Posts by Bryan

1) Message boards : News : Happy Holidays! (Message 1902) Posted 24 Dec 2011 by Bryan Post: Thanks to Santa for the present. You've worked hard all year so time to kick back, relax, and enjoy the holidays. No more work on the project until after the New Year ... it's running nicely right now :)
2) Message boards : Number crunching : (2) 5870's (Message 1689) Posted 12 Dec 2011 by Bryan Post: There is a problem with multi GPUs in BOINC/Windows/Moo that depending on the phase of the moon the time that shows up in the project "Tasks" does NOT match the wall time showing in either BOINC or the STD_ERR report for workunits. I have always had this problem with my 5970s both here and DNETC. Open the STD_ERR report and look at the start time and stop time noted there and you will find that that value is roughly 1/2 of what is showing under your TASKS time. (I looked at your machine). 2 weeks ago I put a new 6990 on a friends account and crunched Moo with that card for the 1st time. The wu STD_ERR report, and BOINC showed them as completing in 784 seconds. It ran that way and reported that time for 2 days. After 2 days I detached from the friends account and moved it over to my account. The time being reported now is double what the STD_ERR file shows. Same machine and GPU but under 1 account it showed 784s and under another account it shows 1570s. The "wall time" is identical in both cases. The time showing under tasks is about the same time I get from a single 5870 BUT THE CREDITS awarded on an update or daily basis is double with the dual card. Before you ask, YES I did set the "core" to the same value as I was running on the friends account. I also ran a benchmark to see that was correct and nothing had changed when changing accounts.
3) Message boards : Number crunching : WU hang & don't exit atrer time-out (Message 942) Posted 28 Aug 2011 by Bryan Post: Until the past 3 days I had not experienced a hangup on my dual card machines. In the last 3 days I've seen 3 ... 2 on 1 machine and 1 on the other for a total of 18 hours. Some folks on DNETC had the same problem (I never experienced it) and I believe they put in an auto abort feature that triggered if the wu went over xxx.xx time.
4) Message boards : Number crunching : Which GPU is Better (Message 890) Posted 7 Aug 2011 by Bryan Post: Good to know about the 68XX, One would thunk that the new version of the 58XX would be better. AMD changed their numbering scheme the 68XX is a replacement for the 57XX NOT the 58XX. The 68XX does not support DP math. The 6970 IS the replacement for the 58XX.
5) Message boards : News : Different workunit sizes added (Message 455) Posted 24 May 2011 by Bryan Post: I'm not having any trouble since the very start of the new format. Originally one of the 5970s was getting nothing but 3 packet wu that absolutely kill throughput :) I detached/reattached and now that machine is getting even packet wu and the large ones. All is fine at this time. I really appreciate the wu optimizaation. On DNETC we kept asking for 1 wu per GPU because our GPUs were sitting idle but we were IGNORED. At least now machines with 1, 2, or 4 GPUs can have everything running full time! THANK YOU!
6) Message boards : News : Different workunit sizes added (Message 427) Posted 22 May 2011 by Bryan Post: Hi, This should minimize idle cards at the end of wu. We'll see how well it works, but regardless, I for one appreciate the attempt at maximizing our hw! Thank you
7) Message boards : Wish list : Credit update (Message 418) Posted 21 May 2011 by Bryan Post: THANK YOU very much!!! This is greatly appreciated!
8) Message boards : Wish list : Credit update (Message 381) Posted 19 May 2011 by Bryan Post: Any chance the stats could be exported more than once per day? Sure would be nice if 4x or 6x like MW and PG.
9) Message boards : News : Switched to granting static credit (Message 357) Posted 18 May 2011 by Bryan Post: It looks like the original DNETC is waking back up. 8.05 credits per stat unit IIRC!
10) Message boards : News : Switched to granting static credit (Message 324) Posted 15 May 2011 by Bryan Post: I contend that anyone running this project (me included) is a "cw" regardless of how many credits they produce. There is zero scientific value since it is statistically a 100% given that a solution will be found if you try all combinations. It is like buying every combination for the lottery, you are guaranteed to have a winner. So anyone contributing processing power here is doing it for the credits. It's hard to take a moral high ground when everyone is here for the same purpose. The question gets to be how cheap a whore you are and at what level you are comfortable participating. For me if the credits are not better than other GPU projects then I will move elsewhere. If I can get the same credits on MW with the ATIs running cooler then why would I play here? At least something like MW does have a scientific purpose other than decoding a 5 word message saying "yep, you broke the code".
11) Message boards : News : Switched to granting static credit (Message 288) Posted 13 May 2011 by Bryan Post: I can't speak of others machines but on my computers w/ 2 5870s and 2 5970s for the past 9 hours my daily run rate projects to 1.06M. On MW those same machines produce 1.35M per day and run cooler to boot. At this credit level it will make a nice backup project for when MW is down.
12) Questions and Answers : Windows : Multi-GPU task takes longer than single-gpu task (Message 108) Posted 5 May 2011 by Bryan Post: As some one mentioned, we would DEARLY love to have the project crunch on individual GPUs rather than use all in the machine. It is more efficient for the computers and more work would get done for the project. Hi, I understand that request but unfortunately this is not up to me. Distributed.net Client doesn't allow selecting which GPU they use and they detect and use all of them. Until they support this switch, there's nothing much I can do. Sorry. :( -w I thank you for an honest reply :) DNETC basically blew us off and never addressed the request/problem.
13) Questions and Answers : Windows : Multi-GPU task takes longer than single-gpu task (Message 98) Posted 4 May 2011 by Bryan Post: As some one mentioned, we would DEARLY love to have the project crunch on individual GPUs rather than use all in the machine. It is more efficient for the computers and more work would get done for the project. DNETC blew us off and refused to address this. It is frustrating to see a GPU sit idle because there are an odd number of packets to process.
14) Questions and Answers : Windows : Observations (Message 29) Posted 2 May 2011 by Bryan Post: I have a couple of observations running dual 5870s on an I7: I was getting low GPU loading (44%) until I turned off all CPU crunching. The load immediately went to 99% on both cards. 2 wu are using 25% of the processor which in itself isn't a problem but different than the old DNETC. The temperature on the cards at 99% loading is FAR below what the old app ran. In fact the temps are as cool as Collatz which uses only SP math. They run much cooler than MW at the same load. In the old version both GPUs would crunch simultaneously until it got to the end of the computation and if there was 1 remaining packet then 1 GPU dropped out until completion. On this one both GPUs run for a period and then they seem to start alternating with one sitting idle. In the old app occaisonally one GPU would drop out for some reason and the run times were quite a bit higher than normal. When this happened the CPU time skyrocketed. I'm seeing CPU times that equal my completion times ie 2200s.
15) Questions and Answers : Windows : Driver Requirements? (Message 8) Posted 2 May 2011 by Bryan Post: I also have a dual 5870 machine and the GPU utilization is only showing 44%. Is there anyway to increase the GPU loading?