[QUOTE=Prime95;392299]While Chris is healing, you can use [url]http://mersenne.org/manual_gpu_assignment[/url][/QUOTE]

This form required specifying an exponent range when choosing "TF for double-checks" and "Lowest exponents" with a bit level of 70. Is that a bug?

[QUOTE=Prime95;392299]While Chris is healing, you can use [url]http://mersenne.org/manual_gpu_assignment[/url][/QUOTE]

No disrespect intended, but choosing "TF for double-checks" with "What Makes sense" results in "Error code: 40
Error text: No assignment available for GPU trial factoring, cpu_id: 248633, user_id = 802"

Very possible I'm being stupid.

[QUOTE=Mark Rose;392303]This form required specifying an exponent range when choosing "TF for double-checks" and "Lowest exponents" with a bit level of 70. Is that a bug?[/QUOTE]

Not really, but we are open to improving the form.

What makes sense hands out the most priftable TF assignments on exponents that are expected to be handed out for DC or first-time LL in the next 90 days. That 90-day range is listed at the bottom of the page. One can override this by specifying there own range, bit levels, etc.

At this point in time, all the DC exponents in the 90-day range are either already assigned for TF or reserved by GPU72.

Short-term memory loss?
 

[QUOTE=chalsall;392289]GPU72 has no thoughts as to 75685481.

It doesn't "own" it, nor has it "worked" it.[/QUOTE]

Maybe. But it assigned it to me. I just want it to accept that it's done.

[QUOTE=Bdot;392321]Maybe. But it assigned it to me. I just want it to accept that it's done.[/QUOTE]

Hmmm... Weird...

The "GPU" table had it listed as "Returned without having been worked", but you are correct, the assignment table indeed had it assigned to you. Fixed this, and seven others.

Let me know if you, or anyone else, sees this situation again.

Is Spidee Stuck?
 

All DCTF Assignments I finished since 14:00GMT Jan 14 (about 60 or so) are still on my assignments page.

spot instances
 

[QUOTE=Mark Rose;388023]The g2.2xlarge is not such a good value. You only get a single GPU at 240 GHz-d/d instead of two GPUs at 280 GHz-d/d. For mfaktc, the g2.2xlarge's fair value in comparison is $0.05/hr.

The cg1.4xlarge is only available in us-east-1 and eu-west-1. Of those, only in one available zone in us-east-1 has the spot price been hovering around $0.135.[/QUOTE]
This discussion helped me a lot to cut down on the curve. Thanks, Chris, Mark!

Hmmmmm... However, along with GPU(s), you get a 3-yrs younger Xeon E5-2670 [I]with AVX [/I](and larger cache and faster memory interface). For a typical prime (or PRP) search project, E5-2670 is as as fast as the two quad-cores that come with the cg1.4xlarge. Also, lesser availability leads to more frequent node drop-outs (even if you set your bid to $0.20, the history shows that you will lose nodes, maybe as frequently as once a week).

So, ~$0.135/hr pays for:
a) 2x M2050 + 8 cores of X5570 CPU (cg1.4xlarge), or
b) 2x K520 + 8 cores of E5-2670 AVX CPU (2 x g2.2xlarge).
I've run both variants for some time. A little more management for variant b.

I think I'll pick up 10-20 of g2.2xlarge spots for my pet project (need to run some [I]Cyclo[/I]; and will pick up some slack on my G.-M. search and/or k=5 Proth that I had dormant for a while). For Cyclo, M2050 = K520, but the CPU power in variant b is twice as much:
[CODE]@cg1.4xlarge:
86990^262144 - 86990^131072 + 1 is composite [Res=11411658520]. ([COLOR=Blue]2132.3 sec.,[/COLOR] err=5.74e-03)
85217^262144 - 85217^131072 + 1 is composite [Res=11150025403]. ([COLOR=Blue]2128.4 sec.[/COLOR], err=5.69e-03)
10^249010-21 is base 3-Strong Fermat PRP! (249010 decimal digits) Time : [B][COLOR=DarkRed]591.348 sec.[/COLOR][/B]
Starting Lucas sequence
10^249010-21 is strong-Fermat and Lucas PRP, Starting Frobenius test sequence
10^249010-21 is strong-Fermat, Lucas and Frobenius PRP! (P = 4, Q = 2, D = 8) Time : [B][COLOR=DarkRed]2303.341 sec.[/COLOR][/B]
(Factored part = 100.00%)

@g2.2xlarge:
82517^262144 - 82517^131072 + 1 is composite [Res=10786891370]. ([COLOR=Blue]2109.0 sec.[/COLOR], err=5.14e-03)
82512^262144 - 82512^131072 + 1 is composite [Res=10823198411]. ([COLOR=Blue]2108.6 sec.[/COLOR], err=5.10e-03)
Starting probable prime test of 10^249010-21
10^249010-21 is base 3-Strong Fermat PRP! (249010 decimal digits) Time : [B][COLOR=DarkGreen]295.374 sec.[/COLOR][/B]
Starting Lucas sequence
10^249010-21 is strong-Fermat and Lucas PRP, Starting Frobenius test sequence
10^249010-21 is strong-Fermat, Lucas and Frobenius PRP! (P = 4, Q = 2, D = 8) Time : [B][COLOR=DarkGreen]1262.299 sec.[/COLOR][/B]
(Factored part = 100.00%)[/CODE]Here is how can prep any of these nodes, in a few lines:
[CODE] sudo yum erase nvidia cudatoolkit
sudo yum install gcc
wget http://us.download.nvidia.com/XFree86/Linux-x86_64/340.46/NVIDIA-Linux-x86_64-340.46.run
sudo yum install kernel-devel-`uname -r`
sudo /bin/bash ./NVIDIA-Linux-x86_64-340.46.run
# answer prompts
nvidia-smi
# ... and ready to run, either CUDA or OpenCL[/CODE]

The GTX 660 Ti is going offline for now. I am moving it to the new box I made. Until I get it up and running, the GTX 660 Ti won't be doing anything.

Now I get to fiddle with drivers and all that for a good little while.

Trying not to be a nag....
 

I know typing is still an issue ... so for some future reference ...

We seem to be losing more and more ranges from GPU Visualization.
Looking at this report ...

[url]http://www.mersenne.info/trial_factored_tabular_delta_7/0/0/[/url]

Everywhere that you see a -20,000 or more in the unfactored column (and there are several) a drill down will soon lead to a missing range.
Obviously we have NOT factored 20,000 in any range in the last week but rather we have lost entire ranges ... they appear as blank lines on the report.

Those are the ranges were the "not yet known" mersenne primes are, so they are hidden on purpose, Chris is doing everything possible to impede us finding them... He might be cooperating with curtisc for that, behind of our backs.

I was especially pissed off about disparition (why is this redlined? disappearance? aha!) of the first million exponents from those tables, but I gave up after a while, trying to explain to myself that getting rid of the 3-4 columns on the left of the table was useful, as it relieved other 3 more important columns on the right, so after some initial cry, I kept quiet. Well, the same effect would be achieved using columns labeled "<63" and ">74" for example, and everything would be more accurate, and the first mio won't be pushed out, but well, that would require some programming, and Chris [STRIKE]was grumpy and didn't want to do it[/STRIKE] was busy/not available.

Now that you mentioned it, I want my first mio exponents back! Grrrr... But Chris said already that gimps-visu site was never intended to be something serious, so I think we can frame our hopes for now, and put them on a wall..

[QUOTE=petrw1;393368]

Obviously we have NOT factored 20,000 in any range in the last week but rather we have lost entire ranges ... they appear as blank lines on the report.[/QUOTE]

They appear as blank lines on the Factored Depth report because they are counted as "Factored". If you go to the "Exponent Status" report and drill down to e.g. < 1M range you´ll see them all standing there as Factored.

I hear what you say about these reports. They were also a strong point of interest for me, and I still check them, although I know they are incorrect. Most ranges are still OK, though. It would be very nice to have them back in shape, but there are many things in life we would like to have and at the end of the day we have to do without... :whistle:

All in all, I guess LaurV´s explanation is the right one: heaps of Mersenne Primes are being hidden from us mere mortals, to keep us focused on finding the next one, with the wrong belief they are rare :smile:.