|
I just deleted one of the lower ones, and moved everything down one.
|
|
|
|
It shouldn't matter much if you don't have higher straps. If your memory is clocked higher than the highest frequency strap, it will just use that one. At least it should. Now that I think about it, my Gigabyte would insta-crash if I clocked my memory to 2001, until I added a 2250 strap. Dunno wtf was up with that but that's not what's supposed to happen. The real downside is that your card's default mem clock will end up using your experimental timings and you might keep soft bricking your card. For undervolting the core you can try setting the voltage per dpm state but it probably won't go lower than 950mV without identifying the controller and adding some shit in the VOI table. Can't help you with that though 'cuz I haven't gotten to that point in my own modding yet. Still fighting with a damn Armor card with MJR to stop giving incorrect shares at 2150  I think I have to give up and go down to 2100. |
|
|
|
|
I'm not a coder or engineer so I don't think that's going to happen any time soon. Just looking for a condensed version of simple rules so I can accelerate the process of tweaking the BIOSes of a few cards.
If it's going to take several months of study to get 3 cards running 1 or 2% faster then I'm obviously not going to bother.
Anyone have some more practical advice?
|
|
|
|
|
Does anyone know off the top of their head which timings depend on others. Like timing x must be >= (timing y + 2), etc.?
For example I noticed lowering ACTRD gives a little hash rate improvement, but seems to add instability, unless I also lower ACTWR....but I don't know if that's actually what's happening, or the errors I am getting just appear to be higher because of variance.
I am trying to make a custom timing strap but it is very tedious...and timings that work in pairs just add way too many variables. Without knowing exactly which ones work this way I am basically just fumbling around in the dark. I actually downloaded the GDDR5 spec and skimmed through it but it pretty much just goes way over my head.
|
|
|
|
580W@85MH for 3 GPU seems irrational. HAHAHAHA and here I am mining with Hawaii cards! |
|
|
|
I have searched so apologise if this is a noob question, just got a GTX 980 to add to a system with an rx 480 in it. Currently testing it in a separate system that normally runs a 1070 no problem doing the same thing I'm trying to do with the 980 which is boot and use the computer with the iGPU of the CPU and mine on the GPU.
If I boot with the display plugged into the 980 I have no problems but when I'm using the iGPU I get the error "CUDA error: cannot allocate big buffer for DAG"
my bat file, which I run as admin
EthDcrMiner64.exe -epool etc-eu1.nanopool.org:19999 -ewal *My-Wallet-Address*.*My-Worker-Name*/*My-E-mail* -epsw x -mode 1
I'm On windows 10
382.53 drivers
Claymore 9.6
Parameters I have tried changing
-gser 2
-lidag 3
Virtual memory to 16gb
No-Overclock
I'm pulling my hair out trying to get this to work ( what little I have left!), I have no problems doing the same thing on the 1070 but the 980 will just not play ball. Any suggestions would be appreciated.
You need to use "-di <GPU #>" So if your IGP is GPU 0, your 480 card is GPU 1, and your 980 is GPU 2; you need to use "-di 2" to use the 980 only, or "-di 12" to use the 480 and the 980 |
|
|
|
- You can always switch back as soon as he updates it.  - Yes....yes it can. |
|
|
|
Mr. Claymore, do you need my log file of crashed ETH mining with AMD Vega for bug fixing in your miners?
I tested your ZCash miner too, same issue.
It is not a bug. My miners do not support Vega yet. I will add it in 2-4 weeks, it requires much more than a few code lines. Is there any chance to add it to epools, its pool related parameter so it makes sense to be there i think...
Ok I will add it in next update (not in upcoming one). My wife is going to kill me if my vega cards sit idle that long! They say patience is a virtue, but that won't save my life! In the mean time have you tried sgminer? I know I've seen a video with vega mining and I'm pretty sure that's what they used. Only low 30s hash rate but better than zero (depending on your power cost  ) https://github.com/genesismining/sgminer-gm/releasesAlso I hope you have some earplugs! |
|
|
|
Today I got the beast - AMD Radeon Vega FE 16GB. Does somebody know why mining with Claymore's ETH miner isn't working? I've set -etha 0 as reported by miner. I need to know which hash rate we can get with this newest GPU. Thank you in advance for a reply. -swisstrader https://s12.postimg.org/8iffawn31/IMG_3170.jpgWhat happens if you disable assembler mode? |
|
|
|
|
From what I understand, VDDC and VDDCI have to match, otherwise it will use the highest one for both voltages. But I don't have first hand knowledge with these cards or that mod (yet).
