|
Aren't you using the latest drivers with 2.6 sdk, OP?
So far, here's a list of what OP is doing wrong:
(1) mining at deepbit - any good reason to choose a pool with mining fees?
(2) using Windows for dedicated miners - why?
(3) most probably using the latest driver/sdk combo which is really bad for VLIW4 and VLIW5 cards.
OTOH, here's what OP is doing right:
(1) using cgminer - good choice!
A few suggestions:
make sure you specify intensity. Use intensity 8 or intensity 9 with those cards.
Use the default values for gpu-threads.
If you're planning to stick to 12.1/2.6 driver/sdk, vectors 4 and worksize 128 should help your hash rate a bit.
BTW, what temps and fan speeds you mining at?
|
|
|
|
Garbage in, garbage out. Calling ECS stuff substandard would give them more credit than they deserve... "Fucking POS" was just about right. There are many extender designs, perhaps some other extenders would do the job. Still, not worth your time and trouble with 775 boards costing as little as they do now. Just grab any decent board and never look back. Thanks for the link Deepceleron, I really enjoyed that piece of ancient history. Hey, if you can't build decent boards just make sure the labels are all in order  |
|
|
|
|
Insufficient airflow. Try adding a fan or two to your case.
Which part of the mobo is the heat being measured on? VRM? Northbridge? Southbridge?
In any case, while comparatively high, 63°C is still far from dangerous.
|
|
|
|
I dunno, DaT...
One thing you seem to have overlooked: I'm not even going to blast your 50/50 comsumer-to-business gpu ratio for lack of precise data and lack of motivation to dig up the correct numbers (consumer market is much larger - if less lucrative - than business market), you've come up with 1.78%. Fine, except you're measuring the whole population of miner GPUs (some of them quite old) against a single quarter's worth od AMD's GPU production. If you want to make a valid apples-to-apples comparison, sum up a couple of year's worth of 57xx..79xx cards production and measure that against the meager 20k..50k of units being employed for mining purposes, like that: I'll accept your estimate of 1.125 million high-end GPUs per quarter. Assuming the production levels per each quarter has been constant over last two years, I arrive at 9 million GPUs made over last two years. For simplicity's sake let's replace the 20k..50k range with an average of 37.5k devices, ok? 37500 is 0.4% of that 9 million GPUs but the bad news doesn't stop here. I should take into consideration the number of high-end GPUs originally purchased for applications as trivial as gaming and re-purposed for mining later on. Shall we assume another wildly inaccurate 50-50 ratio yet again? Uh-oh, we seem to have dropped down to a meager 0.2%A drop in the ocean it is. |
|
|
|
I love Bitcoin but I think some people massively overestimate (by 3 or 4 orders of magnitude) the influence the network has at this point.
Thank God for some common sense here. I laughed out loud reading We are important! AMD acknowledges Mining!As if a marketing bullet point costs AMD anything... Thanks for driving the point home with some real-world numbers, DAT. |
|
|
|
|
Wow, 0.08% stales... I think I'm in love.
Thanks for the hard work, Doc.
|
|
|
|
|
Mania, just make sure you are using the latest (ver. 0.5.2) client. The download speed has been significantly improved.
I don't want to imagine how long an old client version would take on a 128kbs link... ugh!
|
|
|
|
Like it or not but you are doing it all the time - power supplies have plenty other things to fail at...
Agreed, that's why I always try to minimize the danger by eliminating unnecessary points of failure. |
|
|
|
That's another thing but 8-pin connectors will always (?) have at least two molex connectors, there are many 6-pin connectors with two molex connectors so it should be enough.
To put is simply, the sata connector was designed as a low-current connector. The issue is, that connector was engineered with ease of insertion/extraction taking precedence over high reliability of the contacts, that's why there are three contacts per each wire. By design, it is assumed that some contacts will connect cleanly and reliably. I'm discussing deploying the full 2x sata -> 2x molex -> pcie-8-pin suite of connectors. If the card actually needs 150 W at the connector, you are overloading each sata connector by 21 watts. Doesn't seem that much but if any of those 3 sata contacts fails to connect cleanly the problem escalates very fast. Some sparking, the connector heats up, possibly to the point of melting, the least of your worries is rig downtime. Same situation for the 1x sata -> 1x molex -> pcie-6-pin combo. 2x sata -> 2x molex -> pcie-6-pin should be reasonably safe as long as you don't exert pressure on those poor sata connectors. Treat my post as Public Service Announcement, I'm trying to make sure that pitfalls of developing a cavalier attitude to GPU wiring are on everyone's radar  As for the built-in protections only cheap PSUs will take anything along with them at their death.
