Okay thanks. Tried Wattman for a bit and was getting 35MHs at 900/1000 clocks, but temps were getting out of control. Need to get the card into the water loop before mining with it.
Okay thanks. Tried Wattman for a bit and was getting 35MHs at 900/1000 clocks, but temps were getting out of control. Need to get the card into the water loop before mining with it.Just Wattman so far. It's still a bit limited and buggy, but I don't think any third party tools like afterburner are working with Vega yet.
The hashrate is better with the block chain drivers than it is with the normal ones, but yeah. The lowest Wattman seems to allow the voltages to go is 950 mV which should work fine for those clocks, so that might help a little. On my liquid cooled Vega 64 I run 1100 MHz core and memory, both at 950 mV when mining. Although the core frequently jumps to a higher clock speed anyway.Okay thanks. Tried Wattman for a bit and was getting 35MHs at 900/1000 clocks, but temps were getting out of control. Need to get the card into the water loop before mining with it.
What hashrate do you get with the normal driver at those clocks?The hashrate is better with the block chain drivers than it is with the normal ones, but yeah. The lowest Wattman seems to allow the voltages to go is 950 mV which should work fine for those clocks, so that might help a little. On my liquid cooled Vega 64 I run 1100 MHz core and memory, both at 950 mV when mining. Although the core frequently jumps to a higher clock speed anyway.
Only reason to mine now, is if you think the coins you get will go up later. And at that point you might be better off just buying them directly.
Mid to upper 30's.What hashrate do you get with the normal driver at those clocks?
Awesome, glad I could help! I'll have to play around with lowering the core clock and power limit further, but I agree there's just something incredibly satisfying about seeing one card push 44 MH.Feld,
So I changed to mining drivers and now hashing away at 44MH/s. Thanks again for the settings, worked perfectly. Power consumption is a little higher than I like (adds about 50 to 60W over mining at 38MH) but there's just something awesome seeing one card push out 44MH/s. Sold off one of my older 28nm rigs and used the proceeds to order two Vega 56's last night. These VEGA cards are still silly expensive but I figure the VEGA 56's make the most sense since you can flash and tweak them (I'm doubtful aftermarket 56's will allow for the same type of memory overclocking via voltage tweaks making them overall slower - for mining anyway).
New metric I'm going after is Megahashes Per Cubic Foot (MHPCF)
Saw a new YouTube video where someone set the power limit of their cards to negative 50%, memory at 1100Mhz and core at 900 with dual VEGA 64's. From the wall the entire system only pulled 350W with each card mining at 38MH/S. So roughly 140 - 150W per 38Mh seems to be right up there with Geforce 1070 cards in performance per watt.
Most likely, gravity assisted due to softening. You could probably rig up some kind of mounting system internally to prop up the plastic parts (or at least lift the PCB, though that still doesn't solve the softening issue). For now I'd at least put a cardboard spacer in there and let the bottom one prop up the top one.Well, I had an "issue" with my primary rig, that mines with two Polaris GPUs, one XFX RX 470 4GB, one XFX RX 570 4GB.
They both have backplates, the top card has "Hard Swap" fans on it.
Well, for whatever reason, gravity, plastic shroud on top card softening due to constant 89C temps, whatever, the fans were actually dropping down and scraping the bottom card's backplate.
There's some nice circular and semi-circular gouges in the bottom card's backplate. Kind of scary that this could happen. I mean, they're both XFX cards with backplates. There was no disclaimer that they couldn't be used next to each other. I had their rear expansion brackets screwed down evenly.
Motherboard is an ASRock AB350M Pro4, which has the primary PCI-E 3.0 x16 slot in the second position, and the secondary PCI-E x16 (electrically x4) in the fourth position. (It's a micro-ATX board.)
Case is an Azza ATX case.
I noticed that the fans on the XFX with the "Hard Swap" fans, has quite a bit of "give" to them, you can sort of move them back and forth at angles, somewhat. They seem kind of loose.
Could this be due to the plastic softening up at those temps (89C constantly)? Or why would they use such plastic, knowing that their cards could get that hot?
This really needs a YouTube video or some pictures, that might be coming up.
And also, don't let the card run at 89C. Undervolt/underclock to at least keep that upper card cooler. Preferably below 80.Most likely, gravity assisted due to softening. You could probably rig up some kind of mounting system internally to prop up the plastic parts (or at least lift the PCB, though that still doesn't solve the softening issue). For now I'd at least put a cardboard spacer in there and let the bottom one prop up the top one.
That's WITH Wattman configuring freq. -50% and power limit -30%.And also, don't let the card run at 89C. Undervolt/underclock to at least keep that upper card cooler. Preferably below 80.
Yes, turn off auto for the voltage fields right below the clock speeds for the card and you can put in the voltage manually. I only have one 470 and it's an 8GB, but I run it at 960 MHz, 820 mV core. If you're running at -50% clock speed (600-700 MHz?) You can probably set your voltage a good bit lower than that.That's WITH Wattman configuring freq. -50% and power limit -30%.
I don't know how to change the voltage, can you do that with Wattman?
It really sounds like you're not getting enough airflow to the top card. Most motherboards, especially those smaller than ATX, aren't designed to run more than one graphics card with an open air cooler. It's generally assumed that blower style coolers or liquid cooling will be used. If undervolting doesn't get your temps down to a reasonable level, you may want to consider a way to run one of the cards off of a powered USB riser so that both cards can get adequate airflow to their fans.
