configure -> slots -> your gpu slot, scroll down, extra slot options -> add 'cpu-usage' with a value of your choice ( hit and miss, test the right value for your configuration ). Need to have FAHControl on expert or advanced.
Edit:
I removed the limit's in settings, paused and restarted the slots, slots are started without -cpu x settings but usage keeps limited
Quit and restart FAHControl ( which quits and restarts fahclient ) same result. Only after a reboot my core usage is normal.
Last edited by MtM on Thu Feb 23, 2012 7:04 pm, edited 1 time in total.
Hardware configuration: Intel i7-4770K @ 4.5 GHz, 16 GB DDR3-2133 Corsair Vengence (black/red), EVGA GTX 760 @ 1200 MHz, on an Asus Maximus VI Hero MB (black/red), in a blacked out Antec P280 Tower, with a Xigmatek Night Hawk (black) HSF, Seasonic 760w Platinum (black case, sleeves, wires), 4 SilenX 120mm Case fans with silicon fan gaskets and silicon mounts (all black), a 512GB Samsung SSD (black), and a 2TB Black Western Digital HD (silver/black).
MtM wrote:
configure -> slots -> your gpu slot, scroll down, extra slot options -> add 'cpu-usage' with a value of your choice ( hit and miss, test the right value for your configuration ). Need to have FAHControl on expert or advanced.
toTOW wrote:This thread is all about design consistancy : it is useless to have a powerful GPU is the rest of the machine is not capable of handling it. This applies to both CPU and MB which have to be powerful enough and to have enough bandwidth to feed the GPU and to the PSU that have to be powerful enough to power everything in the machine ...
I've never been a fan of multi-rail PSUs, and I think my point has been proved by this thread ...
Sorry, but your assumption that it's just poor quality hardware causing the problem is incorrrect.....
My single rail 1KW XFX (based off of a seasonic platform) is not the cause of the problem... more than likely it's the use of two factory oc'd cards in SLI trim. Even downclocked, and running at 100% fanspeed I'm still seeing one card hitting 80c+.
More than likely I'll end up dropping one of the GPU clients if these WUs are going to end up being the norm.
@sswilson toTow was only referring to OP's shutdown problem, not the temperature issue ( psu's and temperature's are only related if your psu is wired to exhaust hot air into your case instead of out ).
@7im yeah thought I'd give the quick reference.
How about the gpu core usage issue I just showed screens of? Anyone else able to reproduce that?
To sswilson, how about increasing your airflow inside the case? A strategically placed fan fan zip tied to the right place inside your case could make a big difference for that one card.
I run a pair of 9800gt's, and the card closest to the bottom of the case doesn't get very good airflow. Tying a 120mm fan to the bottom of the case off the end of the card blowing across the entire length of the card made a huge difference. Fans are cheap.
I'm running inside a TT level 10GT with a side 200mm fan damn near strapped to the top of the cards... next step would be to move the sli hardware out to my open testbed....
I'm also running SMP at the same time so attempting to downclock CPU usage would be counter productive. What I'll probably end up doing is using Afterburner to downclock the cards in order to keep the temps down... not that big of a deal, but it means that less work will end up being done when/if the hot WUs dry up (machine runs 24/7, but I don't check on it all that often). What we really do need is the ability to tweak the GPU usage, especially since WUs like this are more than likely to become the norm. I know that's on the wish list, but hopefully these WUs serve as incentive to get that done.
You can tweak gpu usage through cpu usage has been said a couple of times already, why not try if it works for the fermi core? I have no idea if it does.
7im wrote:Think of the data stream as a fire hose full of water flowing to the GPU. In the first example, the fire hose full of water has some air introduced, like 1 in 10 liters of water is replaced by an equal volume of air. It still runs full speed 9 out of 10 cycles. In the second example, the water valve (cpu usage) that feeds the fire hose is cranked down so that the total volume of water flowing to the GPU is reduced by a steady amount. 1/10th of every liter is withheld, not 1 in 10 liters.
It's a nice analogy, but I just tried setting with -configonly CPU usage requested and it seems to give same type of behaviour as GPU_IDLE setting. I set to 50% and I still get the good old 9 seconds of lag and 1 second of freedom.
MtM wrote:You can tweak gpu usage through cpu usage has been said a couple of times already, why not try if it works for the fermi core? I have no idea if it does.
Nope it doesn't. My card GTX560Ti seems to be too good at queuing for this to work. As I wrote above, the card just goes from 100% usage to a momentary 0% usage and back again.
I get the same with my low end ati with certain values, but some other seem to be more stable. I can't check against lag, just checking core usage in openhardware monitor. Might be worth it to try different values. Btw, I do see it jumping up and down, but selecting 90 for my gtx275 results in jumps between 75-95, not 100 and 0, so if it does the same on fermi ( and why wouldn't it? ) it's a matter of finding a good value for your system/configuration.
As previously noted, the fah_gpu_idle flag introduces large blocks of sleep time into the GPU jobstream, so it's chunky at best. (Maybe good for AVERAGE heat.)
Also note: It doesn't work on all GPU cores. Don't be surprised if it does nothing.
In all honesty, until such time as there is an actual built-in GPU usage limit switch (as opposed to workarounds), IMO the best solution (rather than just stopping the client all together) is to use a program like MSI's afterburner, or EVGA's Precision to downclock your GPUs. Unfortunately, what this solution means is that we'll lose processing power for the lower temp WUs as we'll have to assume that all WUs have the potential to run this hot.
Guess I would be looking at aftermarket heatsinks and fans. Bigger proteins seems to be the way the PG is headed. The stock intel heatsinks and fans aren't worth crap either for their cpu's.
Well, sounds like this might hasten my decision to go watercooling. The 680x units already cause my factory OC'd MSI to hit lower to mid 80's, I can only imagine what these will do :/
).