Posts by JHMarshall

21) Message boards : Number crunching : Nvidia Geforce GTX 650 Ti slower than GTX 285 ? (Message 56509) Posted 15 Dec 2012 by JHMarshall Post: Try these http://en.wikipedia.org/wiki/Comparison_of_Nvidia_graphics_processing_units http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units
22) Message boards : Number crunching : Nvidia Geforce GTX 650 Ti slower than GTX 285 ? (Message 56500) Posted 14 Dec 2012 by JHMarshall Post: I have just bought a new PC with an NVidia GeForce GTX650 Ti video card. I have averaged the run times for the last 10 workunits for each card ... GTX650 Ti - 24.7 minutes GTX285 - 17.7 minutes Does anyone have any ideas why the GTX650 Ti is taking longer to finish workunits than the GTX285 ? I thought it should be faster. The GFLOPS reported by BOINC is the single precision performance. You would see the 650 Ti faster on single precision projects such as Einstein. The DP performance on these cards is not so good. However the power requirements of the 650 Ti is considerably less than the 285. So even though the 650 Ti takes longer you are getting more performance per watt. Joe
23) Message boards : Number crunching : GTX670's and the MilkyWay project (Message 56487) Posted 14 Dec 2012 by JHMarshall Post: You're welcome. Crunch away!
24) Message boards : Number crunching : GTX670's and the MilkyWay project (Message 56478) Posted 13 Dec 2012 by JHMarshall Post: I am running GTX670's, driver 301.42, cuda 4.2.1. The system has an i7-3930k processor and 16 gigs DRAM. I was advised that this computer might run MilkyWay very inefficiently in GPU processing. I am running the project on CPU on four machines; but I was looking to add to my GPU work where I only have Einstein and SETI on GPU. My ScienceSprings blog has been tilting toward Astronomy lately, so I am beefing up that side of BOINC in my projects. Any advice? Your GTX 670s will run fine on MW. The Nvdidia GPU apps will run many times faster than your CPU. I run several AMD 7950s and one Nvidia GTX 560 Ti. The 560 Ti has no problems but is definitely slower than the 7950s on MW. The AMD high end cards have much faster double precision units than the Nvidia cards and MW requires DP, so the AMD high end cards are several times faster than the Nvidia cards. This may be what you heard about inefficiency. The Nvidia cards do perform on MW and will give you much more throughput than the CPU apps. Joe Edit: Additional information. GTX 560 Ti ~500 seconds AMD 7950 60-64 seconds (slightly underclocked) I think your 670 would be twice as fast as my 560 Ti
25) Message boards : Number crunching : Validator stopped. (Message 56408) Posted 6 Dec 2012 by JHMarshall Post: Same here. Over 3000 waiting for validation. I've stopped all MW tasks and moved to other projects for now. Why no feedback from the project? What's happening? Joe
26) Message boards : Number crunching : GPU card update? (Message 56250) Posted 20 Nov 2012 by JHMarshall Post: An HD 7950 will run 1 WU in 1 min. They are around $299.99 with mail rebate from Microcenter. Haven't checked prices elsewhere. I run 3 of them spread across 2 systems. They also perform well SP on other projects that have OpenCL WUs such as Einstein. Joe
27) Message boards : Number crunching : GPU card update? (Message 56248) Posted 20 Nov 2012 by JHMarshall Post: Hi all, I have been crunching with a GTX 260 for the past two years but think it is finally time to upgrade. I settled (pending community input) on a GTX 660ti. GTX 660 ti model: http://www.microcenter.com/product/397952/02G-P4-3660-KR_NVIDIA_GeForce_GTX_660_Ti_2048MB_GDDR5_PCIe_30_x16_Video_Card Will this card meet the requirements for crunching Milkyway@Home & is it by best option in the 200-300 price range? Thanks! Jack, The Nvidia cards don't have very good double precision throughput. They are fine single precision crunchers, but MW requires DP calculations. AMD cards are much superior for DP work. I would recommend an HD 7950 for the price range you quoted. Joe
28) Message boards : Number crunching : GPU Requirements (Message 56075) Posted 4 Nov 2012 by JHMarshall Post: Michel, MW requires Double Precision arithmetic. The HD 67xx series doesn't support DP. Joe
29) Message boards : Number crunching : GPU Requirements (Message 56052) Posted 1 Nov 2012 by JHMarshall Post: John, I have 3 Gigabyte AMD HD7950s crunching between Milkyway and Einstein. I run Win 7 Pro and Ultimate 64 bit systems, BOINC 7.0.28, and Catalyst 12.6. Performance on both Projects is great. The DP on the 79xx series can't be beat. I don't overclock and run a single task per GPU for stabiilty and to keep system heat down. The HD 7950s will complete a single 1.02 OpenCL task in less than 40 secs. Joe
