20140830, 17:57  #156 
Jun 2014
1111000_{2} Posts 
Throwing my cores at some ECM on M7508981, are B1=50000 and 250000 already reserved? In an LMH way, that is, not necessarily on the server.

20140830, 19:17  #157  
Aug 2002
Dawn of the Dead
5×47 Posts 
No problem and don't worry about reservation. I work through the manual system and it hands out ECM to any takers (obviously not many) and the number of curves required permits this behavior.
What bounds might I use for the suggested range? How many curves? I need to test this, but a good approximation is a recent find of an ECM factor for M9178789, in about 150 minutes per curve on a machine a third as fast as my current one. I have to finish pending assignments but could start on two cores in perhaps two weeks. After that, who knows? I got more cores and like risk ... Quote:
The manual system wouldn't give me this, even though it hands out multiple assignments in my current range, M25000, in batches of 15 20 which I then recombine into a single entry. Perhaps it sense the "official" reservation? Regardless, I can do this the LMH way and eventually get a valid reservation to get banked credit. I'm ready, need bounds ... Last fiddled with by PageFault on 20140830 at 19:32 Reason: Not being sloshed enough 

20140831, 01:03  #158 
Aug 2002
Dawn of the Dead
EB_{16} Posts 
A little tomfoolery ...
I grabbed M7500049 and it runs at 81 minutes per curve, using a maximum of 8142 MB ram in Stage 2 (more was offered, she didn't bite). Yikes! The performance delta is awesome. On a workbox, HP blandware i3 3.4 GHz, similar work takes 160 minutes. Now running is i5 4690 (why does the client report the speed as 3255 instead of the detected 3500??). Ram is HyperX 1600 cl10, 16 GB. All at stock on Asus Z87 Pro (I haven't looked under the hood yet). I get a huge performance hit, with this ECM running, on my LL tests, 2 running in M69xxxxxx range: 20 % slower! Yikes! I see in the output for the ECM: FMA3 FFT length 384k, Pass 1 = 256, Pass 2 = 1536 whereas I normally have 2 cores running much smaller ECM (the void), but these (M25xxx) show the line: FMA3 FFT length 1280k Anyways, the LL will be done in the next few weeks and I don't want any more at this time. Hence, two cores on some of this potentially interesting work and the other two "in the void", where I have found a dismal 1 factor in about 35 000 curves. Ahh grasshopper, computer performance ... I recall the Age of AMD, and my Axia's running at 1400 +, even prime95 stable ... this Haswell might outwork one of those in a week, compared to a full year. Back then, we were doing M13xxxxxx and TF was at M15xxxxxx. What a difference a decade and a half makes ... 
20140831, 01:11  #159 
Aug 2002
Dawn of the Dead
5·47 Posts 

20140831, 02:14  #160  
Apr 2014
7·17 Posts 
Quote:


20140831, 02:16  #161 
Apr 2014
7×17 Posts 

20140831, 02:53  #162 
ἀβουλία
"Mr. Meeseeks"
Jan 2012
California, USA
3^{2}·241 Posts 

20140831, 08:44  #165 
Jun 2014
170_{8} Posts 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
New P+1 record factor  akruppa  Factoring  5  20071101 16:47 
Greg Childers finds Record P+1 Factor  wblipp  ElevenSmooth  9  20051227 20:18 
Record ECM factor found  philmoore  Factoring  10  20050227 09:38 
ECM Server for Record Size Factors at 8195  wblipp  ElevenSmooth  1  20031125 15:47 
Record ECM factor found by Prime95  philmoore  Lounge  0  20030624 20:41 