EVGA RTX 3060 12GB: $175 Sold!
Note: One of the fans on the 3060 _does not work_. It's not locked up (spins freely with your finger) but doesn't spin with the motor. Card obviously runs warm, but has run fine like this for something like a year. You can either ignore it, or replace the...
RAM Prices got you down? Well, for less than the cost of 32GB of DDR5, you can get this entire combo, which comes with 32GB of DDR4.
Included:
- 2x Xeon E5-2699v4 22c/44t CPUs. The very best LGA2011-3 CPUs you can get
- Supermicro X10DRI eATX motherboard with IO shield.
- 4x8GB DDR4 RDIMMs
-...
Yup I get that...I forgot to get my 9950Xs migrated over to this, then I was looking at my points and thought, "Huh, that looks lower than expected" and figured out my mistake. At least it's seven days so a day doesn't matter so much as for the three-day ones.
I've got it running on most of my systems. Not on the Zen 2 or the 9734 since they don't have a cache layout to make it work. But it's doing well on the other systems. And good showing for the Team so far. Third place!
For those not in the PG discord, I found a SR5 prime today. 3.8 million digits; currently in the queue over at T5K for final verification before it shows up: https://t5k.org/primes/page.php?id=140585
After running almost nothing but SR5 for about two months, I'm really excited for this :)
Still no power here, and ETA has been pushed back to tomorrow night at this point. Still clinging on to my jerseys for now, but I'm going to need my power back if I'm going to hold them much longer.
I don't know exactly what's up. Things started acting up yesterday morning while it was merely raining. And then went out pretty much completely in the afternoon once the wind picked up. The utility is reporting a lot of weather-reported outages across their service area, which given the...
Guess I was overly confident. Power has been out at home for about 24 hours, and the electric company says it'll be another 2.5 days before the expect to have it back. Might still be able to sneak out a jersey at the end of the month, but having everything off for three days does not help...
I think you're not gonna find a lot of people running the Sieve stuff during TdP. Since there are time-limited badges to get, most of us are pursuing those instead.
I pulled a GFN 18 this morning, on top of a bunch more PPSEs over the past couple days. At this point I'm feeling pretty confident about landing at least one of the jerseys that are up for grabs. And getting highest prime count + largest total score seems decently likely if someone doesn't...
My guesstimate is that's (on average) about 110K tasks for SR5 to get a prime. I'm at a bit under half that, so once TdP concludes, I'll be switching most of my CPUs back over there to keep grinding it.
Almost certainly 1 task running on all 8 cores will be ideal. My Zen 2 EPYCs have the same size cache (16MB per group of cores) and that's how they run best. You can try other arrangements but I would be surprised if they beat that.
Negative. All of the cores will pitch in. On my 7950X...
Preferences page, then select from the dropdown:
For best results, also turn off HT or set BOINC to use no more than 50% of the CPU cores so it uses "real" ones and not the SMT cores.
If you have a multi-CCD AMD CPU or a dual-socket system, you'll also want something to keep tasks pinned to to...
That's the way the morning ought to start off :)
And also gets me up to 4 GFN17 total, so I'm going to switch the CPUs all over to PPSE and push that badge count up as far as I can get it.
Which Linux are you running? I have noticed that on Linux Mint 22, the client won't start properly headless. On 21.X, it works fine. So I downgraded one system from 22 -> 21.3, and I will do so again with the other two after TdP.
Different PG projects have wildly different cache requirements. PPSE needs just over 1MB/task. SR5 10-11. SoB needs 32. So it's useful to know how much cache your CPU has, and often run only one subproject at a time so you can set exactly the right number of cores to run it optimally...
If you're going to keep running SR5, you almost certainly want to give it more than 1 CPU core per task. These tasks take 10-11MB of cache each, and while it can work by fetching everything from main memory, it's (much) faster if it can get what it needs from CPU cache instead. Probably...
I got an email about another GFN17 this morning, but still not a single PPSE. I'm up to about 125K PPSE tasks since the challenge started, so that's odd stuff -- it "should" be about 5 or 6 of them by now. I'm going to keep pushing for one more '17 to bring my Mega count for this year up above...
Wow. That's a far bigger ordeal than it was looking like from the Slack messages. Sorry to hear about that. Shit timing with all of the other fires out there doubtless making rebuilding slower. Yet, I'm sure you will persevere and make it through it!
And, the result so far is I got all of nothing from all of that bunkering. About 55k tasks submitted, a mix of PPSE & GFN17, since the event started. 10k are still pending validation, but the odds of finding anything in that if the first 45k didn't, not ideal.
My hacked up version of BOINC...
Five hours in, and the damn things are still uploading! One of my GPUs was sitting idle and two CPUs were approaching it, so I hacked up BOINC a bit to keep fetching work even while it's swamped with uploads. So all of the hardware is back to work again. I did not see it taking anywhere near...
Trying to upload all of these tasks is beating my internet connection to a pulp. The GFN 17s are "just" 4MB each, but when trying to do tens of thousands of tasks...well, there's still a lot to go. Guess that's what I get from the very cheapest tier internet access :rolleyes: