Please visit donation page to help the project cover running costs for this month

Sponsored by:

Hosted at and sponsored by Rackspace.

Join PrimeGrid

Returning Participants


Leader Boards



1) Message boards : Aggie The Pew message board (Message 115278)
Posted 21 hours ago by Scott BrownProject donor
Michael, with the Mountain Stage now over, are there any primes to be announced for this part of TdP?

Yes...almost as many as are showing. So the suspense continues...

2) Message boards : Aggie The Pew message board (Message 115248)
Posted 1 day ago by Scott BrownProject donor

We then debated the meaning behind the poem. I said as you that it was poem about death. She pointed out that Thomas' father did not die until 1952 and the poem was written in 1947. She opined that it was about Thomas' father's blindness.

The confusion may come from the fact that, although written in 1947, it was not published until 1951/1952.

In any case, nothing quite compares to the poet reading the verse himself (from a BBC broadcast).

3) Message boards : Aggie The Pew message board (Message 115238)
Posted 1 day ago by Scott BrownProject donor
"Do not go gentle into that good night,
Old age should burn and rave at close of day;
Rage, rage against the dying of the light.

Though wise men at their end know dark is right,
Because their words had forked no lightning they
Do not go gentle into that good night."

Dylan Thomas of my favorites! As I recall, Thomas wrote it to his father as the latter was on his deathbed.

...and for one of the stranger, but oddly funnier readings of this poem, I refer you to Rodney Dangerfield's lines as Thornton Melon in the fairly average movie "Back to School".

4) Message boards : Aggie The Pew message board (Message 115236)
Posted 1 day ago by Scott BrownProject donor
As much as I enjoy Tim's (twisted) humour, I have a tangential question for the group: if someone were to study computer science (undergrad and/or grad), which one between Cambridge and Oxford would you advise as being the better institution?

You are not getting many replies.

It is many years since I went to university. I didn't read computer science but I did decline a place at Oxford for my chosen subject.

There are various league tables for undergraduate courses published and both Oxford and Cambridge are highly ranked eg Imperial College was set up as the original UK technological university. As an added bonus you get to use the letters DIC after your name.

I would have thought that any of the top dozen UK universities would be superb places to study such a course. At that level I think it is somewhat subjective as to which is best and also most people won't be overwhelmed with pleas from such universities asking them to come and study. Oxford, Cambridge, Imperial and Durham perhaps have the added cachet of being internationally well known names.

For post-grad study I would have thought it far more important to think about the precise subject of interest rather than the general undergraduate reputation and head for the key people in that area.

No matter what the subject is, the most important aspects of graduate work are 1) the reputation of the graduate program rather then the university, 2) that the graduate program is known for the specific subject (that is, in something like computer science, a program might be #5 overall, but #1 in database programming...if interested in that subject, then you would preferably attend the latter), and 3) having a good match with an advisor (both to write good letters of recommendation and to also enable you to complete the graduate work to begin with). Even if #1 and #2 are in place, lack of a good advisor can negate both of them.

5) Message boards : Aggie The Pew message board (Message 115200)
Posted 1 day ago by Scott BrownProject donor

On the GPU side it is more interesting with GFN16. My two 980Tis seem to be not doing too bad. They're doing units in around 180-200 seconds, and are first over half the time, maybe 2/3 to 3/4, give or take. The 970 though takes around 240 seconds and is 1st currently 20% of the time.

This is likely more a a random draw of wingmen issue. My 970 does better than 50% on average, but for any given three day run even my 1080 can dip below that just by being randomly matched with some 1080s and 1080Ti cards, etc.

In general, I have found that GTX 960s are about at the 50% level (a GTX 660 Ti also comes close) with times in the 300 to 350 second range. Heck, even my pair of GT 440 cards wins about 2 out of 20.

6) Message boards : Aggie The Pew message board (Message 115191)
Posted 2 days ago by Scott BrownProject donor
Warming up for the mountain stage... literally. Deciding what to run, in what configuration, and it certainly is getting warm now...

With now over 100 PPSE tasks completed, I am running at a pace of about 73% in being the first to report based on confirmed and pending tasks.

As for our friend with the farm of 32-core machines, he is much easier to handle than I expected. He had me thinking about monopolizing prime finds but I am certainly beating him to the punch when reporting completed units.

People worry way, way too much about being first. You don't need fast computers. You don't have to cheat. There's only a few simple rules...

1) set your cache to 0.

Rule #1 by itself is sufficient because so many people do not follow rule number 1.

2) Turn off Hyperthreading.

3) (maybe) Use LLR multi-threading.

4) Sell the AMD CPUs and buy Intel for LLR.

Even with horribly slow CPUs you'll get a decent share of "firsts". By "horrible", I mean "PPSE takes almost an hour." With anything faster than that you should be getting better than 50%, so you're on the winning side of "fair".

That's for short tasks. For long tasks, the effects flip around and CPU speed becomes more important as procedures become less important. Of course, rules 2, 3, and 4 affect CPU speed, so those are always important. But if you're running with, say, a 1 hour cache, it will matter more when running PPSE than PSP.

Rule #2 should be taken with a little more detail. It applies to all CPUs I have been able to test if 1) you don't use rule #3 and 2) you are not running a GPU.

In almost all cases that I have experience with, you will do better overall when running a GPU by leaving HT on to feed the GPU and running your CPU% to match actual core totals (e.g., 50% to match 4 real cores on a Sandy Bridge i7, etc). Leaving those threads open also tends to reduce the effect on normal computer use for machines that are not dedicated crunchers (i.e., checking your email, web browsing, etc.).

I have also seen some reports and have some limited testing experience where, at least on certain CPUs, crunching across threads can increase throughput. My experience with this has been with older, pre-AVX servers where, for example, running 12 or 14 threads on a 10-core/20-thread CPU actually gives greater throughput than running on only 10 threads.

In other words, while rule #1 and #4 are pretty much always true, rule #2 and #3 require you to test your system to verify optimal settings.

7) Message boards : Aggie The Pew message board (Message 115154)
Posted 2 days ago by Scott BrownProject donor
Oh, yep, right so anyway my credit card PIN is the last 4 digits of pi.


I may just have to steal that as a signature line. It's brilliant.

It does beg for theft!!!

3.141592653589793238462643383279502884197169399375105820974944592307816406286208998628034825342117067982148086513282306647093844609550582231725359408128481 ...

Give me a minute... :P

8) Message boards : Number crunching : 2018 Tour de Primes (Message 115017)
Posted 5 days ago by Scott BrownProject donor

All the big hitters running SR5 for the red jersey are now shattered by Genefer.

While that's true, an SR5 find coupled with select smaller GFN16, PPS, and/or PPSE could still quite easily take the green jersey since SR5 is only a bit smaller than GFN18.

9) Message boards : Number crunching : 2018 Tour de Primes (Message 115005)
Posted 5 days ago by Scott BrownProject donor
Everyone switching to GFN-18 now or GFN-19?

This may give you pretty good idea about GFN in general.
GFN-15 is fading due to recent disabling of GFN15 on CPU.
GFN-16 has gained more attraction.
GFN-17-Mega as well, GFN-18 also but GFN-19 is lower.
In general, participants moved into middle ground: GFN-16 to GFN-18 where chances are best (realistic).

Actually, the GFN19 fade may be somewhat due to me moving off of that project after I found one near the end of January (all my GPUs were pretty much there for a while).

10) Message boards : Aggie The Pew message board (Message 114931)
Posted 6 days ago by Scott BrownProject donor
ooooh, crossing my fingers that I might have found something... I have that mysterious condition where the count of GFN primes found on my account page does not match the number shown on my primes page... come on, big money, no whammies!!

It was a pleasure being your DCer. :P

Next 10 posts
[Return to PrimeGrid main page]
Copyright © 2005 - 2018 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 5.33, 5.12, 4.76
Generated 20 Feb 2018 | 21:53:27 UTC