I don't know what might cause that. There's a newer Boinc client available. Maybe that is worth trying:
We are having some very unpleasant experiences with BOINC 7.8.2 on Linux and OS X. I don't know if it applies to Windows. But Richard Haselgrove is on it, so it will get fixed.
I have restarted crunching on the Vega 64 (weather here is finally colder).
For comparison, current units x2 take 560 s. Power (reported by the card) is 200-220 W (220W being the current limit).
The Fury X takes 440 s, but for 1 WU only. I'm still unable to run more in parallel.
Here some more results. Power is the number reported by GPU, which doesn't cover the entire GPU card.
x1 x2
time / power time / power
-----------------------------------------------------
Vega 64 560 / 200-220 W
Fury X 440 / 150 W
RX 480 660 / 80-100 W
R9 280 1300 / ~100 W
1050 Ti 1450 / 65 W
Thanks for the reports, everybody. Mumaks results (280s/WU running 2x) are a significant improvement over the ~360s/WU reported so far for 1x. Unfortunately the power draw close to the chip limit should mean the card is drawing close to its total TDP, i.e. 260 - 290 W total, judging by the experiences with other software. did you touch clocks & power limits yet? Memory OC, maybe chip OC if you want, and reduced power target will improve the efficiency - maybe significantly. In games Vega 64 can draw 25% less power for hardly any performance loss.
Thanks for the reports, everybody. Mumaks results (280s/WU running 2x) are a significant improvement over the ~360s/WU reported so far for 1x. Unfortunately the power draw close to the chip limit should mean the card is drawing close to its total TDP, i.e. 260 - 290 W total, judging by the experiences with other software. did you touch clocks & power limits yet? Memory OC, maybe chip OC if you want, and reduced power target will improve the efficiency - maybe significantly. In games Vega 64 can draw 25% less power for hardly any performance loss.
MrS
I'm still running with stock settings. I don't I'm going to OC it, as this air-cooled model is running 75-80 C even with increased fan levels, which is more than I like. Will probably try some undervolting, but according to other reports this doesn't work well and certain levels don't seem to be properly applied (probably a bug in Wattman).
I'm still running with stock settings. I don't I'm going to OC it, as this air-cooled model is running 75-80 C even with increased fan levels, which is more than I like. Will probably try some undervolting, but according to other reports this doesn't work well and certain levels don't seem to be properly applied (probably a bug in Wattman).
I have moved my little friend to another host here.
I'm also running stock settings and don't plan to overclock due to increased heat - the blower style fan is barely adequate! That said, with an increase to the fan speed settings and reduced temperature targets I have this card running at 65C with a supplementary case fan directly behind the card. I have also set the power limit to maximum.
Running tasks x2 I'm getting average core clock speeds of 1598MHz (and have hit 1670+MHz peaks), memory is at the default 945MHz. In this config. runtimes are in the order of 460 -470 seconds. Average GPU load is around 90%.
I have played around with undervolting, it helps increase core clock and reduce power consumption (a little). Sadly as Mumak has suggested, Wattman is not there yet with voltage control and amd settings will crash at a moments notice and return all settings to default... Not much use.
Power consumption at the wall will make NVidia owners faint! Averaged wall power comes in at 406Watts.
I'm still running with stock settings. I don't I'm going to OC it, as this air-cooled model is running 75-80 C even with increased fan levels, which is more than I like. Will probably try some undervolting, but according to other reports this doesn't work well and certain levels don't seem to be properly applied (probably a bug in Wattman).
I have moved my little friend to another host here.
I'm also running stock settings and don't plan to overclock due to increased heat - the blower style fan is barely adequate! That said, with an increase to the fan speed settings and reduced temperature targets I have this card running at 65C with a supplementary case fan directly behind the card. I have also set the power limit to maximum.
Running tasks x2 I'm getting average core clock speeds of 1598MHz (and have hit 1670+MHz peaks), memory is at the default 945MHz. In this config. runtimes are in the order of 460 -470 seconds. Average GPU load is around 90%.
I have played around with undervolting, it helps increase core clock and reduce power consumption (a little). Sadly as Mumak has suggested, Wattman is not there yet with voltage control and amd settings will crash at a moments notice and return all settings to default... Not much use.
Power consumption at the wall will make NVidia owners faint! Averaged wall power comes in at 406Watts.
Gav.
Thanks. 65 C would be really great. What is your maximum fan speed setting and what GPU power do you see reported with max power limit ? In HWiNFO sensors this is the "GPU Chip Power" value; make sure to use the latest Beta build of HWiNFO.
Fan speed is set to min. 3000 max 4000rpm. Temperature settings are Max. 75 and target 65C. With max power limit set to +50, GPU power draw is in the order of 226Watts - measured using and reported by GPU-z (sorry!).
Well, this is interesting.. I've been playing with the settings and came up with this:
Fan Max: 3200 RPM (anything above this is just too loud)
Temp. Target: 70 C
Power Limit: -20 %
Default power limit (0%) is 220 W here, reduced by 20% readout gives 175W.
This allows to maintain a reasonable temperature (~70 C) at a reasonable fan speed (~2600 RPM), maximum clocks are reduced by ~50-100 MHz.
But now the interesting part - checking the performance (run time) didn't show any slowdown. In fact most run times were even shorter by 10-30 seconds.. Now how to explain this :-) Perhaps this has something do to with the HBM2 - somewhere I've read that they dislike running at high temperatures.
Richie_9 wrote:I don't know
)
We are having some very unpleasant experiences with BOINC 7.8.2 on Linux and OS X. I don't know if it applies to Windows. But Richard Haselgrove is on it, so it will get fixed.
http://www.gpugrid.net/forum_thread.php?id=4625&nowrap=true#47880
Thanks guys. Looks like I
)
Thanks guys. Looks like I have received some WU's this morning. YAY!
I have restarted crunching on
)
I have restarted crunching on the Vega 64 (weather here is finally colder).
For comparison, current units x2 take 560 s. Power (reported by the card) is 200-220 W (220W being the current limit).
The Fury X takes 440 s, but for 1 WU only. I'm still unable to run more in parallel.
Here some more results. Power is the number reported by GPU, which doesn't cover the entire GPU card.
x1 x2
time / power time / power
-----------------------------------------------------
Vega 64 560 / 200-220 W
Fury X 440 / 150 W
RX 480 660 / 80-100 W
R9 280 1300 / ~100 W
1050 Ti 1450 / 65 W
-----
Thanks for the reports,
)
Thanks for the reports, everybody. Mumaks results (280s/WU running 2x) are a significant improvement over the ~360s/WU reported so far for 1x. Unfortunately the power draw close to the chip limit should mean the card is drawing close to its total TDP, i.e. 260 - 290 W total, judging by the experiences with other software. did you touch clocks & power limits yet? Memory OC, maybe chip OC if you want, and reduced power target will improve the efficiency - maybe significantly. In games Vega 64 can draw 25% less power for hardly any performance loss.
MrS
Scanning for our furry friends since Jan 2002
ExtraTerrestrial Apes
)
I'm still running with stock settings. I don't I'm going to OC it, as this air-cooled model is running 75-80 C even with increased fan levels, which is more than I like. Will probably try some undervolting, but according to other reports this doesn't work well and certain levels don't seem to be properly applied (probably a bug in Wattman).
-----
Mumak wrote:I'm still
)
I have moved my little friend to another host here.
I'm also running stock settings and don't plan to overclock due to increased heat - the blower style fan is barely adequate! That said, with an increase to the fan speed settings and reduced temperature targets I have this card running at 65C with a supplementary case fan directly behind the card. I have also set the power limit to maximum.
Running tasks x2 I'm getting average core clock speeds of 1598MHz (and have hit 1670+MHz peaks), memory is at the default 945MHz. In this config. runtimes are in the order of 460 -470 seconds. Average GPU load is around 90%.
I have played around with undervolting, it helps increase core clock and reduce power consumption (a little). Sadly as Mumak has suggested, Wattman is not there yet with voltage control and amd settings will crash at a moments notice and return all settings to default... Not much use.
Power consumption at the wall will make NVidia owners faint! Averaged wall power comes in at 406Watts.
Gav.
Gavin_14 wrote:Mumak
)
Thanks. 65 C would be really great. What is your maximum fan speed setting and what GPU power do you see reported with max power limit ? In HWiNFO sensors this is the "GPU Chip Power" value; make sure to use the latest Beta build of HWiNFO.
-----
Fan speed is set to min. 3000
)
Fan speed is set to min. 3000 max 4000rpm. Temperature settings are Max. 75 and target 65C. With max power limit set to +50, GPU power draw is in the order of 226Watts - measured using and reported by GPU-z (sorry!).
You're forgiven ;-) Is that
)
You're forgiven ;-) Is that with undervolting? +50 power limit gives me ~270 W max.
-----
Well, this is interesting..
)
Well, this is interesting.. I've been playing with the settings and came up with this:
Fan Max: 3200 RPM (anything above this is just too loud)
Temp. Target: 70 C
Power Limit: -20 %
Default power limit (0%) is 220 W here, reduced by 20% readout gives 175W.
This allows to maintain a reasonable temperature (~70 C) at a reasonable fan speed (~2600 RPM), maximum clocks are reduced by ~50-100 MHz.
But now the interesting part - checking the performance (run time) didn't show any slowdown. In fact most run times were even shorter by 10-30 seconds.. Now how to explain this :-) Perhaps this has something do to with the HBM2 - somewhere I've read that they dislike running at high temperatures.
-----