1600pro 256mb cards arrived - getting ready for crossfire

[saaya]

1280x1024 high quality

2005-12-27 04:20:12 - Quake4
Frames: 8346 - Time: 189092ms - Avg: 44.137 - Min: 0 - Max: 62

2005-12-27 04:28:18 - Quake4
Frames: 6543 - Time: 189388ms - Avg: 34.548 - Min: 0 - Max: 62

[saaya]

1024x768 high quality

2005-12-27 04:39:06 - Quake4
Frames: 10212 - Time: 187948ms - Avg: 54.334 - Min: 1 - Max: 62

2005-12-27 04:45:36 - Quake4
Frames: 8425 - Time: 188052ms - Avg: 44.801 - Min: 4 - Max: 62

looks like it scaled best at 1600x1200 with around 30%...
maybe crossfire only scales well with aa and af?
but theese fps are pretty low, so aa and af would drop them even more...
or maybe im cpu lmited with my 2ghz a4?
imma try 2.5ghz

[sabrewolf732]

you're getting quite the weird results

[baracus2k]

can you run doom 3 timedemo @ 1024x768 in high quality mode on 1x1600pro @ 600/800? id really appreciate that... im trying to compare your result to a x1300pro

[saaya]

well the intros i am benching are more demanding than the timedemo i think.
when i walk around in q4 in 1024x768 im always around 40-60fps, most of the time around 50/60. and i cant disable the 60fps cap on q4, wich is really annoying...

but the intro benchmark gives me 44fps average... ingame average fps is 55 or higher, i dont know as i cant see how much above 60fps i could hit if this stupid limit wouldnt be there...

how come com_fixedtic -1 doesnt work in q4?

[saaya]

can you run doom 3 timedemo @ 1024x768 in high quality mode on 1x1600pro @ 600/800? id really appreciate that... im trying to compare your result to a x1300pro

as i said, the timedemo only works in the full version of doom3, and i dont have doom3... is there a way to get the timedemo working in the d3 demo version?
can you email me the demo001.demo from your doom3/base/demos folder?
maybe i can run it... saschakrohnremovethispart@yahoo.de

[baracus2k]

do you have farcry?

EDIT: the doom3 engine is designed for 60fps... if you remove the 60fps cap the game will overspeed... in the timedemos the cap is removed

EDIT 2: ok how about this.... FEAR stress test 1024x768 max details with softshadows with 1 card @ 600/800... just to make sure does the FEAR demo have the same stress test as the retail? the retail version stress test has some guys shooting at each other and then goes to some place with pipes and water then has some room blow up

[saaya]

do you have farcry?

EDIT: the doom3 engine is designed for 60fps... if you remove the 60fps cap the game will overspeed... in the timedemos the cap is removed

EDIT 2: ok how about this.... FEAR stress test 1024x768 max details with softshadows with 1 card @ 600/800... just to make sure does the FEAR demo have the same stress test as the retail? the retail version stress test has some guys shooting at each other and then goes to some place with pipes and water then has some room blow up

yes i have far cry and will bench it tomorow, if theres a specific setting youd like let me know

the doom engine isnt designed for 60fps, thats just a lock they applied to not run into bugs and anormalies at high fps like in quake3
and no, the game doesnt overspeed at all, who told you this?

lol, yes the fear demo stress test is the same as the fear retail stress test
and why 600/800? max oc theese cards do rockstable is 600/900, i ran it with one card at 500/800 already and posted the result:
minfps 11
average fps 25
max fps 57


here are bf2 results, i found out that a single player map usually always gives the same average fps if i play it. so i played two rounds with crossfire and one without it and both cf runs and both single card runs are almost identical.

here are the results:

2005-12-27 06:20:08 - BF2
Frames: 38742 - Time: 606718ms - Avg: 63.855 - Min: 0 - Max: 101

2005-12-27 06:51:05 - BF2
Frames: 33666 - Time: 599996ms - Avg: 56.110 - Min: 0 - Max: 80

a measly 12% boost... yuck... :/

oh, this was run with 2.5ghz and the mem to 200mhz
i only used 512mb memory but the game actually worked fine, the first round was always stuttering a litle, but the second round and following ran flawlessly smooth.
i found out that on this es board using 4 sticks somehow sets it to single channel...
so im back to two 256mb sticks

