I've been doing a lot of research on airflow and CFM in regards to cooling annoying GPU rigs.

Anybody who has MSI 7950's, or any 7950's with similar 80mm fans, would like to take a stab at how much air those things move? (CFM)

Estimates at 50% 75% and 100% power.

I can't imagine these things move more than 20-35 CFM if not a lot less.

I have an idea to pretty much enclose my currently open design and then use several 120mm fans to intake on one side and exhaust on the other. Moving air works the best to cool these GPU heatsinks.

A little bit of testing quickly lets you know that just randomly blowing fans on these things can really be pointless.

For example I see a lot of people putting fans on top of there cards blowing down in between the cards. Now, I can't speak for every card design, but I have seen people do this specifically with MSI 7950 twin forzer III cards, which will in fact raise in temperature if air is blowed down on them.

It is more effective to blow air in between the cards from the back and front.

Due to the nature of my apartment (small 550sqft for college students really) I don't have much location for my rigs. I was trying to figure out how I could possibaly run them in my closet, I went so far as to get a 7,000 BTU portable air conditioner that was being exhausted through an air vent, that was no luck at all lol high gpu temps, ambient temperature was ridicoules.

I can run them in my living room, at 66c overclocked to high hell with just a single honeywell fan blowing on them with no issues. However, there is a lot of noise associated with that. and due to the fact that my cat needs to be able to go from room to room in the evening, I can't just shut the door. My bed is also 10ft away. And I'm a light sleeper. When I first built the rig I actually slept with ear plugs for a couple weeks lol

Anyhow, all advice and suggestions are welcomed. If anyone has any awesome secrets to cool temps and low noise levels, please share!!!

I reinstalled drivers, didn't reinstall the OS, do you think that could make a huge difference? I did go from AMD to Intel, so I guess it wouldn't be to far fetched. And yeah all the voltages are the same, the only way to lower the voltage is in windows with afterburner, and I'd rather use Ubuntu so lower voltages arent an option right now.


My thermostat says 68 but I highly doubt that. The old design was tested though in an open room that I would say is a lot cooler. The cards were a lot closer in the first design and ran at 66c so I don't think thats it, I think it has more to do with the entire rig conducting heat, which I noticed it is, and quite a bit. Also the fact that its not in a closet, on the top shelf near the ceiling I think is the real killer.


Yeah I think you're right with that hot air evacuation. The fact that it sits 3 feet from a ceiling in a closet I think is creating a lot of extra heat/hot air with no where for it to really go, though I do have cold air vent right above it blowing on it.

I started trouble shooting further, if I run just one card by itself, it stays at about 64/65 so I dont think that throws out the OS reinstall idea. I think it works fine in that sense. I added a second card, both cards ran at about 72, so I deft think its a heat/space/fan issue. I added a third card and then came a lot of problems, the third one runs a little warmer then the other two at temps 78 78 82 then after >30 min one will go dead/sick from over heating.

I currently have all the gpu fans turned up to 90% to help. It helped a little.

I currently have little computer fans on top of all three cards and testing to see how it helps. right now I'm at 73 73 78 so that's not bad at all. I also took the vent off the ac duct, so that it blows straight down on the rig. I think adding a 12/15" honeywell fan well improve it a lot, but realistically how much could that fan help? not more than 5 degrees I would imagine.

I haven't even gotten to adding the other 3 cards for a total of 6. That should be fun...

I also want to move it to the floor and see what kind of temp improvements I get.

Also, is it possible that the heat sync from the cards touching the aluminum it's sitting on, in turn making the aluminum almost to hot to touch, keeping the cards warmer than they should me?

Thanks for all the input. 

So I had a pretty neat open air design for my x6 7950 rig, and all the temperatures stayed at about 66/67c but I wasn't satisfied with the design and wanted to re do it. Here is a pic of the old design.

http://distilleryimage1.s3.amazonaws.com/928e0168cef911e2bf3622000a9f1886_7.jpg

Not bad, but I could do better. So after many beers and a long weekend with some fellow engineers. I have this.

http://distilleryimage4.s3.amazonaws.com/e9f5f114dc9011e28cc022000ae80ec6_7.jpg

Borrowed the design on here from the thecryptokeeper

Anyhow, we were handling all the cards kind of rough I guess and the motherboard, in the process I either shorted the cpu or mobo or both, either way the damn thing wouldn't boot, it would come on, spit some LED Debug error, and the cpu fan wouldn't even crank up.

So i swapped out the mobo and cpu for the MSI z77 65 with a basic celeron processor, the notorious mobo everyone likes to use because it doesnt need a presence pin short and etc etc.

