Topic: Any word on amd vega hash rates? - page 70. (Read 202748 times)

Are these Vega 56 or 64? When I was using Stak on my 56s, I had my second thread worksize at 1600.

I might suggest testing out CastXMR, even if just long enough to rule out a configuration issue versus a system issue.


Interesting. What's your physical RAM usage when you have a workload? You prompted me to look at my own system, and I am noticing that with 4 GPUs, my VirtMem commit is only 32GB... maybe we only need 8GB Virtual per card?

Edit: Actually I'm even more confused. Maybe I'm doing something wrong, but following this guide to check my pagefile usage reveals 0 actual use. I wonder how we got here to thinking that we needed a high Virtual Memory counter for things to work properly.

Hmm, I have 5 vega 56s running no problem on a 64GB SSD.  Virt size is 32GB, and I have 16GB of physical.   Getting 1800+ on ea

Hasn't fixed my problem sadly.

Have the Virtual Memory set to 48GB with 3 Vegas. Set my threads to:

   { "index" : 0, "intensity" : 2016, "worksize" : 8, "affine_to_cpu" : false },
   { "index" : 0, "intensity" : 1700, "worksize" : 8, "affine_to_cpu" : false },
   { "index" : 1, "intensity" : 2016, "worksize" : 8, "affine_to_cpu" : false },
   { "index" : 1, "intensity" : 1700, "worksize" : 8, "affine_to_cpu" : false },
   { "index" : 2, "intensity" : 2016, "worksize" : 8, "affine_to_cpu" : false },
   { "index" : 2, "intensity" : 1700, "worksize" : 8, "affine_to_cpu" : false }

And am getting "cl out of host memory" errors.

On a dedicated mining rig, that's easily done with a 128GB M.2 chip, like my favorite  <$60 ADATA.

It's only 118GB usable, but after Windows itself, and setting my VirtMem to 80GB (adding a 5th GPU this week), I still have 7 GB left. You'd have to be a little creative (stripping out Windows components) to get enough for 6, I suppose, but I'm sure it could be done. Of course, the 256GB is only $90 (and traditional SSDs are $80, but I really love the compact 'package' by using M.2).

Tried that repeatedly..


So I just ran out and grabbing 16GB of physical RAM (up from 4GB) and now I'm getting 1 running steady at 2000+. Holy cow 16 per VEGA? I think I set like 40 at the moment for 2 but I'll check. Now I'm on trying to get the second one to work..

Hello all,

I am building my first mining Rig with the below components.

6x Sapphire RX Vega 56 air cooled
Asrock Fatal1ty B250 Gaming K4 motherboard
Intel Celeron G3930 Kaby Lake Dual-Core 2.9 GHz LGA 1151 Processor
2x G.SKILL NT Series 4GB DDR4 SDRAM 2133 Ram
1x EVGA 600w Bronze PSU for motherboard, cpu, and risers
2x HP DPS-1200w with Breakout Board for the cards
6x UBIT PCIE risers
Windows 10 Pro 64-bit

For simplicity sake lets number the pcie slots on the motherboard 1-6 right to left starting with the one closest to the CPU, slot 2 and 5 are full size PCI slots the rest are PCIE extension slots.  Same thing with the GPUs, 1-6 right to left, each one is meant to be plugged in to the corresponding PCIE slot on the motherboard. I assembled everything except I left the PCIE risers unplugged from the Motherboard for now and loaded windows onto the HDD.  Once that was done I plugged in all the cards and loaded AMD driver 17.30.15.  It recognized 4 of the cards after restart but 2 of them were getting code 43. 

At this point I realized I had forget to set my Motherboard Bios so I followed this guide: https://blockoperations.com/gpu-miner-bios-settings-asrock-b250-gaming-k4-motherboard/ Afterwards I was getting 1 card.  I decided to start over and try to load each driver individually. 

Card 1 worked intermittently in PCIE slot 1 I was able to reload the driver on to it.  When placed into slot 2 the first fullsize slot it worked everytime with no problem.  Thinking it may be a power issue to explain the intermittent connectivity I swapped PSU and still nothing, then tried cables, still same issue.  I resigned myself to the intermittent connectivity and moved on to card 2.  Card 2 slot 2 no issues at all I tried it in PCI slot 1, 2, and 3 and it worked in all of them.  Card 3 was not recognized in any slot except slot 2 and when it was I attempted to load the driver and the screen would intermittently go in and out.  I would get a white noise screen from time to time as well.  Loading the driver failed multiple times and gave me error 1803.  I followed several steps to rectify that issue and finally had a successful load albeit with the screen still intermittently going in and out. 

At this point I figured I would try to run with just the 3 cards.  Card 1 in slot 1, card 2 in slot 3, and card 3 in slot 2.  On boot the rig would power up then powerdown after a few seconds, it did this twice and the third time it finally booted.  When i was finally in windows only card 2 and 3 were showing activity lights and the machine only recognized the 2.  Device manager showed 3 hidden cards claiming they were not connected. 

At this point I was informed that Driver 17.30.xx was corrupt on the Vegas and I needed to turn off quick boot and sleep mode.  I also made it so windows update will not download hardware drivers.  So I turned them off and plugged in all 6 cards.  I then rebooted in safe mode and ran DDU.  Once done I restarted and leaving all cards plugged in I ran AMd driver 17.8.2 as I was told that was a stable build.  The motherboard continued to reboot several times on start up before it booted fully.  With all cards in it would start to boot then the screen would go black.  Additionally 3 cards would show activity lights.  I let it sit with a black screen for 30 minutes before I manually shut it down and tried again.  I did this several times sometimes it would get to the windows 10 welcome screen then go black, sometimes it wouldn't.  At this point I unplugged 3 cards from the Motherboard and reset the motherboard bios back to default.  It would still start to run then stop on power up.  After resetting the bios it was only recognizing 1 card and Nicehash Miner was not recognizing a single GPU. 

At this point I stopped in frustration.  I am probably going to start over entirely.  First run DDU on all cards.  Then wipe the HDD and reinstall windows.  I will adjust windows before ever putting a GPU in.  Then update drivers on each GPU 1 by 1 only ever having 1 gpu in at a time.  I am at a complete loss as to what I am doing wrong or what is not working.  Any help at all will be appreciated.

This is possible. Current community understanding is that you should have 16GB Virtual Memory for each Vega. I made the mistake of getting 32GB of physical RAM, and I'm only using 10% (more if I have other programs running like HWinfo64) - never more than 8GB so far.

It sounds like your devcon is not resetting the driver.  Are you running as admin?

Maybe it was just coincidence, but I had a 1300 hash rate at one point until I upped virtual memory to 32GB.

In device manager disable and enable cards one by one.

I'm about to tear my hair out. I have been fighting with these damn VEGAs all weekend!

After nearly a dozen hours I have a single Vega card working in my board.

Fresh install of Fall Creative Update
Aug 23rd Blockchain Drivers
XMR CAST 0.6

HBM: 990MHz

Mining XMR at 1,400h/s

Does anyone have any ideas why I'm so low?

Yes, that one.  Set "EnableCrossFireAutoLink" to "0" after installing the drivers but before rebooting.  You have to close out of the AMD installer and not reboot; go into RegEdit.

But I don't think that is your problem.  Try a different SoftPowerPlay table -- find one that let's you reboot. 

Much appreciated - that's exactly what I needed. I shall dig into this thread with great appetite and hopefully digest and assimilate it accordingly!

https://www.reddit.com/r/MoneroMining/comments/74hjqn/monero_and_vega_the_definitive_guide/

All you need, just a little registry mods in order to take full advantage of your vegas.

Dont mind the RAM size and HBCC mem part. 4GB ram is enough and HBCC has nothing to do with mining, just disable and enable your cards from device manager after every reboot.

just read the last 20-30 pages here, and all your questions will be answered!

Ok so I'm about to get a 6x Vega56 build done on that Onda D1800 board.
I'm a little unsure about all the stuff I'm reading above. What are these mods you keep referring to? I thought oc/voltage settings could be done within the Claymore miner directly? Also, I was under the impression there was no need for BIOS mods on Vegas (esp. 56's) if you intend to mine Crypronight?

I might be misguided... Please help me clear the confusion!

apply the reg files one by one and reboot after each one. This way you will identify the card that is not accepting the settings. Make sure you do restore points before each apply, so you can easily revert it.
After you identify a problematic card - disconnect it, and continue with the rest untill you know which work with the mod and which not. After that - try a different mod, usually with lower memory settings.

thanks for sharing your experience! I have some problems with my 3rd rig. 6 x Vega on a Z170 Pro Gaming Board. in Windows is all ok, after installing my Blockchain driver -> Reboot. Everything ok. Use the reg keys for softplaytable -> cant boot into windows.

i will try to disable crossfire in front of using the softplay table reg

but i have only "Enable"crossFireAutoLink" in the reg keys, did you mean thad one?

Reinstall your drivers.

I have 3 rigs. All stable, no problem.

Use DDU to uninstall driver. Reinstall blockchain driver.