|
99.3 and 99.1 to slush's via proxy, otherwise about 97 to getawork btcguild
|
|
|
|
I'm not confident to run those beauties in HIGH clock because I'm not sure what friedcat means by "if you are having heat problems".
Does anyone knows what configures a heat problem?
What is recommended operating temperature?
Where is the recommended place to measure temperature? Radiator? PCB? On top of hashing chip?
Run them in high. If they get too hot, they'll restart and go into low clocks - that's your warning. Too hot means the boards are resetting, although arguably we'd want these boards as close to room temp as possible. I'd say radiator is the only place you can really measure. How have you mounted the mini fans onto the heatsinks? I've mounted the fans using liquid epox, they withstand high temperatures. With the clock in high they get so hot you can only keep your hand over it for 15 seconds or less. To overclock you need 2x120mm fan on each. |
|
|
|
I'm not confident to run those beauties in HIGH clock because I'm not sure what friedcat means by "if you are having heat problems".
Does anyone knows what configures a heat problem?
What is recommended operating temperature?
Where is the recommended place to measure temperature? Radiator? PCB? On top of hashing chip?
Run them in high. If they get too hot, they'll restart and go into low clocks - that's your warning. Too hot means the boards are resetting, although arguably we'd want these boards as close to room temp as possible. I'd say radiator is the only place you can really measure. How have you mounted the mini fans onto the heatsinks? |
|
|
|
What option does he have if he wants to mine right now though? ASIC stuff is INSANE prices now--plus most are delayed till who knows when. Unless you got in on that first buy, ASIC is pretty tough to jump into at this point.
There were FIFTY of these blades auctioned only a week or two ago. They were delivered in 4 days across the globe. It is hard, but its not impossible. But even FPGAs would have made more sense than such a large GPU farm than this, and they're readily available. |
|
|
|
Wooo. Great improvement from 7950/7970s then.
It is a giga hash card... a 1-1.26 giga hash card, however it runs hot and it is pricey. Its ROI at todays prices would be 7-10 months if mining bitcoin. Litecoin should be shorter time. Too long for me to wait really. Also another update on temps: gpu1 @ 88C, gpu2@ 79C. The engineering of the cooler when in closed case is very poor and unbalanced. I am thinking of getting all sapphire 7950 and sapphire 7970 or sapphire 7870xt now. I have another sapphire 7990 coming, however I think it will do the same as the gigabyte because it has the same design. My next step is to test Litecoin mining. And my Club3D 7990s do 1.2GHs and stays under 84C with the hotter core. I guess the 3 slot design wasn't that unnecessary after all.
As much as I hate to say it, I told you so  My 3 slot cooler struggles like hell on the hottest 2nd card when overclocked, I shudder to think how that 2 slot is coping. Because of the 3 slot I'm able to hit 1350ish per card without going above stock voltage. I'm interested to see the scrypt results although I think we're going to see some power (and/or thermal) throttling as it bounces off the edge. Personally I'd be wary about attempting it at all as if the card isn't software limited to 375W then you might damage some of the power regulation on the board, or even worse, on the mobo. |
|
|
|
good luck with your asic rig & SHA-256 algo ONLY  Good luck with your 8KW of heat and constant problems. I've GPU mined, trust me it will drive you to insanity. |
|
|
|
You would have been so much better off putting that $15k into buyitnow ASICs or even FPGAs. You're going to have to power and cool 8KW of cards, and with the amount of planning demonstrated so far its going to go terribly wrong My 26GH system cost the equivalent FIAT today of about $11k, only generates 250W of heat and sits on my windowsill. Which machine do you have that can have 26 gh/s on 250w. 2x ASIC blades. You would have been so much better off putting that $15k into buyitnow ASICs or even FPGAs. You're going to have to power and cool 8KW of cards, and with the amount of planning demonstrated so far its going to go terribly wrong My 26GH system cost the equivalent FIAT today of about $11k, only generates 250W of heat and sits on my windowsill. Which machine do you have that can have 26 gh/s on 250w. hes probably talking about 250w of excess heat, not wall power usage. I just dropped $500 to get another gighash/megahash on some mid level cards. I think what the other poster said, wait for BFL to catch up and buy one of their ASICs. A large GPU farm is risky in the BTC world.. LTC would be more realistic. No, 260W from the wall [230V]. Going down to 20GH I can get power down to 160ish |
|
|
|
You would have been so much better off putting that $15k into buyitnow ASICs or even FPGAs. You're going to have to power and cool 8KW of cards, and with the amount of planning demonstrated so far its going to go terribly wrong My 26GH system cost the equivalent FIAT today of about $11k, only generates 250W of heat and sits on my windowsill. |
|
|
|
|
I have my entire computer top and back enclosed with 20 sheets of paper, taped up going to an exhaust pipe which goes to the window. Exhausts at least 80% of the heat, without any fans in the actual tubing.
