A puzzle about 3870 in C/C

[Chewy]

I remember hearing that for a gtx to have no bottleneck you need a c2d @ 3.2 or 3.4ghz.. now a 2.66ghz c2d is faster than an Amd x2 3.2ghz black box edition.. and the black box edition could only hit about 3.4ghz I think.. its not an overclocker, it just comes at a decent speed and put out alot of heat.

I think its your cpu, your cpu is prob equivilent to a c2d @ 2.8ghz.

 

[cefurkan]

that site gave fake IQ comparssiong scenes
and told ati IQ was better

but when i gave same scene all of us saw that no difference

so dont trust their scores

 

[trog100]

http://www.legitreviews.com/article/605/10/


interesting how a 1200 dollar intel cpu only gets 15000 odd in crossfire..

bang for buck i would definitely win.. he he he

my single card score is pretty good comparatively as well.. my bang for buck is even better there..

trog

ps.. they are only showing a 30% or so increase in crossfire as well.. not that clever from where i stand..

 

[Chewy]

Never go with a $1200 chip unless your rich and dont mind paying that much your 6000 cpu is a great bang for your buck but a e6750 is still faster for the most part @ stock clocks.. Im not sure if the 6400 black edition is 3.2ghz or 3.4ghz but the e6750 takes that by a little bit @ 2.66ghz.. oh the 6400 is 3.2ghz stock.

so you think the 20,000 point score was fake? it should be you have the best amd board and should be able to match that.. unless thier scoring with a massively overclocked c2d @ like 4ghz+ or something.

 

[Deleted member 24505]

Thats some cpu in the test system.

 

[trog100]

my problem at the moment is i seem to be only dude on the planet with a pair of 3870s sat in an amd 790x chipset board..

i have nothing similar to compare with.. even futuremark dosnt help it just says nothing to compare with.. and no one has reviewed a similar set up..

basically the 790x board dosnt perform with a single card in it as well as the (very) entry level abit na52 it replaced..

in crossfire mode the results are at best disappointing.. at worse disastrous.. he he

soooo the second card and 790x chipset board has to go.. its back to one card and the abit na52 board for a while..

i could keep the board with just one card in it but there are several things i dont like about it..

trog

ps.. course i might try the 3780 X two when it arrives in my single slot.. he he

 

[Mussels]

theres a vista multi GPU hotfix on the main page here at TPU, check it out and see if it helps.

if you're under XP... perhaps the reviews was in vista?

 

[bacon612]

my problem at the moment is i seem to be only dude on the planet with a pair of 3870s sat in an amd 790x chipset board..

i have nothing similar to compare with.. even futuremark dosnt help it just says nothing to compare with.. and no one has reviewed a similar set up..

basically the 790x board dosnt perform with a single card in it as well as the (very) entry level abit na52 it replaced..

in crossfire mode the results are at best disappointing.. at worse disastrous.. he he

soooo the second card and 790x chipset board has to go.. its back to one card and the abit na52 board for a while..

i could keep the board with just one card in it but there are several things i dont like about it..

trog

ps.. course i might try the 3780 X two when it arrives in my single slot.. he he

I think we are CPU bound, my friend.

See Link.

http://www.xtremesystems.org/forums/showpost.php?p=2606191&postcount=215

Or, perhaps its a driver issue with us on ATI chipset mobos? See this link for an intel guy with a lower score too.

 

[Mussels]

lol! i'd forgotten about that - 3dm06 gives a HUGE points boost to quad core systems, beacuse of teh CPU test - the FPS isnt much higher in the game tests, or in real world gaming.

Compare to other DUAL CORE users, for accurate results.

