@Claymore, few questions:
1) -cclock: you wrote "you can overclock only", but it seems not to be true for, at least, Windows and Polaris. My 470x can be underclocked (GPU set to 1150 for all states).
2) Any chance to support -cclock and -mclock on Linux for amdgpu-pro drivers (for Polaris)? It is not said "unsupported on Linux", but does not work for me (mem can't be set to 1950 using this).
3) Any chance to support -cvddc and -mvddc on Linux for amdgpu-pro drivers?
4) Are older cards like 7950/7970 supported using amdgpu-pro drivers on Linux?
5) There is an issue on Windows when used -*clock commands: they are applied at some init stage, but for my 470x it instantly changes hashrates only for the card with monitor connected. I can stop and resume other cards for changes to take effect (and show correct/new hashrate), but probably makes sense to fix it somehow, so miner shows correct hashrates after applying values. Of course, if you run it for 2+ time, drivers keep settings and there is no that effect. Still, it is surprising when you run minr with some value and it is not reflected in hashrates.
1. AMD disabled underclocking in theri drivers in past. May be it enabled it again in recent drivers, I did not check it. 2. Linux requires root access for it, I remember someone said that even sudo is not working. 3. Did not check if it is possible. 4. I did not check it. 5. So if I take two 4xx cards, start miner with -*clock option and it will change clocks only for one card? 5. No. In short: changing memory clock using any tool does not change displayed hashrate until miner restarted or card stopped/resumed. Except a card with monitor connected: for it new hashrate is shown instantly. When I change memory freq using any tool (WattTool) for a single card while miner is running, it instantly changes displayed hashrate only if the card has connected monitor (tested on Rx470). For headless cards the hashrate shown is not changed until miner restarted or card stopped/resumed. That seems to be a feature (@All: please confirm, but I saw someone complained for the same issue). When you change freq using miner cmdline, and previous freq was not the same, miner sets freq but shows hashrate for old freq until restarted. 5. Some cards I have, wattman doesn't allow changing clocks/voltages unless I pause miner. I didn't think of if monitor was connected. I have some display emulators I got for $15 that helps with VNC, but not sure it would be worth it for more than that... It is very strange for me because sometimes displayed hashrate changing when I change ram frequency, sometimes it doesn't react sometimes it react for gpu frequency though  So, I have been using CDM 7.2 for about 5 days now on my W7 system. I was finally able to reduce my wattage pull through Wattman setting (and disabling dual mining), but am still stuck on a paltry 20.75 Mh/s on ETH per each of my MSI RX 470's. I have tried everything on the CDM side, such as setting ETH intensity to the max of 16 in my command line, adding in all 5 of the recommended setx lines, set my virtual memory to 24GB, etc. None of this moved the needle at all.
So, I am now trying the overclocking route before getting into modding the bios on these. A few things are perplexing though. In my Wattman, each of the 4 GPU's can only be increased from 1650 default to 1700. These seems like a very low limit to me and setting all 4 to 1700 had zero impact.
So, using Afterburner, it actually gave me the option to raise Memory Clock all the way to 2100. I only tried 1850 and applied to all 4 GPUs but yet again, absolutely NO impact... all 4 still mining at exactly 20.75. So, my question is, am I doing something wrong? Shouldn't raising the the clock speed on these yield at least a 5-10% bump in mining hash-rate? Or am I completely off base here?
I know of many people using these cards that have achieved 23+ right out of the box, so still trying to figure out why I cannot get close to that mark.
Please update your driver version.. I have same issue with my rx 480 before update drivers I was able only to increase RAM frequency from 2000 to 2100Mhz, after update I can set 2250Mhz in Wattman  Thanks, but I updated to Crimson 16.9.2 over a week ago and it had zero impact. However, I was finally able to at least get a little HR increase by setting to 1700 in Wattman. Unfortunately, this put my wattage usage through the roof! Increased it 50-60W at the wall (although strangely, Afterburner showed steady wattage at around 74 with no increase after moving to 1700 and I can't figure out why). And, this was for only 1 GPU... doing to all 4 pushed my up to around 750 (from 625 before)... all for a measly increase from 20.75 to 21.45, so clearly not worth the extra power drain. I since reset everything in Wattman, tweaked down the GPU mV settings in states 6 & 7 and I'm back down in the 620 range with each of the 4 470's hashing a pretty pathetic 20.7. Bottom line is that a bunch of others here got 23+ out of the box on these cards... no modding, no OCing, nothing. So, I guess I just got really unlucky with my batch. I guess a pseudo-silver-lining is that this makes my decision really easy on modding the bios on these. I had previously thought that if I got 23 out of the box and could OC 1-2 more, then I wouldn't risk modding. But now, it's a no-brainer since there is no way I'm going to ever be happy running this rig getting less than 21 per card, nor will it likely ever be profitable that way. So, hopefully after modding I can at least get in the 25-26 range. If not, then I don't see much point in adding a 5th card to this rig. My next rig though will not be built with MSI cards, that's for sure! What card do you have? The better 470 4GB cards have 7GHz (1750), will do 22MH memory, the 8GB cards have 8GHz (2000) memory (except the terrible MSI ones....) and will do 24.6 (stock, I don't OC them). The best are the Sapphire Nitro, good memory, best undervolting from the ones I have tested. The 470 4GB with 6.6 or 6.7GHz memory you should avoid, also MSI 8GB, only has 6.7GHz, should be 8GHz, since it costs same as the Sapphire, but it has bottom of the barrel memory. Edit, I see you have MSI 470s.... well, that's unfortunate. I didn't buy any of them, even early on and supply was hard to find, because their memory is slow (always look at specs on cards... newegg usually has all the details). Nitro 8GB Sapphires on sale right now @ newegg... Thanks for the feedback, I will definitely consider this while purchasing to build my next rig. I will see what I can squeeze out of those MSIs through the commonly used strapping mod. Others with those cards have been able to achieve close to 5W/MH on a 6-card rig, so theoretically, I should be able to at least get close to 6 on a 4/5 card rig, even if i got a slightly bad batch of them. Thanks for the tip on the sale also, I will check that out! |
|
|
|
|
Thank Lone & Aries! I saw a bit of the video last week but watched all last night and it seems pretty straight forward and even made sense for a noob like me lol. I will be taking a crack at doing both the strapping and power reduction on 1 card today as a test and see how it goes.
