This would be to power 1070s on risers. It’s a 850w power supply so I thought the power is fine I’m just Short on connectors to get the risers and GPUs together. One dual would power the riser/GPU combo for each card. If I’m good o power is it okay? What power supply has enough connectors for 4 cards without some sort of mod?
Most 850w PSU's have 4 PCIe 6+2 connectors. The GTX 1070s have only one 8 pin power connectors. So what seems to be the problem? Why are you running short of connectors? Are you using PCIe power type risers as well? Just use Molex type risers and youll do fine. |
|
|
|
|
Gigawatt is somewhat trustworthy, and that product listed seems to be a generic product I have seen at several Chinese vendors, so i think it exists.
The ETH Hashrate seems in line with p106 but their ETC hashrate is weird, should be same as eth.
Also no way in hell are they getting 800+ on XMR with those GPU's, so the numbers look a little off.
|
|
|
|
I’ll have to resort to that if I can’t find enough dual molex connectors. I was hoping someone has used them before just trying to be as safe as possible.
Dual Molex are not safe. Single Molex are DEFINATELY not safe. Do not use ANYTHING other than PCIe to power GPU's if you do not want to risk burning the PSU's, ive burnt two using dual molex conectors. If you do not have any more PCIe connectors, do not add more cards, this is why people buy 1300w psus even when theyd on tneed the extra power, for the connectors. |
|
|
|
Should i try cheap one which linked first?
What is difference between powered and unpowered riser, which i should take care about,
The cheap ones should be fine, the description looks a bit off, these ribbon risers are supposed to be x16 to x16, thats the whole point of them. Get them from aliexpress, get the ones with the best rating there. I also build and offer complete mining rigs, you can drop me a mail at deadsix@gmail.com. |
|
|
|
in afterburner Clocks for 3 middle cards 100-150mhz slower then end cards....
with any setting... anyclock settings or powerlimit.... all card cool 50-60c
but middle cards run cooler due to less clock?
Nvidia cards auto-overclock based on temperature/power%. The cooler the card is, the higher clocks it will hit regardless of over/under clocks that you may have manually set. The middle 3 cards I am assuming get less airflow and so run hotter, so they clock lower. Nothing to worry about. |
|
|
|
|
Veere dono taan ho ni sakde na. Ja SLI chala k gammein khedd, ja mining kar la cool temps te.
You can try using 16 x 16 ribbon risers for the two 1070's but they are not powered and definitely not recommended
|
|
|
|
Straight to Linux. No issues really getting them recognized. Runs too slow, upgrading to F4620 and 6gb ram soon. Stable enough, got it up to 540 Mh on ETH only. Can't mine any other coins as it crashes, assuming it has to do with CPU but it's stable now.
Running all 19 cards? What cards did you use to get to 540? |
|
|
|
|
Put the 1070's in the physical slots on the motherboard, that way they get 8x -16x pcie lanes and will run in SLI.
Use the remaining slots to connect GTX 1060's with 1x to 16x risers. The 1070's may block some of the slots, so you may need to use an m.2 to pcie adapter plugged into the m.2 slot to get the 3rd 1060 running.
|
|
|
|
I changed other pool but seems not working. it still show 0.00M.hs. here's my log, many thanks 09:42:46:958 4f4 Setting DAG epoch #3...
09:42:48:314 4f4 Setting DAG epoch #3 for GPU0
09:42:48:321 4f4 Create GPU buffer for GPU0 The miner tries to place the DAG file into graphics memory, but never returns a successful result. You should get such message: "Setting DAG epoch #3 for GPU0 done GPU0 DAG creation time - xxxx ms" Both your logs miss this part and it means DAG cannot be created. You said your card is AMD Polaris RX. What is your OS and driver version? My OS is win10 64bit and for AMD driver I use 17.1.1 I have both tried use default GPU clock and OC clock but not working And also tried update to latest too but also not working so back to 17.1.1 thanks Do you have pagefile set to atleast 16gb? |
|
|
|
Maybe I've been unlucky with my cards, but I have started mining last year in July, and I have 4 dead rx 470s on my hands. I got my money back from RMA process so it's not so bad, but it seems like my failure rate has been higher than anyone else. Prior to mining I have never RMA'd a single piece of hardware... And the cards were kept in good condition or so I thought, always below 70c in an open air bench.
Ive had a similar run. Of the AMD cards ive so far bought, ive had a failure rate of about 10% over the first year. Of the Nvidia cards, they haven't failed during usage, but about 5% were DOA. All of these were under warranty so RMA'ed but a pain nonetheless. |
|
|
|
Hi, I just want to start with mining.
Are msi's NVIDIA graphics card good?
or do I need to venture a bit from my local market to get AMD's product.
For Nvidia cards you cant mod the bios, so you are stuck with whatever card you get. Check with your retailer if he will allow you to check the card in a system before buying it - you will need to check the memory type in GPUZ. Samsung memory is the only one thats good for eth mining for Nvidia 1060. For 1070 and above any memory is equally good. In my opinion Samsung memory based Gtx 1060's are the best value for money cards right now. And are risers completely necessary, I'm not finding them in my local market.
And what motherboard would you suggest for upto 4 GPUs?
