I thought I heard the cards will all be reference design/cooling. If there's going to be limited supply there's little desire for partners to redesign the board/cooling.

Actually that does ring a bit of a bell. Gee that kinda sucks.
Looks like there's also bugger all around...at least for now.

https://www.techradar.com/news/amd-radeon-vii-launch-stock-looks-worryingly-thin-in-uk-and-europe

If this is replacing the Vega 64, surely partners will jump on board? I would have thought so.
Guess we'll wait and see.

---

We need a bigger limit allowed for WUs in progress for GPUs with this daily database issue. 600 doesn't cut it with multiple Tahiti based GPUs in 1 PC.

You could reconfigure BOINC to say you have more GPUs than you actually do. Some SETI users do that since there is a weekly server maintenance.
https://setiathome.berkeley.edu/top_hosts.php

[63] NVIDIA GeForce GTX 1080 Ti (4095MB) driver: 410.66

[64] NVIDIA GeForce GTX 1070 (4095MB) driver: 415.27

[64] NVIDIA TITAN V (4095MB) driver: 410.57 OpenCL: 1.2

Doesn't work anymore so much. 200 per GPU up to 600 mad per host, it will only care about 3 of those fake GPUs. Used to be able to get around 1100 or more.

Longer WUs for GPUs with a matching credit adjustment sounds like a good plan. Server getting beat up constantly from these short request times (1m30 between reguest in BOINC) can't be good on top of that.

We need a bigger limit allowed for WUs in progress for GPUs with this daily database issue. 600 doesn't cut it with multiple Tahiti based GPUs in 1 PC.

You could reconfigure BOINC to say you have more GPUs than you actually do. Some SETI users do that since there is a weekly server maintenance.
https://setiathome.berkeley.edu/top_hosts.php

[63] NVIDIA GeForce GTX 1080 Ti (4095MB) driver: 410.66

[64] NVIDIA GeForce GTX 1070 (4095MB) driver: 415.27

[64] NVIDIA TITAN V (4095MB) driver: 410.57 OpenCL: 1.2

Doesn't work anymore so much. 200 per GPU up to 600 mad per host, it will only care about 3 of those fake GPUs. Used to be able to get around 1100 or more.

Longer WUs for GPUs with a matching credit adjustment sounds like a good plan. Server getting beat up constantly from these short request times (1m30 between reguest in BOINC) can't be good on top of that.

Run multiple clients. If you're running 4x concurrently with 1 client then run 4x clients with 1 task. 600 per client then.

I have a patch ready for Radeon VII. If anyone wants to send me a video card, I'm more than happy to test it out. :)

https://github.com/neofob/milkywayathome_client/commit/c9111eb2dd90865cc2a4451260f6613714610277

There already is a Radeon VII crunching:

https://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=797216

Looks like ~22 seconds for one wu - 29 seconds for two wu

Nice.. about 19 sec per one.. will be better run whole boinc from ram disc, becouse short task, upload,, also run multiple task at once..10+
600 task quote per day .. ?

Still not sure I'd like this HOT, noisy beast in my study crunching away 24/7.

---

Wait for the aftermarket products and put it under water.

---

Look like this card needs some serious cooling if you want to run 24/7.
Maybe a Kraken G12 adapter and a Thermaltake Water 3.0 would be sufficient.

Look like this card needs some serious cooling if you want to run 24/7.
Maybe a Kraken G12 adapter and a Thermaltake Water 3.0 would be sufficient.

Runs fine on stock cooling here at an ambient of 73F. Actually it's much cooler than I thought. Currently running Moo AND Milky at the same time.Fan noise can be an issue for some folks at this configuration, but not me.
Use Wattman Auto undervolt if you want silent and save A LOT of power consumption.
Here's my host running 3x on 1 instance and the other instance is running Moo.
http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=797896

How come the count shows (8) Radeon VII's VietOZ?
Man.... I need your job! The gear you have is worth squillions! It's what I'd do if I won lotto lol. (just saying what we're all thinking ;)

---

FYI, I have a developer friend that has put together a Linux utility for controlling AMD/ATI cards. It can control clocks, power levels and p-states. Seeing requests here frequently for the ability to undervolt AMD cards for power usage savings, I thought some might find Rick's utility helpful.

https://github.com/Ricks-Lab/amdgpu-utils

He would like feedback and requests for new features.

[Edit] This is what the utility interface looks like.

---

How come the count shows (8) Radeon VII's VietOZ?
Man.... I need your job! The gear you have is worth squillions! It's what I'd do if I won lotto lol. (just saying what we're all thinking ;)

only one card, not 8 :)
Just a modification to get the max cache.

I have a patch ready for Radeon VII. If anyone wants to send me a video card, I'm more than happy to test it out. :)

https://github.com/neofob/milkywayathome_client/commit/c9111eb2dd90865cc2a4451260f6613714610277

I tried to cross-compile this using MinGW and it failed. I have sense checked the git periodically and status remains "failing" for MinGW and "error" for Linux. Are there any plans to continue with getting this to compile? I do not own a Vii I am trying to do this for a couple of friends.

Thanks!

I have a patch ready for Radeon VII. If anyone wants to send me a video card, I'm more than happy to test it out. :)

https://github.com/neofob/milkywayathome_client/commit/c9111eb2dd90865cc2a4451260f6613714610277

I tried to cross-compile this using MinGW and it failed. I have sense checked the git periodically and status remains "failing" for MinGW and "error" for Linux. Are there any plans to continue with getting this to compile? I do not own a Vii I am trying to do this for a couple of friends.

Thanks!

What is your error? Compiling Milkyway at home could be tricky, at least on Linux system.
Here is the steps (you might need to to modify the directory path to your system)
github gist

#!/bin/bash

mkdir /tmp/MW /tmp/MW/build
cd /tmp/MW
git clone --depth=10 https://github.com/Milkyway-at-home/milkywayathome_client
cd milkywayathome_client
git submodule init
git submodule update --recursive
cd ../build
cmake -DBUILD_32=OFF -DSEPARATION=ON -DNBODY=OFF -DSEPARATION_OPENCL=ON \
  -DOPENCL_LIBRARIES=/opt/amdgpu-pro/lib/x86_64-linux-gnu/libOpenCL.so.1 \
  -DOPENCL_INCLUDE_DIRS=/opt/amdgpu-pro/include/CL/ ../milkywayathome_client
make

Note: I updated the Radeon patch with another commit from eyeballing some posts here. So you could try to patch that. Something like this after you git clone/checkout from the official repo.

git remote add neofob https://github.com/neofob/milkywayathome_client
git fetch neofob
git merge neofob/tp-radeon7

Have fun!

Thanks for the reply!

It finished this time after starting over, deleting everything and re-giting it. ;)
But I only got a linux binary. I cannot find a windows .exe anywhere. Below is the command line I used.

cmake -DCMAKE_TOOLCHAIN_FILE="../milkywayathome_client/cmake_modules/MinGW32-Cross-Toolchain.cmake" -DBUILD_32=OFF - DSEPARATION_OPENCL=ON -DSEPARATION_STATIC=ON -DOPENCL_LIBRARIES=/opt/amdgpu-pro/lib/x86_64-linux-gnu/libOpenCL.so.1 -DOPENCL_INCLUDE_DIRS=/opt/amdgpu-pro/include/CL/ ../milkywayathome_client/
make

This is first time I have tried to cross-compile so I may be doing something stupid. :P


Cheers!

I never compiled on linux, but by looking on the sommand I would guess you compile for linux. Or lets say I would assume to get a linux libary...

---

Yeah, the path to opencl library and others are for Linux.
I've never done this cross-compiling thingy, Linux -> Windows. It's a mess.
I used to cross-compile Linux x86 -> ARM but no more! Farewell to ARM. :)

Goodluck!

btw. Radeon VII just went under 670 € with taxes and shipping in germany ^^
after my relocation to a nother city I will buy one :)

---