I put it on the dual 7763 circuit so I know its good. I only use that box in emergencies, and can find another circuit if its needed !Congrats! For fuel to the FIRE!
Just watch your home power usage so you don't go all out on the fire part
Just make sure to hide it well from the neighborsI put it on the dual 7763 circuit so I know its good. I only use that box in emergencies, and can find another circuit if its needed !
I'm tired of windows 10 at least with 64 core cpus. It is a fight to keep all the cores fired up! Good luck!I have to reinstall windows, Its doing odd things. NO nvidia displays, but F@H sees the Titan, but nothing sees the 4090. I installed the drivers for the motherboard, as during the initial install, it did not detect the Intel nics. I installed the Intel drivers from ASRock, but they should have been auto-detected. All very strange. Will mess with it tomorrow.
Yep! Have Linux Mint running now. Much better for CPU use for having any kind of D.C. tasks that is running at the same time and over 64 threads at a time. I stil like having a GUI to play with, but I am reading that a GUI sucks some cpu cycles?What a high powered setup @Markfw has going with that 4090 and Rome CPU! Installing all those drivers can certainly be a pain. Windows can be finicky sometimes, hopefully it gets sorted out for you. @cellarnoise totally agree that Windows isn't always the best at utilizing many cores efficiently. Curious, have either of you tried a Linux distro on those systems instead? Maybe it could get even better performance out of all that hardware.
For some context: The aarch64 kernel doesn't get broken, it simply doesn't enumerate all of the CPUs. And it's an almost trivial issue of the aarch64 kernel only. (Correction after a look at the patch: I am downgrading this from almost trivial to trivial, because all what's needed was to enable compile-time config switches for already existing architecture-independent code.) Why they are sending a patch for this now rather than, say, half a year ago, is basically a question of product time-to-market, not a question of getting the software implementation done. For the x86-64 architecture, support for 4096 logical CPUs was implemented in 2009 and actively used on Silicon Graphics (SGI) Altix UV servers.Ampere's 192-core chips break ARM64 Linux kernel in two-socket systems, company requests higher core count support.
probably not the issue here. the system it's running on (single EPYC 7542) is only a 32c/64t part.Yes, you can have too many CPU cores - Ampere's 192-core chips break ARM64 Linux kernel in two-socket systems, company requests higher core count support.
Sometimes, it's the order of the drivers. Do lots of reboots after each hardware driver install. I'd download CPU-Z see if that detects what kinda video card you got.
You still have both GPUs in the system, and you did check which one of the two GPUs was selected for video output, right?Now I have problems even worse, as I get the "black screen" problem.
The nomodeset fix allowed it to boot with a visible display, but now the card is defective. It shows err under nvidia-smi for that card. The other shows fine.You still have both GPUs in the system, and you did check which one of the two GPUs was selected for video output, right?
Did you check whether or not a newer BIOS is available for the mainboard?
I had dual-GPU computers with unequal GPUs in the past (AMD + Nvidia with Windows 7; Nvidia 1080 + 1070 with Linux), and while that worked, I find dual-/triple-GPU computers with a single GPU type much friendlier to handle. (Think of BOINC's task duration estimations, for example.)
A fan speed reporting error doesn't look like a PCIe and BIOS related error at least. Still, checkThe nomodeset fix allowed it to boot with a visible display, but now the card is defective. It shows err under nvidia-smi for that card. The other shows fine.
dmesg output for possible PCIe error messages. Also, you do already have X, Cinnamon, and FahCore running, so you seem to be mostly set now. Edit, o wait, the reported Wattage is low, and utilization of 0% isn't plausible, obviously...Except for the fact that the motherboard was running 3 Titan V's the last time it was running, and its a server motherboard ($500)sounds like a motherboard problem.
dmesg command? I can only test it remotely on my Mint computer right now, but it's likely that you will get a nicely colored output in the terminal, and anything which *might* indicate a problem would be colored in red, for you to spot immediately.Think about the parts you’re using and it will make sense. The problem is the motherboard.Except for the fact that the motherboard was running 3 Titan V's the last time it was running, and its a server motherboard ($500)
it appeared totally sealed from Amazon. But... Here is windows screenshot with the Titan V removed. linux was running F@H, but very slowlyAnd more thoughts: Did you get the card new in a sealed box? Or was it an open box offer? Or even used?
Can you see whether or not the fan cables (should be two, plus maybe one or another lighting cable) are actually attached to the graphics card?
I have not rebooted linux after trying to update the GPU drivers, so no dmsg yet. The mothrboard is the ASRock EPYCD8So what about the output of dmesg?
What is the model name of the motherboard?