makes.. and they run tests on em at different timings and speeds.. and at different cpu speeds..

they are trying to find out which is best.. loose timings and fast speeds or the other way around.. slower speeds and fast timings..

at the end of it all they fail to reach any definite conclusion as to which is best.. he he

but at the end of it all what they do say is.. in real life performance tests.. none of it seems to matter much..

they finish off by saying.. if gaming is your thing the money is best spent on a better grfx card not fast memory..

if gaming aint your thing the money is better spent on a faster cpu not fast memory..

thats about as close to saying expensive ram is a waste of money as u will get from any major hardware review sight..

course that wont stop em writing oodles of stuff comparing which over priced memory is better than which other over priced memory even thow none of it matters.. cos in real life tests the difference is so marginal it cant be detected..

they have to write about something..

trog

 

[yogurt_21]

okay.. time for some myth destroying..

three 2006 runs.. with amd 64 single core based system.. single x19000xtx grfx card.. ccc overdrive at 690/800

run one cpu at 2.9 gig.. memory about 420 at cas 3 and 1T..

2006 score.. 5237..

run two cpu at 2.9 gig.. memory about 420 at cas 2.5 and 1T..

2006 score.. 5239..

run three cpu at 2.4 gig.. memory about 370 at cas 2.5 and 1T..

2006 score.. 5166..

soooo.. it seems faster memory timing make bugger all difference in 2006..

slowing the cpu speed down by a massive 500mhz and slowing the memory down by 50mhz or so loses me about 70 odd 2006 points..

sooo 2006 aint gonna prove me wrong at all.. assuming my machine aint some kinda wonder machine that differs from all other basic amd 64 systems.. its all pretty much down to the grfx card just like it is with 2005..

that loss of 70 point would be mostly down to a lower cpu score.. soo it seems messing with memory timings with the amd 64 chip is a waste of time..

trog

you do realize all that shows it that you've a cpu bottleneck at 2.9 just like you have one at 2.4

check this out
http://service.futuremark.com/compare?3dm06=188552
my score witha single x1800xt and a single core cpu.
http://service.futuremark.com/compare?3dm06=171397
eva 2000's score with a dual core but single xt clocked lower than my xt you'll not that hi cpu makes over a 600 point difference both in the cpu score and the 3dscores as his 756/900 is almost identical to my 778/927 which should create a much bigger gap but doesn't because his dual core cpu at 3.1GHZ doesn't create a bottleneck like my single core does at 2.9.