GPU to 72 status... - Page 471

chalsall · 2021-05-31, 21:03

Quote:

Originally Posted by Aramis Wyler

I can't get gpu72 to come up in the browser.

Please expand on your error report.

I just now accessed https://www.gpu72.com/ from two different locations, and it returned the expected page.

What are you seeing? Have you tried pressing F5 (or clicking on the reload button)?

Are the APIs failing? Etc.

Aramis Wyler · 2021-05-31, 21:10

I have the assignments page bookmarked; I go there a few times a day. Today I hit the button and it just spun. I tried just the root: https://www.gpu72.com and it just spun. Eventually I got the error message in the title, I'll attach a screenshot here.

I'm glad to hear you were able to get to it 0 Maybe it's just me, but obviously I can get to the forums ok. And yes, I have tried refreshing the page. Even now I just hit F5 again and it is spinning. Maybe it's a cache on my side,, I'll check.

chalsall · 2021-05-31, 21:22

Quote:

Originally Posted by Aramis Wyler

Maybe it's just me, but obviously I can get to the forums ok.

Hey. Thanks for that screenshot. Useful.

I've been seeing similar messages recently on many different sites. Even Google stuff. Initially, I thought that perhaps it was just me.

I've found that clicking on "Retry" immediately works. But this something which is making me go "Hmmm...

P.S. Yes, I'm paranoid. The question is, am I paranoid enough?

P.P.S. Seriously.

P.P.P.S. Has anyone else observed this kinda thing recently?

Aramis Wyler · 2021-05-31, 21:31

Huh, I shut down all of my firefox windows in both desktops, and restarted firefox. Unexpectedly, I still can't get to gpu72. I'm not worried, I was just checking to see if my last assignment had dropped off, but it is interesting that I can't get to it in firefox. - Here I'll try Chromium....

Wow, I can't get it to load in chromium either. Weird.

chalsall · 2021-05-31, 21:36

Quote:

Originally Posted by Aramis Wyler

Wow, I can't get it to load in chromium either. Weird.

Hmmm... I /do/ have some fairly serious IDS on this site. Perhaps this is the problem. Can you reach it from another IP?

If it's not too much to ask, could you please PM your public IP of where you can't see the server?

James Heinrich · 2021-05-31, 22:00

Quote:

Originally Posted by chalsall

P.P.P.S. Has anyone else observed this kinda thing recently?

GPU72 is working fine from here.

Probably unrelated, but last week I had trouble accessing Dario's ECM calculator page, possibly some kind of Cloudflare-related redirection weirdness of some sort. Worked fine on my tablet but not on my desktop. I eventually deleted the site cookies (easy in Chrome, more hidden in Firefox) and it all worked perfectly again.
As I said, likely unrelated, but you asked.

Aramis Wyler · 2021-05-31, 22:07

For reference, I was able to get to the site when I VPN'd into work and tried from there. I sent Chalsall my local IP and the work one.

RichD · 2021-06-11, 00:55

I just happened to latch on to a covenanted V100 GPU but got the following.

Code:

20210611_004220 ( 0:01): selftest PASSED!
20210611_004220 ( 0:01): Fetching initial work...
20210611_004222 ( 0:01): Running GPU type Tesla V100-SXM2-16GB

20210611_004222 ( 0:01): running a simple selftest...
20210611_004229 ( 0:01): Selftest statistics
20210611_004229 ( 0:01):   number of tests           107
20210611_004229 ( 0:01):   successfull tests         107
20210611_004229 ( 0:01): selftest PASSED!
20210611_004229 ( 0:01): Fetching initial work...
20210611_004231 ( 0:02): Running GPU type Tesla V100-SXM2-16GB

20210611_004231 ( 0:02): running a simple selftest...
20210611_004237 ( 0:02): Selftest statistics
20210611_004237 ( 0:02):   number of tests           107
20210611_004237 ( 0:02):   successfull tests         107
20210611_004237 ( 0:02): selftest PASSED!
20210611_004237 ( 0:02): Fetching initial work...
20210611_004239 ( 0:02): Running GPU type Tesla V100-SXM2-16GB

20210611_004239 ( 0:02): running a simple selftest...