 

[trog100]

2006 is out of date.. we need a new one which slows down frame rates and loads the grafix card far more..

the only thing futuremark are doing at present is selling intel CPUs.. he he he..

when high end scores reached 10.000 futuremark released a new benchmark slowing frame rates down..

high frame rates test the cpu and system.. low frames rate test the grafix card..

basically futuremark and 3Dmark 2006 are helping kill off AMD and giving a false (from a gaming perspective) boost to intel sales..

deliberate or not i dont know and wouldnt care to speculate..

in real life u do not need an overclocked intel system to play games well.. in real life at the level of frame rates we play our games at the super fast cpu makes no difference..

in the benchmark would yep.. in the gaming world nope.. they are two different worlds with two different objectives..

trog

ps.. C/F does work in things that are optimized for it and dont run frame rates so high as to be system limited..

an example.. 3Dmark 2005.. at its default setting of 1024 x 768 and no AA.. my single card sores 18000..

at the max resolution my monitor lets me run.. 1680 x 1050 everything fully maxed out 8 x AA.. it scores 16670 in crossfire mode.. a very decent performance increase indeed..

so assuming the game/driver is optimized for crossfire (most dont seem to be) it will let u run a higher resolution (if u can) at the same game settings.. the only game i have that would benefit from this is crysis.. but crysis aint optimized for C/F..

so i get no benefit from crossfire..

 

[Mussels]

2006 is out of date.. we need a new one which slows down frame rates and loads the grafix card far more..

the only thing futuremark are doing at present is selling intel CPUs.. he he he..

when high end scores reached 10.000 futuremark released a new benchmark slowing frame rates down..

high frame rates test the cpu and system.. low frames rate test the grafix card..

basically futuremark and 3Dmark 2006 are helping kill off AMD and giving a false (from a gaming perspective) boost to intel sales..

deliberate or not i dont know and wouldnt care to speculate..

in real life u do not need an overclocked intel system to play games well.. in real life at the level of frame rates we play our games at the super fast cpu makes no difference..

in the benchmark would yep.. in the gaming world nope.. they are two different worlds with two different objectives..

trog

ps.. C/F does work in things that are optimized for it and dont run frame rates so high as to be system limited..

an example.. 3Dmark 2005.. at its default setting of 1024 x 768 and no AA.. my single card sores 18000..

at the max resolution my monitor lets me run.. 1680 x 1050 everything fully maxed out 8 x AA.. it scores 16670 in crossfire mode.. a very decent performance increase indeed..

so assuming the game/driver is optimized for crossfire (most dont seem to be) it will let u run a higher resolution (if u can) at the same game settings.. the only game i have that would benefit from this is crysis.. but crysis aint optimized for C/F..

so i get no benefit from crossfire..

speaking as a gamer with intel and AMD systems, i can say that the AMD system certainly holds back my GTX, as theres really no speed difference (games and 3dmarks) between the GT and GTX - 5% at best. (10% with 8x AA and above)

its not so much that you need intel to game, its that you need a massively OC'd amd or intel to make use of the latest ATI cards... they really ARE CPU limited... and intel just have the best overclocking chips atm.

(for reference, the AMD is my mini system - FX-62 @ 3GHz)

 

[trog100]

your card is powerful enough to play all games at the maximum resolution your monitor will let u fully maxed out.. crysis being the exception..

bascially any grafix card once the frame rates start to get high will be held back by any cpu no matter how fast.. a faster one will always run more frames.. so in that sense one can say the system is always the limiter..

but in the gaming world where low frame rates are (or can be) the problem a faster system wont help.. only a faster grafix card will do that..

so in one sense u are correct in another u are not..

so a faster system with any card at whatever resolution the card will run say 150 fps at will make it go faster.. but at a resolution the card can only run 20 fps at the faster system wont help it in the slightest..

the faster system will turn 150 fps into 200 fps but it will not turn 20 into 25..

supreme commander with a million units all in action at once might break this rule but it is an exception..

trog

 

[ROE_HUNTER]

In 3dMark06 set at default I got the same kind of scores you did (9680 single,11900 Crossfire), but if I upped to my native(1680x1050) and set 4xAA and 4xAniso I got 5648 single & 9696 Crossfire. So being CPU limited might be a problem.

