512MB Gecube X1900XTX - I'm back

[eva2000]

Yes i'm back and you get to watch me continuously post benchmarks with my new toy

I couldn't wait for Gecube to provide me with my X1900XTX/XT Crossfire cards for my testing on DFI CFX3200-DR http://i4memory.com/cfx3200dr/ so decided to fork out for half the setup myself, in the form of a 512MB Gecube X1900XTX card

The card arrived yesterday but since i have the flu and bad cough keeping me up at night I thought I'd give the card a whirl

System:

• FX-60 CCB2E 0536RPMW
• Kayl Black Frozen SS phase change for cpu cooling
• DFI NF4 UT SLI-DR R.AB0 406BTA
• 512MB Gecube X1900XTX stock HSF
• 2x 512MB OCZ PC3200 Gold UTT BH-5
• 4x 80GB Hitachi 7K80 SATAII Raid 0
• Pioneer DVD-RW + FDD
• 1KW PCP&C SLI power supply

WinXP Pro SP2 fully updated
6.70WHQL Chipset drivers
6.5 Catalyst Official drivers

Right now stock heatsink air cooled with fan speed at 100% I managed on GPU core using ATI Tool 0.24 beta 14.

GPU Overclocking

1.175v = 655mhz
1.400v = 684mhz
1.425v = 702mhz
1.450v = 712mhz
1.450v = 720mhz with ~63C load temps as monitored via Rivatuner RC16.
1.500V = 725mhz with ~66C load temps as monitored via Rivatuner RC16.
1.500V = 729mhz with ~66C load temps as monitored via Rivatuner RC16.

Here's one of my first 3dmark05 runs with ATI Tool and Rivatuner RC16 running in background.



Video Memory Overclocking

As to memory clocking, seems like my 512MB HIS X1800XT, part of the puzzle is GPU voltage, as you increase GPU volts, mem clocks higher at some point. Here it seems to be at 1.525v GPU volts.

GPU/VDDC/VDDQ/VDDCI = GPU/Mem

Note: PASSED means ATI TOOL continued to run at that stated speed only.

1.500v/2.086/2.086/1.486v = 725/801-805.5mhz PASSED
1.525v/2.086/2.086/1.486v = 725/810-814mhz PASSED
1.525v/2.168/2.086/1.486v = Locked up in ATI Tool at 725/814mhz heat up phase
1.525v/2.135/2.006/1.486v = Locked up in ATI Tool at 725/814mhz heat up phase
1.525v/2.140/2.121/1.486v = Locked up in ATI Tool at 725/814mhz heat up phase
1.525v/2.105/2.102/1.510v = Artifacts at 725/850-855mhz
1.525v/2.113/2.102/1.510v = Locked up in ATI Tool at 725/814mhz heat up phase
1.525v/2.121/2.102/1.510v = Locked up in ATI Tool at 725/814mhz heat up phase
1.525v/2.113/2.102/1.510v = Locked up in ATI Tool at 725/814mhz heat up phase
1.525v/2.113/2.102/1.510v = PASSED dropping GPU by 5mhz = 720/846mhz
1.500v/2.113/2.102/1.510v = Locked up in ATI Tool at 720/823mhz heat up phase
1.500v/2.113/2.102/1.493v = Artifacts at 720/846-850mhz
1.500v/2.119/2.108/1.493v = Locked up in ATI Tool at 720/814mhz heat up phase
1.525v/2.113/2.102/1.493v = Artifacts at 720/837mhz
1.525v/2.113/2.102/1.510v = Locked up in ATI Tool at 720/859mhz heat up phase
1.525v/2.116/2.105/1.524v = Locked up in ATI Tool at 720/855mhz heat up phase
1.525v/2.119/2.105/1.510v = PASSED at 720/846mhz
1.525v/2.119/2.105/1.540v = Locked up in ATI Tool at 720/868mhz scanning artifact stage
1.525v/2.119/2.105/1.563v = PASSED at 720/855mhz load temp 75C

Haven't found the max yet for GPU or Mem, but at this point it seems the memory needs more GPU voltage to clock higher than 801-805mhz.

Seems for 3dmark05 at least 720/850 is achievable with 1.525/2.119/2.105/1.563 with load temps of ~67-75C.

3dmark05 = 13,885
http://service.futuremark.com/compare?3dm05=2077846

While Rivatuner RC16 and ATI Tool were running in background.

@720/850 - 13,885 http://service.futuremark.com/compare?3dm05=2077846


@720/850 - 43,894 http://service.futuremark.com/compare?2k1=8998288


@720/850 - 6,962 http://service.futuremark.com/compare?3dm06=353082


Doesn't seem to be the best clocker, but not bad at all

[cantankerous]

Wow, very nice job eva! As usual I love to read your bench results as you are very thorough in describing all details such as clocks, volts and method of tests as well as pictures to backup your claims. Seems like you have a fairly strong card there. My card does 685/810 with 1.40v and it seems yours does the same. I will take some of your results as a test to see how much higher I can get on core with more volts. Keep up the great work as usual and have fun with your very fast new card.

