Ethereum GPU mining?

Page 36 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DrMrLordX

Lifer
Apr 27, 2000
21,808
11,164
136
reb0rn, if you don't trust Vitalik and/or Ethereum then don't invest time/money. I'm not saying you're wrong (or right), it's just that everyone knows that something like Ethereum is speculative, so there's risk involved. At least for some of us, it's a hobby, and the profit is uh, secondary? Sort of? Mostly.

Anyway here's some data that hopefully someone can use.

Card baking
Despite running "only" 4 GPUs per system on open-air rig #1, I managed to damage at least one of my cards (MSI 390) by putting it in a position where it ran hotter than it should. It had unexpected instability issues, and for whatever reason, I wasn't checking its operating temperature like a hawk. Apparently one of my cards (an XFX 390) is kicking a lot of head out of its backplate, enough that there was a zone of hot air between it and the adjacent MSI 390. The hot air was going into the fans of the MSI card, and it was running hot and crashing/seizing up a lot (it did not do this in my Win10 daily driver). Also the MSI 390 cooler generally stinks. Running up the fans does not produce much extra cooling.

Aaaaanyway repeated overheats seem to have damaged the VRMs or something on the MSI 390 such that it sucks up maybe 20W more power than it should running the "max" stable settings for the card (-100mv, 1070 MHz GPU, 1250 MHz VRAM). All that extra power consumption is overwhelming the poor cooling solution causing it to run around 80C solo back in the Win10 machine, which is much hotter than it should be running. I've been forced to detune it (currently running 940 MHz GPU, 1250MHz VRAM -50% power limit, -100 mV) to get power consumption down in order to keep temps in an acceptable range. I have yet to deal with the hotspot, though moving the miner itself and/or the cards may solve the problem. For now, my old Pitcairn is in the hotspot, and it is not mining. So -17MH/s until I fix the problem. Bleh.

I also have a reference XFX 290 that has spontaneously developed a similar problem, despite the fact that it does not seem to have operated at high (75C+) temps on a consistent basis. It's at ~900 MHz GPU, 1250 MHz VRAM, -100 mV, -50% power limit. The sad thing is it's still pushing 72% on the blower fan to stay at 72C with those settings. I could probably tune it to run better at around those settings but I do not have the time for that right now.

It's also possible that some of these cards have TIM that is hardening/cracking which is something I saw on a pair of Sapphire r9 290s I have. That might explain some of the diminished cooling performance, though not the obnoxious extra power draw (observed on MSI 390).

Old miner boards
It might be useful to know which boards will and will not work as mining platforms, and to what extent. I'll provide some data on the boards I have used:

Asus A88x-Pro
I have used this board for mining in Win10 only. If you need to use 1x PCIe risers, only the 1x slots on the board will work in Win10 on their own. You might be able to do it in the 16x slots with extra wiring. Powered ribbon cable risers might be a better bet for the 16x slots in Windows. If you try 1x risers in the 16x slot, Windows simply refuses to acknowledge that a video card (or anything else) in "in the slot". Haven't tried Linux on the board for mining. I had to move my Linux drive to make room for a second full-length GPU.

Hooking up a card to the 1x slots via 1x PCIe risers does work, and you can even use that setup to flash the card BIOS. I have used my daily driver as a test-and-tune system to get cards prepped for use in other rigs. The iGPU on my 7700k means I am not dependent on dGPUs for display, which is nice. The downside is that having a GCN dGPU with a program like ethminer means that it's difficult/impossible to mine with the -t option without getting the dGPU involved in the mining. The solution to this is to mine with --opencl-device, and to run one client per card in separate logins.

Gigabyte GA-MA790X-UP4
I've used this board to mine with a Phenom 9750 + 2x4Gb DDR2 under Linux. All PCIe slots work with risers except the "last" 1x PCIe slot. Attempting to boot with a riser/card plugged into that slot causes Linux to hang while loading the desktop. Not exactly sure why. In any case, it's good for four cards, and this old piece of hardware has been (thus far) bulletproof. Also it has good enough voltage controls that you can undervolt your CPU with it. Sort-of a must with a Phenom 9750. The lag during mining startup is epic. Mining with the primary display device (a 2Gb R9 270) can cause the entire system to lock up for a minute or longer at the start. After that it's fine.

