GPU to 72 status...

[storm5510]

I just looked at the recent results on mersenne.org. There are a lot more wavefront P-1's being ran now than in previous days and weeks. Somebody must have stirred the pot a bit.

About P-1, I run them with gpuOwl. I seem to recall an instance where gpuOwl did not want to run a P-1 because it did not have an assignment ID. Of course, only Primenet assignments have this, I will try a single P-1 from here and see how gpuOwl handles it. If it refuses, I can always run it with CUDAPm1 or Prime95.

[James Heinrich]

Originally Posted by storm5510

I seem to recall an instance where gpuOwl did not want to run a P-1 because it did not have an assignment ID

I recall seeing somewhere in gpuowl output the AID set to "0" (zero) when there was no assignment. I have no idea if this would also work on the input, but you could try it if gpuowl refuses to run without it. Or ask in the gpuowl thread.

[Awesomeotts]

Originally Posted by chalsall

One thing I've noticed, though, is P4s appear to no longer be in the fleet. I haven't seen one for weeks.

Really? I've had one or two over the past week or so

[storm5510]

Originally Posted by James Heinrich

I recall seeing somewhere in gpuowl output the AID set to "0" (zero) when there was no assignment. I have no idea if this would also work on the input, but you could try it if gpuowl refuses to run without it. Or ask in the gpuowl thread.

The test from here had no problems running. gpuOwl probably expects a specific number of comma separated values in an assignment line. It said "N/A" where the ID would have been.

[kladner]

Originally Posted by Awesomeotts

Really? I've had one or two over the past week or so

I saw a P4 last night when I was running some extra notebooks on a free login. It still seems pretty easy to get T4s on free runs.

[storm5510]

I just checked my assignments on GPU72.com. I have, at least, 50 TF assignments which seemed to have slipped by and did not get ran. I need to get them out in a list or some other format so I can run them. How?

Edit: Disregard: I found it at the bottom.

[kriesel]

Originally Posted by James Heinrich

I recall seeing somewhere in gpuowl output the AID set to "0" (zero) when there was no assignment. I have no idea if this would also work on the input, but you could try it if gpuowl refuses to run without it. Or ask in the gpuowl thread.

yes it works to enter 0 as AID, just so the input parser has enough commas to put k,b,n,c, etc into the right variables. There's a reference post for that.

[storm5510]

About my missed assignments. There were 59 in all. The fault was my own. I use Mitfit's batch files, but not Misfit. I can reserve them automatically. They must be submitted to Primenet manually. I forgot to submit them. Thud!

[kriesel]

Originally Posted by storm5510

I would have never thought of that.

About Alaska. Municipalities must have their own private power stations. Fairbanks has a population of around 31,000. Nome is around 3,600. Neither appears to be on a grid system. Elmendorf AFB is near Anchorage. Both are on a grid. I have a brother that was in the USAF back in the '60's. He said it got so cold at Elmendorf at times that the jet engines would not fire. Something about the air density at really low temperatures.

Vacationing many years ago driving south from Cancun, following the progression paved roads and electric overhead wires, to standing bare posts, to posts lying next to postholes, to posts lying on the ground, to unpaved gravel, jeep with 4 soldiers checkpoint at an intersection, village with a bit of paving and small grid, back to gravel and then dirt, and finally an oceanside resort with a windmill and ~12hr/day generator. The main building was wired and provided first and last meals of the day. Guest quarters were small one room structures; walls with shutters to close the window openings, a door and roof, no wiring or plumbing or glass.
Over the decades I've noted a lot of voltage creep in the US midwest. I've seen old equipment marked 108V, 100, and 115. Now it's nominally 120V but a voltmeter in an outlet will often indicate up to 124. That creep is very hard on incandescent bulbs still in use, since life is inverse ninth power of voltage.

Jet engines operate on something close to diesel fuel. I suspect as an engineer with some engine research background that the issues with jets starting related to fuel vapor pressure or the lack of it, and fuel gel temperature. It's hard to pump a gel out of a fuel tank. https://en.wikipedia.org/wiki/Jet_fuel
The residence time of fuel in a jet engine combustor section is very short. Fuel vapor pressure is almost exponential with temperature while well below atmosheric pressure (boiling). The air density varies much less rapidly, as the reciprocal of absolute temperature. It probably requires more starter power with cold dense air.

Getting the engine to start and run is not the only issue. Assuming the vehicle has tires not skis, there's also the consideration of the tire compound's glass transition temperature. I recall hearing many years back of a particularly cold spell in Ohio, where instead of car tires flexing as they rolled, they broke. https://en.wikipedia.org/wiki/Glass_...ion#Polymers_2 Guess which we'd use for Orings on projects headed for the Antarctic's South Pole Station where outdoor temps may be well below -100F: https://mykin.com/rubber-temperature-range. During the colder supply runs there, the military pilots don't turn the engines off and chance getting stuck there for months. Sometimes they don't even taxi to a stop, instead just pushing the cargo out the back while rolling lightly on or flying a few feet over the bulldozed snowpack runway. It's called "drifting" the cargo. The crates slide to a stop eventually.