Yes, there is no other way to connect multiple graphics cards to one system. You can buy risers on amazon or ebay if you cant find then locally. Almost any motherboard out there will support 4 graphics cards, but I would recommend the Biostar TB250 boards. Also as a special case, check out the Biostar GT7 Z170 motherboard, you can run 4 cards on it directly without needing risers. |
|
|
|
This looks mighty inspiring, maybe ill switch back to smOS, undervolting was the one thing keeping me away  Would you care to share your clocks/volts too? Also by offset, you mean a hardcoded bios offset, right? All GPU's are different and such results are not guaranteed. Most of my rigs are running at around 800 watts. By offset yes I mean modifying bios and manually adding and offset (for some fee I can do that for interested parties). Here are my settings for this one particular rig  Nice. So help me understand what the Bios offset is doing in this scenario since the core is already at 850mv? |
|
|
|
I thought I'll share my results. Thats combination of os undervolting + offset. We are back to old results That's 6x Sapphire Pulse RX 570 with hynix memory.   This looks mighty inspiring, maybe ill switch back to smOS, undervolting was the one thing keeping me away Would you care to share your clocks/volts too? Also by offset, you mean a hardcoded bios offset, right? |
|
|
|
A RX 580 properly tuned meaning under-volt mod applied with clocks and voltages properly adjusted will draw 130 watts when mining ETH only and ~150 watts if dual mining. So for ETH only we could do 5 x 130 = 650 Watts, dual mining it would be 5 x 150 or 750 watts, well within reason. Even if he went 6 GPUs (which I only speced 5 for the AMD RX 580, the 6 was for Nvidia 1060's) it would be 780 and 900 watts respectively. While a 6 x RX580 dual mining configuration might be starting to push the PSU, it is still well under the maximum. However, a 6x RX580 just ETH only would be well within reason at 780 watts, or under the 80% load guideline I like to run around.
The key is as you said as i said is : " properly tuned meaning under-volt mod applied with clocks and voltages properly adjusted" So do you think when someone asking a question like : "I dont have the slightest idea how will i build one " , that person know how to properly tune , mod the cards ?  PLease ! dont give bad advice for newbies who dont have "slightest idea how he need to build ", they can burn down they house ! I have SEVERAL rigs of 7 x Rx 570's running at ~115W Dual mining, for a total system draw of about 850W on 1000W platinum PSU's so ... I agree with Za1n, if your'e getting into mining, you need to learn how to optimize, or you stay away. |
|
|
|
[
After I took out your nefarious code Falcon
the shit works good
thanks for the 2%
Care to elaborate what kinda nefarious code you found in there? |
|
|
|
I'm able to get about 760-790 h/s with 10.2 on my cards. Mix of RX 570s/580s 4GB.
On a 580 8GB I have it gets slightly above 800 h/s but thats about the best I can get so far.
If anyone has gotten higher and have any tips let me know.
Thanks
On my RX 580 8Gb(hynix) 1380/2200 I can reach 840-850 h/s -h 1800 -dmem 1 There is some peoples have 1k+ h/s on 480... I don't know how. with ubermix 3.1 straps and samsung mem I didn't rtied ubermix straps on my hynix memory. It's possible that they are will work on hynix memory? No, please do not try to use memory straps from SAMSUNG memory on HYNIX because you risk breaking your card. I'm not 1st day works with computer parts. And I know what is memory timings. Not only on VRAM. And I know that they are can't make damage on card, it can't even damage video memory. Maximum that wrong timings can do - it's BSOD or system hang during initialization of video memory on frequcency were timings are modified. And you must know that almost half of modified straps in the internet for hynix, elpida, micron are maked by modifying memory timings from other memory. F.e. best memory timings for RX 580 8Gb with hynix that you can found in internet taken from 1750 micron memory strap of RX 480 8Gb without any changes. You asked for advice, I gave you advice. I see that you have little to no knowledge, and want to defend it instead of wanting to learn more. You are more than welcome to use any strap on any card, be happy wrecking your cards. |
|
|
|
I'm able to get about 760-790 h/s with 10.2 on my cards. Mix of RX 570s/580s 4GB.
On a 580 8GB I have it gets slightly above 800 h/s but thats about the best I can get so far.
If anyone has gotten higher and have any tips let me know.
Thanks
On my RX 580 8Gb(hynix) 1380/2200 I can reach 840-850 h/s -h 1800 -dmem 1 There is some peoples have 1k+ h/s on 480... I don't know how. with ubermix 3.1 straps and samsung mem I didn't rtied ubermix straps on my hynix memory. It's possible that they are will work on hynix memory? No, please do not try to use memory straps from SAMSUNG memory on HYNIX because you risk breaking your card. |
|
|
|
question
when using this OS with nvidia cards, and the NV build.. we are only limited to ewbf or skunk/eth based mining programs right?
because my 1080TI's on ewbf seem to get half of what they should be hashing at
the claymore zec miner is not working with the NV build... why is this?
... Because Claymores miner is AMD ONLY ... it literally there in the name "Claymore's ZCash AMD GPU Miner", and if that doesn't clear it up for you, it helps to go through the Readme file occasionally. |
|
|
|
|
Great setup. i just hope you're not running 13 Rx cards with just 2 x 1000W PSU's ... Under linux with whatever limited tweaking you can do to voltage, that's a combined load of about ~1800W which is too damn close to the peak.
|
|
|
|
The wolfamdvolt/ohgodavolt? Full automation - it dynamically opens up the I2C interface of each GPU and runs a detect function for every supported VRM, which will find it if it's there. Once found, that VRM's functions are used to get/set shit as requested.
How is that different to solution presented by ethos? I mean does your tool give more control? ethOS bought the tool FROM wolf to implement in their OS ... |
|
|
|
|
Show Posts
