Discussion in 'NVIDIA' started by elf1, Jun 12, 2017.
Again it has nothing to do with the CPU
So hopefully a final update. I think I had two concurrent issues simultaneously. Originally my EVGA motherboard was definitely problematic. However, one of my Nvidia Graphics cards seems also to be bad. It may have started failing soon after i got it because originally they all seemed functional, which confused my situation with the second EVGA board. I'm in the process of RMA'ing the card now, too, in addition to my original RMA of the EVGA Mobo. Now it seems I have an EVGA board that can take, stably, more than one card at a time. Sadly, I'm not terribly surprised that another graphics card was bad. In a total of 8 cards brought home, two now were defective--not a great percentage. The first was more obvious because it made a sound like cards in bike spokes from the outset. This second one was much more subtle. Nvidia thinks it is something wrong with the memory of the GPU.
want to jump in and clear this part up: risers use the usb 3 cable only to carry signals from the mobo slot to the gpu socket. it does not use the usb 3 protocol, it uses the usb 3 cable. it could use any cable but usb 3 cables are common and shielded so they are commonly used.
so it is not hot swap in fact its a good way to blow pcie lines if you plug/unplug it while the rig is running, as its the same as unplugging/plugging in the card directly into the slot whiles its running.
How weird I have never had an issue hot plugging them...
usb plug has ground connect 1st, then everything else. so as its plugged in there is a chance for the signals to have a proper reference to ground. but the electronics on pcie slots dont expect hot plugging as its not in the specs. so no proper buffers to handle hot plugging.
so it doesnt mean it wont work. just a bit risky electronics wise. at some point something may die.
do you do that often? the computer recognizes the card when hot plugging?
I have pulled them out a couple of times and it recognizes they are gone. Not often just when I am fighting with cards.
After some other things I have seen posted "above 4g encoding" seems to be an issue on some or the z170/270 stuff.
Yes. Each are 150W at stock. Total system consupmption won't break 750W stock. Overclocked, with power limits in place, you aren't looking at breaking 900W...
do math gents.
I have pulled working PCIe cards, PCIe supports hot swap.
So have I which is what has me confused
maybe im wrong then.
in addition to the shock to both the card and the circuits connected to the slot (no buffers to protect the electronics on either AFAIK) i thought resources for video cards were allocated in the bios at boot time so they wouldnt be able to be added to the system once it was running.
so you can add a card that has not been in there before and it finds it? or does the card have to of been in that slot before?
certainly would make troubleshooting balky cards easier.
Balky Bartokomis? Or Bulky?
What resources ? Until the drivers are loaded the card only executes display commands issued from the main BIOS which at boot copies the contents of whatever is contained in the ROM of the video BIOS into system memory. Since those basic commands are typically identical between modern cards , I suppose they are hot swap-able since that information is already present in system memory.
I do not know how if there are any power surge protections though.
Why then would there be bios options on newer boards like "above 4G decoding" or whatever it is. My admittedly limited understanding of that option is it moves memory around to make room for more resources for video or other pcie cards. So resources apparently are limited at the BIOS level.
Perhaps just for later OS use? Dunno.
Would love to know for sure. Alas real life intrudes on my research time
I looked it up and from what I gather it basically enables 64bit wide addressing at the BIOS level for the GPU memory. 32bit is the default , makes sense since like I said those basic display commands need to be loaded into memory at boot and be recognizable by the BIOS in all situations to not brake compatibility. I am guessing that internally the drivers enable extensions that allow for a larger memory address space once you get inside an OS. Having those extensions be usable at boot is probably useful in some way if you use virtualization.
Separate names with a comma.