What flags should I use for 5830s?

[P4man]

If your temps are anywhere near 90C with the fans at 100%, you are doing something wrong with case flow or you should really invest in a better cooler. If the DispIO (the default) temperate sensor is at 90C, more often than not the shader temp will be near or over 100C and god knows what your VRMs are at. Run GPU-Z and find out.  None of my cards (5850s and a 5870) are over 65C, none of them are anywhere near 100% fan (either fixed speed 7v or ~50%). At 90C, I wouldnt bet on them lasting a year, let alone two.

[1713927226]

1713927226

[1713927226]

1713927226

[1713927226]

1713927226

[m3sSh3aD]

If your temps are anywhere near 90C with the fans at 100%, you are doing something wrong with case flow or you should really invest in a better cooler. If the DispIO (the default) temperate sensor is at 90C, more often than not the shader temp will be near or over 100C and god knows what your VRMs are at. Run GPU-Z and find out.  None of my cards (5850s and a 5870) are over 65C, none of them are anywhere near 100% fan (either fixed speed 7v or ~50%). At 90C, I wouldnt bet on them lasting a year, let alone two.

i oc cards. 8x5850s(2 rigs which are tables not cases) A 6950(930/300),5970(899/300) & 5970 vapour X (900/300) in one case (well, 2 outside the case ) and a 6870(1025/300 work in progress) all below 80 degrees with 100% fan. VRM's stay well below 100 degrees (120 is there max without effecting there lifespan) tested with IR temp sensor

My point is 90 degrees is perfectly fine, 110 IS DANGER DANGER. Also, auto fan ISNT RELIABLE

[P4man]

i oc cards. 8x5850s(2 rigs which are tables not cases) A 6950(930/300),5970(899/300) & 5970 vapour X (900/300) in one case (well, 2 outside the case ) and a 6870(1025/300 work in progress) all below 80 degrees with 100% fan. VRM's stay well below 100 degrees (120 is there max without effecting there lifespan) tested with IR temp sensor

My point is 90 degrees is perfectly fine

No it isnt. Read up on electromigration and current crowding:
http://en.wikipedia.org/wiki/Current_crowding

The impact of temperature on chips life expectancy is exponential:
T_failure = A exp (E / k T) / (J^2)

Moreover, read some VRM specs.



To give you an idea, increasing VRM temperature for the above VRM from 65C to 105C decreases average lifespan by a factor 100x.. Most ATI VRMs are specced for 130C, so replace 105 by 130 and do your own math.

Its not because a card wont instantly fail at 90C core or 120C VRM, or because none have failed after a few months that it means its safe, because it isnt. You are MUCH more likely to kill your cards at those temps. But hey, its your money.

[m3sSh3aD]

i oc cards. 8x5850s(2 rigs which are tables not cases) A 6950(930/300),5970(899/300) & 5970 vapour X (900/300) in one case (well, 2 outside the case ) and a 6870(1025/300 work in progress) all below 80 degrees with 100% fan. VRM's stay well below 100 degrees (120 is there max without effecting there lifespan) tested with IR temp sensor

My point is 90 degrees is perfectly fine

No it isnt. Read up on electromigration and current crowding:
http://en.wikipedia.org/wiki/Current_crowding

The impact of temperature on chips life expectancy is exponential:
T_failure = A exp (E / k T) / (J^2)

Moreover, read some VRM specs.



To give you an idea, increasing VRM temperature for the above VRM from 65C to 105C decreases average lifespan by a factor 100x.. Most ATI VRMs are specced for 130C, so replace 105 by 130 and do your own math.

Its not because a card wont instantly fail at 90C core or 120C VRM, or because none have failed after a few months that it means its safe, because it isnt. You are MUCH more likely to kill your cards at those temps. But hey, its your money.

And they paid for themselves already.... in 3 months haha ATI make solid hardware, always have whereas nvidia's as always been sketchy. Nvidias drivers are better, thats about it. People worry too much. If a VGA card last 3-5 years, Who cares? really lol. Find cards with lifetime warrentys if your so concerned, most come with 3/5 years anyways

my 6 series are replacements for fans that failed on XFX cards, avoid them like the plague. 6 series just isnt as good as 5 series I've had massive success with sapphire who have really impressed me over the last year or 2. And there trixx overclock is the best software out there (although only windows).