|
|
|
|
Alright, I got 6 full cards hashing at stock at the moment. (not for a long time, last time it took 10 mins to crash) After uninstalling and reinstall amd driver a lots, I downloaded their cleanup utility. I manually removed all rx470 from the device manager with driver's delete option. Launched the cleanup and rebooted, launched the clean again and rebooted. Downloaded the latest driver, I always used 16.9.2, installed the 16.10.1 drivers. Rebooted. I'm going to bed, but it looks good ! Thanks to everyone for all the inputs. Hoping I can OC them a bit tomorrow  EDIT: And GPU hangs in OpenCL call, exit... It tried to restart but it's not working I get those from time to time too. Mostly now when. I updated to claymore 7.0 and above Not saying it is calymore but I see it more with 7.0 or newer Great infos ! Thanks you. What's your frequency ? I run 1500 strap with 1130core and 1945 mem all modded bios I make 1 more setting change for my MSI RX 470 Gaming X: Start WattTool - Set Voltage offset to "-15" and click set for each card. This will decrease your overall watts TREMENDOUSLY; therefore, reduce cost! This will make you smile. Hey Aries... I've been studying your posts in the hopes of at least partially re-creating the power/watt ratio you've been able to achieve on your 6x470 rig and was hoping you might be able to answer a few questions for me  1- You are using MSI RX 470 4GB, correct? Or, are you using the 8GB version? 2- When you say "1500 strap", what are you referring to? I googled it but couldn't find anything that made sense 3- I looked up WattToll and will download it on my rig later, but had a question about the command you used "set Voltage Offsett to -15". I see this field at the lower right of the WattToll window (shows "0 x 6.25mV" where user can over-write the "0"). Just to confirm, you simply enter "-15" in this field and click Set? Thanks! |
|
|
|
@Claymore, few questions:
1) -cclock: you wrote "you can overclock only", but it seems not to be true for, at least, Windows and Polaris. My 470x can be underclocked (GPU set to 1150 for all states).
2) Any chance to support -cclock and -mclock on Linux for amdgpu-pro drivers (for Polaris)? It is not said "unsupported on Linux", but does not work for me (mem can't be set to 1950 using this).
3) Any chance to support -cvddc and -mvddc on Linux for amdgpu-pro drivers?
4) Are older cards like 7950/7970 supported using amdgpu-pro drivers on Linux?
5) There is an issue on Windows when used -*clock commands: they are applied at some init stage, but for my 470x it instantly changes hashrates only for the card with monitor connected. I can stop and resume other cards for changes to take effect (and show correct/new hashrate), but probably makes sense to fix it somehow, so miner shows correct hashrates after applying values. Of course, if you run it for 2+ time, drivers keep settings and there is no that effect. Still, it is surprising when you run minr with some value and it is not reflected in hashrates.
1. AMD disabled underclocking in theri drivers in past. May be it enabled it again in recent drivers, I did not check it. 2. Linux requires root access for it, I remember someone said that even sudo is not working. 3. Did not check if it is possible. 4. I did not check it. 5. So if I take two 4xx cards, start miner with -*clock option and it will change clocks only for one card? 5. No. In short: changing memory clock using any tool does not change displayed hashrate until miner restarted or card stopped/resumed. Except a card with monitor connected: for it new hashrate is shown instantly. When I change memory freq using any tool (WattTool) for a single card while miner is running, it instantly changes displayed hashrate only if the card has connected monitor (tested on Rx470). For headless cards the hashrate shown is not changed until miner restarted or card stopped/resumed. That seems to be a feature (@All: please confirm, but I saw someone complained for the same issue). When you change freq using miner cmdline, and previous freq was not the same, miner sets freq but shows hashrate for old freq until restarted. 5. Some cards I have, wattman doesn't allow changing clocks/voltages unless I pause miner. I didn't think of if monitor was connected. I have some display emulators I got for $15 that helps with VNC, but not sure it would be worth it for more than that... It is very strange for me because sometimes displayed hashrate changing when I change ram frequency, sometimes it doesn't react sometimes it react for gpu frequency though So, I have been using CDM 7.2 for about 5 days now on my W7 system. I was finally able to reduce my wattage pull through Wattman setting (and disabling dual mining), but am still stuck on a paltry 20.75 Mh/s on ETH per each of my MSI RX 470's. I have tried everything on the CDM side, such as setting ETH intensity to the max of 16 in my command line, adding in all 5 of the recommended setx lines, set my virtual memory to 24GB, etc. None of this moved the needle at all.
So, I am now trying the overclocking route before getting into modding the bios on these. A few things are perplexing though. In my Wattman, each of the 4 GPU's can only be increased from 1650 default to 1700. These seems like a very low limit to me and setting all 4 to 1700 had zero impact.
So, using Afterburner, it actually gave me the option to raise Memory Clock all the way to 2100. I only tried 1850 and applied to all 4 GPUs but yet again, absolutely NO impact... all 4 still mining at exactly 20.75. So, my question is, am I doing something wrong? Shouldn't raising the the clock speed on these yield at least a 5-10% bump in mining hash-rate? Or am I completely off base here?
I know of many people using these cards that have achieved 23+ right out of the box, so still trying to figure out why I cannot get close to that mark.
Please update your driver version.. I have same issue with my rx 480 before update drivers I was able only to increase RAM frequency from 2000 to 2100Mhz, after update I can set 2250Mhz in Wattman Thanks, but I updated to Crimson 16.9.2 over a week ago and it had zero impact. However, I was finally able to at least get a little HR increase by setting to 1700 in Wattman. Unfortunately, this put my wattage usage through the roof! Increased it 50-60W at the wall (although strangely, Afterburner showed steady wattage at around 74 with no increase after moving to 1700 and I can't figure out why). And, this was for only 1 GPU... doing to all 4 pushed my up to around 750 (from 625 before)... all for a measly increase from 20.75 to 21.45, so clearly not worth the extra power drain. I since reset everything in Wattman, tweaked down the GPU mV settings in states 6 & 7 and I'm back down in the 620 range with each of the 4 470's hashing a pretty pathetic 20.7. Bottom line is that a bunch of others here got 23+ out of the box on these cards... no modding, no OCing, nothing. So, I guess I just got really unlucky with my batch. I guess a pseudo-silver-lining is that this makes my decision really easy on modding the bios on these. I had previously thought that if I got 23 out of the box and could OC 1-2 more, then I wouldn't risk modding. But now, it's a no-brainer since there is no way I'm going to ever be happy running this rig getting less than 21 per card, nor will it likely ever be profitable that way. So, hopefully after modding I can at least get in the 25-26 range. If not, then I don't see much point in adding a 5th card to this rig. My next rig though will not be built with MSI cards, that's for sure! |
|
|
|
|
Well, I continue to try and tweak my clock speed and voltage settings on my 4 RX 470 rig. At last check, I had finally got all 4 cards up slightly from 20.75 to 21.45 Mh/s by setting clock speed to 1700 (max amount in my Wattman for some reason...Afterburner allows all the way up to 2100 so not sure why the disconnect). Unfortunately, this increased my wattage at the wall to ~730-740 (from ~625 before the tweak). Obviously, this was a net negative trade-off, so I went back in and tweaked some things a bit earlier today.
First thing I did was reset everything on all 4 cards. I wanted to get back my power savings first and then start to mess with clock speeds. Looking only at the GPU mV by state, I first took a crack at lowering state 4-7 of my 2 hungrier cards. For what ever reason, these 2 suck like 15% more power across the board than the other 2, so I chopped them down a bit. Unfortunately, I must have lowered it too much as 1 of the cards crashed, then my system crashed and rebooted. So, I dialed back the reductions a bit, as well as lowering the other 2 cards. This time all 4 cards were stable and I was actually showing a bit lower total wattage than my original low at about 615.
From here, I upped the clock-speed to 1700 on only 1 GPU to test the result. The HR did go back up to 21.45 and surprisingly, Afterburner was showing almost no increase in wattage pull! BUT... when I checked my kilawatt, the wattage had jumped up about 50W! So... 2 questions here:
1: is a 50W bump for only a measly 1650 to 1700 over-clock reasonable? I never OC'd before but this seems high for such a nominal increase
2: why would Afterburner still show the wattage on this card at around 74 both before and after the OC, but my overall wattage pull increased by 50 at the wall? Does this sound "normal" or is there possible something wrong elsewhere in my system?
I also did some benchmarking on other system resources finally. At boot up with nothing running (except windows of course), I am at 95W (not including monitor that is not plugged into my Kilawatt unit). Then I launched MinerGate, that I have 7 cores of CPU mining XMR. This added about 45W. That CPU is hashing about 260 on XMR so that's a semi-decent side-profit of about $10/month lol.
|
|
|
|
If you turn one thing up, you have to turn something else down, so the power stays the same.
Memory frequency is the biggest positive hammer for hash rate, and voltage is the biggest way to reduce power.
Up the memory speed by 100 MHz, and reduce the voltage from 1000mV down to 950. And then you can down clock the core frequency, too.
You need to keep it all balanced
Thanks Lone... I just had what at least feels like a breakthrough... or at least did for a brief moment lol! I did a reboot after setting all cards to 1700 in Wattman and 1851 in Afterburner. Still no movement at all on HR. Then I noticed I was back up to ~690 watts vs the ~625 from before the reboot. After looking at Wattman, I realized that the State 6 & 7 lowerings I made yesterday were reset. So, I changed them back to roughly what saved me the 75 watts. At this time, I also move the memory back up to 1700 as that was reset too. In addition, this time I moved the GPU frequency slider bar to -10% on all 4 cards. Started up Claymore again and was happy to see that all 4 GPU's were now up in HR, albeit slightly (from 20.75 to 21.45). But unfortunately, I then checked my wattage and it had ballooned up to 770! So, I obviously need to do some more tinkering because I am pretty sure that a gain of 3 Mh/s is not worth an extra 140 watts lol. But, at least good to know something has finally affected the HR. Is it possible that the GPU Frequency set to -10% increased my HR? Seems counter-intuitive but I am still learning this stuff and that was the only new change that I made. |
|
|
|
|
So, I have been using CDM 7.2 for about 5 days now on my W7 system. I was finally able to reduce my wattage pull through Wattman setting (and disabling dual mining), but am still stuck on a paltry 20.75 Mh/s on ETH per each of my MSI RX 470's. I have tried everything on the CDM side, such as setting ETH intensity to the max of 16 in my command line, adding in all 5 of the recommended setx lines, set my virtual memory to 24GB, etc. None of this moved the needle at all.
So, I am now trying the overclocking route before getting into modding the bios on these. A few things are perplexing though. In my Wattman, each of the 4 GPU's can only be increased from 1650 default to 1700. These seems like a very low limit to me and setting all 4 to 1700 had zero impact.
So, using Afterburner, it actually gave me the option to raise Memory Clock all the way to 2100. I only tried 1850 and applied to all 4 GPUs but yet again, absolutely NO impact... all 4 still mining at exactly 20.75. So, my question is, am I doing something wrong? Shouldn't raising the the clock speed on these yield at least a 5-10% bump in mining hash-rate? Or am I completely off base here?
I know of many people using these cards that have achieved 23+ right out of the box, so still trying to figure out why I cannot get close to that mark.
|
|
|
|
Ah oh thanks... I just had a bit of success lower wattage use before I saw your post... I did stop Claymore while I messed with Wattman this time, which I think helped. I made the same kinds of changes as before that crashed the system but this time it was fine. The slight difference is that instead of using the slider bare to lower the entire curve, i manually over-wrote only state 6 & 7 with lower values than default (I also re-set wattman on every card before doing anything. I did not use the same values for each card because some were over 1100 in state 7 while others will right around 1000. Still, after lower these values a bit for each GPU, I checked my Kilowatt and found that wattage dropped from 695 to 625, so that's definitely a mover in the right direction! And, this change had NO effect on hash-rate, so I would bet that I can fine-tune a bit more and get to around the 600 mark, which would already be 25% lower than what I was at with dual mining yesterday. Still trying to figure out why I can't get better than 20.7 hash-rate no matter what I do, but one thing at a time  |
|
|
|
Big shout out to Phil for sending me the extra molex cable I needed to get my 4th 470 up and running! Thanks also for the extra goodies you included Phil... really appreciate it!
So, as it stands now, I am getting just shy of 82 Mh/s on ETH and about 1300 Mh/s on DCR. I am happy to finally have this rig running with 4 GPU's but I am still baffled at how these MSI 470's are only getting 20.7 per unit on ETH. I have tried pretty much everything suggested in this forum and in the Claymore readme file but nothing notably effects the speed of ETH mining. I've done the following:
- Added the 5 setx lines to my .bat file
- Added the "-ethi xx" command within the command line of the .bat file. Tried it at 10, 12, 14 & 16.
- Added the "-ethi xx" command to the config file. Tried it at 10, 12, 14 & 16 in various places.
- Increased my Virtual Memory from 16GB to 24GB
None of these moved the needle on ETH mining at all... still all 4 GPU's averaging 20.7 per unit. The only thing that effected ETH speed at all, was to ratchet down DCR intensity significantly but the minimal gain of .4% (from 20.72 to 20.82) would actually be more than offset by the dramatic decrease in DCR effectiveness. On the flip-side, bumping up DCR to 35 increased its hash-rate by ~15% but only nominally effected ETH mining (less than .1%). Bottom line, nothing I've have tried so far is getting me even remotely in the ballpark of the ~23 Mh/s that others are getting from those GPU's out of the box.
I would almost just think I got a bad batch, but my 4 GPUs came from 2 different vendors (3 from Amazon and 1 from NewEgg), spread over 3 orders over 2-3 weeks, so this seems unlikely. I really wanted to close this gap before proceeding with modding, but it's looking bleak at this point. I am putting off purchasing a 5th GPU for this rig until I can figure out exactly how much I can improve its current performance. Frankly, at less than 21 per card, it's barely going to be profitable, especially at the ridiculous power draw it is currently at... which reminds me...
So, I have my rig plugged into a KilaWatt unit. While running the rig with 3 GPUs, it was pulling around 630-640 watts. This seemed high to me because I was thinking max of 330 for the GPUs +140 for the i7-6800k, which left about 160 watts for the Mobo, SSD, heatsink, etc. Again, seemed a bit high but not out of the realm of possibilities.
Well, here's the perplexing thing... after hooking up the 4th GPU, KilaWatt now says I am using 810 watts! That's a 170 watt bump over the 3 GPU set-up. So, how exactly can a GPU with a max TPD of 110 watts add 170 watts of power usage? At this rate, I couldn't even add another GPU to this rig as my PSU is 1kw. I know I need to really dig into the power settings that have been mentioned here recently but was wondering if anyone could explain how a 110w GPU could add 170w to my power draw.
Thanks again for any tips!
I would push down the core 20% Up the memory to 1850 That will up the MHs and lower the power without bios mod 120tdp isnt the max power draw. Also msi cards have a high power set than the other 470s If you want help with bios mod I will help anyway I can If you pay 10cents or more in power I wouldn't dual mine. Your only talking about maybe 10-20$ profit Pushing the cards harder and hotter for only 20$ ehh I'll save them Yeah, my electricity is 11 cents, so doing rough math for 110 watts, cost me about $9/month. At best with 4 GPU's on DCR, I could mine maybe 15 in a month, so at DCR current rates, maybe $5 profit/month at best so really not worth it. Now, if it goes $2 and above then certainly worth looking at but not really at $1. I am curious on your 2 suggestions... can these settings be done within the AMD WattMan App? I haven't really dug into that yet but if it is indeed managed there, I will certainly try that ASAP! Yes those settings are for wattman. It'll be the easiest way to play around with OK, so I attempted to make the recommended changes this morning but not really seeing any results, so I am probably doing something wrong. Here is what I tried in Wattman: - decreased Power Limit (in the Temp section) by 20% - increased the Memory Frequency from 1650 to 1700 (I know 1850 was recommended but the max value I could set in Wattman was 1700, and only "State 1" was enabled (on all 4 cards) so, only 1 State moved and only up 50 Mhz I applied this changes on all 4 GPUs. There was NO impact to hash-rate. There was impact to wattage usage but in the wrong direction... went from about 695 to 710. When I moved the Power Limit back to anything between -10% and 0 (flat), I got back down into the 695 range. After this, I opened Afterburner. Also my first experience with it so still early in the learning curve on all of this stuff. - Did further testing with Power Limit but same results as with Wattman... no impact between -10% & flat but increased wattage pull with -20% - In this app, Memory Clock was movable all the way up to around 2000, so I tried the 1850 and applied to all 4 GPUs, but alas, again, zero impact to hash-rate on any of the 4 GPUs. The other thing I briefly played with (but with scary results) was the GPU frequency in Wattman. I saw a youtube video where they were pulling down the state 6 & 7 frequency max values to around 970 & 1000 respectively. I tried that on GPU#1 (not a large decrease from the reflected values) and it actually lowered my wattage a bit. But, when I did the exact same thing to GPU#2 (this was actually a larger % decrease from the reflected values as this card seems to run hotter & pull more wattage than the others), it ended up stopping that GPU and crashing my system. After that, I didn't mess with the GPU settings again. So, given all of this, can anyone provide some advice on where I am going wrong? Where exactly should I be trying to lower the power/core settings? And, am I missing something on the Memory increase side? Thanks again for the help! 20% was on core not power limit. Memory only has one state. Did surprise me tho it doesn't let you push more Did you reset wattman before you started using afterburner? - ok, how do you decrease core by 20%? not sure what that term means in relation to the fields in Wattman or Afterburner - yeah, me too, but I Afterburner did let me increase memory to the recommended 1850... but still no change in hash-rate - no, I have not done anything like a reset to wattman either before or after using afterburner, but can give that a try if needed Also wanted to clarify something... I have been making these changes real-time while the GPUs are mining. Is this OK or do I need to stop Claymore while changing then re-launch every time to see if it had an impact? |
|
|
|
Big shout out to Phil for sending me the extra molex cable I needed to get my 4th 470 up and running! Thanks also for the extra goodies you included Phil... really appreciate it!
So, as it stands now, I am getting just shy of 82 Mh/s on ETH and about 1300 Mh/s on DCR. I am happy to finally have this rig running with 4 GPU's but I am still baffled at how these MSI 470's are only getting 20.7 per unit on ETH. I have tried pretty much everything suggested in this forum and in the Claymore readme file but nothing notably effects the speed of ETH mining. I've done the following:
- Added the 5 setx lines to my .bat file
- Added the "-ethi xx" command within the command line of the .bat file. Tried it at 10, 12, 14 & 16.
- Added the "-ethi xx" command to the config file. Tried it at 10, 12, 14 & 16 in various places.
- Increased my Virtual Memory from 16GB to 24GB
None of these moved the needle on ETH mining at all... still all 4 GPU's averaging 20.7 per unit. The only thing that effected ETH speed at all, was to ratchet down DCR intensity significantly but the minimal gain of .4% (from 20.72 to 20.82) would actually be more than offset by the dramatic decrease in DCR effectiveness. On the flip-side, bumping up DCR to 35 increased its hash-rate by ~15% but only nominally effected ETH mining (less than .1%). Bottom line, nothing I've have tried so far is getting me even remotely in the ballpark of the ~23 Mh/s that others are getting from those GPU's out of the box.
I would almost just think I got a bad batch, but my 4 GPUs came from 2 different vendors (3 from Amazon and 1 from NewEgg), spread over 3 orders over 2-3 weeks, so this seems unlikely. I really wanted to close this gap before proceeding with modding, but it's looking bleak at this point. I am putting off purchasing a 5th GPU for this rig until I can figure out exactly how much I can improve its current performance. Frankly, at less than 21 per card, it's barely going to be profitable, especially at the ridiculous power draw it is currently at... which reminds me...
So, I have my rig plugged into a KilaWatt unit. While running the rig with 3 GPUs, it was pulling around 630-640 watts. This seemed high to me because I was thinking max of 330 for the GPUs +140 for the i7-6800k, which left about 160 watts for the Mobo, SSD, heatsink, etc. Again, seemed a bit high but not out of the realm of possibilities.
Well, here's the perplexing thing... after hooking up the 4th GPU, KilaWatt now says I am using 810 watts! That's a 170 watt bump over the 3 GPU set-up. So, how exactly can a GPU with a max TPD of 110 watts add 170 watts of power usage? At this rate, I couldn't even add another GPU to this rig as my PSU is 1kw. I know I need to really dig into the power settings that have been mentioned here recently but was wondering if anyone could explain how a 110w GPU could add 170w to my power draw.
Thanks again for any tips!
I would push down the core 20% Up the memory to 1850 That will up the MHs and lower the power without bios mod 120tdp isnt the max power draw. Also msi cards have a high power set than the other 470s If you want help with bios mod I will help anyway I can If you pay 10cents or more in power I wouldn't dual mine. Your only talking about maybe 10-20$ profit Pushing the cards harder and hotter for only 20$ ehh I'll save them Yeah, my electricity is 11 cents, so doing rough math for 110 watts, cost me about $9/month. At best with 4 GPU's on DCR, I could mine maybe 15 in a month, so at DCR current rates, maybe $5 profit/month at best so really not worth it. Now, if it goes $2 and above then certainly worth looking at but not really at $1. I am curious on your 2 suggestions... can these settings be done within the AMD WattMan App? I haven't really dug into that yet but if it is indeed managed there, I will certainly try that ASAP! Yes those settings are for wattman. It'll be the easiest way to play around with OK, so I attempted to make the recommended changes this morning but not really seeing any results, so I am probably doing something wrong. Here is what I tried in Wattman: - decreased Power Limit (in the Temp section) by 20% - increased the Memory Frequency from 1650 to 1700 (I know 1850 was recommended but the max value I could set in Wattman was 1700, and only "State 1" was enabled (on all 4 cards) so, only 1 State moved and only up 50 Mhz I applied this changes on all 4 GPUs. There was NO impact to hash-rate. There was impact to wattage usage but in the wrong direction... went from about 695 to 710. When I moved the Power Limit back to anything between -10% and 0 (flat), I got back down into the 695 range. After this, I opened Afterburner. Also my first experience with it so still early in the learning curve on all of this stuff. - Did further testing with Power Limit but same results as with Wattman... no impact between -10% & flat but increased wattage pull with -20% - In this app, Memory Clock was movable all the way up to around 2000, so I tried the 1850 and applied to all 4 GPUs, but alas, again, zero impact to hash-rate on any of the 4 GPUs. The other thing I briefly played with (but with scary results) was the GPU frequency in Wattman. I saw a youtube video where they were pulling down the state 6 & 7 frequency max values to around 970 & 1000 respectively. I tried that on GPU#1 (not a large decrease from the reflected values) and it actually lowered my wattage a bit. But, when I did the exact same thing to GPU#2 (this was actually a larger % decrease from the reflected values as this card seems to run hotter & pull more wattage than the others), it ended up stopping that GPU and crashing my system. After that, I didn't mess with the GPU settings again. So, given all of this, can anyone provide some advice on where I am going wrong? Where exactly should I be trying to lower the power/core settings? And, am I missing something on the Memory increase side? Thanks again for the help! |
|
|
|
Big shout out to Phil for sending me the extra molex cable I needed to get my 4th 470 up and running! Thanks also for the extra goodies you included Phil... really appreciate it!
So, as it stands now, I am getting just shy of 82 Mh/s on ETH and about 1300 Mh/s on DCR. I am happy to finally have this rig running with 4 GPU's but I am still baffled at how these MSI 470's are only getting 20.7 per unit on ETH. I have tried pretty much everything suggested in this forum and in the Claymore readme file but nothing notably effects the speed of ETH mining. I've done the following:
- Added the 5 setx lines to my .bat file
- Added the "-ethi xx" command within the command line of the .bat file. Tried it at 10, 12, 14 & 16.
- Added the "-ethi xx" command to the config file. Tried it at 10, 12, 14 & 16 in various places.
