Topic: Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux) - page 989. (Read 6590718 times)

For the ETH, there is little difference between the 470 and 480. However, the RX 480 has faster memory.

So what's better? 470, 480, or 480 8GB? I've been checking around and can't find a definitive answer on this one.

I had a second ssd of 32gb in stock.

This is why I wanted to raid, for precisely, not to have to buy another one.

I allocated 10go on the primary ssd c: and 16go on the secondary ssd d:

We will say that it works.

I wouldn't worry about raid0.. it would only save you about 0.25 seconds on virtual memory allocation on miner startup. Just replace your 32gb with a larger drive.. at least 60gb. I have my vmemory set for 25GB and have zero problems.

IF you did insist on a raid config you'd have to get a PCI-e RAID card, but then you'd be taking up one of your slots :\

ETH+PASC is the most profitable for my modded RX480s ATM... I use '-dcri 15' and it only drops my ETH by about 5% on my 4gb cards. Nearly 0% on my 8gb cards.

Checked it but none of the card is running in Crossfire mode :x

Hello

I will have a question about the virtual memory,

I wanted to put a 2nd hard disk of 32go on my astrock btc h81. To make a raid0. To have a 64GB hard drive.

To increase virtual memory.

But asrcok btc h81 is not compatible raid.

I can choose the hard drive for virtual memory?

Eg choose 16go from my hard drive 2.?

Hi all,

Can you suggest what is best for mine eth.

siacoin, DCR, pascal  

I try DCR but I finally switched on only eth, large increase in consumption and a small profit, perhaps siacoin is better?

I have 9x 470 low power bios

I think you missed the point. I know Polaris is more efficient. I was just using Hawaii as an example to illustrate that there is VDDC, and there is VDDCI or AUX voltage, which are two different things. It is dangerous to set VDDCI higher than 1050mV or so, whereas the core (VDDC) can be safely higher....so to say to just set them equal because it will use the higher of the two means if he is already running at, say, 1150mV+ on the core, and then he sets his AUX voltage to 1150mV also, he can fry his card.

Also, it seems like you are saying that neither AUX nor VDDC can be set at a value lower than 1000mV. That may be the case for AUX, but I'm pretty sure it's possible to set VDDC to whatever you want in DPM 7, as long as you lower the other DPM states to a value the same or even lower. If DPM 6 is set to 1000 and you try to set DPM 7 to 950, sure, it probably won't take...but as long as you lower DPM 6, 5, etc. to 950 or less as well it should work.

EDIT: I should add that AUX/VDDCI is not actually the memory voltage, but rather the memory controller voltage, which is actually on the die itself. AFAIK the memory voltage is about 1.5V and is not configurable even with a BIOS edit. You would have to do a hard mod to change that which is a bit extreme and would require some serious knowledge. The effects of changing AUX voltage are small compared to a change in VDDC so even if you could get it working below 1000mV it wouldn't save you much power in comparison to lowering the core. I'm pretty sure the 480 hash rate is bottlenecked by its memory performance anyway, so it would be best to try and get the core voltage down because even if you had to downclock the core a bit, it shouldn't affect the hash rate too much compared to downclocking the memory.

Forget the GPU of the notebook, break only the PC, and you do nothing, that has not enough memory to create the dag

Hi!

Anyone getting error like this?

Im setting it up on my laptop just to test. Can anyone help? Thanks!


=========================================================================
15:46:32:529   1aa0   Create GPU buffer for GPU #0
15:46:32:530   1aa0   CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.

15:46:35:047   b0c   NVML: cannot get fan speed, error 3
15:46:35:532   1444   Setting DAG epoch #112 for GPU #0
15:46:35:534   1444   GPU 0, CUDA error 11 - cannot write buffer for DAG

15:46:38:048   b0c   NVML: cannot get fan speed, error 3
15:46:39:794   22ec   ETH: checking pool connection...
==========================================================================

Im using nvidia GTX 645M.

Hi everyone. Could you please tell me can i do more to lower my power consumption ?. I v got 6*rx470 hashing eth only at 170Mhs and 880watt from wall. I d like to lower power consumption even with less hashrate.