I do have two cards in my main desktop too. The bottom card is a 480 that doesn't go above 70C, but the primary card above it is liquid cooled. Currently a Vega 64, which replaced a Fury X. Prior to that I had a pair of aftermarket air cooled 390s in there, and I had to keep the primary card significantly underclocked to keep it from getting too hot.It surprised me how easy it is to disrupt airflow to a video card. I installed an ASUS GTX 1080 Ti ROG Strix into a Fractal Design Define R5, and it was doing about 69C at around 40-50% fan speed. That's not too shabby. I took an EVGA 1080, removed its air cooler (ACX 3.0) and put on the Hybrid cooler. I placed that below the 1080 Ti, and I figured it would be fine since AIO card wouldn't be dumping its heat onto the air-cooled card. Yeah... the ASUS card is so thick that it ended up very close to the AIO card, and the thermals were significantly worse. I was getting pretty close to 80C with a higher fan speed.
I ended up completely changing that build. I laid the case on its side, removed the side panel, installed a PCI-E extension cable, put a MDF board on the side, put a laptop cooler on that, put the card on the laptop cooler, and run it that way. (The card heats up too much if it's just on the MDF since wood acts as an insulator.)
I do have a dual card system that works fine. It's got a reference GTX 1080 Ti modified with a Hybrid cooler and a GTX 1060 6GB (single fan). I did slow the 1060 down to get its temperatures a little better. It was running around 75C, and I dropped it down so it runs at about 62-65C now.
I'm announcing that 100% of my @omise_go + @kybernetwork *advisor shares* will be either (i) donated to charity (AMF, GiveD, SENS etc) or...
(ii) used to privately fund Ethereum second-layer infrastructure (state channels, multisig wallets etc), or some combination of the two
General ETH question here (not necessarily related to mining):
Recently, the team behind the open-source Raiden development announced an ICO - RDN - to further fund Raiden development. To incentivize people to invest in RDN, they announced that there would be a fee structure set up requiring the expenditure of RDN tokens in order to make use of the Raiden feature-set.
Buterin has come out against the ICO, offering up $35 million of his own ERC20 tokens (mostly OMG and Kyber) to try to convince the Raiden team to cancel the ICO; in the event that they refuse, he'll donate to charity instead. Buterin is out $35 million either way.
How do you feel about this development? Personally I think Raiden is a major step forward in the development of the Ethereum ecosystem. I was very interested in RDN as a great profit opportunity, especially considering how important it is for improving transaction scalability.
But Raiden was supposed to be part of the base Ethereum ecosystem. I can understand the argument against adding another fee layer to the Ethereum public blockchain. Transaction fees should entail ETH alone, not ETH + RDN + whatever. Plus Raiden is currently open-source, so it would be possible to just fork the entire project and complete it without a token layer, albeit with delays since the original dev team is obsessed with adding an additional fee layer.
https://twitter.com/VitalikButerin/status/911217245094686720
Yeah I agree with Vitalik too. The functionality of Ethereum should be kept as simple, efficient, and inexpensive as possible, because otherwise sooner or later it will be replaced. A Raiden that only uses ether and minimizes fees is much better for Ethereum and the price of ether in the long run, greatly outweighing the possible opportunity to flip a new token for some short term profit. I hope the Raiden team takes Vitalik up on his offer.General ETH question here (not necessarily related to mining):
Recently, the team behind the open-source Raiden development announced an ICO - RDN - to further fund Raiden development. To incentivize people to invest in RDN, they announced that there would be a fee structure set up requiring the expenditure of RDN tokens in order to make use of the Raiden feature-set.
Buterin has come out against the ICO, offering up $35 million of his own ERC20 tokens (mostly OMG and Kyber) to try to convince the Raiden team to cancel the ICO; in the event that they refuse, he'll donate to charity instead. Buterin is out $35 million either way.
How do you feel about this development? Personally I think Raiden is a major step forward in the development of the Ethereum ecosystem. I was very interested in RDN as a great profit opportunity, especially considering how important it is for improving transaction scalability.
But Raiden was supposed to be part of the base Ethereum ecosystem. I can understand the argument against adding another fee layer to the Ethereum public blockchain. Transaction fees should entail ETH alone, not ETH + RDN + whatever. Plus Raiden is currently open-source, so it would be possible to just fork the entire project and complete it without a token layer, albeit with delays since the original dev team is obsessed with adding an additional fee layer.
https://twitter.com/VitalikButerin/status/911217245094686720
Some pics of my Vega 56, custom watercooled:
I'm mining at 900/950 (stock BIOS) and getting about 35MHs. I flashed to a 64 BIOS and the mem would go up to 1100MHz, but at 900/1100 I was only getting about 37.5mhs (standard driver), so I don't think it's worth the extra power consumption just for 2.5mhs. I ended up flashing back to the 56 BIOS. Is the mining driver still iffy to use for daily gaming/mining purposes?
HBM is definitely not level with the die on my unmolded core. The block itself would pivot slightly if you tried rocking it, so I had to put extra TIM on the HBM side. I think the molded die cards have it level so are easier to cool.That's a clever mod, allowing use of a VGA waterlblock + existing blower fan + shroud assembly.
Are the HBM dies level with the GPU core? Curious how much cooler the HBM runs, and if it impacts overclockability and/or performance due to better cooling.