Crossfire impressions

[G H Z]

First things first. Being pretty much an '01 addict I can say that it was quite a dissapointment for me to see that X1800's in Crossfire is basically useless for 3DMark 01. You are better off running a single X1800 at this point with current drivers and CF options.

And after struggling for a day trying to get CF running (on a Windows 2000 installation), I was starting to think this was all a mistake. Turns out that CF will not work on Windows 2000 as CCC never gives the CF enable page even when drivers are loaded and devices are started. As soon as I switched to XP everything worked fine.

Anyway I played around with '03 on a 2.5Ghz X2 3800. Not a big overclock by any means on this chip, I still need to tame the RDX.

http://service.futuremark.com/compare?2k3=4516939
http://img.photobucket.com/albums/v31/canny/CF.jpg

[k|ngp|n]

Setup looks great. 2 bad about 01' :(
So x1800xt cf on xp is slower than single card 1800xt on 2k1?

[blind_ripper]

nice setup GHZ ;) the dual-core cpu's withe the 5.12 drivers give big boost remeber thate .
the RDX is a strainge mobo , some tijds i can run 270mhz 2-2-2-5 and some times ican not even run 260mhz @ the same settings , there is a new bios ;)

[G H Z]

So x1800xt cf on xp is slower than single card 1800xt on 2k1?

Well I'm not 100% on that but running 700/800 is only worth 380FPS in Nature similar to single card numbers.

Hey blind :)

Ya I'm on the 5.13's this is all fresh. RDX is different for sure, anything past 255 HTT bluescreens for me and I really need some high HTT spedds with this 3800.

[blind_ripper]

k flash the bord to a new bios and ul get for sure above 255htt , best htt link is 3x en ldt volts myn likes 1.4v ;)
http://us.dfi.com.tw/Support/Download/bios_download_us.jsp?PRODUCT_ID=3669&STATUS_FLAG=B&SITE=US
hf , i can not test the new bios dont have a pci card now its out to mod :D

[eva2000]

sweet cards there :D

[SIOUX]

nice crads:)...but DAMN on the 01 :(

[zakelwe]

The big question with Crossfire before it came out was whether it would have as much impact on cpu tests in 3dmark2001 as the nvidia does. It seems from the little evidence I have that it is worse than that and doesn't seem to work at all .. or does it and I have missed the results ?

Ln2 + single X1900 + Dothan looks like it will be the only option come the end of January.

Regards

Andy

[Willis]

The big question with Crossfire before it came out was whether it would have as much impact on cpu tests in 3dmark2001 as the nvidia does. It seems from the little evidence I have that it is worse than that and doesn't seem to work at all .. or does it and I have missed the results ?

Ln2 + single X1900 + Dothan looks like it will be the only option come the end of January.

Regards

Andy
"Ln2 + single X1900 + Dothan"
it seems you forgot the fact that there is an 2*PCI-e dothan board is coming :banana:
and by the end of january i'm afraid that 3dmark01 wil stil be Nvidia's playground...
does the X1800Xt comes even near the gtx 512MB?

[lawrywild]

it seems you forgot the fact that there is an 2*PCI-e dothan board is coming :banana:

eh? I know a company had some plans, but I wouldn't say there's one coming. got any proof?

[Willis]

eh? I know a company had some plans, but I wouldn't say there's one coming. got any proof?
Fugger is your proof :)
perhaps i should change it into this to be more specific..
2*PCI-E and S480 ;)

[scampi]

can you do a 3dmark05 with 2x1800xt in crossfire, please?

[Hicks]

Why the hell have we still not got a decent overclocking tool for these X1800's?

Bloomin crap.:mad:

[biohead]

Why the hell have we still not got a decent overclocking tool for these X1800's?

Bloomin crap.:mad:
because we all are tweakers, I think the "how to" is only necessary, which is now discovered. Disable all ati services (unnecessary anyways), open up ATi tool and set the clocks of each card individually. How hard is that?

[STEvil]

new version of atitool came out a few days ago iirc.

[saaya]

nice results ghz :)
can you check if the latest ati demos work for you?
the assesin and toyshop demo dont work on my x1600s :confused:

oh and do you also get a weird lag in gt3 2k5?
a tripple hickup at the start of the test...
i made 4 clips with my cam, 2 with a single card 2 with cf:
http://www.filefactory.com/get/f.php?f=51fd6cefe493e7d53e2b204d

[biohead]

nice results ghz :)
can you check if the latest ati demos work for you?
the assesin and toyshop demo dont work on my x1600s :confused:

oh and do you also get a weird lag in gt3 2k5?
a tripple hickup at the start of the test...
i made 4 clips with my cam, 2 with a single card 2 with cf:
http://www.filefactory.com/get/f.php?f=51fd6cefe493e7d53e2b204d
FileFactory - 100% FREE unexpected end of archives!

[saaya]

argh.... gimme your email :D

[G H Z]

Sure I'll try the demo's a bit later, and yes I do get the same thing in GT3. Happened on single card also. Killing Explorer seems to help a little.

This is a 2.5Ghz dual core only. While these few '01 tests I ran do not look very good that may change with a 3.5Ghz FX-60 or 57. Still I was expecting more from this setup.

[blind_ripper]

instal microsoft framework thate whille fix it ;) had it befor to ;)
GHZ those the new bios helps ?

[saaya]

framework fixes the hickup? not for me...
and i only get this with two cards, not with one...

gt3 starts, a few frames run, then it looks like somebopdy fast rewinds the szene and it plays back again, then the same thing happens 2 times more.

1. when the ship appears and moves towards the camera, 2nd then its going into the curve to pass the camera sideways, 3rd when its passing the cam sideways.

as you can se on the small clips i made it doesnt happen at all with a singe card, ad is clearly visible with 2 cards.

i remember i had a similar thing with my 6600gt and x850 though... weird.