[chalsall]

Originally Posted by kriesel

Over the decades I've noted a lot of voltage creep in the US midwest. I've seen old equipment marked 108V, 100, and 115. Now it's nominally 120V but a voltmeter in an outlet will often indicate up to 124. That creep is very hard on incandescent bulbs still in use, since life is inverse ninth power of voltage.

I *love* discussions like this!!! "Normals" just have no idea...

To talk about voltage creep... Barbados has been quite slow in migrating to RE, because the incumbent monopoly provider (BL&P) has hundreds of millions of dollars they need to amortize off their books. It's taken years of pressure before we /started/ getting serious about PV.

Attached are three temporal graphs of Voltage plotted using an SNMP stream out of a Dell 208V UPS. As you can see, BL&P *never* provide the voltage actually contracted for. It is, however, *mostly* within the allowed +/- 10% range.

For context, this was at a client's location in Bridgetown, Barbados. Across the street from the Queen Elizabeth Hospital, if anyone's interested in doing some Google Maps.

Several days out of the month the voltage would actually drop below the contracted minimum. I complained to BL&P on behalf of my client, because it was not the ideal situation for the poor little UPS to have to deal with. They sent out their engineers with their own instruments (they wouldn't trust my data, of course), agreed that it was out-of-spec, and then said there was nothing they could do about it. (!)

As is clear from the graphs, the minimum voltage was always seen right around noon, but only on weekdays, not weekends.

The reason, of course, is all modern office buildings here in Barbados have electrical HVAC kit, working like crazy trying to cool the buildings. And, because we didn't have much PV back then, Mr. Sun was working solely against us at that time of the day, rather than helping out.

Thankfully, private enterprise is beginning to seriously invest in PV here, and feeding it back into the grid. Although (IMHO) it's still not happening as fast as it should be, because our regulator is completely incompetent, and so the incumbent is still in a more favored position than they should be.

(Friends of mine and I call the BB.FTC "Favouring The Company" and/or "Fsck The Consumer".)

[kriesel]

Originally Posted by chalsall

The reason, of course, is all modern office buildings here in Barbados have electrical HVAC kit, working like crazy trying to cool the buildings. And, because we didn't have much PV back then, Mr. Sun was working solely against us at that time of the day, rather than helping out.

One of UW-Madison's newer buildings has motorized vertical blinds. It's more efficient to run lighting in the daytime than pump the broader solar spectrum's heat load out with HVAC. So the building automatically closes and opens the light colored blinds to manage radiant heat load and total energy consumption through its aluminum and glass facade.

agreed that it was out-of-spec, and then said there was nothing they could do about it. (!)

Of course they could do something about it. Either trigger any out-of-spec price discount clause, or install corrective equipment on the client's premises.
What does line voltage at the meter look like at these times?
How interested in rooftop solar is the client, since the local utility can't be relied upon to keep its promises?
(Redundant power switch; utility as backup to their own plant. https://www.treelinepowersystems.com...rATS_p_24.html)

https://downloads.dell.com/manuals/a...ide5_en-us.pdf 5600W high efficiency line interactive UPS, yum, I'll take two, as reward for this idea:
Get a suitable autotransformer to step UPS input up to spec, or get an electrical shop to add some handy 85, 88% and 92% taps on something like this: https://www.solaris-shop.com/outback...r-w-enclosure/ Then use it as a step-up transformer.

175V (your plotted min)/0.85=206V; monthly (max+min)/2=191.355; / 0.92 = 207.995V.
Monthly max 202.79/0.92=220.42V, 1.06- times nominal; 179.92min/0.92=195.57V, 0.94+ times nominal 208V. So 92% looks right.(If load grows, or utility power worsens, with time, lower taps may come into play.)

Probably less costly than a whopping big VARIAC https://variac.com/staco_240_1ph.htm which would need mechanical safety stops but could have slow motorized servo adjustment of output voltage.
(As a mechanical engineer with electronics and farm-kid background, I once balked at the price of a Varian Ti-ball titanium sublimator pump power supply/control unit, bought a cheap stick arc welder from the local Farm & Fleet, and modified the output winding myself by unwinding about a third, saving enough to buy additional Ti-ball heads as spares. It was used intermittently under manual operation. It lasted the 25-year life of the vacuum system it was used on. The only issue was it still had enough power to melt part of the inside of a Ti-ball and permanently short the Ti-ball when it flowed and froze around the internal filament lead. Slow ramp on a new ball, no problem. Lesson learned in one costly attempt. http://www.emcgrath.com/product/vari...-ball-control/ http://schlom.mse.cornell.edu/sites/...dfs/%23355.pdf)
Veering more than a little off topic though from GPU72 status.