[sadpandatech]

  To get back to the point Mess was making. AUTO FAN IS BADDDD.  I don't know how it is handled in cgminer or whatever you are using. But, the auto fan feature built into these 58xx cards will fuggin let them get extremely hot, 90+ before kicking any real RPMs up.....  I set it to manual 80 for each card, then lower from there for the cooler cards. Most of them end up about 60% at 70c

[m3sSh3aD]

  To get back to the point Mess was making. AUTO FAN IS BADDDD.  I don't know how it is handled in cgminer or whatever you are using. But, the auto fan feature built into these 58xx cards will fuggin let them get extremely hot, 90+ before kicking any real RPMs up.....  I set it to manual 80 for each card, then lower from there for the cooler cards. Most of them end up about 60% at 70c

That was my point, Auto is a bad idea if your using your cards 100% load 24/7. You'll get failures for sure. I use trixx in windows. My miners are win7 just for ease of use. *Im actually on a linux debian netbook that i monitor everything on haha. And android for that matter. Multi platform

If you set 100% fan (90 if your worried about 100) then OC you know where you stand, less variables

[sadpandatech]

  To get back to the point Mess was making. AUTO FAN IS BADDDD.  I don't know how it is handled in cgminer or whatever you are using. But, the auto fan feature built into these 58xx cards will fuggin let them get extremely hot, 90+ before kicking any real RPMs up.....  I set it to manual 80 for each card, then lower from there for the cooler cards. Most of them end up about 60% at 70c

That was my point, Auto is a bad idea if your using your cards 100% load 24/7. You'll get failures for sure. I use trixx in windows. My miners are win7 just for ease of use. *Im actually on a linux debian netbook that i monitor everything on haha. And android for that matter. Multi platform

If you set 100% fan (90 if your worried about 100) then OC you know where you stand, less variables

  aye, and fans are cheaper to replace than GPU's ;p  Just make sure to monitor the fan or ahve it set to idle down if it does die.  Mine will alarm and cut power tot he cards if fan dies. It also helps that the ones not plugged into a board I've built individual enclosures for that force air through them. if fan dies and cut-off fails they would likely still live long enough to take action. =)

  as for flags, VECTORS DEVICE=0 BFI_INT AGGRESSION=8 FASTLOOP=false WORKSIZE=256 -k phatk2

   change device= appropriately

[P4man]

The "auto fan feature" being referred to was cgminer's autofan feature which is not "BAD" its superb, it works beautifully. Just set a target temperature and let cgminer figure out what fan speed is needed. No need to kill the fan in order to save the card, you can do both (if you are not stacking your cards with 2mm spacing like I suspect m3sSh3aD is doing).

BTW, in cgminer you can also set overheat temperature and cutoff temperatures at which points cgminer will underclock and stop the card.

[P4man]

Since the OP is using cgminer, this may help. Here is what I use for cgminer:

cgminer -c config

with config being  a text file containing:

Code:

{
"pools" : [
	{
		"url" : "http://mint.bitminter.com:8332",
		"user" : "P4man_boven2",
		"pass" : "***"
	},
	{
		"url" : "http://api2.bitcoin.cz:8332",
		"user" : "P4man.boven",
		"pass" : "***"
	}
],

"intensity" : "9",
"gpu-engine" : "825-900",
"gpu-fan" : "40-85",
"gpu-memclock" : "300",
"gpu-powertune" : "0",
"gpu-vddc" : "1.088",
"temp-cutoff" : "85",
"temp-overheat" : "75",
"temp-target" : "62",

"algo" : "c",
"auto-fan" : true,
"expiry" : "120",
"gpu-threads" : "2",
"log" : "5",
"queue" : "1",
"retry-pause" : "5",
"scan-time" : "60",
"submit-stale" : true,
"temp-hysteresis" : "3",
"worksize" : "256",

"donation" : "0.50",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
}

Note it starts the fans at the maximum speed you defined in the config file (85% for me) and will slowly slow them down to reach your target temperature.

Defining two or more pools also allows you to do failover, if one pool is DDoS-d or not reachable, it switches over to the next one.

Here is a screenie:


ignore the avg MH/s, I just restarted it. Also the RPMs are almost certainly incorrect, its an aftermarket cooler with dual fans. 50% is whisper quiet, not what youd expect from 3000rpm. Its probably 1500 RPM

[m3sSh3aD]

Im have not used cgminers auto fan so i cant comment on its internal one, I still wouldn't trust it though. But thats just my personal opinion. I need have a play/look around all the options in CGminer as i only really use the pool fallover feature at present. I like that you can set it to kill the device if it does stop, Will have to look into that

[ragingazn628]

That's what it should look like, except your fan speeds are through the roof...should be no more than 85% for 24/7 use.

