-
Notifications
You must be signed in to change notification settings - Fork 17
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
GTX 1070 - Version 1.3.8 alpha neoscrypt not work #78
Comments
on GTX 1080 - system crash. |
Are you running anything else on your cards beside excavator? That would explain the 'out of memory' error. You can try to decrease the B parameter, the size of memory needed to run will decrease as well. Let me know how it goes. Also, could you please post config file here? Thanks. |
Only the excavator has been launched, no other miners or applications Stable earned only with parameters "worker.add","params":["0","0","B=1230"] And a very large consumption of computer RAM (Not GPU RAM) - at the start of the miner is 500 megabytes. And after the start, the memory consumed is reduced by 1 megabyte per second. Several mines passed and consumed memory of 216 megabytes. |
This doesn't seem ok. Which version of drivers are you using? I will try to reproduce the problem on my machine. FYI. You're resetting workers that don't even exists. You have 5 workers (not 10). |
Win 10 Pro 64bit, nvidia driver 382
Yes, I forgot ... Config copied from another file, fixed the worker, and forgot the worker reset ... But it could not affect - the excavator is smart enough to reset the non-existent workers ;-) |
Try GTX 1080 Ti - all fine :
|
Try GTX 1080 Ti - all fine. |
@UselessGuru We're still working on optimisations, especially for 1080 and 1080 ti cards. We'll release an update soon. |
works fine for me on 1070cards with newest nvidia drivers |
Not working for me as well on both my rigs ... |
@ebbbang driver version? Which cards do you have? |
@obit8 It's definitely not ok to get this error. We're working on it. Thanks for the info. |
@dropky my card is a 1060 msi nvidia. Some times this error shows up |
@obit8 driver version? |
Can't it be the problem that sometimes algorithms use > 5GB of GPU memory and your card have only 3GB ? |
I personaly use excavator on ubuntu with gtx 1070 cards and never saw similar issues |
@kgiedrius Neoscrypt requires quite a lot of memory, on 1070 cards it's 5GB, on other cards it's less than that (usually allocates around 70% of the GPU's memory). The cuda 'out of memory' error happens because at the moment there isn't enough free memory on the GPU, the system might be using it (primary cards) or something hasn't been freed completely yet (nhm2 switching). We're working on a way around - if there won't be enough free memory the block parameter will be modified (the intensity will be decreased for a bit). Note for everyone: |
@dropky Driver version 388.4 However now i see that the escavator reboot frequently! Also without any CUDA memory error!... There is something i can do to solve this? I'm using the Nicehashminer v 2.0.1.5 BETA |
@Kayfolom i've disabled the neoscrypt from the config file! Is just annoying that you need to edit every single gpu file! But at the End the excavator started the work without error mining the "Lyra2Rev2"... Today it worked well for me for almost 10 hours with CryptoNight and Equihash...then i see 0 worker online and when i checked i saw that it continued to give me the CUDA MEMORY error trying to mine the neoscrypt! |
Got the same 'out of memory'. |
Well, for Now al is gone with nicehash.
Al BTC Wallet easy stolen.
I’m looking for another altcoin. ;(
Il giorno gio 7 dic 2017 alle 09:24 Meeh <notifications@github.com> ha
scritto:
… Got the same 'out of memory'.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#78 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ACvtKE7XDaWKOOvQ57lazByHsHw8f9JMks5s96C2gaJpZM4Qp4Fz>
.
|
Is this issue still on the radar now Nicehash is down? Have the same memory errors on my dual GTX1070 setup. |
@hapklaar Yes :) |
I assume this is not a common scenario yet, but I have a rig with 13xGTX1070 and 4xP106-100 Nvidia GPUs. Before Dec 5. it was running only the 13x1070-s, and with that config NHM2 and excavator was running perfectly stable. Now with the P106-100-s (6GB) I'm getting this same CUDA error, out of memory. Also some other crashes without noticable error message. Win10 1709, Nvidia driver 388.43, - and with all fairness all ccminer incarnations also have the same problem with neoscrypt and lyra2rev2 |
One of my miner just hit this issue too, had to restart NiceHash in the morning and missed out on 6 hours of mining. Please do one of the following:
|
I have Same error neoscrypt 'out of memory' |
To me it happens on a 1080 too. Even if I run nothing else, and simply try to start a single excavator on a single GPU. And it fails right away. Seems like a problem with parameters or extremely screwed up memory allocation, if I had to guess. |
Hi, |
Hi all and Happy New Year!!! |
Hi. Also refer to this thread. Seems there might also be a fix on the way. |
Hi Tottom! I have 8GB RAM and 40GB virtual memory. |
Just checked RAM loading under Lyra2rev2 and Equihash, interesting effect - Lyra2rev2 loads the RAM up to 48% (3.8GB) and works ok. When running Equihash it runs ok several seconds before the RAM load reaches 50%+ (4GB) and then Excavator restarts. I raised virtual memory to 80Gb but unfortunately Excavator continues to restart. |
Hi, I am experiencing the same issue of running out of memory especially with neoscrypt algorythm on one of my rigs . The second rig is the exact same except for OC settings 70% / +100/ +700 The first rig is working perfectly without any issue even with neoscrypt ( take appart if I run the 6 gpu on neoscrypt it crashes after minutes). the second rig (the strix) is stop working using neoscrypt on nicehash 2.0.1.5 with only two gpus mining in neoscrypt. i will try to clone the G1 OS and run the strix with it to see if the problem is due to software problems because seems weird that the two rigs aren't behaving the same way |
So the 11x 1080ti cards has 11GB mem each? My 8x 1070 has 8GB each. at 64GB vRAM it was still giving issues. Now also fixed on mining only Lyra2rev2, Equihash and Blake2s. |
Understood your point! Thanks a lot! |
How long before the excavator restarts? Maybe you can keep the 80GB vRAM and enable 6 cards, then 7 and then more to see how long it before it becomes unstable? Last night i started see the excavator restarting. so it seems the garbage collection is not clearing the vRAM efficiently and when i restarted the Rig, the excavator started and has been running since without the same issue. so i guess i have to restart the Rig every 24 to 30 hours... This will help to determine if you really need an additional SSD for just the Pagefile. |
M1r00, what vMEM have you set for each your Rigs. Since you have 2x Rigs with 6x 1070. I have 8x 1070(2 of them ti) and had tried it with 64GB vMEM and it still gave me Out of Memory... So i changed it to 80GB and it was stable for just over 32hours. If you are on 16GB or 32GB vMEM then make it 48GB or 64GB. see how it runs then? |
In my case if i disable all the algo exept one the miner is stable, if i enable 2 or more the miner continue to restart! I have x13 1060... Now i'm only mining lyr2v2, and when i see it on whattomine i change with Equihash!!! |
Before I got to 11x 1080ti I run 7 of them and Excavator worked stable (several days without issues) on any algo even with 2-3 algos at the same time. |
Don't think 256GB is what you should set it rigt now. |
I am setting 20 GB for each. my ssd is 120gb and still have 30gb free on my ssd so I assume I can go to 40GB. But my G1 rig is rock stable with 20 GB unlike the strix. |
might be some app config perhaps. what version of excavator are you running? |
Using nicehash 2.0.1.5...tried multiple drivers 384.76 then 388.31....the G1 is running stable with the 384.76 ....will post screenshots when @ home |
Tried 128GB and then +32GB up to 244GB unfortunately only Lyra2rev2 stays stable, at Equihash Excavator still restarting. |
That is unfortunate. I am running nVidia driver: 388.71 Excavator_Server 1.38 - https://github.com/nicehash/excavator/releases Apart from the above, i am at a loss why you are experiencing the issue :( |
Running the same versions driver and Excavator - will keep on digging) |
I had the same problem when I was mining neoscrypt solved do not try to cheat the benchmark by doing the test with a configuration and then change it to mine |
Increasing virtual memory size should solve the problem. #104 |
So I can finally reproduce this issue. When running standalone excavator everything works fine, but if there is NHM running on the system (doesn't have to be mining, it's enough that it's opened), then 'out of memory' crash occurs on the primary GPU. To avoid such issues, I suggest you to disable NeoScrypt on your primary card (only primary GPU is problematic, at least in my case) until this bug is fixed by the NHM team. Of course, the crash can still happen when using a standalone Excavator, since third party applications that are GPU consuming can crash the primary card as well, in such case just do the same - don't run neoscrypt on your primary card. If this issue persists in the future, we'll work on a different implementation of the NeoScrypt (that won't need as much memory as current one). I'm closing this (will reopen if needed), thank you all for your feedback! Keep on mining :) Btw, we released new version of excavator today - https://github.com/nicehash/excavator/releases/tag/v1.3.9a |
@dropky I have started seeing this error on v2.0.1.11 even though my primary card is my separate onboard graphics card. Has something changed in the latest build? Edit: I've not yet tried the VM increase. About to try it now but through I would mention the issue. |
@dropky Apologies, it appears that my primary card has switched to one of my mining cards. I'll apply the fixes suggested and try again. Thanks for the info. |
[17:11:24][0x00001518][info] Log started [17:11:25][0x00001518][info] core | Found CUDA device: GeForce GTX 1070 [17:11:25][0x00001518][info] core | Found CUDA device: GeForce GTX 1070 [17:11:25][0x00001518][info] core | Found CUDA device: GeForce GTX 1070 [17:11:25][0x00001518][info] core | Found CUDA device: GeForce GTX 1070 [17:11:25][0x00001518][info] core | Found CUDA device: GeForce GTX 1070 [17:11:25][0x00001518][info] api | Listening on 0.0.0.0:33333 [17:11:25][0x00001518][info] http | Listening on 0.0.0.0:38080 [17:11:25][0x00001518][info] core | Initialized! [17:11:26][0x00002300][info] net | Connecting to 169.50.175.203:3341 (neoscrypt.eu.nicehash.com) [17:11:26][0x00002300][info] net | Connected! [17:11:26][0x00002300][info] net | Authorized as 19M9g3vvfyDfUhdekwokAVanKdiTtSCmF2.node07 [17:11:26][0x00002300][info] algo-neoscrypt | New job_0 '000000096d78dbc0', diff=0.0078125 [17:11:28][0x00000774][info] wrkr1-1 | Algorithm: CUDA-neoscrypt parameters: B=1410 [17:11:29][0x00002b8c][info] wrkr3-3 | Algorithm: CUDA-neoscrypt parameters: B=1410 [17:11:29][0x00000d58][info] wrkr2-2 | Algorithm: CUDA-neoscrypt parameters: B=1410 [17:11:29][0x00002008][info] wrkr0-0 | Algorithm: CUDA-neoscrypt parameters: B=1410 [17:11:29][0x00002814][fatal] wrkr4-4 | CUDA error 'out of memory' in func 'cuda_neoscrypt::init' line 1258 [17:11:29][0x00001518][info] Shutting down [17:11:30][0x00001518][info] api | Closing [17:11:30][0x00001518][info] http | Closing
Overclocking disabled :
{"id":1,"method":"device.set.tdp","params": ["0","100"]},
{"id":1,"method":"device.set.core_delta","params": ["0","0"]},
{"id":1,"method":"device.set.memory_delta","params":["0","0"]},
The text was updated successfully, but these errors were encountered: