X1900 Series Overclocking and Feedback thread

[Dynamic]

Besides the internal fan cooler inside the R580 intakes air from the back of the card and exhaust air through the front. Why don't you relocate the fan unless you're happy with the performance now....


BTW a successful flash with the Sapphire X1900XTX bios...so i got me a Sapphire X1900XTX for only $508 bucks...

 

[yogurt_21]

not sure about the VDDCI you'd have to ask w1zzard about that. but if you can manage to keep the core volts at say around 1.45 and leave the mem volts at stock it seems you might be able to reach ~730/855 which would be a more recomended voltage setting for the long term. that would still give you plenty of performance. well from my perstective anyways seeing as you 05 score is around 2000pts higher than my best.

 

[Super XP]

Okay so I've got an ATI x1900xt, bios flashed to x1900xtx

I've overclocked with ati tool 0.25 beta 14

I'm managed to reach

747/864 artifact free
my voltages are:
VGPU 1.575
MVDDC 2.157
MVDDQ 2.157
VDDCI 1.575

Temps idle/load:
Voltage regulators 51/84
GPU Temp 30/41
Temp of temp chip 33/40

Cooling
Swiftech MCW55 waterblock 1/2
Swiftech BGA memory sinks
arcticsilver applied to mosfets heatsink

3DMark05 13017
3DMark06 5691 seems low for clocks

Any words of wisdom?

I'm concerned that my core voltages are too high, what you do folks think, am I shortening the life of my card?

Or is my cooling sufficient to the volts?

Any tweaks so I can get my memory to 900 the rated spec?
cheers,

a.

If your card is running too hot, then you will not get a good score. I don't trust the CCC's temp's also.

 

[Super XP]

O.K. here are three links so that I can show you the problem I have with F.E.A.R - and I am now using the CCC Beta 6.1 1/2 drivers, because the stock from the CD, the 6.1's & 6.2's always gives me an error message, and the 6.1 don't even know what card I have.

http://img467.imageshack.us/img467/2265/fearproblem024jk.jpg
http://img48.imageshack.us/img48/5620/fearproblem012vl.jpg
http://img48.imageshack.us/img48/4554/fearproblem037ko.jpg

Why is it artifacting anyway? this also happends to Dungeon Siege 2, and it don't matter what quality settings I use. Also my card is Bone Stock with a Fresh format & instal of Windows XP SP-2 - and no ATI Tool installed, and still does this crap?

Is my card defective? And does ATI offer "Advanced Replacement" I mean this card does cost a shit load of cash.

 

[yogurt_21]

seems like overheating memory to me, that or simply faulty mem, ati doesn't do an advance rma, you will have to pay to ship it to them, but they will usually resolve it and get you your card fixed or a new one in under 5 business days.
https://support.ati.com/ics/support/default.asp?deptID=894

 

[Super XP]

seems like overheating memory to me, that or simply faulty mem, ati doesn't do an advance rma, you will have to pay to ship it to them, but they will usually resolve it and get you your card fixed or a new one in under 5 business days.
https://support.ati.com/ics/support/default.asp?deptID=894

Thanks for the info & link

Do you think I should wait for the new CCC 6.3's or just get another card? ALso someone else I know plays FEAR with the XT no problems with CCC 6.2.

 

[ViperJohn]

I read a post by houserate (sp) who suggested that we maybe using way to much voltage when overclocking. I did some subjective testing, here are the results...

X1900xtx
P5WD2-E Prem
660 3.6 OC to 4.2
X-Fi Platnium
PC Power and Cooling 510 Express/SLI
*120MM Fan on Vregs and Mosfets (with copper heatsink...results will very)
ATI TOOL OC to 711/810 that's only 11MHz on core and 10 MHZ Mem higher then CCC Overdrive

During the initial OC at 1.40/2.160/2.160 the Vreg reached as high as 58C.
During the recent OC @ 1.35/2.130/2.130 the Vreg reached as high as 55C.
It appears that the lower voltage increased HDR/SM 3.0 scores. This is subjective opinion.