Long story short, I booted this thing up, re installed drivers and updates, made everything exactly the way it was before and I'm only running 5 cards right now, 6th one is buggy, but of those five 3 of them stay at 63,66,68 which is great. However these two other ones, are both currently at 80. Even though I have there fan speeds cranked up to 90% while the other cards only are at 70%.

Any ideas why on earth these temps are so much higher? Theres no way its running hotter in the new design, so I'm thinking maybe the heatsink came loose and I need to tighten the screws on the back of the cards? Or maybe fresh thermal paste?

Any insight and ideas are much appreciated!!!

Greg

Hey guys, so I'm running a LEPA 1600w PSU and a gigabyte FM2 a85x up4 board, with a single corsair 8gb stick of ram, 80gb sata hard drive, and an A4 processor.

This rig was originally running x5 MSI 7950's without a problem, but two were a little warmer than the others, 5 degrees higher, and the 6th card was dead. So I said the hell with it, I'll have all 3 replaced...

Well, when the new ones got here, before I swapped them in I plugged in a kill a watt meter and realized I was pulling 1520w give or take with only 5 cards and there was no way I'd hit 6 cards with the current settings.

I tried undervolting in cgminer (running ubuntu 12.04) and no luck, even though I know for a fact afterburner in win 8 or 7 will do it in a second. Very frustrating. So the next thing I tried was gpu bios flashing, I figured I'd try them on a card that was running warm that was going to be replaced just in case something went south. Well it turns out it did, I wasnt able to actually edit any 7950 bios, because you can't just edit 7000 series bios these days. And so in an attempt to get a lower voltage was flashing other 7950's bios, stupid, I know. This led to a 15% decerease in hashing power, and the same damn voltage. So no progress, in attempt to grab the bios off another card, no issues there, I wasn't able to flash it back to the card I messed up. So at that point I just pulled it out and said this one is going back, period.

So I swapped in two new cards with the bad bios one and the other warm one. I also switched the powered risers I was using, some of you may have seen my last thread about how they were poorly made/soldered and caught fire. I was about to go away for 5 days and leave these running, so I thought lets not take any chances, and I switched out the powered risers.

At this point all sorts of interesting problems started happening. Long story short, only 4 would run stable now, and I had to leave for a trip and didn't have time to mess with it. I just got back the other day, and still really don't have to much time till this weekend and have just been running it with those 4 cards to get some kind of work in.

But NOW a new serious problem has arrived. After about 12-15hours the whole rig just shut off ,cold turkey, this happened sunday night into monday monring, and then monday morning into tuesday morning, and now this morning it is off again. However, after checking all connections and re plugging everything including the PSU and power cable. IT WON'T TURN ON.

It turns on for like a second the cpu fan spins a few times and not a second later it just stops.

My question is, is this the PSU failing, or have I shot the motherboard.

With 4 cards running I was only pulling 1150 on the meter, no where near the 1600.

I know that's quite a story, but members usually want as much detail as possible to give as educated/accurate answer as possible.

Thanks,
Greg

UPDATE: If I pull all the cards and risers out, it does boot. So maybe the risers?

UPDATE2: Swapped one of the powered 16x to 16x risers with one of the older ones from the batch that had one catch fire. I added some hot glue around all the contacts to keep them in spot, system booted and is mining now with all x4 cards.

I think these other powered risers I was using aren't pulling any power through the 12v molex adapters, and to much power is getting pulled through the board. That's my best guess. If someone could link a good step by step how to make your own powered risers, I think I might give that a shot.

+1   

How is payout time for a GPU 400 days? Wow, that number is really off. At a modest estimate 180 days, at a good estimate 120, anywhere in between is fair game.

If you have serious GPU power, then you just hop on every morning, check how the alt coins are doing, have a scrypt/bat file setup for each type of coin, I just hop on my ssh tunnel from my laptop, 3 lines later I'm mining whatever coin is most profitable, and I ride it till it changes, I usually stay on a coin for 24-48 hours. If some of the alt coins are being flakey with exchange prices (ie digitalcoin) then maybe I'll hop on LTC for some stable income to pass the time, and then when something flys back up in profitability, hop over.

Obviously mining isn't as glorious as it was say 45 days ago, when I could have easily made $120/month per 7950 with LTC, now that number is about $84 if I hop alt coins every day which is pretty sad, but hey, I'm still making plenty.

Especially because I don't pay for electricity and have infinite access to all the power I could ever need lol

Put it this way, $12,500 deep. 5 rigs, x6 cards each, each card hashes at 675. That returns $2500-3000

And even if I was paying for power, that would shave off about %35.

I don't think that's bad at all.

Yeah that is the exact same guy, believe it or not the ones from china are built better Lol

Would you be interested in linking him to this page and asking for a refund together? Because I honestly don't want to use the other 2 I bought from him.