|
|
|
|
Avalon .. they are becoming the next BFL =(
its sad
How on earth do you start comparing 6 months of "ITS COMING NEXT WEEK BUY BUY BUY BUY BUY BUY MORE ITS HERE GOGOGO OR LOSE AND DIE" to 3 weeks of "yeah sorry team has been ill, off schedule by a few weeks." |
|
|
|
I hope you get it right
happy mining
Don't get me wrong guys, I do appreciate your input and help - its just I'm the only one in my room groping these particular boards with this particular airflow. Finally got round to setting up a stratum proxy (that was easy) to slush. Tried mining_proxy.exe -o stratum.btcguild.com -p 3333 to btcguild but it spams errors. I can't make sense of them but either way it doesn't send info to btcguild. |
|
|
|
3rd attempt at mounting to reduce temps, ready for the new fans. This seems to be the approach most are taking, flat with a 1" gap underneath. This will be ventilated with the old 120mms once the new ones arrived.
Did you read this? https://docs.google.com/document/d/1VMO3VfIBy3KIwUaaQ8XhKrSVEbXeeMEqsVhhk6Bz9io/edit?pli=1quote: "Please make it vertically placed if you could. The chips are packaged via QFN, but not all the heat are absorbed by the PCB. The back side will also become hot. And if you have no mechanical setting to lift it and have to lay it on the supporting table, please make sure that there are no heat isolation material under it." Why did no one read what I actually posted :/ "flat with a 1" gap underneath." The gap still traps the hot air without any ventillation. It's easy to mount it on it's edge (again, check lightbox's comments) and you can easily create an air channel, using cardboard or plastic. here's a crappy sketch: https://dl.dropboxusercontent.com/u/8310142/cooling.png That was my initial config and it wasn't even close to sufficient because of the inability to force the fans onto the heatsinks. There is actually quite strong airflow underneath and the mini fans on the side blow under there. Because its quite a confined area a small amount of conventional airflow pushes all the way through as there is no where for the pressure to escape. I can feel the flow underneath and its as strong as when it was standing. I'll check underneath temps and temps of the box walls etc after 4 hours of uptime. |
|
|
|
|
B2 is identical to B1. B3 might have a different, smaller case, don't think it was finalised. B4 will likely have the half size avalons.
|
|
|
|
3rd attempt at mounting to reduce temps, ready for the new fans. This seems to be the approach most are taking, flat with a 1" gap underneath. This will be ventilated with the old 120mms once the new ones arrived.
Did you read this? https://docs.google.com/document/d/1VMO3VfIBy3KIwUaaQ8XhKrSVEbXeeMEqsVhhk6Bz9io/edit?pli=1quote: "Please make it vertically placed if you could. The chips are packaged via QFN, but not all the heat are absorbed by the PCB. The back side will also become hot. And if you have no mechanical setting to lift it and have to lay it on the supporting table, please make sure that there are no heat isolation material under it." Why did no one read what I actually posted :/ "flat with a 1" gap underneath." |
|
|
|
Dogie, take the blade out of the box. The chips on the opposite side of the board need air circulation. I would try to set the blade on its edge, heat-sink on the bottom. Get two 120mm case fans and park them in front of the heat-sink with air going away from the sink. I added a third fan today and have it pointed across the power supply module, it was getting a little too hot for comfort. The third fan blows air across the length of the module, not directly on it.
The blades aren't meant to be resting on anything that would stifle air flow across it.