As for Crysis the Hotfix drivers allow Crossfire to work very well for me . 7.10 & 7.11 caused a slowdown like the one you are experiencing .I really don't notice any other games that benefit from Crossfire. Maybe future drivers will help us 3870 Crossfire users out . Being that the Crysis Hotfix "fixed" my Crossfire for that game I do have hope.

 

[zOaib]

my actual 2006 scores

one card
3DMark Score 11180 3DMarks
SM 2.0 score 4867 Marks
SM 3.0 Score 5411 Marks
CPU Score 2577 Marks

two cards
3DMark Score 13037 3DMarks
SM 2.0 Score 5650 Marks
SM 3.0 Score 7174 Marks
CPU Score 2534 Marks

i see a puzzle that with one card their system is about 10% faster than mine..

with two cards its 70% faster than mine..

the other puzzle is why am i only getting something like a 15% performance increase with two cards over one card.. lets not forget that if anything on this planet is fully optimized to take advantage of two cards 3Dmark 2006 is..

their single card score at actually 12500-ish isnt that much faster than my single card score of 11180.. this is fully explained by their system being faster than mine.. no probs here.. we see a difference of about 10%..

what we are seeing is their overclocked intel system with one card is 10% faster than my overclocked amd system.. 10%

what we are also seeing is their overclocked intel system with two cards is 70% faster than my overclocked amd system.. 70%

now i see a puzzle even if some of u guys dont..

nobody seem to want to bench with anything but overclocked intel systems any more.. which is a shame.. cos if what i am seeing is the reality amd are dead in the water which isnt something i want to accept just yet..

is what i am seeing the reality or just a fluke with my system.. ???

i do find it a bit amusing that folks see no puzzle and just accept its an intel amd thing.. also a bit frightening concerning the fate of amd.. he he

trog

i scored 15k + with my setup , with 2 cards and my quad is at 2.61 ............. its definately the amd cpu ............... also the x38 chipset runs cf great.

AMD in cpu si DEAD till they figure out their crap is stinking now. Ati is doing okay ! =P

 

[jpierce55]

O.K. Z what was the single card speed. What is your ratio of increase on the 3dmark?

 

[DOM]

O.K. Z what was the single card speed. What is your ratio of increase on the 3dmark?

i say he getting about 12k almost 13k with one card and if his running at 2.6Ghz im sure the cards are stock

 

[jpierce55]

It seems many people miss the point he is mad about the ratio of 2 vs 1 not the number itself.

 

[DOM]

It seems many people miss the point he is mad about the ratio of 2 vs 1 not the number itself.

well I know that, he more mad that intel is better then amd i think and 3dm06 is all about cpu power

he should try games but that dual core is holding the cards back in 06 and intel is a better gamer cpu you can still play games with amd at the same settings but with intel your going to get better fps

 

[trog100]

folks dont read the thread do they..

intel scores way higher in 2006.. yep i agree entirely..

intel plays games better.. nope i disagree..

and yep it is the ratio between two and one i am on about.. not the fact amd scores worse in 2006..

i have just the one game i cant play fully maxed out at 16080 x 1050 on my amd system.... we all know what that one game is..

assuming the folks that say intel plays games better have decent grafix cards i wonder just how they know this.. what game do they have that plays better on an intel system than an amd one..

crossfire even with the hotfix drivers plays worse in crysis than with just the one card..

now intel plays better guys.. try and convince me that an overclocked intel system would make my crysis playing experience better..

we can put it to the test.. anybody have an overclocked intel system.. a 3870 card and crysis.. ????

trog

ps.. how about calling it the amd versus intel crysis challenge.. he he he.. crysis being the only thing we have that actually brings a modern high end grafix card to its knees..

 

[DOM]

do you have that cysis bench tool thing cuz I can play it but not on all high @ 1680 X 1050 ?

