nicehash
"out of memory" while using Neoscrypt Algorithm
Hello there! Nicehash is working fine for me except when it's mining on Neoscrypt algorithm. My rig only has 24 gb ram and I can't use more than 6 gtx 1070 for Neoscrypt mining. Whenever I try more than 3 gpu it always end up with the same Error message : Out of memory. I tried changing the virtual memory but it seems it doesn't change anything ? How much RAM does Neoscrypt require to run smoothly? I have 13 GTX 1070's running on Asus B250 Mining Expert with 24 gb ram (8+16) Regards
Hi
I have been struggling with the same issue for more than three weeks. My rig has 06 gtx 1070 and whenever Itried to mine on neoscrypt i have got errors such out of memory , CUDA error on majority of my cards and sometimes random reboot or freezes. I am running on windows10 x64 1709. and nvidia driver 384.76 (this one seems to be the most stable for me) and my bios is up to date .
I decided to track the issue cause i was suspecting a riser default. so I launched each GPU one by one and tested with equihash, Lyrarev2 and then neoscrypt to detect any drop of performances or power.
When testing my GPU3 which is mounted on riser connected on pcix8 n°2 with neoscrtypt ....I just noticed that the Vram usage was huge, around 5.7GB ....per card!!!!! note that the five other cards were on idle and not mining but their Vram just jumped to 5.7GB
Tested the other cards on neoscrypt the Vram didn't exceed 96MB per card . So I tested nicehash 2 with 5 cards on neoscrypt and GPU 3 on equihash and Lyrarev2 (disabling neoscrypt). the Vram usage was 1.139GB and the rig was more stable. However after two hours found the rig freezed.
I was really disappointed. I tried then Nicehash Legacy with disabling neoscrypt from GPU3 ....EUREKA the rig has been mining for 3 hours now and seems stable ....with very lownusage in cpu 40% (i5 6400).
Hope this will solve your problem with trying to detect which GPU is over using Vram...don'tknow if this is driver, hardware(GPU, motherboard or riser ) or Nicehash/neoscrypt related issue . but 5.7GB vram (* 13 in your case) is unmanageable even with more virtual memory settings.
Regards
Hi, there.
I think I found a solution this problem, at least for my rig....hope this gonna solve your problem also here is my config: (the one causing problems) MSI Z270 A PRO bios 1.51 (latest update) I5 6400 12GB DDR4 6x GTX 1070 (1 Aorus+ 5Strix OC) on latest nvidia driver at present time 390.xx Windows 10 x 64 pro build 1709 SSSd 120GB 2x V750 cooler master PSU.
the other one is the same only but the cpu G3950(7th Gen) this one was rock stable on 24/7 mining
So I decided to use the same settings for soft (BIOS /windows/ driver) that are onthe stable onthe faulty one
My first step was to export the bios settings from the healthy to the sick rig
several changes were made (see attachement)
I think (and Ican be wrong ) the crucial parameter is the FCLK as it was originally set to 800Mhz on the faulty rig ...which if we consider the southbridge frequency can be determinent for such heavy usage on the motherboard. Anyway these changes made my rig rock stable and it is running now for 48h nonstop mining on neoscrypt
My only question was why the two rigs were not set up in the sam eway as they have the exact same bios and I don't recall when I set up the FCLK??? may be the CPU generation had arole on that ? Regards.