MSI k9a2 Platinum
These boards are rare and expensive for the most part. I got mine as a package deal which is the only reason I have one. Anyway, jury's still out on this thing. I haven't (yet) gotten a video signal out of cards plugged into PCIe slots via 1x risers, though granted, I haven't tried much. I can confirm that Linux will detect and configure individual cards connected via risers, though I have yet to get the system to actually mine through a riser. It mined once through a 290 plugged directly into a 16x slot, but now it just sort of loads the DAG and then sits there? I can't figure out what's wrong with it. Might be a problem related to my Sempron 140 which has had the crap beaten out of it and is now a bit on the unstable side, though at 2.2 GHz it should be fine. Nothing appears to be crashing, in any case . . .

NewEgg r9 290 refurb situation
I talked to a NewEgg rep on the phone yesterday and got a slightly more-positive resolution: the rep indicated that within a week or so, they expected more Tri-X refurbs to go into stock for their US site. She intimated that a direct exchange would be possible in which I returned my reference 290 for a Tri-X, though I can not confirm that they'll actually do that. She did say that if I call back and initiate a return over the phone, given the cirumstances, they'll replace the card and give me a paid shipping label (no restock fees either). So I'm going to wait a short while and look for the Tri-X refurbs to become available to see what happens.

(just an FYI, but the reference 290s they were selling marked down to $179 sold out on the US site)

Seeing as how the 'Egg is selling the Tri-X refurbs on their .ca site for the equivalent of ~$190, I see no reason why I should not be able to return my item 1 for 1 considering I paid $199 for it.
 
Last edited:

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
ZERO facts for you claims ethereum is future, why??
I would never invest in something which is 90% premined, at same time there is ZERO distribution when someone control almost whole primine, Vitalik is smart he managed to hide this fact as not important

here is some info about what is ethereum:
http://www.vox.com/2016/5/24/11718436/ethereum-the-dao-bitcoin

Don`t get me wrong I still don`t see any value to 1 billion cap!, also as text say DAO is controled buy ppl who has premine, don`t forget that

Here's what the cofounder of Coinbase thinks about Ethereum.

https://medium.com/the-coinbase-blo...t-of-digital-currency-5300298f6c75#.h70a2q6s6

What is Coinbase?

https://en.wikipedia.org/wiki/Coinbase
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
reb0rn, if you don't trust Vitalik and/or Ethereum then don't invest time/money. I'm not saying you're wrong (or right), it's just that everyone knows that something like Ethereum is speculative, so there's risk involved. At least for some of us, it's a hobby, and the profit is uh, secondary? Sort of? Mostly.

Anyway here's some data that hopefully someone can use.

Card baking
Despite running "only" 4 GPUs per system on open-air rig #1, I managed to damage at least one of my cards (MSI 390) by putting it in a position where it ran hotter than it should. It had unexpected instability issues, and for whatever reason, I wasn't checking its operating temperature like a hawk. Apparently one of my cards (an XFX 390) is kicking a lot of head out of its backplate, enough that there was a zone of hot air between it and the adjacent MSI 390. The hot air was going into the fans of the MSI card, and it was running hot and crashing/seizing up a lot (it did not do this in my Win10 daily driver). Also the MSI 390 cooler generally stinks. Running up the fans does not produce much extra cooling.

Aaaaanyway repeated overheats seem to have damaged the VRMs or something on the MSI 390 such that it sucks up maybe 20W more power than it should running the "max" stable settings for the card (-100mv, 1070 MHz GPU, 1250 MHz VRAM). All that extra power consumption is overwhelming the poor cooling solution causing it to run around 80C solo back in the Win10 machine, which is much hotter than it should be running. I've been forced to detune it (currently running 940 MHz GPU, 1250MHz VRAM -50% power limit, -100 mV) to get power consumption down in order to keep temps in an acceptable range. I have yet to deal with the hotspot, though moving the miner itself and/or the cards may solve the problem. For now, my old Pitcairn is in the hotspot, and it is not mining. So -17MH/s until I fix the problem. Bleh.