I wanna see what fps you get on 1 vs 2 it shoulnt be much as the game hasnt fixed the CF on it

 

[DOM]

but heres a review that has crysis with a Q6600 @ 3GHz

1680x1050 0xAA 0xAF, Shadows/Post Processing: Medium, Other settings: High, DX9

 

[trog100]

it has to me the same grafix card DOM to make any sense all..

running the gpu benchmark i get 40 average at the settings i chose to play at..

all on medium except textures on high and shaders on high.. at 1608 x 1050 = an average of 40 fps..

i cant play on high either i think from memory i get about 27 fps in the cpu benchmark.. not playable by my standards thow some would claim it was..

now u have the fast intel system with not so fast grafix card.. i have the slow amd system with a faster grafix card.. we are opposites..

if my system out scores yours thow it does kinda make my point..

as for crossfire in crysis it actually makes the frame rates worse.. it has a negative effect.. the 40 fps drops to 35 fps..

trog

ps.. looking at the link u posted it does seem that sli works in the game thow its just crossfire that dosnt.. the hotfix driver did improve the drop off but i had stopped using them for the figures i quote above

i dont think the fps i see on that website are playable either at the resolutions they claim.. thats just my opinion thow..

 

[yogurt_21]

lol don't get hung up on 3dmarks they really don't tell much these days other than who can make the fastest cpu to score the record. You've completely forgotten what bottleneck means though, take your 2 lane highway load it with moderate traffic and then drop it to one lane, bottleneck right? now you're gona half to call the wife and tell her you'll be late for dinner.
now do the same thing with twice as much traffic, hmm major bottleneck, it's gonna take you hours to get home call the wife and tell her not to wait up.

the point being with w single card you're bottlnecked, which means if you add more power behind the bottleneck you're going to become even more bottlenecked. the problem isn't so much the cpu as the resolution used, theres no point in buying two 3870's to run at 1280x1024, miogh as well stick with one. but at 1680x1050 with aa turned on, or even at 1280x1024 with the coveted 24x sample aa. you'll notice much more of a difference. especially in real world applications.

so while the 6000 is slowing you down at the default 3d06 resolution, it won't matter at higher resolutions. easiest way to tell, run crysis on the maximum detaisl you can and put at your current settigns, then take off your cpu oc and run it at stock and do the same test. you'll get the same resuslt as the gpu's are now the bottleneck.

 

[trog100]

lol don't get hung up on 3dmarks they really don't tell much these days other than who can make the fastest cpu to score the record. You've completely forgotten what bottleneck means though, take your 2 lane highway load it with moderate traffic and then drop it to one lane, bottleneck right? now you're gona half to call the wife and tell her you'll be late for dinner.
now do the same thing with twice as much traffic, hmm major bottleneck, it's gonna take you hours to get home call the wife and tell her not to wait up.

the point being with w single card you're bottlnecked, which means if you add more power behind the bottleneck you're going to become even more bottlenecked. the problem isn't so much the cpu as the resolution used, theres no point in buying two 3870's to run at 1280x1024, miogh as well stick with one. but at 1680x1050 with aa turned on, or even at 1280x1024 with the coveted 24x sample aa. you'll notice much more of a difference. especially in real world applications.

so while the 6000 is slowing you down at the default 3d06 resolution, it won't matter at higher resolutions. easiest way to tell, run crysis on the maximum detaisl you can and put at your current settigns, then take off your cpu oc and run it at stock and do the same test. you'll get the same resuslt as the gpu's are now the bottleneck.

u are pretty much saying what i say.. several times over to be exact.. he he he.. someone else that aint read the thread..

the real problem with crysis thow is the game screams "how its meant to be played" at u as soon as u load it.. its written around nvidia and its as simple as that.. sli scales +80% crossfire scales -10%.. the figures say it all..

as nvidia and intel get better this is gonna happen more and more.. amd ati are fighting a losing battle it seems..

trog

 

[-=CrAnSwIcK=-]

trog, i'm a regular at PCstats.com a guy on our forum broke 21k in 06 with a CF 3870 setup, but he has a Q6600 GO running at 4ghz...i definately think it's your CPU...most likely the poor increase is due to a) CPU bottlenecking and b) low CPU score...also his 3870's have been to 920mhz core