Changed the wording above to make it clearer. What I meant was, the lowest stable clocks that you found for memory should be applied to the core as well, and to from there to undervolt the core you will need to introduce an offset as directly changing mvs below that wont work.
Also, Polaris is an efficient architecture so it uses much lower core values than Hawaai/Tahiti so stock clocks are stable at 1000mv on almost every chip, while memory if pushed to 1900+ will need the full 1000mv that it defaults to, so in most cases the higher mvs are on the memory not the core. Unfortunately the higher mvs that you key in for memory are also used by the core (ignoring what you specified it to run at) unless you introduce an offset.

So i installed Boysies v2 rom for red devils 8gb.

But now there is no way to keep it hashing, every setting that i try results in my screen going off...

Has anybody got a solution for this? or maybe can share the setting that they are using for this rom, please.

Help

Umm....you sure about that? That is weird as fuck if it's true. Haven't used Polaris myself but any other card they are two separate voltages. For Hawaii...mem voltage above 1.05V can be dangerous and yet is probably too low for the core under load even at stock clocks. Meanwhile the core can easily do 1.25V and not harm anything.

@Swiped

I recommend you try the overclock.net forums for answers. There are lots of helpful and knowledgeable people there that will gladly supply accurate info.

http://www.overclock.net/t/1604567/polaris-bios-editing-rx480-rx470-rx460/0_100
http://www.overclock.net/t/1605802/official-radeon-rx480-470-460-owners-club/0_100

If you have found the lowest stable voltages for the clocks you want and you're already running modded mem timings, then there's probably not a whole lot you can do except downclocking either the core (if you are bottlenecked by memory performance) or memory (if you are bottlenecked by the card's compute performance). If you can downclock one of those parameters without significantly lowering your hashrate then you can probably find a lower stable voltage for your new lower clocks and thus save power.

Try to Run your cards at factory settings, that means flashing back stock BIOS, and reverting to stock clocks. If the cards are stable, theres your problem. its usually high overclocks or tooo tight timings in the BIOS mod.
If that doesent work, replace the card, then replace the riser. If neither of those work, switch around the risers on the PCIe slots to see if a particular PCIe slot is causing the failure, which could mean a faulty PCIe slot on the motherboard.
Also, change your attitude, if you whine about the community being unhelpful, its a self fulfilling prophecy. Your EXACT problem has been encountered and reported by no less than a 100 people here, and NOTHING that I or anyone else will share now has not been said before ... if only every new poster could bother to spend the time needed to look for solutions already posted.


Seems like you already have found stable voltages for Core and Memory. I am assuming this was at stock clocks? I am also assuming you ahve modded your bios memory timings already. I recommend having core and memory both set at the lowest volts that you can get your memory to work on (because they will each use the higher of the two anyways).
Then you reduced core and increased memory, are those stable at the voltages you figured out earlier? Did the power draw now go up or down?
The next step would be to introduce a voltage offset for the core in a tool like watttool Because you increased Memory clocks you can not reduce the volts to it, infact you may need to increase them if higher memory clocks arent stable. Then because you lowered core, do not change the core volts but put in an offset, lets say -50mv (directly changing core volts to below memory volts will make no difference) and see if that is stable. Repeat with a lower value till you hit the lowest stable setting.
Remember, with every change in clocks the volts needed for it to be stable would also change so its a hit and try to find the perfect balance.

crazyearner do you have a reboot.bat in the Claymore folder? Inside the .bat should be:

Have had this happen alot.  The best fix is to simply increase the voltage on the GPU by 10 mV at a time until it stops.  If you are confident you drivers are good, and it only happens on one card out of a set, it means you are undervolted for the application.  Not all cards are equal, and I have had a number of new cards struggle with this and I have had a number of 480s with this problem, and I can get rid of the OpenCl every time with a simple increase in the power.

Question for the experienced manipulator of GPU card parameters.  I want to retain my ETH mining rate on my Sapphire rx 470 8g cards but lower the power.  I have played with the voltages on the memory and GPU and figured out where they run stable.  My question is around the GPU clock and memory clock settings.  I have the GPU set at 1150 based on some posts I read, but am I trying to get it as low as possible to reduce power?  My memory is set at 2050 and should I being trying to raise that up as much as possible?  Won't that elevate my power consumption?  I have a Mod'd bios on my cards to reduce overall power use.  But I am wanting to more fully understand the objective around changing the clock speeds.  Any insight would be appreciated.