[Dynasty]

Same as SLI, CrossFire needs Win XP. No go with win 2K. :cool:

[saaya]

hmmm the odd thing is if i run fraps and record the fps then the hickup is almost completely gone :confused:

but its still slightly notable, i checked the frame duration and found that after around 1/3 of the first szene wich i recorded there are two places where the frames took 2x as long as they usually took in that szene.
usually my 1600s in cf give me 14fps average wich means they fire 1 frame every 70ms, of course this varies between 50 and 100ms, but there are two situations where it took a frame over 200ms to appear wich is 2x the max time a normal frame can take... odd

why does it run better with fraps enabled?
i think its a problem with how 2k5 counts the frames... or its a cpu issue because i noticed that if my cpu is also running something else in the background wich steals some cpu cycles there is no lag either.

[saaya]

i made an excel with the frames captured to see how linearily the fps get fired out, and around 70% of the frames come at almost the same time.
this means that although the average fps is 15, it actually looks more like 10fps, depending on how the frames look... if the frames have the right szene time then itll all look right, but it doesnt look and feel like 14 fps, it feels more like 10fps or less, wich makes me think the frames that get shot out almost at the same time are almost identical.

too bad i cant capture every frame in that test... i will ask the fraps author if he can allow a series of frames to be captured instead of having a min delay of 1sec.

since cf is only around 50% faster i will downclock both cards in cf and then run vs a single card.
i will try to push them both to the sam average fps and then check whether theres a diffference in the linearity of how the frames get put out

[blind_ripper]

myes the hickup is some thing to go withe windows if i have it good , can also be some bug in 3dmark . but framework helpt me , if the hickup are gone ul get a beter score to :cool:

[G H Z]

Wow '05 is very CPU limited with Crossfire. I managed to get the 3800 going a bit better, not bad for air eh ;)

Comparo (http://service.futuremark.com/compare?3dm05=1642483)
http://img.photobucket.com/albums/v31/canny/15063-715-877.jpg



Here's what CF does with '01 (http://service.futuremark.com/compare?2k1=8848140)

Also 33K in '03 (http://service.futuremark.com/compare?2k3=4519156)

@ blind,

Hey man you can't run CCC without Net Framework :)

[blind_ripper]

nice clocks on thate x2 3800+ ure faster thane my x2 4400+ :eek: on water .
how much vcore on thate baby , cpuz reads it wrong ;)
try 01 withe 1 card at thate cpu speeds ur car low has to be around 500fps
http://service.futuremark.com/compare?2k1=8821116
i dont use CCC :p:

[flopper]

is it just me but dont they need to upgrade testing to 1280x1024 since that resolution seems more acurate today?

Crossfire looks sexy though, a double Ruby sight is for sour eyes to see ;)

[scampi]

Wow '05 is very CPU limited with Crossfire. I managed to get the 3800 going a bit better, not bad for air eh ;)

Comparo (http://service.futuremark.com/compare?3dm05=1642483)
http://img.photobucket.com/albums/v31/canny/15063-715-877.jpg



Here's what CF does with '01 (http://service.futuremark.com/compare?2k1=8848140)

Also 33K in '03 (http://service.futuremark.com/compare?2k3=4519156)

@ blind,

Hey man you can't run CCC without Net Framework :)


very good :clap:

what voltage have you given to core and mem?

[saaya]

actually i made a run with fraps wich somehow prevents the hickup and i got .1fps less, so with the hickup the system scores better...

nice scores!
and what did you expect? with two powerfull cards like that your going to be pretty cpu bound in almost every app now i guess ^^

[G H Z]

how much vcore on thate baby , cpuz reads it wrong ;)
i dont use CCC :p:

1.55v

Plus you can't run CF without CCC ;)

@ Scampi,

1.4v core, stock memory.

[G H Z]

I got tired of trying to tame the cold bug on the 3800 and switched to the FX. I was a little leery of the CPU swap :D

But no problems at all and I actually picked up some Mhz running on the RDX. I really like this board actually, ATI & DFI have done a very nice job. I think there is still some room on these. This Gskill is pretty good but I need to try the old standbye.

03 Compare (http://service.futuremark.com/compare?2k3=4526382)
http://img.photobucket.com/albums/v31/canny/34055-722-882.jpg

05 Compare (http://service.futuremark.com/compare?3dm05=1654988)
http://img.photobucket.com/albums/v31/canny/16231-722-882.jpg

[blind_ripper]

very nice scores GHZ but @ 3.3ghz x2 is stronger thane a 3.7ghz single ;) , RDX is very nice bord but some times it crappy.
the RDX will oc beter cpu on low vcore ;)

[G H Z]

Had some fun benching tonight on Dual pelt setup, cards are clocking very well.

I'll try to find out what these will really do tomorrow, I ran out of ice tonight :)


Compare (http://service.futuremark.com/compare?2k3=4559596)
http://img.photobucket.com/albums/v31/canny/35720-796-913.jpg


Compare (http://service.futuremark.com/compare?3dm05=1717468)
http://img.photobucket.com/albums/v31/canny/16657-805-922.jpg

[Sampsa]

G H Z, nice scores! Although looks like FX-57 @ almost 3,7 GHz can't compete with FX-60 @ 3,4 GHz: http://service.futuremark.com/compare?3dm05=1694804

Your GT3 score is better :)

[G H Z]

I know, I only got you with clocks in GT3. I seriously doubt if I can make up the score difference with clocks alone :)

When are you gonna try some '03?

[Cpt Twitchy]

Very nice card clocks. I got plenty of ice over at my house and its not a far drive =P