I know right!! It was on a tile floor, so at worst maybe a small burning piece of cable could have fallen? But nothing flame-able nearby, and its funny because I was getting read to move these into a highly flame-able closet Lol

I got some new powered risers the other day from china, these are much sturdier and the connections look solid. The one's I got from LA though, deft. no good

I didn't build these, they were ordered from a vendor.

Thanks for the insight.

Update:

Hephaestus was able to help me get them going in Ubuntu, discovered one of the cards was actually bad and possibly a second one.

Currently have x5 7950's stable in Ubuntu 12.04 at about 670 each temps are all below 70c one of the cards likes to spit out a handful of HW though, so I'm having that one replaced by amazon and the 6th card is also sick/dead/faulty so they both will be replaced, and I'll post back here and see where I'm at.


It would be nice to run them in windows strictly for access to afterburner, but if the lepa 1600w can handle them with under volting it, then I would much rather stick to Ubuntu or any linux based OS, SSH and byobu are glorious. 

Hey it's all good man, can't control everything. None the less, great site!!

Maybe post on this thread when it's back up.

Hey! great site! love it.

Though right now I can't get it to display any data, the whole chart of coins/rates etc is missing.

Is the site down for repair?

Thanks,
Greg

That particular riser was plugged into a card, but the other end wasn't plugged into the board, card wasn't plugged into the psu either. Had been sitting like that for a while now, then on a reboot it just decided to catch fire.


Yeah capacitors are cheap to replace, that is an option. Does anyone have a good guide/link to making your own powered risers/extenders? I'd be interested to see a good close up/diagram of where the wires should be soldered on.

Sorry, software engineer, not electrical lol. I corrected it

I'd say very dangerous, and no safter markings at all on the "power filter"



You got it, and I don't know what the hell is a power filter either lol but thats what the ebay title calls it.

In the picture it's the little transistor capacitor that says 16v, it sounded like it popped.

You can see the cable is completely burnt out from the being on fire for a sec lol

Hey if anyone has ordered these "PCI-E Extension Cable 16x to 16x Slot Extender Riser Card Adapter Power Filter" off of ebay from California, they advertise same day shipping.

Please be aware, one of there cables literally just had the power filter pop, and the entire 12v yellow cable burst into flames...

Just a heads up.

Greg

When you tested each GPU, did you test them individually or with the others plugged in. Or one by one. I'll give that a try both ways in just a moment. So should I just count on only using x5 with a 1600w? If so thats fine for now.

Any ideas how I could find the 12.6 trick for windows 8? I had windows 8 with 13.1 earlier and device manger saw all 5 of them that were plugged in at the time just fine. However, when cgminer started I would see all the hash rates jump up where there suppose to be, but then the screen would either freeze or go blank and I would loose all control, it was like the driver was crashing.

Not having to short any pcie slots would be nice, I had to short one of the 1x on my gigabyte board. It has no problem though seeing all 6 in ubuntu or xubuntu, I just cant run cgminer with a thread concurrency above 1792 or all the cards but one report as off, on the flip side if i run at 1792 they all hash at 150 or so, but i get 10 HW error a second a card so thats issue as well.

I'd prefer to stick with windows 7/8 so that I can utilze afterburner to turn down the voltage, though I suppose a bios flash is always an option?

Yeah sorry if that came off as a bit harsh, If you don't have any credit cards this might be a good time to give discover a call and get one lol. Yeah I mean if someone is really ready to do it and according to the last post it was resolved? Then best of luck to you. Though really, do consider grabbing at least one credit card for emergency's at least

If you're not a kid behind a screen and actually have bills, proof of income, and own things in your name, 1.85 BTC ($225 at best) is easily manageable on a credit card? they give college kids credit lines in the thousands, employed adults have on average a credit line of $10-30k, if you're confident that you can pay it back quickly, why not put it on a card? They won't charge 30% interest either... 18% at worst...

I think resorting to basically begging a forum publicly, (one of which you recently joined) to lend you $200 bucks is beyond absurd.

Also, you have to be out of your mind if you think any right person would hand over there SSN... online... to anyone for anything...

There's just literally a million better ways to go about, getting your hands on $200

if cgminer doesn't change it for you with the gpu vdc option then your next best bet is flashing the bios.

I have the same PSU and Graphics cards as you and am struggling to get all six working. How did you figure out that one of your rails was malfunctioning? Also, did windows 8 see all six right off the bat? Just installed 13.1 and they worked?

I'm using an A4 FM2 processor with a gigabyte F2A85x-UP4 but so far its been nothing but a pain in the ass lol. If you have 6 msi r7950 twin frozr working good in windows 8 with that msi board, I might be making a return to amazon soon lol