Topic: help with setting up 5970 gpu (Read 1964 times)

All this talk about running 5970 fans at 100% reminded me why I started undervolting. At 0.9V and 625MHz you can barely hear the fan on mine and the cards stay at around 80C, getting 3,2MH/W is a pleasant side effect as well. Pretty much you get the same efficiency and hash rate as a stock 7950 with those settings, wish I had more of them    

My cards have gone from mid 70s to 40-50's temp range. No running my cards with fans at 80+% just to keep them in the 70's.  It sits under a desk.  Its silent.  I can overclock them and get full gain out of them as much as their chips will allow.  And $42 over TWO video cards.  Not having to deal with noise, re-oiling my fans to keep the bearings from going, dust, replacing a heatsink/fan if it does fail, etc.  How do you consider that NOT a good investment?

why not?  i'd consider 0.01c wasteful

Well seeing how I have spent a grand total of $42 on water cooling products on 2 video cards, I wouldn't really consider it wasting money.

My screenshot was to demonstrate to the esteemed Mr ssateneth that I indeed have a 5970 and as a matter of fact was running it on the same PC where I read his smug comment.

I had already demonstrated earlier that the fan can be controlled easily for OP by changing one number. In the case of the 5970 if OP was not sure which device is the correct one to control the fan then all he needs to do is change 85 to 100 two times, or whatever setting he wants if not 100%. I thought this would have been clear.

My mining rig is a HTPC/gaming computer in my living room in a tiny apartment. You'll know how loud those fans are from your own experience and that is why I have watercooling: because it is almost silent. And "wasted money" on it? Not really. Have you seen the premium that watercooled cards go for on eBay? Especially in my country where there are far less cards going around than say in the USA. In fact the completed listings feature of eBay shows cards still doing for essentially the same price I paid 12 months ago for just reference cooler fitted cards so please explain to me how I have wasted my money.

When the 6870 and 5970 in my rig were not watercooled, in summer the noise from the fans was so loud that people could hear it from outside the building, and a few people from neighboring apartments had made complaints about it to our body corporate committee. So something had to be done: watercooling or give up mining. Unlike a lot of people here I don't live in a giant property where I can hide a bunch of GPUs in the basement, garage or study room. How am supposed to explain to family and friends who come inside why the computer next to my front door sounds literally louder than some vacuum cleaners?

Not only suffereing from noise issues but with the reference cooler running at 100% fanspeed and stock settings I was getting 90-95 deg C temps on both cores and the some of the voltage regulators were exceeding 128 deg C regularly. Where I live it is common to have 30-40 deg C ambient temps in summer and I don't have 'proper' AC. Volterra's recommended maximum temperature for the voltage regulators on the 5970 is 125 deg C but surely over 100 deg C we'd expect poor lifetimes.

As for the waterblock itself well it is a syscooling block. As you'll see these days there are some reviews on various sites and they are all pretty scathing. At the time I purchased it was the cheapest available commercial solution and I thought I'd give it a try. Indeed some of those scathing reviews are written by me. Not only does the block have no coolant flow channel(s) over the VRM area of the card, but it has terrible machining tolerances above these areas. On any single GPU card having no direct coolant flow over the VRM areas might be acceptable but on a dual GPU card it is not. In hindsight a swiftech or danger den block would've been the way to go.

My overclock is also probably crap because my motherboard is half dead. 2 of the SATA connectors permenantly failed, 2 of the ram slots can no  longer be used as the PC wont even boot with a single stick in them, and it has a host of other problems as well. The board is 5 years old now running essentially 24/7/365 for that half a decade, and will run some time more until I upgrade. The 5970 and other cards might overclock better if I played around with the settings more but I don't trust this motherboards stability so I see little point. I can't in good conscious try to even sell this board as a board+CPU+RAM combo so I am essentially going to run it until it fails completely or it is just so slow I am forced to upgrade.