They're actually on 1" stilts and can raise them up more. Will have 2x120mms blowing sideways across/under the boards. I've tried 3 different configs and this is by far the coolest, and will get even cooler with the new fans tomorrow. Power boards have their own mini fans, will have bigger ones soon. That's going to be an epic fail...
..why? |
|
|
|
As it has been said before, this approach is the worst, because the fans are just circulating the heat around the blade. The goal is to transfer the heat from around the blade. The problem with turning the fans into a pull config is that there isn't a seal, or anything close to a seal on the fans. They'll just spam all their pressure sucking in the non heatsink air because its so much easier to move. At least with a push config it has to go into the heatsink. |
|
|
|
this is the old 7990, not the new 7990 malta..i would not pay $1000 for this as it under-performs the new 7990 maltas. No, no it doesn't. Identical cores, identical specs. This new one just has way worse cooling and is power limited to 375W. It can NOT mine scrypt. yes identical tahiti cores, but not same specs. the new one is clocked higher than the old one by at least 25mhz each core. what sources do you have to say the new one has worse cooling? tomshardware and a few other websites say the new one has much better cooling. it boggles my mind why a new card that came out about a year after the first 7990 would have poorer cooling. So please provide evidence. Who cares about stock clocks when you're spending $1k :/ We all buy 800-900Mhz 7950s and slap them at 1100. Full comparison was done here: https://109.201.133.65/index.php?topic=187328.0Oh dogie, where do I even start? Your thread only states pre 7990 numbers and what you think theoretical new 7990 #s are? If you are gonna do a unbiased and factual review, get numbers for both sets of hardware: In this case 7990 pre malta and 7990 malta (new). You only have real #s for 7990 pre malta and the other numbers are what you theorize / speculate they will be for a new 7990 malta. This creates a huge lack of credibility in my mind. Its common sense.... I'll state it step by step for you though... 1) The cores are IDENTICAL. Stock clocks are different, but clock for clock identical. 2) Because they are identical, power consumption is going to be VERY close. +/- 15W. 3) Hence, clock for clock these hash identically. 4) The cooling solutions are designed for their TDP. One has a TDP of 375W and one at 525W. What more do you want? Why would I spend $2k on these shit cards just to prove you wrong? I'm not the moron who thinks power usage and cooling is the same. 7990 malta uses 375-389watts at full load for both cores total. The 7990 Malta runs cooler than previous 7990, so again you are wrong. If a card runs cooler then it will clock even the same core higher simply if the design makes it run cooler. This means that hashrate will be higher. DURR? Also the cards sell for $900-$1100, not $2000. So again you are wrong. Finally go look at tomshardware, guru3d and anaandtech for reviews before you go spouting off that you are right. You give bad advice and information to the community when you just start blowing numbers and misleading information out your ass. links below to re-educate yourself. http://www.tomshardware.com/reviews/radeon-hd-7990-review-benchmark,3486.htmlhttp://www.anandtech.com/show/6915/amd-radeon-hd-7990-review-7990-gets-officialhttp://www.guru3d.com/articles_pages/radeon_hd_7990_review,8.htmlCan't be bothered arguing with morons again. Enjoy your life. |
|
|
|
|
Clear to see this guy has a hand wrapped around BFL's .... and is furiously vibrating
|
|
|
|
THIS HEAT SHALL NOT DEFEAT ME. Ordered 4x120mm 100CFM fans + a controller, will remount for the third time in an even more heat friendly configuration. Noticed something really weird and very unsettling though.
I'm drawing 262W for 2 units + 500W PSU 80+ PSU, 230V. So assuming worst case PSU is at 80% and burning 52W.
So raw board + fans of 210W. Each small fan is just over 1W, and assume the large but rubbish ones are 2.5W.
Thats 195W of power draw from 2 boards. I thought we were expecting low 80s for overclocked clocks? Something seems wrong here.
Our experiment data: 70-75W on 1.03-1.05V (no overclock) 83W on 1.10V (no overclock) ~100W on 1.20V (no overclock) ~120W on 1.20V (overclock) My mistake, thanks  |
|
|
|
|
Show Posts