Still, I why put a rig full of perfectly good 5970 cards at the PSU's mercy? All hardware will fail, the protection circuit is no exception. |
|
|
|
And another 6 extra GPUs from sata to molex + molex to pci-e:  ... Would've forgotten.. even if you can it doesn't mean it is good to run 6 GPUs through sata, additional adapters increase resistance so it might not be very stable. It's not the resistance that worries me, you're going far out of spec with the sata connector. A sata power connector has three 12V contacts. THe manufacturer's load limit per contact is 1.5 A. That's just 4.5 Amps for the 12 line resulting in the connector's max load at mediocre 54 Watts. If (heaven forbid) you try the sata -> molex -> pcie 8-pin combo be sure to have a fire extinguisher at the ready. The melting connector can short circuit and kill the PSU. If the built-in protections are inadequate, the PSU can kill the whole rig. Definitely not the prudent path to take. |
|
|
|
|
P4, boards designed with multi-GPU configurations on the radar do use additional connectors for powering the PCIE slots.
The reason for the additional connectors is that the standard 24-pin connector only provides two 12V lines and the low-grade contacts might max out 8 Amps. The best commercially available contacts (Molex's High Current System) can handle up to 13 Amps.
With 8 Amp contacts, the connector can only deliver up to 192 W to the motherboard.
While the HCS contacts should provide a safe 312 W, the mainboard manufacturers can't know what grade contacts the PSU uses.
I wouldn't worry about motherboard traces, I think manufacturers can be trusted to design their boards right.
Keep in mind that PSUs are built on PCBs as well and handle a great deal more current with no ill effects.
In case you're still worried use Gigabyte's "Ultra Durable" boards boasting twice the normal thickness of traces.
|
|
|
|
|
I was checking out an unread topic, believe me I was stunned when I saw those posts...
Being no programmer is one thing, being unable to get even your miner up and running would explain a lot of questions about this pool's history and - more importantly - its future.
It's not as if wasn't obvious you were new to Phoenix, that's the whole point: all one needs to do is read the help screen and specify the parameters...
I believe this to be a valid, if perhaps painful, criticism of your ability to run this pool.
Not by a single word did I imply you are dishonest.
|
|
|
|
I'd like to touch on another issue, an issue seriously affecting the pool op's credibility as far as I am concerned. JackRabiit did a fairly good job pointing all the typos out but I can't get past these posts by a1coinpool in a GUIMiner topic: Does anyone know how to setup the backup pool on this.
What setting are you using on phoenix. Mine are running slow on phoenix.
Along with A1's other posts, this raises serious doubts about his tech credibility. This is ludicrous: another (apparently half-baked) attempt at creating a pool by an individual unable to get his guiminer config right. How am I to trust a guy apparently unable to figure out the easiest of miners?!? Confidence level zero, exit stage left! |
|
|
|
PCIE is the more powerful connector as PSUs don't enforce the 75W limitation, therefore are able to supply insane amounts of current should it be required. Also, being able to get rid of the old-fashioned molex daisy chains is a boon from elegance standpoint. Connector choice should make no difference - I'd trust the mobo manufacturer to know what they are doing. That having been said, I don't like the sata power connector in my Gigabyte ga990fxa-ud7 as intuitively it feels substandard on a high-end board. |
|
|
|
Ok thanks, and relax
Roger that, I may have been a tad harsh. Previous post has been updated. |
|
|
|
If you have linuxcoin running, just launch cgminer with "--verbose -D -T" parameters and it'll output the driver and SDK versions. Wish I could give you a direct answer but it's been ages since I've looked at linuxcoin, say sorry. |
|
|
|
Yes I understand this, but there is no safeguard?? Like Putting the coins back in to the community when they have been unused for lets say 2 years.
Uhmmm... how about putting your dollars back in to the community when they have lain unused in a bank vault for two years? There is absolutely no problem with an amount of coins being lost. Since bitcoins are virtually infinitely divisible, the problem of price increase of the remaining coins is moot. |
|
|
|
They would probably charge in like 2 minutes lol.
I sincerely hope you're not being serious. You are aware that lithium-* batteries require strict current and voltage levels? You can't just throw more voltage/current at them to make the recharge faster - the protection circuit board won't allow that. Circumventing the protection unit is a surefire way for a magnificent fireshow. |
|
|
|
|
All by itself, the clock does not influence the longevity in any way.
However, power consumption and dissipation rise linearly along with clock speed and those two factors are the chip killers when the microscopically small traces inside the chip "burn up" (actually, it's electromigration that kills them - a process where electric current displaces molecules of the conductor it's running through).
Load (heat and current) also leads to degradation of transistors inside the chip to the point where they become unstable at speeds they used to run at without any issues. If your card suddenly starts crashing where it used to run stably for months dropping the clocks is the only solution.
95% fan speed (or even 85%) is something you should worry about. When manufacturers cut corners, fan quality goes first.
Chances are that when purchasing inexpensive 10$ case fans you are getting much higher quality devices than the ones mounted on your 300$ card...
70..79°C is not a bad temperature range for a battery of overclocked 6970s locked in a PC case. If it's an open rig, try increasing the distance between cards or pointing more airflow directly at them.
It's a bloody damned shame that your manufturer seems to have cheaped out on the VRMs - undervolting should drop the temperatures just enough for you to be able to reduce the fan speed to sane levels.
TL;DR:
(1) && (2) No clock - all by itself - is too high provided the card is stable. 79°C isn't exactly low but it'll do.
(3) Yes.
(4) Independent of current and temperature, no.
|
|
|
|
|
Show Posts