- Increased my Virtual Memory from 16GB to 24GB
None of these moved the needle on ETH mining at all... still all 4 GPU's averaging 20.7 per unit. The only thing that effected ETH speed at all, was to ratchet down DCR intensity significantly but the minimal gain of .4% (from 20.72 to 20.82) would actually be more than offset by the dramatic decrease in DCR effectiveness. On the flip-side, bumping up DCR to 35 increased its hash-rate by ~15% but only nominally effected ETH mining (less than .1%). Bottom line, nothing I've have tried so far is getting me even remotely in the ballpark of the ~23 Mh/s that others are getting from those GPU's out of the box.
I would almost just think I got a bad batch, but my 4 GPUs came from 2 different vendors (3 from Amazon and 1 from NewEgg), spread over 3 orders over 2-3 weeks, so this seems unlikely. I really wanted to close this gap before proceeding with modding, but it's looking bleak at this point. I am putting off purchasing a 5th GPU for this rig until I can figure out exactly how much I can improve its current performance. Frankly, at less than 21 per card, it's barely going to be profitable, especially at the ridiculous power draw it is currently at... which reminds me...
So, I have my rig plugged into a KilaWatt unit. While running the rig with 3 GPUs, it was pulling around 630-640 watts. This seemed high to me because I was thinking max of 330 for the GPUs +140 for the i7-6800k, which left about 160 watts for the Mobo, SSD, heatsink, etc. Again, seemed a bit high but not out of the realm of possibilities.
Well, here's the perplexing thing... after hooking up the 4th GPU, KilaWatt now says I am using 810 watts! That's a 170 watt bump over the 3 GPU set-up. So, how exactly can a GPU with a max TPD of 110 watts add 170 watts of power usage? At this rate, I couldn't even add another GPU to this rig as my PSU is 1kw. I know I need to really dig into the power settings that have been mentioned here recently but was wondering if anyone could explain how a 110w GPU could add 170w to my power draw.
Thanks again for any tips!
I would push down the core 20% Up the memory to 1850 That will up the MHs and lower the power without bios mod 120tdp isnt the max power draw. Also msi cards have a high power set than the other 470s If you want help with bios mod I will help anyway I can If you pay 10cents or more in power I wouldn't dual mine. Your only talking about maybe 10-20$ profit Pushing the cards harder and hotter for only 20$ ehh I'll save them Yeah, my electricity is 11 cents, so doing rough math for 110 watts, cost me about $9/month. At best with 4 GPU's on DCR, I could mine maybe 15 in a month, so at DCR current rates, maybe $5 profit/month at best so really not worth it. Now, if it goes $2 and above then certainly worth looking at but not really at $1. I am curious on your 2 suggestions... can these settings be done within the AMD WattMan App? I haven't really dug into that yet but if it is indeed managed there, I will certainly try that ASAP! |
|
|
|
Thanks for the tip generalt... I just tried that and it didn't really impact ETH speed... went from like 20.72 to 20.75 per card... BUT, along the lines of what Phil just mentioned, shutting off dual-mining cut my power draw from 810 to just under 700! I will have to do the math on the DCR piece but I doubt it makes sense with its current rate at only $1. Also, seemed that my card temps were down as well so yet another reason to probably solo mine. Yeah thanks again Phil, I especially liked the fan and will put it to good use. Don't need the extra SATA cable at the moment but could certainly put that to good use in the future when I build my dream personal system with multiple SSDs & HDDs I hear you on the heat issue... not sure how you can sleep in 82 lol, I have trouble sleeping in anything over low 70's. I have my current rig in my garage, which has no ventilation but is very cool most of the time so no issue, at least with 1 rig. Since I am in a cool weather climate in Oregon, I am really considering building a 2nd rig and placing it in my main living area for heating. Maybe it won't 100% offset what I would normally pay for extra heat but I bet it would at least come close, which means better profit margin. Fact is that I need heating of some degree 8-9 months of the year here, so definitely something to factor into the equation. |
|
|
|
|
Big shout out to Phil for sending me the extra molex cable I needed to get my 4th 470 up and running! Thanks also for the extra goodies you included Phil... really appreciate it!
So, as it stands now, I am getting just shy of 82 Mh/s on ETH and about 1300 Mh/s on DCR. I am happy to finally have this rig running with 4 GPU's but I am still baffled at how these MSI 470's are only getting 20.7 per unit on ETH. I have tried pretty much everything suggested in this forum and in the Claymore readme file but nothing notably effects the speed of ETH mining. I've done the following:
- Added the 5 setx lines to my .bat file
- Added the "-ethi xx" command within the command line of the .bat file. Tried it at 10, 12, 14 & 16.
- Added the "-ethi xx" command to the config file. Tried it at 10, 12, 14 & 16 in various places.
- Increased my Virtual Memory from 16GB to 24GB
None of these moved the needle on ETH mining at all... still all 4 GPU's averaging 20.7 per unit. The only thing that effected ETH speed at all, was to ratchet down DCR intensity significantly but the minimal gain of .4% (from 20.72 to 20.82) would actually be more than offset by the dramatic decrease in DCR effectiveness. On the flip-side, bumping up DCR to 35 increased its hash-rate by ~15% but only nominally effected ETH mining (less than .1%). Bottom line, nothing I've have tried so far is getting me even remotely in the ballpark of the ~23 Mh/s that others are getting from those GPU's out of the box.
I would almost just think I got a bad batch, but my 4 GPUs came from 2 different vendors (3 from Amazon and 1 from NewEgg), spread over 3 orders over 2-3 weeks, so this seems unlikely. I really wanted to close this gap before proceeding with modding, but it's looking bleak at this point. I am putting off purchasing a 5th GPU for this rig until I can figure out exactly how much I can improve its current performance. Frankly, at less than 21 per card, it's barely going to be profitable, especially at the ridiculous power draw it is currently at... which reminds me...
So, I have my rig plugged into a KilaWatt unit. While running the rig with 3 GPUs, it was pulling around 630-640 watts. This seemed high to me because I was thinking max of 330 for the GPUs +140 for the i7-6800k, which left about 160 watts for the Mobo, SSD, heatsink, etc. Again, seemed a bit high but not out of the realm of possibilities.