I also have a reference XFX 290 that has spontaneously developed a similar problem, despite the fact that it does not seem to have operated at high (75C+) temps on a consistent basis. It's at ~900 MHz GPU, 1250 MHz VRAM, -100 mV, -50% power limit. The sad thing is it's still pushing 72% on the blower fan to stay at 72C with those settings. I could probably tune it to run better at around those settings but I do not have the time for that right now.

It's also possible that some of these cards have TIM that is hardening/cracking which is something I saw on a pair of Sapphire r9 290s I have. That might explain some of the diminished cooling performance, though not the obnoxious extra power draw (observed on MSI 390).

Old miner boards
It might be useful to know which boards will and will not work as mining platforms, and to what extent. I'll provide some data on the boards I have used:

Asus A88x-Pro
I have used this board for mining in Win10 only. If you need to use 1x PCIe risers, only the 1x slots on the board will work in Win10 on their own. You might be able to do it in the 16x slots with extra wiring. Powered ribbon cable risers might be a better bet for the 16x slots in Windows. If you try 1x risers in the 16x slot, Windows simply refuses to acknowledge that a video card (or anything else) in "in the slot". Haven't tried Linux on the board for mining. I had to move my Linux drive to make room for a second full-length GPU.

Hooking up a card to the 1x slots via 1x PCIe risers does work, and you can even use that setup to flash the card BIOS. I have used my daily driver as a test-and-tune system to get cards prepped for use in other rigs. The iGPU on my 7700k means I am not dependent on dGPUs for display, which is nice. The downside is that having a GCN dGPU with a program like ethminer means that it's difficult/impossible to mine with the -t option without getting the dGPU involved in the mining. The solution to this is to mine with --opencl-device, and to run one client per card in separate logins.

Gigabyte GA-MA790X-UP4
I've used this board to mine with a Phenom 9750 + 2x4Gb DDR2 under Linux. All PCIe slots work with risers except the "last" 1x PCIe slot. Attempting to boot with a riser/card plugged into that slot causes Linux to hang while loading the desktop. Not exactly sure why. In any case, it's good for four cards, and this old piece of hardware has been (thus far) bulletproof. Also it has good enough voltage controls that you can undervolt your CPU with it. Sort-of a must with a Phenom 9750. The lag during mining startup is epic. Mining with the primary display device (a 2Gb R9 270) can cause the entire system to lock up for a minute or longer at the start. After that it's fine.

MSI k9a2 Platinum
These boards are rare and expensive for the most part. I got mine as a package deal which is the only reason I have one. Anyway, jury's still out on this thing. I haven't (yet) gotten a video signal out of cards plugged into PCIe slots via 1x risers, though granted, I haven't tried much. I can confirm that Linux will detect and configure individual cards connected via risers, though I have yet to get the system to actually mine through a riser. It mined once through a 290 plugged directly into a 16x slot, but now it just sort of loads the DAG and then sits there? I can't figure out what's wrong with it. Might be a problem related to my Sempron 140 which has had the crap beaten out of it and is now a bit on the unstable side, though at 2.2 GHz it should be fine. Nothing appears to be crashing, in any case . . .

NewEgg r9 290 refurb situation
I talked to a NewEgg rep on the phone yesterday and got a slightly more-positive resolution: the rep indicated that within a week or so, they expected more Tri-X refurbs to go into stock for their US site. She intimated that a direct exchange would be possible in which I returned my reference 290 for a Tri-X, though I can not confirm that they'll actually do that. She did say that if I call back and initiate a return over the phone, given the cirumstances, they'll replace the card and give me a paid shipping label (no restock fees either). So I'm going to wait a short while and look for the Tri-X refurbs to become available to see what happens.

(just an FYI, but the reference 290s they were selling marked down to $179 sold out on the US site)

Seeing as how the 'Egg is selling the Tri-X refurbs on their .ca site for the equivalent of ~$190, I see no reason why I should not be able to return my item 1 for 1 considering I paid $199 for it.

That's some bad luck with a few of your cards. Of the 7 MSI 390s I have, none of them have degraded over time even running at higher temps. Maybe I've just been really lucky so far (well not that lucky, I just got a $110.00 ticket for an expired license plate sticker two days after my bday lol)

Since you're talking about motherboards I just bought one these for a new mining rig I'm putting together.

http://www.newegg.ca/Product/Product.aspx?Item=N82E16813157564

Added a cheap pentium to go with it.

http://www.amazon.com/Intel-Celeron...psc=1&redirect=true&ref_=od_aui_detailpages00

Hard to beat this combo for a base mining platform. I was going to go the Skylake route but it would cost at least double for a similar setup.

Just waiting for Polaris 10 / Geforce 1070 to drive the cost of 390's down another $50-75 CAD and I'll put together a 5x 390 rig for another 150Mh. An EVGA 1300 G2 / 1200 P2 should be enough for five 390's or four 390's + 2 Fury Nano's.
 
Last edited:

Dice144

Senior member
Oct 22, 2010
654
1
81
Finally got my miner going. My 290x water cooled is only doing 24.9 Mhash tho. Not familiar with overclocking/underclocking in Ubuntu yet.

First made the mistake of install Ubuntu 16 and AMD drivers would not work.
 

sirmo

Golden Member
Oct 10, 2011
1,014
391
136
I'll give mining a try, but once summer is over. In winter I need the extra heat in my office
 

DrMrLordX

Lifer
Apr 27, 2000
21,808
11,164
136
That's some bad luck with a few of your cards. Of the 7 MSI 390s I have, none of them have degraded over time even running at higher temps. Maybe I've just been really lucky so far (well not that lucky, I just got a $110.00 ticket for an expired license plate sticker two days after my bday lol)

Yeah I dunno what the deal is. I did get it used so it may be that some previous abuse is finally taking its toll, who knows?

Since you're talking about motherboards I just bought one these for a new mining rig I'm putting together.

http://www.newegg.ca/Product/Product.aspx?Item=N82E16813157564

Please let us know how well that works. Different boards produce different results.

Hard to beat this combo for a base mining platform. I was going to go the Skylake route but it would cost at least double for a similar setup.

I have to say that I'm a little disappointed that a Haswell Pentium was so expensive. The G4400 isn't much more than that.

yeah im taking my 1st shot at a ubuntu rig this weekend.

14.04 lts is the goto?

14.04 is commonly used, though you can use 15.10 if you want something a little more up-to-date. I'm using 15.10.

Finally got my miner going. My 290x water cooled is only doing 24.9 Mhash tho. Not familiar with overclocking/underclocking in Ubuntu yet.

First made the mistake of install Ubuntu 16 and AMD drivers would not work.

The first trick is to make sure you're using the right miner settings with your card. Try using --cl-local-work 256 --cl-global-work 16384 as switches when initiating the miner. Your 290x should pull ~32 MH/s if set up correctly.

As far as overclocking under Ubuntu goes, here's what I normally do (not necessarily for the faint of heart):

1). Flash BIOS on card to set GPU to 900 MHz, VRAM to 1250 MHz, VDDCI to 900 mV, and VDDC to -100 mV under whatever GPUz was reading while the card was mining on a Windows test machine (it can take some trial and error to get the VDDC just right). Use HawaiiBIOSReader to edit the ROM after saving the original, and ATIWinFlash to perform the flash operation following normal "safe" procedures for flashing a video card. This procedure assumes you have a Windows machine somewhere that you can use a test/tune box. Keep in mind that it's important to flash in a low base clockspeed for your card since some Hawaii cards will not allow you to decrease the clockspeed via software.

2). use the aticonfig utility in Linux to "dial in" a stable clockspeed for the card during mining. First you need to enable OverDrive via aticonfig --od-enable. After that, assuming the 290X is the only card in the system, it's aticonfig --adapter=0 --od-setclocks=clockspeed, 1250 where clockspeed is the new clockspeed (1000, 1070, whatever). 1250 is the RAM speed that is normally recommended for mining on Hawaii cards.

3). use the aticonfig utility to "commit" the clocks, which puts them into use. Type in aticonfig --adapter=0 --od-commitclocks .

You should probably read up on the use of aticonfig, though most of what you need to know is accessible via aticonfig --help . Bear in mind that you can use aticonfig to set a static fan speed if necessary, or to monitor fan speeds. Those commands are not well-documented. Here is some info on that.
 
