SLOWER hashrates when virtual memory enabled ...

[theagemaway]

Hello all,

Not sure if this is the right place, so please flame me if necessary

I just started a 2x Vega56 rig using windows 10 and a Biostar TB85 mobo, 8GB of crucial DDR3L RAM, and a 200GB HDD. I'm currently getting about 5 Kh/s (**using cryptonight-fast** so i think this is approximately normal) from just one 56 running with virtual memory disabled. No BIOS mods, nothing fancy, no undervolting, no overclocking, just windows 10 and castXMR. I'm currently getting about 5 "CL_MEM_OBJECT_ALLOCATION_ERROR"s per second, but the gpu seems to be running stable. The other one is idle right now, until I can get the first one functioning properly. I have tried using both gpus, and I get about 100 errors per second (literally pages and pages per second), but it says I'm getting 13 Kh/s running both together. I only did this for a couple minutes.

Now enough details, on to the meat of the problem. I've seen in several threads and several posts that the answer to the allocation error is to increase my virtual memory allocation to 16, 32, even 64 GB. But anytime I enable virtual memory, my hashrate literally goes down to 10%. I'm getting about 400 h/s on each gpu. No errors, so I guess it solved my problem, but I don't think I should be giving up 90% of my hashrate. I've tried 16, 32, 64GB page files on my C drive (HDD), as well as on an external HDD (USB 3.0), but everything kills the hashrate down to 400 h/s each.

Has anyone seen something like this before? Any suggestions?

Thanks in advance! Best regards,
theagemaway

[1715533242]

1715533242

[jmigdlc99]

Welcome to the forums. Your problem is "No BIOS mods, nothing fancy, no undervolting, no overclocking." All of those are needed to ensure you get optimal hashrates. The AMD driver software, especially for Vega GPUs, is buggy and you really can't expect it to work the way you intend it to. So the reason for your problem could be a multitude of errors.

The simplest and probably only solution is to configure it correctly. If you are a fan of CastXMR then follow his guide: http://www.gandalph3000.com/cast_xmr/cast-xmr-optimized-cryptonight-miner-for-rx-vega/

[toptek12]

CastXMR 1.2.5 is out it runs better and make sure the VM Is set to 32000 at least or you might get red  kernel errors it keeps mining but will shut down at some point if that happens set the VM to 32000 so I found out the hard way even tested it below that . 30000 might work just add the extra 2000 it won't really mater .

Updated on June 19th, 2018 to version 1.2.5
http://www.gandalph3000.com/

[SpceGhst]

Running 6 RX 480s, I had to set virtual to 32000/64000 in Cast to keep from getting the red openCL errors.

[Mamasita22]

Virtual memory is needed for well-working amd cards. Try other miner like cast-xmr, as stated above

[theagemaway]

Thanks for all of the replies.

CastXMR 1.2.5 is out it runs better and make sure the VM Is set to 32000 at least or you might get red  kernel errors it keeps mining but will shut down at some point

Just as you said, I left it running a single one for a while, and it looked normal, except it hadn't received any shares after about the first 5 minutes. I am using 1.2.5, and I did try some older versions as well, with no luck.

I'm currently redoing a fresh reinstall of win 10 pro, and following this guide: http://vega.miningguides.com/

I was up til about 2AM getting as far as I could, but I had to stop and sleep. Going to continue today. Praying for a success! I keep updating as the situation changes.

[theagemaway]

Update:

I followed the guide I linked earlier, as closely as I could. Using the "starting values" from OverdriveNTool gave me immediate artifacting and BSOD within a couple minutes, so I went back to stock values on the core/mem clocks, and kept the fan speeds high. I was able to get Cast XMR to do cryptonight-fast (masari coin) at about 3500 h/s, but not stable longer than 5-10 minutes. So I went to xmr-stak and I've had it doing about 3300 h/s and stable for the last 6 hours.

Unfortunately, I could only get this result using one GPU. Any combination I've tried so far of using the second GPU, I get immediate artifacting. So I'm going to enjoy my small piece of victory pie for tonight, let it mine, and tomorrow I'll start experimenting with swapping GPUs into different PCIe lanes, swapping out different risers, etc. If anyone has suggestions, I would love to try them. To be clear, right now my goal isn't to maximize performance of these GPUs. It's simply to OBTAIN performance on both of them. I'd cry tears of joy right now to have them both running at even 75% of what they're capable of. Once I have that, then I can shoot for the stars.

Thanks again guys.
Theagemaway