PrimeGrid: Race to the New Moon challenge starts on May 20th, 19:43 UTC

Following the lunar cycle, PrimeGrid is giving chace to the new moon which should appear later this month. To do so we're hosting a challenge on the Cullen LLR project, on which no challenge has been before. The challenge will start on May 20th, at 19:43 UTC and last 8 days. We invite you all to participate and look forward to seeing you both on the forums and in the statistics.

More...

Turned back on.

For kicks and giggles I thought I would throw some of these WU's and see how quickly I could process through a few. Needless to say the overclocking causing glitches is an understatement, while none have errored out I'm up to 50+ hours on a few of them and am finally considering just voiding them out.

AMD-8350 oc'ed to 4.8ghz

Seems I found an Intel Fanboy. I'll stay civil, I promise :D

The completion times on these WU's are just off the chain IRT trying to use the AMD processor. I guess I know not to give any more attention to Primegrid CPU challenges in the future. The TITAN loves the Primegrid WU's though so I'll throw in if needed on that front.

I have my i3 on them right now.

Sarge104, since the beginning of development of LLR client (with prime95 library) the Intel processors were always faster than AMD ones. We had a rule, sieve with AMD, LLRing with Intel. So nothing against AMD's processors. The problem of PrimeGrid is that those guys don't understand what a hell they are doing, they don't understand the math behind prime search. They don't read the papers of the math involved. Put a technical question and you will see they will fail on answering.

Quote:

---

Originally Posted by pinhodecarlos

Sarge104, since the beginning of development of LLR client (with prime95 library) the Intel processors were always faster than AMD ones. We had a rule, sieve with AMD, LLRing with Intel. So nothing against AMD's processors. The problem of PrimeGrid is that those guys don't understand what a hell they are doing, they don't understand the math behind prime search. They don't read the papers of the math involved. Put a technical question and you will see they will fail on answering.

---

I was actually able to get some useful info regarding the AMD chip process from my discussion with the Mod there, but it is like you said with completion times IRT AMD/Intel. I'm experimenting with just running the LLR's on a few cores to see if they complete any faster. I was curious to see the completion times and unfortunately chose a WU that took forever to complete to find out.

I wasn't going to just dump the WU's but was interested in why the processor didn't work the way I thought it should and the Mod was able to explain it. All and all he was very helpful and I PM'ed him with a thank you message.

Did he talk about the algorithm?

Quote:

---

Originally Posted by pinhodecarlos

Did he talk about the algorithm?

---

Of course not, anything I would've seen probably would've been lifted from wikipedia ;), but no, my questions were directed toward the process AMD uses to allow cores to use AMX instruction sets.

With the 8350 sharing instruction sets over all the cores at once I'm seeing a definite decrease in times with less WU's active at once. Times went from 401,700.23secs/111hours(8WU's at once) to 296,329.46secs/82hours(6WU's at once). Now that I'm down to the last five wu's I'm on course to finish up within 70 hours on the latest one.

With only four physical cores on the 8350 I'm wondering if just trying four at a time would allow for reasonable times on the WU's. I'm not interested in getting down to Intel times but just making the CPU work efficiently. In the past I was mistaken in the fact that there were eight physical cores, so I'm wondering if limiting the WU load would provide similar benefit on other PrimeGrid sub-projects.

WU's worth 743 hours still pending, 464 hours worth have been invalidated. Nothing has been validated atm. No error-ed out WU's but purged the rest due to the long run times and another upcoming challenge .

The FX-8350 really struggled fully loaded at 111 hours per WU, with just the physical cores loaded (4) the times went down to 60 hours at the end. Ran out of time to finish any more WU's so I'll never know the best run times for the chip unfortunately.

Today the tower will be down for maintenance, will load up this weekend after cleaning is done.

These monster long WU's are always a killer.