[ahmad]

14k with stock cooler is not bad

[SIOUX]

Nice work as always eva Very detailed. I reasently got a new xtx for my n ew setup, and havde been testing it a bit. by far the best card i ever owned
this was just testing a bit in my favorit bench with normal water setup. i will test further soon sorry for highjacking

[Silves]

Great post and great results Eva, just like the one u did with the HIS x1800xt, i bought one because of that post.

[Rovtar]

Great post Eva keep us posted ...

[Plywood99]

Lol, Eva is at it again! As always awesome attention to detail. See you got your phase setup. Makes me wanna break out my old ghetto dice gear.

@Sioux - Man! That card is sick! show us what it can do...

[eva2000]

thanks guys more results later today i hope

Ply break out that ghetto DI gear hehe...

SIOUX, what voltages where that clock at ? and water and ambient temps for the GPU h20 cooling ? What did that card do on stock hsf air ?

[rob[GL]]

I'm surprised that an XTX clocks so poorly... I have an X1900CF (625/725) same as XT... I am overclocked 729/828 with stock cooler and stock volts...

I thought XTX was supposed to be far better than XT?

Wish I had an FX processor...

Keep up the good work!

[eva2000]

Yeah not that good an oc'er this card.. GPU can handle lower voltages but mem needs more GPU volts it seems

bad news of sorts.... seems my system with FX-60 just boots up into windows then shuts down

if i try to power it up again, it immediately shuts down ?

i removed the evap and insulation foam to dry things out, put a hair dryer on low heat to the cpu socket to ensure it's not condensation...

left it for a few hours, then installed the insulation foam again and remounted evap....

go into windows at 200x16 at 1.552v and tried stressprime dual prime test it went for around 5mins and then rebooted, and then shutdown itself again ????

the idle and load temps are like normal idle -18C to -20C and load +0C to +1C to +2C....

soooo what can be wrong ? still condensation in cpu socket ? with all the insulation i did of the board, could it still not be enough ?

any ideas guys ?




EDIT: The culprit seems to have been a pool of condensation/water on back side of motherboard http://www.xtremesystems.org/forums/...=1#post1528532 LOL

[fhpchris]

I want to say, thank you!

I have not seen such a nice compilation of x1900 voltages and clockspeeds

It will help me very much, along with your CFX3200 UTT settings

[SIOUX]

thanks guys more results later today i hope

Ply break out that ghetto DI gear hehe...

SIOUX, what voltages where that clock at ? and water and ambient temps for the GPU h20 cooling ? What did that card do on stock hsf air ?

1.55 / 2.34 / 2.151 / 1.575


The water was ambient yes. 2x120mm radi + swiftech pump and a EK X1900 waterhead

I dont know what it does on air...never tryed

[eva2000]

thanks good to have an idea of voltages used although it may differ from card to card heh

I want to say, thank you!

I have not seen such a nice compilation of x1900 voltages and clockspeeds

It will help me very much, along with your CFX3200 UTT settings

no probs that's what my CFX3200-DR guide is for hehe

had to go to air cooling on FX-60 to check if my cpu was okay after condensation on phase....

FX-60 @2940mhz 1.408v on stock heatsink air with X1900XTX @650/774 on stock hsf air too and rivatuner, mbm5 and ati tool running in background

[ahmad]

Lol 12.6k stock clocks is pretty sick.

[mursaat]

Thanks for the benchs

BTW, your '05 scores seems "low" to me, this is what I got with a 165 TTBT @ "only" 3003 and a X1900XTX@725/850 stock cooler 1.550/2.210/2.210/1.501:

3dk05 score: 13616
http://service.futuremark.com/compare?3dm05=1996857

3dk06 score: 6651
http://service.futuremark.com/compare?3dm06=268158

The 3dk01 is impressive, because of the +300mhz in the CPU. This is mine, 40544:
http://service.futuremark.com/compare?2k1=8966641

Thanks again, not trying to hijack your thread your voltage tables are very handy

//EDIT: Now I see the project details, my '05 Firefly is higher than yours and Canyon is really close (70.7 vs 71).
In 3dk06 your score is higher because of the CPU score, the SM2&3 are really close, which makes sense as it's 3D and not CPU dependant.

I selected Texture and Mipmap quality to performance. No LOD.
BTW what speed was your RAM at? I can't find it in the post.
Hope this helps

[scream]

guys, im planning on getting the Accelero X2 for my Powercolor X1900XT.. do you guys think it's worth it?? and also i've tried to push down the palstic clips on the HS of the VR but it does not seem to be going down... i live in singapore so ambient temps are really high.. on idle my X1900XT is doing 55 degrees... lol

[eva2000]

3rd party air coolers for x1900xt/xtx won't do much as temps would still be higher, for max clocks better off with stock x1900 heatsink @100% fan speed or water cooling heh

mursaat, nice scores, you can find my mem timings in my compare url descriptions or clicking the images for larger screenie as i'm running uplock 233 divider

3dmark05 won't change much at around same clocks you have 5mhz more GPU core than me 725 vs 720 hence GT2 higher...

I'm doing more voltage tests soon to see if i can make a break through regarding mem clocks hehe

[eva2000]

Further voltage testing updates

GPU/VDDC/VDDQ/VDDCI = GPU/Mem

Note: PASSED means ATI TOOL continued to run at that stated speed only.

1.525/2.138v/2.009/1.564 = FAILED went up to 720/864-868mhz on memory before blanking out. Seems VDDQ can be lowered without too much detriment to mem clocks, wonder if raising VDDC only will get me higher

1.525/2.151v/2.015/1.564 = PASSED 720/864 ran scanning artifact fine. So it seems keeping VDDQ low is fine and just up VDDC

1.525/2.165v/2.015/1.564 = FAILED up to 720/868 and failed in heat up phase with 868mhz memory.. getting there though

1.525/2.176v/2.031/1.564 = FAILED up to 720/868 and failed in heat up phase with 868mhz memory

1.525/2.181v/2.020/1.564 = FAILED up to 720/868 and failed with 868mhz memory with 16 seconds left to completion in heat up phase!

1.525/2.192v/2.020/1.566 = FAILED up to 720/868 and failed with 868mhz memory with 9 seconds left to completion in heat up phase!

1.525/2.203v/2.020/1.564 = FAILED up to 720/868 and failed with 868mhz memory with 8 seconds left to completion in heat up phase!

1.525/2.203v/2.015/1.564 = FAILED up to 720/868 and failed with 868mhz memory with 15 seconds left to completion in heat up phase!

1.500/2.209v/2.020/1.564 = FAILED up to 720/868 and failed with 868mhz memory with 24 seconds left to completion in heat up phase!

1.550/2.203v/2.020/1.564 = FAILED locked up at 792mhz on mem.. guess 1.550v GPU volt too much for it on air cooling ?? Or maybe with 1.550v i can reduce VDDC volts ?

1.550/2.184v/2.020/1.564 = FAILED got up to 877mhz highest so far, after passing heat up phase at 868mhz, but artifact detecting knocked it back from 877mhz to 873mhz back to 868mhz

1.550/2.195v/2.023/1.564 = PASSED 720/868 - even passed heatup phase @873mhz on memory, then ran artifact scanning for 90 seconds before artifact detecting bumped it to 868mhz

Getting there hehe

[Marioace]

very nice and thx a lot eva for the resume

will try some tips

[eric_r]

I just got a gigabyte x1900xt, too afraid to oc it (im only 18, too much out of pocket if something goes wrong)

nice scores. really helpful.

[eva2000]

Update: For now settled on 720/864 with 1.550/2.195v/2.023/1.564 - backing off on the mem a notch just for safe measure. 3Dmark05 ran perfectly fine at this speed.

3dmark05 = 14,028 with ATI Tool, Rivatuner RC16 both running in background logging temps.



Pretty happy that I managed to get the vid memory clock up from what seemed like a 814mhz wall in ATI Tool 0.25 beta 14 to the 864-868mhz speeds now on stock hsf air. GPU clocks do seem low at 720mhz but guess nothing that better cooling on GPU wouldn't help fix

[mursaat]

mursaat, nice scores, you can find my mem timings in my compare url descriptions or clicking the images for larger screenie as i'm running uplock 233 divider

3dmark05 won't change much at around same clocks you have 5mhz more GPU core than me 725 vs 720 hence GT2 higher...

So that means with a X1900XTX at those clocks 3dk05 is also GPU bound. Which also could mean a A64 DC @ 3ghz is just the point where the GPU starts to lack CPU power, because I get higher scores @ 3000 than @ 2800

[eva2000]

yeah, ya need SLI or CF setup to really see that cpu @3400mhz move some nice figures in 3dmarks

[mursaat]

What's interesting is it's at around 3ghz where 3dk05 starts to be GPU bound. I get ~700 more points at 3ghz than at 2.8, but you don't get real benefit from 3.3ghz.

Looks like the perfect combination, 165@3000 + X1900XTX air 725/850

[eva2000]

3dmark2003 = 22,583



3dmark06 = 7,025



3dmark2001 = 44,974



AM3 = 124,159