Ok here is what happens when I reduce:
VGPU from 1.40 to 1.35
MVDDC/MVDDQ from 2.160 to 2.130
Score on 3DMARK06 5791 (again, subjective however, this is with a lower voltage regardless if its only a 3 mark gain from 5789). This would sugges that we are playing a juggling act with heat and voltage not with the core and mem but with the mosfets and vregs!!! If you water cool you need some serious cooling for those vregs/mosfets! If you do it's possible to see an nice bump using the same core/mem OC's!
ATI Vreg water cooler. <--This is the only water cooled solution I that was given to me that I know of to date. Any suggestions is appreacited.

What you are looking at is normal run to run score variation. Changing the voltages will have no effect on the scores unless you lower them so far that the card is artifacting so badly that it effects the score (always lower when it does). It is clock speed and memory timings that effect the cards through put and score.

Anytime you go to a fanless GPU cooling solutions you need to take care of the loss of on card power supply cooling air. That is relatively easy to take care of. High on card PS area temps can effect a card memory OC as the heat from the PS area conducts through the PCB glass adding heat load to the closet memory chips.

Viper

 

[Dynamic]

Yeah FutureMark scores always very by one to another on the same rig. Doesn't make a difference what you change with the voltages. Anyhow i haven't installed b14 but is it working for you people because i would get stuttering in COD with b13 not using the overclock even.

 

[ViperJohn]

Yeah FutureMark scores always very by one to another on the same rig. Doesn't make a difference what you change with the voltages. Anyhow i haven't installed b14 but is it working for you people because i would get stuttering in COD with b13 not using the overclock even.

Seems to me Futuremark at one time said .05% is the margin of error that can be expected run to run but I have found it closer than that myself.

Viper

 

[EastCoasthandle]

What you are looking at is normal run to run score variation. Changing the voltages will have no effect on the scores unless you lower them so far that the card is artifacting so badly that it effects the score (always lower when it does). It is clock speed and memory timings that effect the cards through put and score.

Anytime you go to a fanless GPU cooling solutions you need to take care of the loss of on card power supply cooling air. That is relatively easy to take care of. High on card PS area temps can effect a card memory OC as the heat from the PS area conducts through the PCB glass adding heat load to the closet memory chips.

Viper

ViperJohn that is not what I am discussing, I am pointing out a fact that the amount of voltage needed to obtain a certain frame rate may not be that much, depending on OC.

 

[ViperJohn]

ViperJohn that is not what I am discussing, I am pointing out a fact that the amount of voltage needed to obtain a certain frame rate may not be that much, depending on OC.

The cards clocks and memory timings determine the card speed and the resulting frame rate in games. If you hold the clocks and memory timings the same and just change the voltages there will be no change in card speed or frame rate.

Viper

 

[EastCoasthandle]

VJ I am not sure you are right about this. I am getting consistant drops in frame rate with voltage higher then needed. If I lower them my frame rates increase. Maybe its the card or setup, whatever the case its happening.

 

[ViperJohn]

VJ I am not sure you are right about this. I am getting consistant drops in frame rate with voltage higher then needed. If I lower them my frame rates increase. Maybe its the card or setup, whatever the case its happening.

Well do in 3-5 cards per week and god only knows how many benchmark passes a week and I have never seen it. Voltages do not effect card speed and frame rate IF the card is running clean. If it is in fact happening to you it is an anomoly and cetainly not the norm.

Viper

 

[hotrippr]

Well do in 3-5 cards per week and god only knows how many benchmark passes a week and I have never seen it. Voltages do not effect card speed and frame rate IF the card is running clean. If it is in fact happening to you it is an anomoly and cetainly not the norm.

Viper

Have you thoght maybe you are both right? My guess would be that it is the higher temps that more volts can cause. Not actual volts. Speed(viper) and heat(EastCH) will affect your score.
My hyp: if you dont increase any speed(clocks) but increase volts(heat) you should have a lower score.

IMO 1900s have been a pain in the rear to oc, very finicky. Hats off to W1zard for atitool.