I just know from reading forums that you can insert things into the ROM to change those voltages, depending on the controller you have (IRxxx vs. NCPxxxxx)...and also change all the table pointers to correct for the changed length. Just pointing you in the right direction for that.
The user "hellm" (as well as some others) on overclock.net has some informative posts for doing this type of mod, but I don't think it's been compiled anywhere into a straightforward guide. It's spread all throughout the Polaris (and other) modding thread(s). From what I gather there should be enough info to do these mods on cards with the IR controller, but I think the NCP controller info is more murky.
|
|
|
|
The Sapphire one is classic - pretty easy to do - although your modded one clearly isn't a proper undervolt. Didn't check the MSI one.
What are the tools required, used commonly? Is there some specific how-to? Why are SRBTool or PolarisBiosEditor not working on those classic old cards? Check here in the first post down at the bottom, "How to edit ROM for data/command table length changes" http://www.overclock.net/t/1561372/hawaii-bios-editing-290-290x-295x2-390-390x |
|
|
|
First post, just joined.
Any news on support for Vega FE? I have two of them on the way.
Did you buy them for mining? Yes. Mining is the excuse I used to buy them, with the hopes they'll pay for themselves, not actually turn a profit. Will continue to game off my GTX 970 until the FE's pay for themselves. Currently getting about 16MH from my Asus Strix GTX 970. No OC, just whatever the card does out of the box. I'm curious what they could do with an optimized kernel too. I would expect them to be 80% faster than Fury given the memory speed advantage, and I'm pretty sure the bus won't be a bottleneck. Last I heard though someone tried and only got low 30-something MH/s with one. Not sure which miner was used. Maybe it was sgminer. |
|
|
|
womp womp @KoSoVaR That second strap you posted has some timings that are looser than the first one. You could try taking the second one and changing tRRD to 5 and use the other tighter timings from the SEQ_RAS_TIMING section of the first one to make a modified version of the second one. |
|
|
|
I can directly control all VRMs on Polaris cards I've come across, and STILL haven't found one that will let me fuck with MVDDC.
Ah okay..thought I read you found some ASUS strix cards that you could do it on. not following on "not PBE shit". VDDC does not work on Linux, regardless of what I set in PBE and flash? And this is where omg comes in? I think those omgtools are all for timings modding. You would need to hex edit the rom for volt mods afaik...still need to learn this shit myself. I don't think PBE can do volt mods but I thought this SRB one could do a basic one. |
|
|
|
Solo mining is definitely still do-able depending on how high your Sol rate is. It would probably take about 2 pretty good rigs though to get 1 block / day. You just have to ask yourself 1 question...'Do I feel lucky?' I just started mining and am using hushpool.cloud. It's been working good for me, although it would be nice to make the auto payout higher than 0.1 to save on tx fees and have less incoming tx spam. It is paying out steady though and amounts consistent with my hashrate/difficulty. Only thing is you should NOT use suprnova...at least until their % of total network hashrate comes way down. Couldn't tell you much about wallets other than the ability to encrypt the wallet.dat from the Windows 0.68.6 wallet seems broken. |
|
|
|
|
The UberMix 3.1 only gets you to 24?
What if you take the 1625 timings and run them through nerdralph's tool?
Keep in mind that some of these guys have highly tuned timings that they've been testing and optimizing...not just some copy/paste jobbie. And some of them have been able to overvolt the VRAM and therefore I assume can get higher mem clocks/tighter timings than us normies. Unless you are willing to pay someone for a mod or learn how to directly control the VRM I wouldn't expect to get 30+MH/s.
|
|
|
|
|
What's the best (fastest) OS/driver/miner combo for dagger/hashimoto on these cards (assuming all your voltage and timing mods have already been done on the bios)?
I am guessing it is linux/gpupro/sgminer, but what about xorg or fglrx drivers or claymore's miner?
Or is there even a noticable difference from win+claymore or win+ethminer?
|
|
|
|
|
Thinking of mining this, but I just installed the 0.68.6 Windows wallet and noticed the 'Wallet > Encrypt...' function is grayed out. Am I doing something wrong or is this a bug?
If I encrypt the wallet.dat file with a 3rd party tool will that break the wallet software?
|
|
|
|
|
Why would big pharma pay for the folding that the curecoin team does when they can get the evga team's and every other team's folding for free?
It doesn't work the way you want it to because it can't unless you can convince Stanford of your arguments, not the curecoin devs.
Curecoin is worth what people are willing to pay for it, just like every other coin. If it's a problem for you, then why not point your rigs at something else and dump that instead?
|
|
|
|
|
Show Posts