Last edited:

VeryCharBroiled

Senior member
Oct 6, 2008
387
25
101
Yeah I dunno what the deal is. I did get it used so it may
14.04 is commonly used, though you can use 15.10 if you want something a little more up-to-date. I'm using 15.10.

thanks!

over on another forum someone mentioned GenEthOs as a more or less turnkey OS. 14.04 based. probably start with that. then mess around with the official 15.10.

right now I just want to build it and run it, too many real life projects going on ATM.
 

Dice144

Senior member
Oct 22, 2010
654
1
81
I am familiar with the bios editing tools. Will try that once in have a backup card to game on. Thank you for the tips
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
I have to say that I'm a little disappointed that a Haswell Pentium was so expensive. The G4400 isn't much more than that.

Yeah the G4400 is a much better value for $20.00 more however try finding a 1151 motherboard with 6 PCIe slots for $65 CAD AR. And we know the CPU doesn't really matter for mining (Sempron 140 is enough for 4x cards) so the G1820 will be plenty and has a 53W tdp before being undervolted which will be great for power consumption. Plus you can pick up a 4GB stick of DDR3 for peanuts these days.

For dedicated mining rigs the 1150 or FX AMD boards with six PCIe slots still make the most sense from a pure cost perspective. The FX boards are more money but come with a lot more features and are better built but you're "limited" to an 83XX CPU after the party is over. With a budget 1150 board and CPU this cheap you can always toss in a quad core or Xeon 1231v3 if you want to make a decent gaming box and reuse the G1820 for a NAS / HTPC. Or just sell for it all $10-$20 bucks less than what you paid for it. No big deal when costs are this low to begin with.

1151 boards are especially attractive when you get up to the Z170 chipsets, boards like the AsRock Z170 Pro4s (5 PCIe slots, and 1 M.2 which can be used) can be found for $99 CAD A/R as you can still use the "SkyOC" feature to ramp up the speed of the locked CPU's, just need to stick with BIOS 3.0 or lower. If going this route ideally you want at least a Core i3-6100 to make for a decent mid-range gaming PC. But add in a stick of DDR4 and it all starts to add up where you could put that $70-$100 towards another GPU or better power supply etc.

However depending on how long you plan on dedicating the box to mining it's kind of a waste to spend more money unless you plan on re-purposing the hardware. That's where the Overclockcable Z170/Core i3-6100 combo can make sense.

Anyway I'll let you know how well the AsRock 1150 works, I found a post or two already praising the board for dedicated mining rigs so I don't think there will be many surprises. I had to order my CPU from Amazon.com instead of .ca (use up gift cards!) so it'll be a few weeks before the CPU arrives. In the meantime I'll toss in another CPU to get started on an open bench.

I really hope Fury Nano's drop price soon as I'm starting to plan on unloading my 390's and Fury X's to cut down on power/heat (at least for the cards that have paid for themselves which is most of them!). I'm actually more worried about the heat over the extra electricity costs as then you have to offset the heat with air conditioning which adds even more to your power bill
 
Last edited:

EightySix Four

Diamond Member
Jul 17, 2004
5,121
49
91
Had a Gigabyte R939G1 die on me last night while mining. Had it running in my rig 24/7 for a few weeks now with no overclock (too hot inside an actual case) at about 75* C. Over the last few days I'd noticed my MH/s becoming very unstable coming out of the rig so I thought something might be up. Last night Claymore froze and when it restarted there were only 2 cards available instead of three.

ATIFlash detects a Generic Hawaii device and the AMD drivers list it as "disabled". Fortunately Gigabyte is RMAing it. Would have liked the extra GH right now though.
 

poofyhairguy

Lifer
Nov 20, 2005
14,612
318
126
It'll depend on power costs ultimately. If you currently get cheap power forget about Polaris and stick with existing cards.

That makes sense. You kinda made me rethink things.

I put both the 290s in my gaming rig, but they are WAY too freaking loud to stay there. So instead I put the 390x and the 280x in my gaming rig (I assume during gaming the 280x will stay idle?) and I am putting the 290s in junker desktops just for the purpose of mining. The only issue I have is in Linux the ref 290s throttle like crazy- I can't get them to go over 915mhz on the clocks. I need to use the link above to ramp up the fan speed to 60% and then I will be set.
 

DrMrLordX

Lifer
Apr 27, 2000
21,808
11,164
136
Madpacket, I find myself looking for 4 cards per rig instead of 6 since that seems to open up more (and cheaper) board options, cheaper PSU options, and fewer thermal problems from having cards so close to one another.