 

[ViperJohn]

Have you thoght maybe you are both right? My guess would be that it is the higher temps that more volts can cause. Not actual volts. Speed(viper) and heat(EastCH) will affect your score.
My hyp: if you dont increase any speed(clocks) but increase volts(heat) you should have a lower score.

IMO 1900s have been a pain in the rear to oc, very finicky. Hats off to W1zard for atitool.

Heat only limits the clock speeds you can run clean not the card speed, benchmark scores and frame rates you will get from a given clock speed.

I did think of one thing that may do what ECH is experiencing if he is using the CCC to OC the cards. If the core temp goes up to high the drivers will lower the OC dynamically to reduce core heating and that will effect the scores and frame rates.

Viper

 

[yogurt_21]

well shoot I don't know what you guys are arguing about, why would anybodu overvolt a card they aren't overclocking? seems like a good way to kill your card without recieving a performance boost. lol

 

[ViperJohn]

well shoot I don't know what you guys are arguing about, why would anybodu overvolt a card they aren't overclocking? seems like a good way to kill your card without recieving a performance boost. lol

This is a discussion not an arguement and I do not think anyone is talking obout over volting witout over clocking at the same time.

A properly modded card can live longer than the same card stock, even with increased voltages as long as you do not go nuts with them (which is not possible with an x1900 without hard mods) due to the hugh temperature reductions the modded card runs with.

Viper

 

[ccokeman]

Sorry to change the subject but i just looked at ati tool comeing out of the GT-1 in 3d03 and it was showing a gpu current draw of 24.2amps. Was a bit shocked to see it pull that much current at 698/810. Gpu 1.475, MVDDC 2.130,MVDDQ 2.102,VDDCI 1.5. How accurate is ATI tool on the current draw. I'm using .25beta13b

 

[ViperJohn]

Sorry to change the subject but i just looked at ati tool comeing out of the GT-1 in 3d03 and it was showing a gpu current draw of 24.2amps. Was a bit shocked to see it pull that much current at 698/810. Gpu 1.475, MVDDC 2.130,MVDDQ 2.102,VDDCI 1.5. How accurate is ATI tool on the current draw. I'm using .25beta13b

24.2 amps is nothing. That isn't even 36 watts GPU input at 1.475 Vcore.

Viper

 

[ccokeman]

So that means its not pulling 24 amps off the 12v line? If not then i should still be ok with my powerstream 520 sli. When i saw that i figured it only left 9a to run the rest of the system.

 

[dolf]

Make difference between 25 amps at 1.475 V and 25 amps at 12V (it is almost 8 times higher power).

 

[JDAdams]

Got a Sapphire X1900XT from scan.co.uk here that won't go above 650/725 (yup, stock memory) without crashing (higher core clock) or artefacting in 3DMark05 Test 3 (higher mem clock). Suspect a problem with my Antec TrueControl 550PGB PSU - didn't have dedicated PCI-E lines, or a 24-pin connector, and only provided 24A 12V. Have now picked up an Enermax Liberty ELT620AWT, will see if that helps.

 

[Jodiuh]

...and only provided 24A 12V. Have now picked up an Enermax Liberty ELT620AWT, will see if that helps.

lol, you answered your own question and solved the problem in the same post. Good times, eh?

So far I've got mine up to max CCC 690/800. Looks good so far as I just jumped from stock XTX speeds and got no issues whatsoever. Also, it's "overdrive testing optimal settings" button auto detected really high too...within 5 mhz of max. Is that pretty much the norm?

 

[ViperJohn]

So that means its not pulling 24 amps off the 12v line? If not then i should still be ok with my powerstream 520 sli. When i saw that i figured it only left 9a to run the rest of the system.

Correct...that is the current draw off the on-card Vcore PS.

An x1900 is no slouch in the "leaning on the 12V rail" thought. A stone stock x1900XTX has peak loaded 3D power draws of about 121 watts off the Computer PS's 12V rail or a tad over 10amps peak at 12 even.

Viper

 

[JDAdams]

Still artefacting on Canyon Flight 05 if I push the memory above stock - 725 it is then.