[saaya]

just found out that theres another sub menu in fear, my bad ^^
all the fear results are with 4aa and 16af!

fear is really fkced up though...
baracus2k, i ran 600/900 in 1024x768 for you again, with and without aa and af, and with all other settings enabled or set to maximum. and i found out that theres something weird going on.

when i just set the videocard setting to maximum 4aa and 16af gets applied.
in 1024x768 with 2.5ghz cpu power, 200mem and the card at 600/900 that gives me: 13/28/66 (min/av/max)

when i disabled aa and af, i get:
13/24/48 (min/av/max)
wich is lower than with 4aa and 16af????

so i enabld 4aa and 16af again and got this:
9/16/31
even lower?
and the funny thing is aa is definately not applied correctly, the fog, the water reflections on the wall, and the flame textures are all pixeled heavily...

i first thought i had confused some settings, i but i closed fear and did the exact same thing 2 times now, and i keep getting the same results...

default = highest score, aa af disabled = lower, enable aa af again = even lower score

[saaya]

i just reran 2k3 and 2k5 with 2.5ghz cpu clocks... no gain, nothing at all... :/

[hipro5]

i just reran 2k3 and 2k5 with 2.5ghz cpu clocks... no gain, nothing at all... :/

In 2k3 you should see some gain with 500MHz more on CPU....

[saaya]

lol, my cpu speed is back to 2000mhz??? this board seems to like to reset everything to default ^^

[fhpchris]

My Laptop is even easily better than these cards!

I am not impressed

a single 7800gt costs as much as both x1600s

[saaya]

My Laptop is even easily better than these cards!

I am not impressed

a single 7800gt costs as much as both x1600s

i know... thats why i said 1600s in cf arent worth the money...
1600xts look btter, but whether they look good... i dont know, id say no, but you better ask althes.

ok, i set the cpu to 2.5ghz gain, maybe i forgot that i set it back?
anyways, i gained......... 50 points ^^

50 points for a 50mhz cpu boost... in crossfire!
now that should give you an idea how theese 1600s scale with cpu speed... and thats why althes couldnt even hit 10k in 2k5 with 3ghz.
this cards are just like g4mx cards when it comes to cpu power...

so the only way to boost their speed is vpu and mem speeds.
te 1600pros are very bandwidth cut, and even the gddr3 on the 1600xt doesnt help it THAT much as its 128bit...
and the core speed doesnt seem to have a big effect either on performence... only in havily gpu intense apps, and those apps run too slow on this card.

its a closed case, 1600s in crossfire make no sense...
and even the gddr3 wich ocs really nicely doesnt help to make the 1600xt an ocers card, because ocing doesnt mean upping the mhz, it mans upping the performence, and thats something wich the 160s are very limited in.

to bad... 256bit memory might do the trick...
this card would have bee a dream if it would have 256bit memory and 8x1 instead of 4x3 configuration of the vpu...
rv530 is half the size of r520 and i think 8x1 would have made it even smaller and cheaper to manufacturer, but would have performed better...

well, lets hope the 1700xt is a better performer!

[althes]

And thats all folks.

[saaya]

your returning your cards?
imma send my cards back in a few days... was kinda fun, but the results are really dissapointing.

the average boost sems to be 30% from using crossfire with theese cards, and thats about the same boost you can get by ocing the videocard, almost all cards can do a 30% oc, and with higher cpu speed you can get even more of a boost... not with theese cards...

higher than 30% gains are def possible, butwith theese cards those gains are in unplayable framrates with and wthout crossfire...
hope the 1700s are out soon!

[althes]

Yeah I sent the x1600pros back.
I will send back the xt's also not worth keeping them.
Futuremark doesnt even have them up to put in some scores.
I may do a mod with a pot to see it works.

[saaya]

fm has em in the orb, but they are hard to spot, i had some trouble finding them as well
but they dont make a difference betwen xt and pro iirc.