You can see my setup in the pictures thread but I doubt you'll care since your comments come across as quite dismissive and/or elitist.

https://bitcointalksearch.org/topic/--7216

You'll notice if you follow the link I mention the extremely long coolant lines running between the waterblocks and the radiators, and that my pumps only deliver <1 liter per minute flow rates, this inherently limits the overclock potential as if the above problems weren't enough. If I wanted extreme performance and blistering overclocks I could have tried to put <1 meter of tubing and had a more conventional setup with superior heat removal capability.

my 5970s have always had first core control the fan, though they get reversed in cgminer

it doesn't matter much anyway, they are always at 100%

^^ person up above, screenshot is essentially irrelevant since it appears you've either wasted money on water cooling or aren't using the stock fan/fan power connector, as both of them read -

oh, i guess i'll also add that the last two 5970's i had, I ran at 835-825-795-845 @ 1.05v,  you're running those at 1.1v at much lower speeds

Cool story bro 

https://i.imgur.com/mP0QTwp.jpg

Judging from the posts in this thread, none of you except the OP actually have owned a 5970. Luckily for the OP, I have.

The fan can only be adjusted from the second GPU. The fan setting in the first GPU is a dummy setting (always shows 30%, can't change it. If you try to, it will fail). The second GPU is the one that controls the fan. Bear in mind that fan control on the 5970 is a little wonky. Speeding up is usually instantaneous, but slowing down is done slowly, so if you try to decrease fan speed and it doesn't seem to be working, give it a minute.

Here is what I have for my machine here at the office: 5870 / 5830

Just remember, the 5970 is essentially 2 cards, each one a 5870.

If you are setting a target temp in your config then you need to use --auto-gpu and --auto-fan

Make sure in your ATI CCC panel that "Overdrive" is disabled.

Then choose to either use a vendors program to modify the fan speed, or, do it with your miner.

If you want cgminer to force a fan speed other than its default 85% speed, you need to edit your config file or add an argument for "gpu-fan"

I use the config file and I have four GPU's so my config file looks like this:


Where the first 3 GPU are watercooled so fan speed is no matter, and the 4th is set to 100% speed.

My config file does not list "auto-gpu" or "auto-fan" nor do I use those arguments.

---

If it's not working then you probably have a 3rd party or vendor program that's actively hogging the state of the fan speed.

How do you get them to run at max because no matter what am trying its always been auto and driving me insane trying to change it.

i've had like half a dozen fans fail, but if you're running reference cards they only cost about $10 to replace anyway

i always run (ran) the 5970 fans at maximum

Hmm, couldnt tell ya.  Then again I never said I was a CGMiner expert. Try posting on the official thread and see what they have to say.

https://bitcointalksearch.org/topic/official-cgminer-mining-software-thread-for-linuxwinosxmipsarmr-pi-4110-28402

This is what I have for this card

cgminer --lookup-gap 2 --gpu-fan 75 --thread-concurrency 8192 -g 1 -I 17 -w 256 they are the settings however gpu fan control change does not work and my temps are hitting 80s

Well yeah, if you do that kind of continual upkeep then yes they can run continually that way.  But, lets be realistic here, 98% of the people (including myself) do not do this, so I was siding with the majority.

I have never had a fan fail, and I have run dozens of cards under heavy load even before bitcoins popularity boom running a 4870x2 farm for distributed computing. Usually when it is time to clean the fan of dust I apply some light machine oil ($2 at hardware store) to the bearings. On some cards you have to drill/dremel a small hole (0.5mm or less) into the front of the fan to access the bearing.

An option, yes, but running your fans that high constantly will burn them out quick.  Trust me, have done it to 3 cards.  But again, you can set constant fan speed with CGMiner, you dont need Trixx, or MSI, or Over Drive or anything.  Those are simple GUI programs that arent doing anything different that what CGMiner can do.

Unless you have a source of air <20 deg C you will want to run those cards at 90 to 100 % fan speed at all times to keep the VRM temps as low as possible (monitor with GPU-Z)

My suggestion would be to:

-Disable fan changes with mining software
and
-Use Sapphire Trixx to set fan speeds OR modify each GPU's BIOS with Radeon Bios Editor (RBE) to give a fan profile that is much more vacuum-cleaner like.

Post all of your settings.  Sometimes you have to have other arguments in order to get one to work (i.e. auto-gpu doesnt work withouth target-temp and auto-fan, etc).