SETI@Home Wow!-Event 2019

[Modular]

well...
View attachment 9748
I obviously need help here. What's the average runtime for RX570 on Windows?

My rx580 on Linux is running 6 to 7 minutes per task. I'm running 2 tasks at once. Not exactly what you were looking for, but at least something to gauge by.

 

[ao_ika_red]

My rx580 on Linux is running 6 to 7 minutes per task. I'm running 2 tasks at once. Not exactly what you were looking for, but at least something to gauge by.

Thanks.
I also run 2 tasks concurrently but it takes 17-20 minutes to complete.

 

[biodoc]

I need to think of something.

I'm sure you will.

 

[VietOZ]

remember, Seti@Home will be down tomorrow. So don't forget to load up enough work for the down time guys

 

[Assimilator1]

I have a RX 580 & run single tasks, for high 80s credit results I'm seeing run times in the region of 880-900s (14.8 minutes).

Modular
Wow! That's some fast times you've got there! . What credit results were you looking at?

 

[Modular]

remember, Seti@Home will be down tomorrow. So don't forget to load up enough work for the down time guys

What's the best way to do this? Should we change local prefs?

 

[StefanR5R]

Too much red, not enough green:
View attachment 9740
I need to think of something.

I'm sure you will.

Well, first I did the easy thing and added a couple of 14-core CPUs to the effort.

Then I "found" my old IvyBridge-E with a GTX 1070 in the basement, plugged a spare disk into it, and installed Linux Mint 18.3. Right away I noticed trouble: The onboard NIC of the Asus X79 based mainboard of the IVB-E has got the same MAC address as the onboard NIC of another PC with an Asus X99 based mainboard.

I new it from earlier times that these two computers wouldn't play together in the same LAN. But that was when they both ran Windows 7, and I chalked it up to a bad installation procedure (cloned from a base image rather than an installation from scratch). But now I realize it must be the MAC address.

Perhaps this was because I inserted a custom flashed BIOS chip in each of them when I bought them: I bought the X79 board and the IVB-E CPU, but the BIOS version with which it came in the box had only support up to SNB-E. This board already has an early form of Asus' BIOS "Flashback" feature, but it simply did not work at the time. Hence I ordered a pre-flashed BIOS chip with an up-to-date BIOS.

Later, when I bought the Asus X99 board and a BDW-E CPU, the board came only with HSW-E support out-of-the-box. And yet again Asus' Flashback feature did not work, so I ordered a pre-flashed BIOS chip again from the same vendor.

In one of the two cases (I don't know whether it was the X79 or the X99 board), the flashing process simply never terminated. In the other case, Asus had broken the process themselves, because they fundamentally changed the BIOS format somewhere between their many BIOS revisions, and the upgrade to the new format had to be done with CPU assistance.

At the time when I built the computers I was extremely relieved to get the updated BIOS via that 3rd party EEPROM flashing service. But now I learn that this service, to all appearances, put a bad MAC address in each of the EEPROMs.

:-(

Apart from the networking trouble, I am stuck just now with Mint 18.3 not switching from nouveau to nvidia-384 correctly. The driver manager says it's on nvidia-384, but lspci shows the contrary (as do some rendering bugs on the Cinnamon desktop).

 

[biodoc]

Apart from the networking trouble, I am stuck just now with Mint 18.3 not switching from nouveau to nvidia-384 correctly.

sudo rmmod nouveau to see if the nouveau module is loaded.

You can blacklist the nouveau module and others with:

sudo nano /etc/modprobe.d/blacklist.conf and append the following at the end of file:

blacklist vga16fb
blacklist nouveau
blacklist rivafb
blacklist nvidiafb
blacklist rivatv

sudo update-initramfs -u

reboot.

 

[StefanR5R]

I found a USB-Ethernet dongle which should get me going.

As for nouveau vs. nvidia: Rookie alert! I forgot to switch off UEFI secure boot in the BIOS.

 

[Modular]

I have a RX 580 & run single tasks, for high 80s credit results I'm seeing run times in the region of 880-900s (14.8 minutes).

Modular
Wow! That's some fast times you've got there! . What credit results were you looking at?

Hey, it's hard to tell. I'm running two WU's at once on this card.

I just took a look at the last 10 WU's to get me between 70 and 90 points and averaged the time it took to crunch them. It's almost exactly 13 minutes on average. Now remember, that's running 2 WU's at the same time though.

13 divided by 2 = 6.5 minutes per WU on average.

I tried running it with just one WU at a time and was getting 7 to 8 minutes on average for the same WU's. So it's just a little faster for me to crunch 2 at once on the RX580.

Running more than one on the nVidia cards actually slows them down from my testing.

 

[Ken g6]

I briefly tried to crunch two at a time on my Nvidia cards but all I got was out of memory errors.

 

[StefanR5R]

I woke up to a beautiful morning.

The one remaining user to overtake me is Mark-4790k, so that's OK with me. :-)

 

[Assimilator1]

Hey, it's hard to tell. I'm running two WU's at once on this card.

I just took a look at the last 10 WU's to get me between 70 and 90 points and averaged the time it took to crunch them. It's almost exactly 13 minutes on average. Now remember, that's running 2 WU's at the same time though.

13 divided by 2 = 6.5 minutes per WU on average.

I tried running it with just one WU at a time and was getting 7 to 8 minutes on average for the same WU's. So it's just a little faster for me to crunch 2 at once on the RX580.

Running more than one on the nVidia cards actually slows them down from my testing.

I know Linux confers some advantage, as does running 2 WUs at once, but it seems your setup is twice as fast as mine! Unless you're looking at different credit WUs, or something's up with my setup (drivers or client?). What credit WUs were those singly run WUs? Times without credit amounts isn't much use I'm afraid.
I've just looked at 2 low credit WUs, ~47, 1 took 315s the other 479s, a wide variance! The high 80s WUs seem more consistent, mostly! lol (just saw one at 327s! But that seems to be an outlier).

 

[biodoc]

For those of you with AMD GPUs, this thread regarding command lines may be useful.

 

[StefanR5R]

The IVB-E + 1070 PC shut itself off shortly before I left for work today (I switched it on again, opened the side panel, and put a desk fan in front of it), and once more while I was absent. Sadly, it does not power up anymore, even though all the mainboard LEDs are lit when the PSU is on.

I can't really do anything about that for the time being. I think I will pull out the disk tomorrow, attach it to another computer, and try to cancel the tasks in progress on it (because of S@H's absurdly long deadlines).

 

[Assimilator1]

RIP a fellow IVB-E rig , although maybe it's just the PSU that's given up? (despite the mbrd LEDs).

For those of you with AMD GPUs, this thread regarding command lines may be useful.

Project down for maintenance
SETI@home is temporarily shut down for maintenance. Please try again later.

Doh!

 

[biodoc]

I can't really do anything about that for the time being. I think I will pull out the disk tomorrow, attach it to another computer, and try to cancel the tasks in progress on it (because of S@H's absurdly long deadlines).

Sorry about the computer failure Stefan. I seem to remember @Ken g6 has posted a method to transfer boinc tasks from one computer to another.

Seti's task deadlines are about 7 weeks which I agree, is way too long. You'd think a week would be more than long enough.

 

[Assimilator1]

Depends how much your PC is on, I bet my mum's PC wouldn't complete many (or maybe any?) WUs with a 1 week deadline (she is on our team btw, and has been for years ), her rig (a C2 Pentium o/c 3.6 GHz) is probably only on a couple of hours per day at most.

 

[StefanR5R]

RIP a fellow IVB-E rig , although maybe it's just the PSU that's given up? (despite the mbrd LEDs).

Perhaps not...

-- When I switch the PSU on, motherboard lights go on, and pressing either the front panel power button or the onboard power button make a relay click in the PSU, but nothing else.​

-- When I unplug the PSU from the mainboard and shorten the green wire with a black one, then the relay clicks, the PSU fan begins to spin, and the DVD-ROM makes its usual power-up noise.​

It was a top-shelf PSU, and it is only about half as old as the mainboard. Still, I might go looking for a spare PSU later this week and test that.

