Message boards :
Number crunching :
Monitoring inconclusive GBT validations and harvesting data for testing
Message board moderation
Previous · 1 . . . 8 · 9 · 10 · 11 · 12 · 13 · 14 . . . 36 · Next
Author | Message |
---|---|
-= Vyper =- Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 |
Petri mentioned SMX units and the unroll value! Should it be the same value as what the manufacturer says it should be? It seems like that a 750Ti should have unroll 5 set! Is that correct? "GeForce GTX 750 The 1Gb GeForce GTX 750 ships with 4 activated SMX units containing 512 Shader Cores and 32 texture units. The core clock frequency will be 1020 MHz while it can boost to 1085 MHz. The memory speed is locked at a 5010 MHz effective data rate based on a 1252 MHz quad data rate for GDDR5 over 128-bit memory bus. GeForce GTX 750 Ti The more interesting product will be the GTX 750 Ti which has 5 activated SMX units containing 640 Shader Cores and 40 texture units / 16 ROPs. The core clock frequency will be 1020 MHz while it can boost to 1085 MHz. The memory speed is locked at an 5400 MHz effective data rate based on a 1350 MHz quad data rate for GDDR5 over 128-bit memory bus. " _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
That line is present when the 750Ti Hangs or Stalls. It hasn't 'Stalled' yet, but it's probably just a matter of time. Looking at your one AP here, Driver version: 367.35 That is a CUDA 8 Driver. In fact, that is the Same series driver I Updated to, that Stopped the Hangs that were happening with driver 352.79. Try it with this driver, http://www.nvidia.com/Download/driverResults.aspx/97645/en-us My 2 EVGA and 1 Zotac 750Ti cards chug along with the zi3 Apps also, as long as I use a CUDA 8 driver. I didn't have that problem with any of the Baseline Apps or the zi Special Apps, and I have compiled and tested Dozens of each App version. It Only happens with the zi3 Special versions. You can't change any voltages on a Mac, and since it works in Linux with driver 367.79 there is No reason to change it in Linux. The fact that it ONLY happens with the zi3 version Apps, and happens on both Platforms, is indication the problem is with the App. When the problem happens on Multiple cards on Multiple Platforms with just one particular App, it's a pretty good indication of where the problem exists. BTW, here are the OSX CUDA Drivers, http://www.nvidia.com/object/mac-driver-archive.html Note there aren't any CUDA 8 drivers. The only place you can find a OSX CUDA 8 driver is in the CUDA 8 Toolkit. I don't think that will go over very well, having to register as a Developer to download the Toolkit so you can install a working Driver. Also, the driver in the Toolkit doesn't work with the current OSX, Darwin 15.6. My guess is you'll have to wait until Darwin 16.0 before you will see a Public CUDA 8 driver. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Also, I see the Autocorrelation Error on your machine as well, Best autocorr: peak=45123.75, time=5.727 Strange it's always around that same exact time. That's a lot of Inconclusives to dig through, you should update to x41zi3f as soon as possible. |
Kiska Send message Joined: 31 Mar 12 Posts: 302 Credit: 3,067,762 RAC: 0 |
Datafile has been archived here |
-= Vyper =- Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 |
Also, I see the Autocorrelation Error on your machine as well, Where can i find it? I would do it as soon as i can get my hands on it. _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
Also, I see the Autocorrelation Error on your machine as well, I could e-mail that to you.. To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
Also, I see the Autocorrelation Error on your machine as well, The exact same time would indicate that something gets overwritten (buffer underflow or overflow) or that the chirping produces an artifact in the data and the autocorrelation check finds a false positive. I'll check if I can reproduce the same autocorr error. If yes then I'll start finding a fix. Howabout the OpenCL version - does it have a cut off peak power so that it does not report overly high values? To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
