Claymore:

I am having some odd problems with my rig suddenly jumping out of the Miner every few minutes, seldom ever runs for 90 minutes like it was.

I added "-dbg 1" to the end of the command line to get more information on what is going on, but I don't see any more stuff?

The only thing that has changed is the ambient temp went from 5-10° to 25+

Read back a couple pages - it has been discussed - twice, actually.

Is that the general expectation these days?

I always thought that if I got Samsung VRAM I was going to be able to do more. Hynix (formerly HYumdai electroNIX) is better?

I know that Elpida (now defunct, but owned by the generally accepted last place Micron), although some cards made 5 years ago have it, or does that brand live on, made by Micron?

I know that much of it is subjective, absent concrete measurements & data. My 6gb (Gigabyte) 1060 with Hynix is not keeping up with what the 3gb generic/HP 1060 which had Samsung memory.

I could OC the Samsung to +280 stable, but the Hynix has to be dialed back to +200 even though it will run with +240.  I prefer to be able to leave my rig running for the week/end as opposed to having to constantly monitor from remote.

It's just being mean.

Look back a couple pages for some hints that I already gave.

Basic one is that many programs hook VRAM and run on a GPU if they can. Check the box in the nVidia driver to show the GPU app helper in the system tray and make sure things like Dropbox, WinZip and Firefox are not using some of the GPU memory. They all will, and there are numerous other programs that do. Logitech SetPoint, etc

Which AV are you using?  Several recently started detecting any/all P2P software as being 'bad', like WebRoot (for me).  I beta test for them, and it's done to be careful for noobs, but in my opinion it's a PITA.  I just remember to catch the new *.EXE (both in the root where I unpack and CUDA9.1 subdir for me) and let them live, and then add them to the 'skip' file list.

There's nothing wrong with it, even 10.5 / 10.6 give me the same message(s).

Now, I can't absolutely say that there will never be a problem, but that's what is going on now.

The 11.0 version runs very smoothly for me, although so far I haven't been able to get "Blake" altcoin stable, but I'm only using a single 6GB GTX1060 at the moment.

I'm plugging in my Asus 8GB GTX 1080 Strix, RX580 8GB Strix and Sapphire 4GB Nitro+ OC tomorrow, now that my Asus Prime Z270-A mobo is back from the repair depot.

We'll see; since the rig I'm using is super-limited mostly because of the factory power supply (don't even know what exactly it is, it's an HP Power Pavilion 580-023w prebuilt jobbie), so I'm throwing 850 watts Platinum EVGA power at the new one, should be able to get them all cranking.

OK, I see them all over the place, but I have zero experience with the AMD RX 580 cards.  What MH/sec should I be able to get from them?  One's a 4GB Sapphire Nitro+ OC and the other is an 8GB Asus ROG / Strix "T8G" model.  I understand that GTX1080 cards ain't the best choice for Ethereum mining, but I have the "A8G" advanced OC version, so we'll see what kind of fun I have tuning that one.

I do have a set of risers, so I might go ahead and plug in one or two more cards once I get those three stable and tuned up.

What's the best card - acknowledging that the prices right now are all over the place.  I assume that the GTX1060 is probably the leader as far as MH/power eaten, but I know that there are plenty of cards that will outperform it.  I would love to stick with Asus and Sapphire brand cards, but of course Gigabyte and others are all under consideration.

False Positive results are not uncommon. I have to rescue my Claymore  .EXE files (root + CUDA9. 1 are the ones that I need) because it blocks them because of the peer-to-peer activity.

I beta test for WebRoot and it is a known problem. Your choice with them, you just uncheck the blocks and enable the "leave alone" side.

Guys, you do know that you can set those five environment variables permanently and eliminate the lag running SETX five times at start-up each time, right?

Go into your Control Panel and set them and they will "stick". I know that on my little 7400 it takes a while to run that you can recoup with this simple process / fix.

The bottom line is that basically Keccak likely has the most ASOC resistance compared to the other algos he might have chosen, so it is likely to survive relative to others.

As we go forward, these ASIC resilient algos will be the best to choose.

I agree, maybe better for an overclocking site?

Mining rig assembly Monday, this rig is limited by the time puny power supply.

I did tune in to over 20Mh/sec.

Odd one, I agree.  I have visual evidence where I screen-shot it 'fixing' the hashing speed, but there is nada in the log about <Ctrl-Break> or anything else for 8 seconds.

11.0 just goes on mining, as if nothing happened - but the speed is back to what it is normally.  Curious.  I'll send the log, don't think it will be enlightening, but the graphic screen dump shows exactly what I described.

Awesome!

I have a technical question that seems to stump my logic: when I hit <Ctrl-C> in my Claymore Miner window, sometimes it exits out, sometimes the hashing rate returns to normal values (when I break, it's usually because it's not doing what it should).

Is it truly dual threaded where a double break <Ctrl-C><Ctrl-C> is required to cycle?  (seems to be)

"Overclocking resistant" graphics cards?

I finally figured out exactly what the problem was with my shiny new Gigabyte GTX 1060 6GB "OC" card.

It's always a good idea not to have more than one overclocking tool installed, but I suppose that I didn't look closely enough at software downloads for the new replacement video card? Their website doesn't seem to differentiate between the VGA Tool (for overclocking) and their Aorus graphics card software which I thought was simply a support and required load, but it turns out that it is simply a different style overclocking tool.

Their basic (now v 1.25) version is a glitzy graphical tool with more bells and whistles, whereas the Aorus tool is more straightforward and easier to use - sort of. Your mileage will vary!