EightySix Four, that sucks! I'm sorry your card went down like that. I've got a few cards that sit around 75C so . . . yeah, have my doubts.

poofyhairguy, I highly recommend that you stop running those 290s in any situation where they will throttle. If it's just the fan problem then yes, just tweak that, and you'll be fine, though running the fan balls-out at 60-70% on those reference cards is so damn noisy. There are other ways to deal with the problem, such as tweaking TDP Max and Power Limit in HawaiiBIOSReader. You can also mandate a lower Max ASIC Temp for what it's worth. By default, throttling starts at around 94C which is a temp that will murder a mining card quickly.
 

reb0rn

Senior member
Dec 31, 2009
222
58
101
The problem is more VRM heating then GPU ... but be happy ethereum is so easy and low optimized that GPU usage is very bad, mining scrypt, xpm, etc would be at least 50% more power usage
claymore eth+dcr is bit better but still not full usage
XPM burned my seasonic power pci-e cables on 290x and that with undervolt and no OC
 
Last edited:

poofyhairguy

Lifer
Nov 20, 2005
14,612
318
126
poofyhairguy, I highly recommend that you stop running those 290s in any situation where they will throttle. If it's just the fan problem then yes, just tweak that, and you'll be fine, though running the fan balls-out at 60-70% on those reference cards is so damn noisy. There are other ways to deal with the problem, such as tweaking TDP Max and Power Limit in HawaiiBIOSReader. You can also mandate a lower Max ASIC Temp for what it's worth. By default, throttling starts at around 94C which is a temp that will murder a mining card quickly.

Thank you for the advice. I decided I am going to keep one 290 (the better one) just for the fun of getting the most I can out of it mining-wise. I like a challenge, but that will reduce my buyers remorse if we see something crazy like a clearance $200 390 in a month. Plus one of them just completely sucks and is unstable mining at stock clocks. Bye bye to it.

So I guess the best way forward is work out what is the optimal mining config (clock speed, voltage, fan speed, ram speed, etc.) and use that app to make me a custom BIOS for Linux use? Is that better than using a software overclock on boot?

Also I have read on the Eth forum that some people have a lot of luck flashing the 390 bios onto a reference 290 (like 10% better hash but runs cooler), do you think I could hack up a 390 bios the same way or should I just pick a poison?
 

DrMrLordX

Lifer
Apr 27, 2000
21,808
11,164
136
I have heard of people flashing 390 BIOSes onto 290s, though I haven't tried it myself. Not sure if that offers any real benefits versus just editing the stock BIOS or using one of The_Stilt's custom 290/290X BIOSes.

Near as I can tell, the easiest choice I make with any card is RAM speed and VDDCI. I set VDDCI to 900mV (down from 1000mV) and VRAM to 1250 MHz on every Hawaii card since that seems to work on all of them. You'll drop power draw by a bit without losing any performance or GPU clockspeed.

Otherwise, I would recommend using both BIOS flashing and software overclocking to give yourself versatility. You will have no voltage control under Linux, so pick a max voltage that you think will be "enough" (typically I use -100mV) and either leave the stock clocks or set the clocks really low to give you the flexibility to set clockspeeds wherever you want them via aticonfig. Bear in mind that some Hawaii cards experience a voltage jump beyond certain clockspeeds, typically around 990-1000 MHz but not always. Usually the post-jump voltage will be closer to what you set in your BIOS edit than what you will see at clockspeeds below the jump. The exception to that rule has been the Sapphire Vapor-X but that's a weird case.

And of course you will see power usage increase with higher clockspeeds, even if voltage stays static.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Had a Gigabyte R939G1 die on me last night while mining. Had it running in my rig 24/7 for a few weeks now with no overclock (too hot inside an actual case) at about 75* C. Over the last few days I'd noticed my MH/s becoming very unstable coming out of the rig so I thought something might be up. Last night Claymore froze and when it restarted there were only 2 cards available instead of three.

ATIFlash detects a Generic Hawaii device and the AMD drivers list it as "disabled". Fortunately Gigabyte is RMAing it. Would have liked the extra GH right now though.

Sounds like you just had bad luck. Sorry for your loss!