Well, here's the perplexing thing... after hooking up the 4th GPU, KilaWatt now says I am using 810 watts! That's a 170 watt bump over the 3 GPU set-up. So, how exactly can a GPU with a max TPD of 110 watts add 170 watts of power usage? At this rate, I couldn't even add another GPU to this rig as my PSU is 1kw. I know I need to really dig into the power settings that have been mentioned here recently but was wondering if anyone could explain how a 110w GPU could add 170w to my power draw.
Thanks again for any tips!
|
|
|
|
|
I could totally be wrong on this, but for some reason I have always assumed -ethi was a value between 1-10. Like I said though, I could totally be wrong.
[/quote]
Actually, -ethi accepts a range of values from 0 to 16, with default being 8. There are others who have the value set at 10, 12, etc, which is what I am trying to accomplish. So, still looking for the methodology to implement the "-ethi" command.
On a side note, the default intensity for DCR seems to be 30. At that setting, I was getting about 310 per GPU and increasing it to 32 resulted in about 331 per GPU... so about 3% increase per notch. I don't expect the ETH impact to be that drastic per notch increased, but these GPU's should be getting at least 2-3 more Mh/s than their current 20.7 before any mods. So, hoping that increasing ETH intensity will do the trick.
Thanks!
[/quote]
Thank you! You learn something new every day I guess. I might have to play around with some of my cooler cards now....Woohoo
Back to your original question though.
"If there are two or more options in the command line, miner will take all options from the command line, not from configuration file."
So what happens if you specify -ethi 14 in your command to start claymore instead of the config file?
[/quote]
Well, I have not tried adding it into the .bat file since I don't know enough about this to know whether that is a viable option or not. I guess I could try it but i really don't want to screw something up by just blindly attempting something.
I would assume that there is probably a way to enter this command directly in the into the command window that Claymore is running in (especially since that is what the + & - signs due to DCR intensity), but it's just a matter of finding someone who has actually done it to learn exactly how.
|
|
|
|
Hello everyone... I just starting using v7.2 on my first build, currently with 3 MSI RX 470 4GB GPUs. I was a bit intimidated in moving away from a GIU-based miner since I am a noob, but through a lot of help from this forum and youtube, was able to get it up and running successfully this weekend!
I am dual mining with DCR, but only getting about 20.7 Mh/s on ETH per GPU. I know of people getting in the ~23.5 range on these GPUs with no mods, so I believe i need to adjust the ETH intensity with CDM. I understand the command to do this is "-ethi xx", the "xx" being the intended intensity... set at 8 default, with a max possible value of 16.
Here's my issue: since I am new to this, I can't figure out how to implement that -ethi command. When I try to type it directly into the command line (cursor at the bottom of the command window where CDM is running), it instead only decreases DCR intensity (since it seems the "-" & "+" are short-cut keys to affect DCR only). I also trying pasting "-ethi 12" into the command line cursor, but it didn't work... that actually shut down 2 of my 3 GPUs.
Through recommendation of another forum member, I added this command into my config.txt file as follows:
#-epool us1.ethpool.org:3333
#-ewal 0xD69af2A796A737A103F12d2f0BCC563a13900E6F
#-epsw x
#-dpool stratum+tcp://yiimp.ccminer.org:4252
#-dwal DsUt9QagrYLvSkJHXCvhfiZHKafVtzd7Sq4
#-dpsw x
#-esm 1
#-mode 0
#-tt 70
-ethi 14
My understanding is that this file is checked when CDM is launched and all lines with "#" are ignored, so I did not type the "#" in front of the "ethi 14"
I then saved this file, and relaunched CDM but it had absolutely no impact, still mining at exact same rate for ETH.
I also added the 5 setx lines to the top of my .bat file, which now looks like this:
setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100
EthDcrMiner64.exe -epool us2.ethermine.org:4444 -ewal 0x1a132440ede1a4b3f4d56d64063ef4cab3f55abc -epsw x -dpool stratum+tcp://dcr.suprnova.cc:2252 -dwal gerrox.jworker1 -dpsw x
So, my question is... how exactly can I use that "-ethi xx" command to increase ETH mining intensity?
Thanks!
I could totally be wrong on this, but for some reason I have always assumed -ethi was a value between 1-10. Like I said though, I could totally be wrong. Actually, -ethi accepts a range of values from 0 to 16, with default being 8. There are others who have the value set at 10, 12, etc, which is what I am trying to accomplish. So, still looking for the methodology to implement the "-ethi" command. On a side note, the default intensity for DCR seems to be 30. At that setting, I was getting about 310 per GPU and increasing it to 32 resulted in about 331 per GPU... so about 3% increase per notch. I don't expect the ETH impact to be that drastic per notch increased, but these GPU's should be getting at least 2-3 more Mh/s than their current 20.7 before any mods. So, hoping that increasing ETH intensity will do the trick. Thanks! |
|
|
|
|
Hello everyone... I just starting using v7.2 on my first build, currently with 3 MSI RX 470 4GB GPUs. I was a bit intimidated in moving away from a GIU-based miner since I am a noob, but through a lot of help from this forum and youtube, was able to get it up and running successfully this weekend!
I am dual mining with DCR, but only getting about 20.7 Mh/s on ETH per GPU. I know of people getting in the ~23.5 range on these GPUs with no mods, so I believe i need to adjust the ETH intensity with CDM. I understand the command to do this is "-ethi xx", the "xx" being the intended intensity... set at 8 default, with a max possible value of 16.
Here's my issue: since I am new to this, I can't figure out how to implement that -ethi command. When I try to type it directly into the command line (cursor at the bottom of the command window where CDM is running), it instead only decreases DCR intensity (since it seems the "-" & "+" are short-cut keys to affect DCR only). I also trying pasting "-ethi 12" into the command line cursor, but it didn't work... that actually shut down 2 of my 3 GPUs.
Through recommendation of another forum member, I added this command into my config.txt file as follows:
#-epool us1.ethpool.org:3333
#-ewal 0xD69af2A796A737A103F12d2f0BCC563a13900E6F
#-epsw x
#-dpool stratum+tcp://yiimp.ccminer.org:4252
#-dwal DsUt9QagrYLvSkJHXCvhfiZHKafVtzd7Sq4
#-dpsw x
#-esm 1
#-mode 0
#-tt 70
-ethi 14
My understanding is that this file is checked when CDM is launched and all lines with "#" are ignored, so I did not type the "#" in front of the "ethi 14"
I then saved this file, and relaunched CDM but it had absolutely no impact, still mining at exact same rate for ETH.