Petri mentioned SMX units and the unroll value! Should it be the same value as what the manufacturer says it should be? I have tried with different unroll values and found that the unroll should be the number of SMX units or at least 75% of them. To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
In prev build such big numbers would trigger sanity check. Currently sanity check for autocorr disabled so any value will pass through to validator and validator will decide if such power valid or not (there is agreement between wingmans or not). EDIT: and no, even with sanity check enabled there is no any "cut-off" - task just would end with computation error. SETI apps news We're not gonna fight them. We're gonna transcend them. |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
Thanks. To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
I ran across a couple machines that may be of interest; Coprocessors: [3] NVIDIA GeForce GTX 1080 (8192MB) OpenCL: 1.2 Operating System: Darwin 15.6.0 In progress (98) · Validation pending (480) · Validation inconclusive (402) · Valid (193) · Invalid (10) · Error (4) http://setiathome.berkeley.edu/results.php?hostid=8018045 Coprocessors: NVIDIA GeForce GTX TITAN X (12288MB) OpenCL: 1.2 Operating System: Darwin 15.6.0 In progress (74) · Validation pending (224) · Validation inconclusive (223) · Valid (114) · Invalid (5) · Error (0) http://setiathome.berkeley.edu/results.php?hostid=7297852 Shame with all that power the machines are forced to use an App that rates around 'Weakly similar' Q= 19.24%. Much better to use the one at Beta that rates 'Strongly similar' Q= 99.82%. Even the Special App does better than Q= 19. In other news I changed the two 750Ti to another machine and installed the latest version of Ubuntu 14.04, along with the updated driver. We'll see how it works now. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
In other news I changed the two 750Ti to another machine and installed the latest version of Ubuntu 14.04, along with the updated driver. We'll see how it works now. That just reminded me that my Linux machine has a 680, which is one compute capability lower than the new code will do at the moment. Will have to think what to do about that, since the 3 platform simultaneous build system automation is coming along. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Kiska Send message Joined: 31 Mar 12 Posts: 302 Credit: 3,067,762 RAC: 0 |
Just uploaded about ~250 results and datafiles. |
-= Vyper =- Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 |
Things seem to improve. Look further down and look at my consecutive valid tasks with Petris latest revision. http://setiathome.berkeley.edu/host_app_versions.php?hostid=8053171 *Thumbs up* _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
The OpenCL Apps scored much worse in Darwin 15.4 even though these Apps worked very well in Darwin 14.5. I booted into Darwin 14.5 to compile x41p_zi3g and decided to try a few of the Apps there. The Above is the App r3346 in El Capitan. Below is the exact same App in Yosemite; KWSN-Darwin-MBbench v2.1.07 Running on TomsMacPro.local at Tue Sep 6 04:13:22 2016 --------------------------------------------------- Starting benchmark run... --------------------------------------------------- Listing wu-file(s) in /testWUs : blc2_2bit_guppi_57403_HIP11048_0006.17091.831.22.45.71.wu reference_work_unit_r3215.wu Listing executable(s) in /APPS : MBv8_8.05r3346_nvidia_ssse3_x86_64-apple-darwin MBv8_8.17r3516_NV_ssse3_x86_64-apple-darwin setiathome_8.10_x86_64-apple-darwin__opencl_nvidia_mac Listing executable in /REF_APPs : MBv8_8.05r3344_sse41_x86_64-apple-darwin --------------------------------------------------- Current WU: blc2_2bit_guppi_57403_HIP11048_0006.17091.831.22.45.71.wu --------------------------------------------------- Running default app with command : MBv8_8.05r3344_sse41_x86_64-apple-darwin 4955.87 real 4942.34 user 11.31 sys Elapsed Time: ………………………………… 4955 seconds --------------------------------------------------- Running app with command : MBv8_8.05r3346_nvidia_ssse3_x86_64-apple-darwin 1131.83 real 126.52 user 261.62 sys Elapsed Time : ……………………………… 1131 seconds Speed compared to default : 438 % ----------------- Comparing results Result : Strongly similar, Q= 99.89% --------------------------------------------------- El Capitan: Result : Weakly similar, Q= 7.881% Yosemite: Result : Strongly similar, Q= 99.89% Quite