I've been running some of my cards at 85C core for months 24/7 for a few months now (2 MSI and one Asus) without issues. The VRM's run in the 70's though. I agree with reb0rn, VRM cooling is definately more important than core cooling for card longevity. Ideally you want to run them as cool as possible so that's why I recommend something like a milk crate miner especially if you're running 4 or less cards. The cards in my milk crate miners run really cool at around 65C with just a smaller radial table fan blowing air across them. Sure it's a little ghetto looking but very practical and easy to setup.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Madpacket, I find myself looking for 4 cards per rig instead of 6 since that seems to open up more (and cheaper) board options, cheaper PSU options, and fewer thermal problems from having cards so close to one another.

Yes currently I'm only running 4 390's per rig for similar reasons but as I transition to Fury Nano's that will change to 6 Nano's per 1KW. I've done extensive testing with my EVGA 1KW Platinum and can confirm 4 390 cards run fine even with Claymore. From the wall I'm pulling up to 1040W which is getting close to the max but they still perform admirably. These high end power supplies can handle 90% loads and still hit around 90% efficiency so it's never good to cheap out on power supplies. I guess that's why they come with 10 year warranties, overkill

I also have a 1300 G2 powering my 4xMSI setup with plenty of spare watts and a 1200 P2 arrived yesterday for the new rig to give me the necessary breathing room for the transition (so I can run 4 390's + 1 - 2 Nano's) per rig. I also ordered one more 1KW Platinum EVGA as they had them on for a good deal at NewEgg It'll be a spare / backup until I can transition over the rest. I would like to end up with 6 nano's per rig x 4 rigs for maximum efficiency but it depends on how much of a price drop happens with them. 390's have been easy to see a ROI but Nano's cost quite a bit more money and the energy savings may not be worth it. Need to spreadsheet this all out first before dropping a shedload on Nano's. I'm waiting on Polaris results as well. They may be energy efficient but the bus speed will likely hold back performance where Nano's will still make the most sense for performance per watt.
 
Last edited:

MrTeal

Diamond Member
Dec 7, 2003
3,586
1,746
136
Sounds like you just had bad luck. Sorry for your loss!

I've been running some of my cards at 85C core for months 24/7 for a few months now (2 MSI and one Asus) without issues. The VRM's run in the 70's though. I agree with reb0rn, VRM cooling is definately more important than core cooling for card longevity. Ideally you want to run them as cool as possible so that's why I recommend something like a milk crate miner especially if you're running 4 or less cards. The cards in my milk crate miners run really cool at around 65C with just a smaller radial table fan blowing air across them. Sure it's a little ghetto looking but very practical and easy to setup.

Yeah, proper airflow is extremely important, and not really even that tough if it's planned out right. At one point back when I was running Bitcoin mining gear I was pulling 80A (@240V) in my mining closet, with all the equipment sitting on a 48"Wx24"Dx72"H wire shelving unit. Airflow was just ducted so that all the air coming into the closet was fresh outside air, which prevented mixing and made it manageable.

It's unfortunate I disassembled it last fall.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Also I have read on the Eth forum that some people have a lot of luck flashing the 390 bios onto a reference 290 (like 10% better hash but runs cooler), do you think I could hack up a 390 bios the same way or should I just pick a poison?

I find this a little hard to believe for ethereum mining. My Gigabyte 290 at the same clock speeds as my 390 cards perform within 0.5Mh of each other.

I found the best balance (performance of 30Mh at the lowest wattage) if you have adequate cooling to be around 1060 - 1100 on 290's at -100mv but the undervolting really depends on the quality of the card. I transitioned most of my builds over to Claymore and had to back off my clocks a bit.

For example my MSI rig ran at -100mv @ 1111Mhz, -35 power tune, -40 VDDC or whatever the memory bus voltage is called when running the standard ethminer. With Claymore I had to back off to 1060Mhz and set one of the cards to -90Mv before I reached stability. But even with the lower clocks seem to have more "stable" results which is nice but still not sure if Claymore is worth the added wattage.
 

EightySix Four

Diamond Member
Jul 17, 2004
5,121
49
91
Sounds like you just had bad luck. Sorry for your loss!

