bigpackets

[ChasR]

I was looking at the server status on the Stanford site and found that p1134, p1135, p1140, p1141 and p1475 have been released to general FAH so that you no longer have to have the -advmethods flag set to receive them. I don't know how long it's been since it changed as I don't look on that page all that often. What got me looking was a thread where someone without -advmethods got a p1475, the best point producer out there. So I deleted the -advmethods flag on several of my bigpacket rigs just to see what happened. Low and behold, instead of nothing but 600 pointers, I got nothing but p1475s (on Friday and Saturday). On sig rig # 2, these WUs produce 30% more points than any other. I thought production would go through the roof but, alas, the server ran out of them on Saturday afternoon.
The bigpacket rigs w/o -advmethods that downloaded WUs yesterday and this morning got a mix of 600 pointers and p131x. The p131x series is a mixed bag point wise. For instance p1312 produces 20% fewer ppd than the 600 pointers while p1316 produces 20% more. Since the server is also out (< 200 WUs) of the p131x series, this won't make much of a difference anyway. For now, with or without -advmethods, if you enable bigpackets, the odds are you'll get 600 pointers.

[mtb856]

Eh, that's okay...... I'd rather have a steady diet of 600's rather than a mix of lower and higher PPD WU's.....plus if Stanford gets some new WU's (high PPD WU's perhaps) then the advmethods people will be the first to see and benefit from them.

[TollhouseFrank]

how do you set up to recieve bigpackets? i've been trying to figure that out. i have access to a MINIMUM of 5 machines with over 1 gig of ram (one of them a dually xeon) and methinks that they can be doin' some major foldin'...

(that doesn't include the other machines i've got folding.... which is another 3 or 4)

[mtb856]

In your client.cfg file, scroll right until you see this part
bigpackets=no

now change that to
bigpackets=yes

and boom, your folding with bigpackets :)

Most BP's take about 100MB of ram per instance of FAH on average, some have been 90MB and once I got a pair that were taking 140MB apiece, so take a look at how much memory each computer has free before you load more than one instance of BP's on them.

[darkknight187]

Eh, that's okay...... I'd rather have a steady diet of 600's rather than a mix of lower and higher PPD WU's.....plus if Stanford gets some new WU's (high PPD WU's perhaps) then the advmethods people will be the first to see and benefit from them.

I had been seeing a nice diet of 300-375 pointers through the weekend and they were quite nice to munch on....but now i'm back on a 600pointer if only more of my borgs had more ram....

[Ven0m]

I had a few nice 600s, but hmm some time ago I started to get some huge packets that take much time to compute and give around 300. That resulted in dramatic points / time ratio drop.

[obsolete]

Hmm, I'm going to have to thread jack, sorry. Here I gooo......

I just set my duallie up for two instances (wedo's one click, multiple proc installation) & I enabled big packets. They are both running with maybe 58mb........Is that normal? On my other two PC's & before on my duallie when I was running 1xproc it used 100+ memory everytime thus far......

[Ven0m]

You don't get packets requiring tons of ram all the time. In my opinion it's pretty normal.

[ChasR]

Hmm, I'm going to have to thread jack, sorry. Here I gooo......

I just set my duallie up for two instances (wedo's one click, multiple proc installation) & I enabled big packets. They are both running with maybe 58mb........Is that normal? On my other two PC's & before on my duallie when I was running 1xproc it used 100+ memory everytime thus far......

Normal. Memory use depends on the WU and several of the current crop of big WUs use as little as 52 MB.

I'm going to have to take back what I said about getting mostly 600 pointers with or without -advmethods. I removed -advmethods from the registry on 7 instances to see how production would be affected. With -advmethods on I got nothing but 600 pointers. With it off I get a mix. Below is the current snapshot of results

Computer WU Frame time ppd ppd on 600 ptrs
XP-M @ 2.556 p1135 31:01 279 279
XP-M @ 2.4 p1312 16:49 217 270
Xeon @3.08 #1 p1475 23:29 223 173
Xeon @3.08 #2 p1316 25:19 195 172
Xeon @3.08 #3 p1322 26:48 163 173
Xeon @3.08 #4 p1475 29:29 178 172
P3 @ 1.4 p1140 1:18:33 110 110
Total 1365 1349

So at the moment it looks like there is no production disadvantage in running without the -advmethods flag.