a difference, wouldn't you say? Here is the App running reference_work_unit_r3215.wu; --------------------------------------------------- Done with blc2_2bit_guppi_57403_HIP11048_0006.17091.831.22.45.71.wu. Current WU: reference_work_unit_r3215.wu --------------------------------------------------- Skipping default app MBv8_8.05r3344_sse41_x86_64-apple-darwin, displaying saved result(s) Elapsed Time: ………………………………… 2198 seconds --------------------------------------------------- Running app with command : MBv8_8.05r3346_nvidia_ssse3_x86_64-apple-darwin -sbs 192 -oclfft_tune_gr 256 -oclfft_tune_wg 128 326.80 real 97.70 user 87.68 sys Elapsed Time : ……………………………… 327 seconds Speed compared to default : 672 % ----------------- Comparing results Result : Strongly similar, Q= 99.47% --------------------------------------------------- Here's the App at Beta in Yosemite; --------------------------------------------------- Current WU: blc2_2bit_guppi_57403_HIP11048_0006.17091.831.22.45.71.wu --------------------------------------------------- Skipping default app MBv8_8.05r3344_sse41_x86_64-apple-darwin, displaying saved result(s) Elapsed Time: ………………………………… 4955 seconds --------------------------------------------------- Running app with command : setiathome_8.10_x86_64-apple-darwin__opencl_nvidia_mac 1086.06 real 128.77 user 241.15 sys Elapsed Time : ……………………………… 1086 seconds Speed compared to default : 456 % ----------------- Comparing results Result : Strongly similar, Q= 99.98% --------------------------------------------------- Much different in EL Capitan; --------------------------------------------------- Current WU: blc2_2bit_guppi_57403_HIP11048_0006.17091.831.22.45.71.wu --------------------------------------------------- Skipping default app MBv8_8.05r3344_sse41_x86_64-apple-darwin, displaying saved result(s) Elapsed Time: ………………………………… 4797 seconds --------------------------------------------------- Running app with command : setiathome_8.10_x86_64-apple-darwin__opencl_nvidia_mac 1444.76 real 117.24 user 257.14 sys Elapsed Time : ……………………………… 1444 seconds Speed compared to default : 332 % ----------------- Comparing results ------------- R1:R2 ------------ ------------- R2:R1 ------------ Exact Super Tight Good Bad Exact Super Tight Good Bad Spike 0 0 2 6 0 0 0 2 6 0 Autocorr 0 0 0 0 0 0 0 0 0 0 Gaussian 0 0 0 0 0 0 0 0 0 0 Pulse 0 0 0 5 0 0 0 0 5 1 Triplet 0 0 0 1 0 0 0 0 1 0 Best Spike 0 0 1 1 0 0 0 1 1 0 Best Autocorr 0 0 0 0 1 0 0 0 0 1 Best Gaussian 1 1 1 1 0 1 1 1 1 0 Best Pulse 0 0 0 1 0 0 0 0 1 0 Best Triplet 0 0 0 1 0 0 0 0 1 0 ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- 1 1 4 16 1 1 1 4 16 2 Unmatched signal(s) in R1 at line(s) 608 Unmatched signal(s) in R2 at line(s) 592 636 For R1:R2 matched signals only, Q= 19.24% Result : Weakly similar. --------------------------------------------------- |
Kiska Send message Joined: 31 Mar 12 Posts: 302 Credit: 3,067,762 RAC: 0 |
A significant difference, perhaps we should stop distribution of the affected versions of apples broken 'driver' |
-= Vyper =- Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 |
Thats insane, and that is without changing a single line of code, only exchanging O/S version? If so i wonder what is going on really. Seems like s@h need to ban El Capitan or what is the suggestion now? _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
That's been talked about for a while, https://setiathome.berkeley.edu/forum_thread.php?id=78569&postid=1811003#1811003 Nothing has happened...yet. It all started back here, https://setiathome.berkeley.edu/forum_thread.php?id=78569&postid=1813801#1813801 Hmmmm, my latest prediction is, there will be more people testing the Mac CUDA Apps at Beta soon. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Thats insane, and that is without changing a single line of code, only exchanging O/S version? Well, don't we seeing just the same with OpenCL NV build over than year (as TBar stated) already ?... SETI apps news We're not gonna fight them. We're gonna transcend them. |
Kiska Send message Joined: 31 Mar 12 Posts: 302 Credit: 3,067,762 RAC: 0 |
Fact is, we are going to see alot more inconclusive from macs due to the fact the checker in the benchmarker is based upon the validator |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.