I also added the 5 setx lines to the top of my .bat file, which now looks like this:
setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100
EthDcrMiner64.exe -epool us2.ethermine.org:4444 -ewal 0x1a132440ede1a4b3f4d56d64063ef4cab3f55abc -epsw x -dpool stratum+tcp://dcr.suprnova.cc:2252 -dwal gerrox.jworker1 -dpsw x
So, my question is... how exactly can I use that "-ethi xx" command to increase ETH mining intensity?
Thanks!
|
|
|
|
Thanks but I think I may have mis-worded something. It is not that I cannot get a driver installed. In fact, i was able to "update" from v16.30 (installed a few weeks ago via the MSI DVD that came with my 470's) to v16.40 without any problem. What I don't get is how the download from AMD that is labelled everywhere as Crimson 16.9.2 can install an old driver version??
Also, yes I am on Windows 7 Pro x64 and I believe I already did a clean install... meaning, I uninstalled v16.30 prior to installing what should have been v16.9.2... but ended up being 16.40.
About the -ethi command in CDM... not sure what you mean by making the change in the config file. I looked at all the contents of the config file and this is mostly a repeat of things that I already changed in the .bat file. There is nothing in there at all containing "-ethi"... but, are you saying that this command should be ADDED to this file?
something like this maybe?
-ethi 16
#-epool us1.ethpool.org:3333
#-ewal 0xD69af2A796A737A103F12d2f0BCC563a13900E6F
#-epsw x
#-dpool stratum+tcp://yiimp.ccminer.org:4252
#-dwal DsUt9QagrYLvSkJHXCvhfiZHKafVtzd7Sq4
#-dpsw x
#-esm 1
#-mode 0
#-tt 70
yes you add it into the config file this is my file -epool us1.ethpool.org:3333 -ewal 0x66d7B21286cbcfE5eEBDA87A8E6Eee22b31F707A.Rig1 #-tt -50 #-epsw x #-dpool stratum+tcp://lbry.pool.mn:8878 #-dwal xleejohnx.rig1 #-dpsw x #-esm 1 -mode 1 #-powlim 0 -gser 2 -etha 0 #-dcri 20 #-dcoin lbc #-ethi 12 this is my start.bat file timeout 60 setx GPU_MAX_ALLOC_PERCENT 100 EthDcrMiner64.exe Thanks for that clarification. So, i went ahead and added "-ethi 14" to the bottom of my config file. I also added that several setx rows recommended in the readme (including the one above that you have). They seemed to load fine at startup. But alas, still maxing out right at 62. I would almost think that maybe its just a difference between brands of GPUs, but others on this thread are using the exact same MSI 470's and getting 23-ish out of the box, so I must still be missing something. I will take a fresh look tomorrow as I am burnt out on this stuff for today. At least I should be getting my molex connector in the mail tomorrow so I can finally hook up my 4th 470! |
|
|
|
|
Radeon Software Version - 16.9.2
Radeon Software Edition - Crimson
Graphics Chipset - Radeon (TM) RX 470 Graphics
Memory Size - 4096 MB
Memory Type - GDDR5
Core Clock - 1242 MHz
Windows Version - Windows 7 (Service Pack 1) (64 bit)
System Memory - 16 GB
CPU Type - Intel(R) Core(TM) i7-6800K CPU @ 3.40GHz
Radeon Settings Version - 2016.0628.2138.37120
Driver Packaging Version - 16.40.2311-160916a-307333C
Provider - Advanced Micro Devices, Inc.
2D Driver Version - 21.19.137.1
Direct3D® Version - 21.19.137.1
OpenGL® Version - 21.19.137.1
OpenCL™ Version - 21.19.137.1
AMD Mantle Version - 21.19.137.1
AMD Mantle API Version - 98310
AMD Audio Driver Version - 7.12.0.7723
Vulkan Driver Version - 9.1.0.10.21.0
OK, so i just noticed something and maybe I do indeed have 16.9.2 installed after all. I was looking at the Software tab with the Radeon settings app and the Driver Packaging version says 16.40.2311. BUT... I just noticed that on the Overview tab, the Radeon Software Version does indeed show 16.9.2. So, does this mean I am ok? or is that 16.40 showing on the Driver label still mean I don't have the latest?
If I am on the latest, I got no bump at all vs my previous drivers and my ETH HR for 3 GPUs is still running almost exactly at 62 Mh/s.
|
|
|
|
Thanks toptek... that 2nd link you have is to the exact file I downloaded twice. But for what ever reason, after installing the display driver in that packet, my driver shows as v16.40, not the v16.9.2 that it should. I honestly can't understand how this is possible but hopefully someone has experienced the same issue and figured out how to fix it. |
|
|
|
|
Thanks but I think I may have mis-worded something. It is not that I cannot get a driver installed. In fact, i was able to "update" from v16.30 (installed a few weeks ago via the MSI DVD that came with my 470's) to v16.40 without any problem. What I don't get is how the download from AMD that is labelled everywhere as Crimson 16.9.2 can install an old driver version??
Also, yes I am on Windows 7 Pro x64 and I believe I already did a clean install... meaning, I uninstalled v16.30 prior to installing what should have been v16.9.2... but ended up being 16.40.
About the -ethi command in CDM... not sure what you mean by making the change in the config file. I looked at all the contents of the config file and this is mostly a repeat of things that I already changed in the .bat file. There is nothing in there at all containing "-ethi"... but, are you saying that this command should be ADDED to this file?
something like this maybe?
-ethi 16
#-epool us1.ethpool.org:3333
#-ewal 0xD69af2A796A737A103F12d2f0BCC563a13900E6F
#-epsw x
#-dpool stratum+tcp://yiimp.ccminer.org:4252
#-dwal DsUt9QagrYLvSkJHXCvhfiZHKafVtzd7Sq4
#-dpsw x
#-esm 1
#-mode 0
#-tt 70
|
|
|
|
|
Show Posts
