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

Toggle Menu

Join PrimeGrid

Returning Participants


Leader Boards



1) Message boards : Number crunching : International Women's Day Challenge (Message 160963)
Posted 84 days ago by Profile bill1024
Thirty four primes will be found.
2) Message boards : Number crunching : Tour de Primes 2023 (Message 160724)
Posted 96 days ago by Profile bill1024
The amount of GFN-16 being done continues to be unprecedented. The first GFN-16 prime of TdP 2023 had a b of 186 million. We're currently above 241 million. That's an increase of 55 million in less than 23 days. By contrast, it took 10 years to do the first 55 million in the project. In fact, we were at b=55 million only 3 years ago

I guess that's thanks to cloud computing.

I would think it is because more people are running it since it is the easiest prime to find for the TDP. Couple that with the fact GPUs have become much more powerful.
Prices of RTX30xx GPUs are way down, used GPU prices are way down.
Isn't the price to rent a GPU a lot more than a CPU core? I never priced it but I would think so.
3) Message boards : Number crunching : Tour de Primes 2023 (Message 160627)
Posted 100 days ago by Profile bill1024
There is only three types of persons on PG
One is searching primes. dont care about badge and credits
Second try to get as more credits as can, dont care about primes ( they usually run PPS sieve) so they cannot find primes as all, but got alot of credits.
Third kind is happy with every prime discovered and every badge they get.

From my point of view: many years PG was unfair, because I have slower hardware then others so only fastest find primes ( there was some exceptions but extreme small percentage)
So I run my own search because at home I was always first.

Then I got few GPU , and PG enter new era: with no DC, everyone is first. I return to PG because now it is fair game: have fastest card or slowest one if I take workunit with prime , prime is mine.

I also use my own hardware, because I already calculate for next year that only mega primes will do in TdP 2024, and for that I will need near whole month on GFN17 or PPS Mega. with unknown price of electricity next year, that badge could be very expensive. ( if rules not changed in meantime)

But, if I had a lot of money, and I wish to win Jersey , then I will rent some computers to archive that goal. Dont blame people who has more money then you have, they can buy more computer power you can. At the end we all live on same globe: some have more money then others. So I am at Scott side. If he can afford something you cannot it is your problem not his.

"Third kind is happy with every prime discovered and every badge they get."

But I use my own hardware and pay my own Electric at home and not at an office or at a school.

That's how BOINC has changed over the years with cloud computing or renting pc's bcavnaugh.
People can effectively "buy" their way to a badge. The mountain stage is a classic example of that. (boss has made a good example of that this year) Sadly that's just the way it is mate.
I don't sweat it. I just accept it.
I'm thankful to have found 8 this year and I'm not even really looking for GFN-16, it's just that my old Radeon VII's aren't good for much else imo. My NVIDIA hardware has done bugger all so far this year. Ironic isn't it! :D
I'll take the win for AMD lol.

So if I go to the store and pick up 5 RTX3090Ti, 3 RTX 4080 and a couple RTX4090 a few motherboards with processors with AVX512, power supplies and spend 20 grand.
And I win a couple jerseys, is that buying my way to a jersey/badge?

Lets not worry about what other people have/do with their money and worry about ourselves.
Time to get over this and get to finding primes.
There is an old saying in drag racing. Race what you brung and hope you brung enough.
I don't care what sport your doing, hunting primes or racing, the deeper the pockets...
And lets not forget lady luck. There is no skill here its all luck getting the prime sent to you in the first place. Either you're lucky or your not.
4) Message boards : Number crunching : Tour de Primes 2023 (Message 160173)
Posted 113 days ago by Profile bill1024
With resource set to 0 I don’t think you can get a cache built up, I can’t anyway. It waits for tasks to finish before getting new tasks in case other projects want work. I guess even though the other projects are set to not get tasks.

I bet the PG servers are getting slammed with all the small 3 second and other small check tasks being constantly sent in .
5) Message boards : Number crunching : Tour de Primes 2023 (Message 160169)
Posted 114 days ago by Profile bill1024
Bill, I have it set to resource share 100, unlimited tasks, 9cores per task (running big prime search) one open for the GPU. 10 core cpu and one GPU.
Then on my computer I have it set to 0.01 cache use 100% and HT is off in the BIOS.I get plenty of GPU and CPU tasks.

The big question is why is your connection slow? Try resetting the router/modem, have a bad wire connection or a bad patch cable swap cables, have a different router to try, connect to the modem directly. I have no idea what's wrong with your connection but I think it is on your end as my up and downloads are fine. Maybe your ISP is having an issue on their end.
Using a VPN? Maybe try it without and see if its any better.
Do a trace and see where the delay is. Something is not right.