30) Message boards : News : Apology for recent bad batches of workunits (Message 55973) Posted 26 Oct 2012 by JHMarshall Post: John, Good to get confirmation that others are seeing the same failures. Thanks, Matthew, Travis, I don't know if this is going to help, but instead of pulling my machines off MW, I'm going to reallocated my resources to MW (1 NVidia GTX 560Ti, 3 AMD HD 7950s, and 7 of 14 CPU cores spread across three machines). I'll either help clear out the bad WUs or add to the confusion. By the way, I haven't seen any failures where my wingmen didn't also have faiures. That brings up a question. If these WUs have bad data or something else wrong how confident are you that the WUs that validate are really good? Joe
31) Message boards : News : Apology for recent bad batches of workunits (Message 55964) Posted 26 Oct 2012 by JHMarshall Post: Matthew, It's not just the "de_separation_22_3s_free_3" WUs that have problems. Many "de_separation_22_3s_edge_3" WUs are also failing on AMD CPUs, Intel CPUS, Nvidia cards, and AMD cards. Joe
32) Message boards : Number crunching : Lots of crunching errors since today (Message 55949) Posted 25 Oct 2012 by JHMarshall Post: Ray, Absolutely NOT a driver problem. I have several instances where a WU failed on 4 completely different configurations (mine and wingmen) (Intel CPU, AMD CPU, Nvidia Card, and AMD card). These are definitely bad workunits. Joe
33) Message boards : Number crunching : Credit drop since starting of the new batch of WU (Message 55930) Posted 23 Oct 2012 by JHMarshall Post: Whoops, I missed the two simultaneously part.
34) Message boards : Number crunching : Credit drop since starting of the new batch of WU (Message 55925) Posted 23 Oct 2012 by JHMarshall Post: Nowi, Your run times for the HD7950 don't seem quite right. I have two HD7950s and they each run a "ps_separation_22_3s_free_3" work unit in 38 seconds. Your 75 seconds is much too long. What else is going on in your machine. I have an i7-3770K cpu and always leave 2 "cores" free for GPU support. Joe
35) Message boards : News : Apology for recent bad batches of workunits (Message 55894) Posted 21 Oct 2012 by JHMarshall Post: Link, Tackleway has some good points. I've been checking the error from my system, less now than before but still coming through. I check the WUs to see what other systems have failed. If I see 4 errors while computing for the work unit, I assume I have no problems with my system. The WU was bad. The failures are across many systems, CPUs, and GPUs. One recent failure on a "free_3" WU showed an error for an AMD CPU computer, 2 opencl-nvidia computers, and my opencl_amd_ati machine for a total of 4 (WU assumed bad). I've seen many combinations of CPU and GPU failures. This is NOT a GPU driver issue. Joe
36) Message boards : News : another test run 'de_separation_22_3s_edge_1' (Message 55793) Posted 14 Oct 2012 by JHMarshall Post: I haven't seen that. My run times prior to the new WUs were consistently 46-47 secs with a single GPU task per HD7950. The run times on the edge_3 tasks are 38-39 secs. I run an i7-3770K processor with several free cores and two HD7950s with the GPUs dedicated to Milkyway (Catalyst 12.6, BOINC 7.0.28, Win 7 Utlimate 64Bit). Three CPU cores run Cosmology. Joe
37) Message boards : News : another test run 'de_separation_22_3s_edge_1' (Message 55767) Posted 13 Oct 2012 by JHMarshall Post: I'm having the same problem with high error rate (?10% +) and communication deferred for a couple of hours. Before these work units I seldom ever had an error (i7-3770K and 2 HD7950 GPUs - the system was rock solid). This system crunches 2 GPU work units in less than a minute. So a 3 hr deferred communication causes a minimum of 2 x 60 x 3 = 360 work units to be missed. I have to manually update to get going again. These recurring errors are causing you to miss a lot of production work! Joe
38) Message boards : Number crunching : numerous validate errors (Message 55662) Posted 5 Oct 2012 by JHMarshall Post: Nevermind! I looked at errors, not "invalid". Duh! Joe
39) Message boards : Number crunching : numerous validate errors (Message 55661) Posted 5 Oct 2012 by JHMarshall Post: These separations are bad. If you look at the individual work units you will see that all computers are having errors. Look for "noCore" in the separation name. These are the bad ones. Joe
40) Message boards : News : NBody Update and New Runs (Message 55391) Posted 18 Aug 2012 by JHMarshall Post: Most of my NBody runs are failing. I've changed my preferences to not run NBody applications. I'm still receiving failing NBody runs. Why?

Previous 20