Could the CMOS battery (if depleted) stop the mainboard from powering up? Although when I was going through some BIOS menu items yesterday, it still held the settings it had several months ago when last used.

 

[crashtech]

One thing I can think of that caused me trouble on an X79 board was one of the pins on the ATX 12v 8-pin connector had developed resistance and partially melted the connector. It's unlikely, but just unplugging the 8-pin and inspecting both halves will quickly and easily eliminate it as a source of trouble.

 

[StefanR5R]

I looked once more with a good light source, and female and male both look healthy.

(Thankfully. I had my share of molten electric equipment already earlier this month.)

 

[Ken g6]

I seem to remember @Ken g6 has posted a method to transfer boinc tasks from one computer to another.

Don't bet on it working, though. You need two very similar computers, and they need to have been running secondary BOINC instances.

Edit: Anyway, I came here to post that SETI is back up.

 

[StefanR5R]

My understanding is:
If there is binary compatibility with the application executables, it should be possible to download tasks on computer A, complete them on computer B, and report them on computer A.
But downloading them on computer A and reporting them in computer B is quite likely to go wrong; although I haven't quite learned yet what factors are involved precisely in the determination of host identity in the client--server communications.

As for my own current use case, right now I don't have another computer into which I'd like to stick the GTX 1070. Assuming that I had another Linux x86-64 computer with any type of CUDA 9.0 compatible GPU in it, the jobs should complete on that one without problem. But reporting from there would incur the risk of everything being canceled by the server, as long as I don't know precisely what it takes to do that (if it is possible at all). However, all I really want is to cancel these jobs, for the sake of my wingmen. (Hoping that the server generates new tasks for these WUs in a timely manner.)

 

[lane42]

1-Phoenicis...……………………...4,920,069
2-biodoc...……………...…………...4,287,272
3-Lane42...…………………………...2,974,707
4-Pokey...……………………………...1,830,291
5-xii5ku...……………………………….1,124,205
6-crashtech...………………………..1,057,506
7-Mark-4790k...……………………..937,460
8-Orange Kid...……………………….857,317
9-Howdy...……………………………...761,328
10-10esseeTony...………………...608,272
11-Modular...………………………….601,369
12-IEC...………………………………….491,034
13-Grumble_D...………………......276,543
14-Ken_g6...…………………………..239,252
15-emoga...…………………………...199,583
16-Tom Servo...……………………..131,550
17-(H)Skillz…............……...……..131,023
18-QuietDad...………………………..130,993
19-Ingleside...………………………..104,561
20-(TA)Assimilator1...………….103,961
21-Kiska...……......….........……...93,467
22-burninator34...………………...79,163
23-MarcelliuslkaP...……...……...66,363
24-ETPhoneHome...……………..35,463


Let's Go Team

Finish from last year

#237



1-xii5ku...………8,755,950
2-Howdy2u2......2,702,127
3-Lane42...…….2,261,828
4-crashtech...….2,180,533
5-biodoc...……...2,105,021
6-OrangeKid...…795,617
7-petrusbroder...765,976
8-Tejas...……….693,120
9-Ken g6 Team Anandtech….632,028
10-10esseeTony...……………420,985
11-iwajabitw...…………………329,039
12-Mark-4790k...……………...285,886
13-CFrench...………………….265,451
14-Ingleside...………………….250,375
15-QuietDad...………………….226,889
16-GLeem...…………………….212,040
17-kiska...……………………….164,674
18-TA Assimilator1...…………..142,382
19-MarcelliusIkaP...……………..113,087
20-Tom Servo...………………….110,160
21-curiousmike...………………...55,325
22-ericip...………………………...19,851
23-emoga...……………………….2,436

 

[biodoc]

@phoenicis , 3.5 M in one hour. Very nice indeed.