GW_8_337 poly
 

Here is a better than average one:
[CODE]# expecting poly E from 2.24e-12 to > 2.57e-12
R0: -1065438607826297238301895836221
R1: 2298089318957281
A0: 12334085463005489537130672600655445528
A1: -42336223293802175473724376312026
A2: -2922508932494730842315681
A3: -9850672644688730851
A4: -1320582894819
A5: 526680
skew 3281156.76, size 3.704e-15, alpha -7.497, combined = 2.466e-12 rroots = 3[/CODE]

W_2_802
 

It has been a while since I have done any postprocessing just wondering will any of the following W_2_802/806/809 or 139_133_minus1 fit into under 16 gig of RAM & approximately how long would it take to complete? I have a Haswell E 5960 X at 3 GHz 8 cores/16 threads

[QUOTE=Speedy51;397865]It has been a while since I have done any postprocessing just wondering will any of the following W_2_802/806/809 or 139_133_minus1 fit into under 16 gig of RAM & approximately how long would it take to complete? I have a Haswell E 5960 X at 3 GHz 8 cores/16 threads[/QUOTE]
You should be fine with 16GB. With target_density of 110-120 I'm getting matrices between 9M and 12M. The LA takes between 60 and 110 hours on a quad-core Ivy-Bridge. With DDR4 and more/faster cores you'll probably see faster times.

[QUOTE=VictordeHolland;397867]You should be fine with 16GB. With target_density of 110-120 I'm getting matrices between 9M and 12M. The LA takes between 60 and 110 hours on a quad-core Ivy-Bridge. With DDR4 and more/faster cores you'll probably see faster times.[/QUOTE]
Thank you for your response. If somebody asks like to take one of these please feel free I will be back in touch by decide what to take one

Looks like USTL-FIL (Lille Fr) moved a few cores to the project running 14e and 15e tasks. Not sure what is LAF goal but please just be aware of the queues. Usually the trend is full power during week and shutdown during weekend for these cores from university.

[QUOTE=fivemack;400252](15000 curves at B1=1e8, ten days on GTX580)[/QUOTE]

How much CPU time did it take to complete stage 2 on all those curves :shock:

[QUOTE=axn;400263]How much CPU time did it take to complete stage 2 on all those curves :shock:[/QUOTE]
From zero to infinity... it depends on B2.
As he formulated it, I don't know if any stage 2 was done.
As I understand the post (maybe you didn't see the text line about matrix) the split was done by NFS, and the complaint was that the ECM should have found the p56.

[QUOTE=axn;400263]How much CPU time did it take to complete stage 2 on all those curves :shock:[/QUOTE]

I used the default B2 (776268975310); it took between two and three minutes a curve, so say 40k CPU-minutes, but I think I used the 48-CPU machine for that and it took a day or so.

[QUOTE=LaurV;400280]From zero to infinity... it depends on B2.
As he formulated it, I don't know if any stage 2 was done.
As I understand the post (maybe you didn't see the text line about matrix) the split was done by NFS, and the complaint was that the ECM should have found the p56.[/QUOTE]

Sure. I get that the factorization is complete, and he was complaining about ECM miss. I am not really interested in all that. Just wanted to see the relative performance of GPU stage 1 vs CPU stage 2. I am assuming that stage 2 was run. After all, the ECM effort listed (15000 @ 1e8) is barely a t55. If stage 2 was not run, there is no point calling this an ECM miss.

For B1=1e8, on a single GTX580, stage 1 for one curve is done in 40 seconds [B]wall-clock[/B]-time.

Stage 2 for one curve is about 150 seconds [B]CPU[/B]-time, so you want at least four threads per GPU to keep up.

GW_3_513
 

I must say it has been quite in this thread of late. Does anyone want GW_3_513? If not and it will fit in 16 gig of RAM & there is no real rush for it e.g. 14 days running at roughly 12 hours a day I I would like to take it on. And could somebody please tell me what target_density to run it at and also where you put it in the bat file (it has been a while since I have done a job)