Side note: I've noticed the GPU uses one full core under Linux but under windows it says a lot less CPU used for the GPU tasks. Just seconds vs. the full run time under Linux. And I am getting no more warnings on my one x299 in win10, with LLR2 tasks, not sure why Linux was acting up.. Think I am switching back over to win10 for PG, Linux was not much faster and was slower when there were errors with LLR2.
6) Message boards : Number crunching : Tour de Primes 2023 (Message 160160)
Posted 114 days ago by Profile bill1024
It seems this year I am spending more time uploading and reporting and downloading then really running tasks.

Get a small cache of day 0.1 day.

I have never gotten a Prime doing a cache of pending tasks.
Most of them end up be others error or aborted takes from days before.
We have to many members that download only to about them at a latter time.

Bill it does not matter any more if you get someone's aborted task since you do not have to be first anymore. Used to be if you received aborted tasks it was a better chance of not getting a prime because the other person would return the task first. That is over with.
The person who aborts 1000 tasks may very well be aborting primes that will be sent to you.
Or better yet me! That person has no way of knowing if it is prime or not unless they crunch it. So again aborted tasks have just as much chance of being prime as any other task not processed.
Getting sent a prime is all luck. But the more to do the better are your chances.

It's kind of hard to error tasks with LLR2 since it seems to correct its self and start at the last point again. You will see the task with all red letters with a warning in red saying there is a problem.
And even if you get their errored out task it still can be a prime since it does not have to be returned first. Either you get a main task that may or may not be prime, or a short check task.

So yeah you want to have a cache of tasks so you do not have to wait for new tasks to upload and download. The more tasks you process the better the chance of being sent a prime. You can process the prime task and turn it in three days later and you're still the founder.

I am seeing several check tasks for every main task I do, to bad it can't be given out so it is one to one.
But its not all that bad, so I do 10 checks a 9 seconds each, that's 90 seconds and one main that runs for 6 or 7 minutes. So I am processing more time on main tasks but the ratio does seem to be out of sorts.
7) Message boards : Number crunching : Great Internet Fermat Prime Search Challenge (Message 158162)
Posted 182 days ago by Profile bill1024
When I click the link on the main page for this challenge it takes me to a different thread.

Some one should correct it.
8) Message boards : Number crunching : Prime Meridian Day Challenge (Message 157653)
Posted 208 days ago by Profile bill1024
My I9-9900k Coffeelake does the woo 33,000 9.1 hrs.
The 4.4 to 5 hour woos are on an I9-7900x 10 core with AVX512.
That makes a huge difference!

The i7-5960x did the woo in 8.78 hours. Same size FFT
Somewhere around 1500 seconds faster.
9) Message boards : Number crunching : Prime Meridian Day Challenge (Message 157651)
Posted 209 days ago by Profile bill1024
My 7900x is running at 4ghz multiplier is 44-4AVX512 offset

The Haswel 5960x is at 4ghz also. No AVX offset.
10) Message boards : Number crunching : Prime Meridian Day Challenge (Message 157647)
Posted 209 days ago by Profile bill1024
My I9-9900k Coffeelake does the woo 33,000 9.1 hrs.
The 4.4 to 5 hour woos are on an I9-7900x 10 core with AVX512.
That makes a huge difference!

Wow AVX512 performance is awesome!! Seems like the 13.752MB L3 cache size in i9-7900X doesn't affect the performance much. As a comparison my i7-6900K ETA for cullen task (8 core) is expected to finish in about 10 hours or so. It has 20MB of L3 with AVX2, more than i9-7900X (13.75MB). So really the AVX512 instruction set is the dominant factor here. I wonder what's the speed difference if both has the same AVX instruction set and how the L3 cache really affects the performance and perhaps the number of memory channels. Both supports quad memory channel. I have four sticks.

The 7900x skylake have a larger L2 per core and somehow it can be added to the L3 cache for a larger total useful cache than it would appear to be, so can't really call it just 13mb, add in the L2 also, there is 1mb per core 10mb + the 13 = 23mb useful Lxx cache??? I do believe that's it.
xii5ku touched on skylake CPUs a little bit in his post about skylake and zen GPUs how they use cache.
I am pretty sure there are threads on here where it goes into better detail than I can explain.
But yeah in a nutshell AVX is awesome!!! It has to be the CPUs with 2 sets of AVX512 I believe some of the lesser CPUs have one instance of it.

EDIT: My 8 core 5960x haswell on a woo 2304FFT is at 68% in 6hrs with 2hr50min to go
11.1% per hour it says in the tasks properties.
just under 9 hrs vs your 10hr? But I not know what FFT size is your task.

And I have four 8gb sticks at 3600 17,21,21, something or other with the 7900x

Next 10 posts
[Return to PrimeGrid main page]
Copyright © 2005 - 2023 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.20, 1.41, 1.54
Generated 31 May 2023 | 2:41:55 UTC