I had to terminate so I would not run out my time allotment. :-(

chalsall · 2021-06-11, 13:46

Quote:

Originally Posted by RichD

I just happened to latch on to a covenanted V100 GPU but got the following.

Damn. Sorry. SPE. Fixed.

I would have *loved* to have seen that throughput!

LaurV · 2021-06-12, 05:26

Quote:

Originally Posted by chalsall

Damn. Sorry. SPE. Fixed.

I would have *loved* to have seen that throughput!

Nope. Not fixed. This is strong related to what I was describing here. Maybe you didn't see that post. Your toy works well as long as each assignment lasts longer than one comm cycle. But for few fortunates who have access to faster cards, and want to work lower bitlevels or higher exponents (i.e. short time assignments), it is a hell. When we are lucky to get a good card where the assignments last less than few minutes (one communication cycle in your script), then the card will run for less than one cycle (i.e. it completes the assignment in less time than the cycle lasts, emptying the worktodo.txt file). Then, because the worktodo is empty, the mfaktc will stop. Then the second cycle of comm will detect that there are "less than one" assignments in worktodo, and it will bring a new line to worktodo.add, but because mfaktc is stopped, there is nothing to handle the transfer from the add file. Then the third cycle will add this to worktodo.txt, so in the fourth cycle mfaktc (which attempts to start every cycle, by the script) will work again. This way the card works less than one cycle in 3, so about 70% or more of the time is lost. The solution is very simple, just change that freaking line to two, or three, or five, or more. I have to do that by hand every time I restart your script, just to have the right flow of assignments. This is bothering, and that's why I moved off from your script for now, handling the assignments by myself. In fact, you should allow us to chose, say, between 2 and 5 assignments at a time (could be more, but 5 is kind of enough to avoid hanging a lot of expos when the Colab dies, however, if I factor large exponents to 73, or so, it will only take one minute for each, so it could be more than 5 per cycle, maybe set to max 10?). So, you should add an input box, same way as for the "access key" and "logging level", to say "max number of assignments at a time" and let me pick between 2 and 5 (or 10), in fact, only 3 options, like 2, 5, 10, would be enough. Having only one assignment is non-optimal, as I explained in the other post. Since you implemented the ".add" step, there will always be dead times between cycles, for short-cycled assignment.

chalsall · 2021-06-12, 15:16

Quote:

Originally Posted by LaurV

Nope. Not fixed. This is strong related to what I was describing here. Maybe you didn't see that post.

Hmmm... No, I didn't see that post. Thanks for bringing this forward.

Somewhat ironically, I had brought it down to one at a time at the request of several Colab users. As you know, it's a single variable in the script.

I will look at increasing it to four for the short term, and make it settable when I'm able to invest some time in the UI code.

2021-05-31, 21:10	#5172
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³×53 Posts	I have the assignments page bookmarked; I go there a few times a day. Today I hit the button and it just spun. I tried just the root: https://www.gpu72.com and it just spun. Eventually I got the error message in the title, I'll attach a screenshot here. I'm glad to hear you were able to get to it 0 Maybe it's just me, but obviously I can get to the forums ok. And yes, I have tried refreshing the page. Even now I just hit F5 again and it is spinning. Maybe it's a cache on my side,, I'll check. Attached Thumbnails

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2021-05-31, 21:31	#5174
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³·53 Posts	Huh, I shut down all of my firefox windows in both desktops, and restarted firefox. Unexpectedly, I still can't get to gpu72. I'm not worried, I was just checking to see if my last assignment had dropped off, but it is interesting that I can't get to it in firefox. - Here I'll try Chromium.... Wow, I can't get it to load in chromium either. Weird.

2021-05-31, 22:07	#5177
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 1A8₁₆ Posts	For reference, I was able to get to the site when I VPN'd into work and tried from there. I sent Chalsall my local IP and the work one.