Familiarize yourself with --auto-gpu & --auto-fan, also the Core and Memory clock switches.

As long as your temps are below 90 put your fans on 100%, auto IS A BAD IDEA, ALWAYS. Ive seen mine throttle the fan down when its 100% workload because of all sorts of things. Hense why i have 6 series cards on some cause my 5 series fans failed due to this. 6 series sucks compared to 5

I always have 100% and as long as it doesnt get too hot, 100+ degrees they WILL seize up. Then again, i plan to replace cards every year or 2 so its not as if im too fussed about thrashing the fans. I wouldn't go below 85% fan speed, Dependant on your location as well though (tempreture outside).

Looks like you got it working sweet. Try playing around a little, you might get slighly different results with say 2 threads, not 4. Worksize 128, not 256 etc etc but Vectors 2 seems to be the best on all of mine. Everyones system behaves differently, cause no card/board is the same

Well I got it running great all night. When I woke up, turned on the monitor and my top card crashed (OC;d at 975mhz) so I did a restart and downclocked it to 950mhz and OC'd my bottom card to 1000mhz!!! both together are giving me ~630 mh/s

I currently have both fans at 85%, should I up it like you said?

Top GPU Temp is at around 60C at night and 66C during the day
Bottom GPU Temp is at around 52C at night and 58 during the day

[m3sSh3aD]

no need, 85% with them temps is sweet

[ragingazn628]

no need, 85% with them temps is sweet

Thanks I'm a first time miner but I learn fast. Here's my specs;

Sempron 145 Unlocked to Dual Core (Getting a Phenom II X4 830 for $50 from MC soon - so I can game when not mining, take a break? lol)
ASRock M3A770DE AM3
2x HIS 5830 CF (used)
Rosewill 1000W Bronze Certified

Whole rig cost me $320

[m3sSh3aD]

Drop the intensity to 6, if you get a bit of judder still in movies/games, drop it to 5. Use the affinity to put cgminer on 1 core but it is a sempron, may need go to I4 but you should get a fully usable system with good mining performance, It basically uses whats available at them intensitys and lets your system roll on

[ragingazn628]

Drop the intensity to 6, if you get a bit of judder still in movies/games, drop it to 5. Use the affinity to put cgminer on 1 core but it is a sempron, may need go to I4 but you should get a fully usable system with good mining performance, It basically uses whats available at them intensitys and lets your system roll on

no this is a dedicated mining rig. When I want to play games I will stop mining.

[P4man]

Well I got it running great all night. When I woke up, turned on the monitor and my top card crashed (OC;d at 975mhz) so I did a restart and downclocked it to 950mhz and OC'd my bottom card to 1000mhz!!! both together are giving me ~630 mh/s

Learn to use cgminer. It will auto overclock (and reduce clocks if needed) and even restart the GPU's if they crash. That doesnt always work, but usually it does. It will also take care of adjusting the fan speed. Your temps are great, feel free to use the autofan option and set the target temperature a bit higher than what you have now, so it doesnt sound like a hairdryer. Do check your VRM temps with gpu-z though, just to be sure they arent burning up while the GPU itself is nice and cool.

[ragingazn628]

Well I got it running great all night. When I woke up, turned on the monitor and my top card crashed (OC;d at 975mhz) so I did a restart and downclocked it to 950mhz and OC'd my bottom card to 1000mhz!!! both together are giving me ~630 mh/s

Learn to use cgminer. It will auto overclock (and reduce clocks if needed) and even restart the GPU's if they crash. That doesnt always work, but usually it does. It will also take care of adjusting the fan speed. Your temps are great, feel free to use the autofan option and set the target temperature a bit higher than what you have now, so it doesnt sound like a hairdryer. Do check your VRM temps with gpu-z though, just to be sure they arent burning up while the GPU itself is nice and cool.

WHere is VRM temp on here?



Uploaded with ImageShack.us

[P4man]

There isnt. Not all cards support it. Yours doesnt. In fact, most non reference cards dont support it, but AFAICT, most non reference cards also have no VRM temperature problem.

[ragingazn628]

There isnt. Not all cards support it. Yours doesnt. In fact, most non reference cards dont support it, but AFAICT, most non reference cards also have no VRM temperature problem.

AFAICT? Never seen this acronym before

[P4man]

AFAICT? Never seen this acronym before

http://bit.ly/vbU2NR