Anyway, apparently having both of them installed causes them to fight over which has control over the overclocking - and neither wins. Frankly they both pretty much lose bumping the opposite tool in the head.

Once I had removed both of them, I installed "nVidia Inspector", more or less a generic overclocking solution for nVidia cards sort of along the lines of CCC for the AMD cards.  I used the parameters that I knew worked well for this specific card, and now it is always mining at 18.7MHm/sec as opposed to the 14.8-MH/sec it ran out of the box.

Just a little tuning and ended up with 18.8MH/s every time. I should note that Claymore Miner seems to be a little temperamental if you adjust the GPU clock while it is running, so use the tool to create a shortcut to your preferred settings, and add it to your START.BAT file.

Now I will futz around and see if I can get back to 20+

Those ASIC folks don't miss a beat: it would be rather silly of them to bake in support for any particular algo and not make it extensible.

Bottom line: if one algo fits multiple coins, it's likely going to be eat up by ASICs.

Windows "Swap" file size = virtual memory settings in the control panel.  Look up "adjust virtual memory size" using Google for you OS (assumed to be Windows 10?) and follow the instructions to increase the amount of available disk space for memory swapping.

In order to mine effectively, the DAG is kept in a ring buffer in VRAM (graphics memory).

You don't usually need to think about it, but graphic cards these days host programs, like "Dropbox", "WinZip", "Firefox" and others. Those reduce the total amount of DDR4 memory available to the mimer.

On nVidia cards (not certain how to do this on AMD cards) you can set a parameter in the driver to show the apps running in VRAM in a 'tray' icon. Now that we went to Epoch 168 today, the size is 2.31GB. - only one hundredth of a gigabyte but still, 10k more than the previous Epoch. Eventually 3GB cards will be unable to hold the DAG in memory.

Ah, got it. Thanks! I knew that I was missing something. I run a website with forums, but I don't think I read the rules hrre.

Quoting myself, I realize - sorry! I try not to quote entire posts.

Anyway, I am not sure what to think, but after running nearly ten hours overnight at about 21MH/sec, the card is back to 15.5MH/s most of the time.

I have seen (once) it start with barely 7MH/s but when I restarted it went back to 15.5MH/s

Nothing seems to click, but I was constantly over 21MH/s with occasional 22.5MH/s using the generic HP GTX1060 with only 3GB DDR4. This 6GB RAM card holds a 300+ overclock where the old one would not be able to get 270.  Actually it will go 333 but I get constant restarts clocked that high.

At 300 OC it runs an hour or more with no issues, but for some reason seldom lasts longer. I wrote a recursive batch file to archive the log files and restart, but I know that 82° can't be great for the card.

I looked at the temp boundaries for the card and nVidia shows that 99° is the throttle point with 102° = time to punt.

Odd one for me, a rookie: I am trying to tune my replacement 6GB GTX1060 in my original rig, which started life as an HP Pavilion 580-023w for $495.  The 3GB graphics card was giving me fits, so I replaced it with a Gigabyte Mini ITX "OC" 6GB version.

Right out of the box it gets 14.9-15.3MH/sec but I was able to tune it to give me 20.8-21MH/sec (see graphic below) consistently for 10 hours yesterday, but can't for the life of me figure out what I did - nor can I reproduce it.  It's back to the consistent 15.4MH/sec I had right after setting the pretty normal OC values.

You can see in the graphic that the single card hit a 36MH/s effective rate in the middle of that consistent 20MH/s+ run.

I double-checked the power supply and my Kill-a-Watt meter thinking something was holding it back, but - nada.

Any ideas from the experienced miners?  I was getting 19-20MH/s from my 3GB card before I swapped them, and now it seems unable to stick with it.  It boots, but it won't stay online, the fans kick into high gear after a few seconds of the miner running and the screen blanks out.



(what's the trick to posting imbedded graphics?   hmmmm...  I hadn't tried)

P S - I found it! I retraced my steps yesterday and realized that I had accidentally saved a change in my MSI Afterburner settings trying to see if this YouTube video was right about tuning the core power to 50-55% (instead of 0).

It doesn't work for me, and it results in losing 40% MH/s.  I'm back to 21MH/sec.

"Effective hashrate" will always go up and down and be inconsistent, because it is affected by luck (i.e. how quickly you solve problems, as opposed to the current around 51 second expectation).  The reported hashrate is the only one you can truly adjust by working with your overclock settings.  Memory clock doesn't help true hashrate that much, but the other clock is pretty much directly tied to the best hashrate you can get.

My 1060 6GB got 18.8MH/sec out of the box, but I have it up to 23MH/sec after playing with the overclocking settings until I was truly weary.  Over 20% of much better than most cards are going to see even with the best overclock adjustment settings, so I am lucky or as some say "won the silicon lottery", since I use a generic HP card.  Fortunately for me, they are built by Samsung and have Samsung memory - it makes a difference, if only a little.  Memory clocks don't make that big a difference, but it does let you make other adjustments you just can't get with Hynix or Micron or Elpida DDR5.  My Asus 1080 has Samsung DDR5X memory, so I expect it to be a great performer, even though I am told that - in general - 1080's don't do as well as others relative to hashrate performance.  We'll see...

While that's true, he must not be keeping up: dual mining is dead for now - until Claymore gets his 10.7 version out (probably a week from his previous statement here?), it will have at least one new algo supported that isn't already being sucked dead by the ASIC miners.

"Give it up!" for now...