I've been running some of my cards at 85C core for months 24/7 for a few months now (2 MSI and one Asus) without issues. The VRM's run in the 70's though. I agree with reb0rn, VRM cooling is definately more important than core cooling for card longevity. Ideally you want to run them as cool as possible so that's why I recommend something like a milk crate miner especially if you're running 4 or less cards. The cards in my milk crate miners run really cool at around 65C with just a smaller radial table fan blowing air across them. Sure it's a little ghetto looking but very practical and easy to setup.

Yeah, it's a trade-off, can't run milk crate miners at the data center where I get "free" power. For now the 3 390s and and one 380 will have to do. The other two are running fine and we'll see what Gigabyte says when I RMA it.
 
Last edited:

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
At one point back when I was running Bitcoin mining gear I was pulling 80A (@240V) in my mining closet, with all the equipment sitting on a 48"Wx24"Dx72"H wire shelving unit. Airflow was just ducted so that all the air coming into the closet was fresh outside air, which prevented mixing and made it manageable.

Awesome, I've been pondering over a similar setup. I know an electrician who I need to bug first to up the amps but the shelving / ducting is a great idea and something I can do on my own.

Heh, too bad we're all spread out geographically. Together we could pool our miners and rent some commercial space and do this properly
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Yeah, it's a trade-off, can't run milk crate miners at the data center where I get "free" power. Oh well, The other two are running fine and we'll see what Gigabyte says when I RMA it.

Nice, you can't beat free power. My last power bill was almost $380 CAD (we pay monthly) and I just turned on my AC for the first time this year last night! Next month is going to be closer to $500 if it doesn't cool down.
 

MrTeal

Diamond Member
Dec 7, 2003
3,586
1,746
136
Awesome, I've been pondering over a similar setup. I know an electrician who I need to bug first to up the amps but the shelving / ducting is a great idea and something I can do on my own.

Heh, too bad we're all spread out geographically. Together we could pool our miners and rent some commercial space and do this properly

It really wasn't too much, it took me a good evening to build it up. The WAF actually wasn't too bad either since everything was enclosed, though once it was sealed up there wasn't any light in from the window anymore.


Unfortunately residential rates in SK are such that I don't want to invest too much into it again.

Let me know if you decide to go big. I have a bunch of PM10i-30a intelligent PDUs and a shelf full of 1300W server PSUs and a handful of 2400W platinum models with PCIe connectors already soldered to them.
 

poofyhairguy

Lifer
Nov 20, 2005
14,612
318
126
I have heard of people flashing 390 BIOSes onto 290s, though I haven't tried it myself. Not sure if that offers any real benefits versus just editing the stock BIOS or using one of The_Stilt's custom 290/290X BIOSes.

Near as I can tell, the easiest choice I make with any card is RAM speed and VDDCI. I set VDDCI to 900mV (down from 1000mV) and VRAM to 1250 MHz on every Hawaii card since that seems to work on all of them. You'll drop power draw by a bit without losing any performance or GPU clockspeed.

Otherwise, I would recommend using both BIOS flashing and software overclocking to give yourself versatility. You will have no voltage control under Linux, so pick a max voltage that you think will be "enough" (typically I use -100mV) and either leave the stock clocks or set the clocks really low to give you the flexibility to set clockspeeds wherever you want them via aticonfig. Bear in mind that some Hawaii cards experience a voltage jump beyond certain clockspeeds, typically around 990-1000 MHz but not always. Usually the post-jump voltage will be closer to what you set in your BIOS edit than what you will see at clockspeeds below the jump. The exception to that rule has been the Sapphire Vapor-X but that's a weird case.

And of course you will see power usage increase with higher clockspeeds, even if voltage stays static.

Thank you for the advice. I might try the Stilt bios.

One problem I am having so far is even at 74% fan speed (which is "I can hear it through a wall" loud) the card throttles down to 950 clocks after about an hour. Even at that speed its way faster than my 280x (about 28 vs 20), so I might just try to find a way to get to 1000 on a lower speed (via a new bios or undervolting or even just replacing the TIM on the GPU) and call it a day.

I find this a little hard to believe for ethereum mining. My Gigabyte 290 at the same clock speeds as my 390 cards perform within 0.5Mh of each other.

Fair enough, I won't run down that rabbit hole then.

This reference card is for sure a challenge, but I like challenges at least for a while.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |