0k, i will take your word for it. edited my post above to strike out that line. guess ive been lucky.
what tools do you use?
what tools do you use?
As I said, I wrote them myself - the closest analog to what I'm working on and polishing is WattTool for Windows, I think. Allows you to see VRM temps, set/get voltages, set/get offsets, change VRM switching frequency, and more - but WattTool can only give you this control over IR3567B and compatible VRM controllers. Mine supports four controllers that are used in AMD cards as main controllers (and sometimes this main controller is used for other purposes in the card besides core volts - for example, XFX RS 470/480 cards use the NCP81022's main output (VDD) for core (duh) but use the other one (VDDNB) for the IMC voltage (VDDCI.)) I've also added basic support for one controller that's more of a secondary one, which I found on my DirectCU II 290X. It controls MVDDC.
sounds like a dangerous tool in the hands of a noob. those are the best kinds of tools IMO
any chance of your releasing it at some point?
At some point? Possible - but it would be a while in the future. I actually sold a couple (limited) copies before I considered it mature, cause they kinda just happened. Right now, I'm focusing on rounding out the feature set.
Another thing that makes it not just dangerous, but possibly deadly: I can not just adjust the thresholds for protections like OCP/OVP/OTP... I can disable them.
Actually, if you run the AMDGPU-PRO drivers, you CAN mix them in a rig. The Fiji card may perform slightly worse than with fglrx, but it works.
Since AMDGPU Pro drivers only support GCN3 or newer cards, I couldn't get it to work with my Pitcarin and RX 480 rig on Arch Linux. I had to switch to Windows 7 to be able to use all my cards
https://wiki.archlinux.org/index.php/Xorg#AMD
The support is labeled "experimental" but it IS there.
some cards do not display vrm temps. but if the core is cool, chances are the vrms are ok also.
Please don't give people who don't know any better terrible advice like that - I was mining on my linux dev/test rig, Naomi, just a day ago, and had strange crashes when core was under 70C.
I used one of my tools to check the temps on the VRM controller feeding the core - it was over 106C.
Also, I assume that under Windows, the tools you guys have should show VRM temps on all cards except those with an NCP81022 feeding the core - because I can't get temps off it either; cheap piece of shit has no sensor.
ok, i will take your word for it. edited my post above to strike out that line. guess ive been lucky.
what tools do you use?
As I said, I wrote them myself - the closest analog to what I'm working on and polishing is WattTool for Windows, I think. Allows you to see VRM temps, set/get voltages, set/get offsets, change VRM switching frequency, and more - but WattTool can only give you this control over IR3567B and compatible VRM controllers. Mine supports four controllers that are used in AMD cards as main controllers (and sometimes this main controller is used for other purposes in the card besides core volts - for example, XFX RS 470/480 cards use the NCP81022's main output (VDD) for core (duh) but use the other one (VDDNB) for the IMC voltage (VDDCI.)) I've also added basic support for one controller that's more of a secondary one, which I found on my DirectCU II 290X. It controls MVDDC.
Are you doing this all via i2c commands on linux? I've been digging a bit but cant seem to get a i2cdump anywhere near what I get with aida under windows, so I don't even know where to start :/. Also, how are you checking that the vddci is getting set on those ncp81022 cards? afaik there's no sensor, so are you manually checking em? How
I don't want to give too much away - but basically this: I access the card using direct HW access (this requires root, of course.) NCP81022 only lacks a temperature sensor. VDDCI is on the second output of NCP81022 on XFX RS cards, so you have to first select the second output by writing to one of its regs, THEN reading the voltage out.
If you're trying to get i2c access by using tools like i2cdump - you're doing it wrong. This relies on the AMDGPU driver, which is a little "special" if you get my drift. Gotta do it by paw.
Thanks for the hints, I appreciate it! At least now I know to abandon my i2cdump attempts. Direct hw access eh..