Message boards :
Number crunching :
Intel® iGPU AP bench test run (e.g. @ J1900)
Message board moderation
Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · Next
Author | Message |
---|---|
Marco Franceschini Send message Joined: 4 Jul 01 Posts: 54 Credit: 69,877,354 RAC: 135 |
|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
http://ark.intel.com/it/products/80818/Intel-Core-i5-4460S-Processor-6M-Cache-up-to-3_40-GHz Wow, did not expect Intel marketing department SO bad... They even can't compare 2 devices with common set of options... What instruction set J1900 supports for example? What could be easier than just to take set of options, form table and fill that table, for both devices. And not to invent new table for each one?.... |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
No, it's not normal. Maybe for Win8 though... Use other tools as recommended. I don't know if the results and %blankeds are ~similar enough for to make the conclusion: They similar enough to say that -use_sleep slowdowns GPU too much in your case. But they definitely not similar enough to make any conclusions on overall performance of loaded device. Blanking should be exactly zero for that. Cause blanking % says nothing about how blanking distributed inside task and how many separate blanking replacements was made.
CPU time means how much time single logical CPU of particular host spent inside process' code. As already stated, if process consists of many threads it can exceed wall-clock time for process. In case of AstroPulse that could happen only on NV hardware w/o -use_sleep and with heavely blanked task. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
http://ark.intel.com/it/products/80818/Intel-Core-i5-4460S-Processor-6M-Cache-up-to-3_40-GHz There have always been lots of holes in the data on the Intel site. You can really tell when you do a compare: http://ark.intel.com/compare/75048,78867 SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I saw something strange on my J1900 board. Your system CPU load in Windows 8.1 with < 4 tasks running seems high to me. For comparison this is my J1900. Running Windows 7 & I have most of the annoying unused background processes turned off. http://www.hal6000.com/seti/images/j1900_1-task.png http://www.hal6000.com/seti/images/j1900_2-task.png http://www.hal6000.com/seti/images/j1900_3-task.png http://www.hal6000.com/seti/images/j1900_3-task_processes.png http://www.hal6000.com/seti/images/j1900_4-task_processes.png In my case since I use vnc to connect to the machine it adds some CPU usage normally 1-2% but sometimes 4-5%. That is reflected in 1 CPU task graph. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Marco Franceschini Send message Joined: 4 Jul 01 Posts: 54 Credit: 69,877,354 RAC: 135 |
Actually the Intel Celeron J1900 it's classified as SoC (by the way all the Core generations with integrated memory controller and gpu belong to this taxonomy). http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/pentium-n3520-j2850-celeron-n2920-n2820-n2815-n2806-j1850-j1750-datasheet.pdf Note the GPIO interface... The processor do have support for SSE4.1 and SSE4.2 s.i.m.d instructions set. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Actually the Intel Celeron J1900 it's classified as SoC (by the way all the Core generations with integrated memory controller and gpu belong to this taxonomy). I think Intel classifies it as SoC because it functions without an external south-bridge. So the whole "system" is on the chip. The only external IC needed is memory. Some of their mobile Bay Trail SoC also include memory. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
From AstroPulse v7 rev 2742 <stderr_txt>: CPUID: Intel(R) Celeron(R) CPU J1900 @ 1.99GHz Cache: L1=64K L2=1024K CPU features: FPU TSC PAE CMPXCHG8B APIC SYSENTER MTRR CMOV/CCMP MMX FXSAVE/FXRSTOR SSE SSE2 HT SSE3 SSSE3 SSE4.1 SSE4.2 - - - - - - - - - - Just curious, HAL9000's J1900 show BOINC as: Intel(R) Celeron(R) CPU J1900 @ 1.99GHz [Microsoft Windows 7] My J1900: Intel(R) Celeron(R) CPU J1900 @ 1.99GHz [Family 6 Model 55 Stepping 8] [Microsoft Windows 8.1] @ HAL9000, you made something 'specially' that the 'CPUs [ ] infos' aren't shown? - - - - - - - - - - Since January '15 the Intel driver do a good job, everything work fine. Contrary to the wisdom 'don't change things on a well running machine' (the following) ... ;-) I use still the v10.18.10.3408 Intel driver from the driver DVD which came with the MoBo. Maybe a newer driver would effect a speed difference (I hope a speed up! ;-)? Where I can download newer (for SETI/AP apps recommended) and how I uninstall correct the old driver first before upgrade (until now I have just experiences with NVidia GPUs in this case)? Thanks. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
From AstroPulse v7 rev 2742 <stderr_txt>: I modify boinc.exe to not display [Family A Model B Stepping C] information. I don't like how it messes up my computers view. I also modify it to show the exact ATI model name instead of a generic series name. This is also why I don't change versions often. As I have to modify boinc.exe each time. On the drivers the saying here is "if it isn't broke, don't fix it". With Intel that has been shown to be good. As they sometimes break the drivers to cause garbage results. So far I have had no problems with release 4061. Display Driver Uninstaller supports Intel drivers too, but I have installed over the top of the old driver the past few times without any issues. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Given iGPU uses system memory. I imagine memory CL times may effect iGPU run times. This is the memory I happen to be using in my system. http://www.hal6000.com/seti/images/j1900_mem-config.png It was from a project at work where we were throwing away things. So I helped myself to the memory instead of it going in the trash. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
I hope you don't get 100+ PMs/day of other members who want to know how you mod BOINC. ;-) It's a big deal? 'Advanced' SETI users like me could do this also? For to prevent your PM inbox overflows, maybe you could explain it how you mod BOINC? (@ all: BTW, why the stock BOINC software don't show the correct ATI/AMD GPU card name (series) like NVidia? It's an impossible feature?) It's a pity that it's not longer possible to do this over the registry. :-( I had an automatically task which were executed at every boot up, so not needed to do it manually after every switch off. Thanks. - - - - - - - - - - It looks like you use same MHz + CL like I do: My ASRock Q1900DC-ITX board is running with: 8GB G.Skill DDR3L-1333 SO-DIMM CL9 Dual Kit (2x4GB) - Low Voltage 1.35V - 1333MHz @ CL9-9-9-24 - F3-1333C9D-8GSL |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
I hope you don't get 100+ PMs/day of other members who want to know how you mod BOINC. ;-) I've been telling people for years that I mod BOINC to do my bidding. Normally after the "How do you do that" & I tell them "With a hex editor" that normally scares them away. If you don't know what you are doing in a hex editor you can screw things up very quickly. To me it isn't a big deal. There are several text string that are available to change so it is just a simple matter of blanking out the things I don't want or replacing the text where I want to change it. Now here comes the "Don't try this at home kids!" section. -Exit BOINC so it isn't running. -Make a backup of boinc.exe so you have the original to go back to when something bad happens. -Open your preferred hex editor application. -Load boinc.exe -Locate the text string you wish to change Here we will be removing the [Family A Model B Stepping C] string from being dispalyed. http://www.hal6000.com/seti/images/boinc_mod_1.png -Replace the hex vales with zeros. DO NOT ADD OR DELETE DATA. Replace. http://www.hal6000.com/seti/images/boinc_mod_2.png -Save the modifications. -Place the modified boinc.exe in your BOINC program directory & run as normal. This is the same process I use for my Pentium III 850MHz machine. As it doesn't have and CPUID name data. The reason they can't get the exact ATI GPU model name has to do with how NV & ATI respond to an API call asking for the GPU model name. They had trouble early on getting the actual GPU name from ATI cards. So they hard code the displayed name into groups to save time as I recall. It seems they do something similar with the Operating System string for Windows as well. Given this machine is running Windows 10 & not Windows 8.1 as it is displayed with BOINC 7.2.42. Where BOINC 7.4.42 will display "Windows 10". SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
Thanks! :-) Is there a way to edit the ATI/AMD GPU name also, for to show the really installed GPU card? Until now I don't have, but soon I will. I'll inform me about the 'hex editors' later ... You could recommend an 'easy' tool for 'beginner'? Thanks. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Thanks! :-) One of the hex editors I use often can be had here. Modifying the GPU strings is just as easy as the other example. Find what BOINC is displaying. http://www.hal6000.com/seti/images/boinc_mod_3.png Then make it say what you would like. http://www.hal6000.com/seti/images/boinc_mod_4.png If you have two machines with different GPUs in the same family. Such as the string "AMD Radeon HD 7870/7950/7970/R9 280X series (Tahiti)". You would have to save different versions of boinc.exe for a HD7950 in one machine & a R9 280X in the other. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
Thanks! :-) Now I have enough 'houseworks' (to learn how to use a 'hex editor') for the next days or weeks ... ;-) |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
OK, I did this: Old string "ATI unknown": http://asteroidsathome.net/boinc/show_host_detail.php?hostid=110 (at the moment of post I didn't still contacted the servers of Asteroids@home which is set to No New Tasks) New string "ATI Radeon HD 6570 (NI TURKS) {ASUS EAH6570/DI/1GD3(LP)}" http://setiathome.berkeley.edu/show_host_detail.php?hostid=4832843 The new string is long - I needed to overwrite also the next 2 strings (which are about low HD5xxx models) Of course the new string ends with 00 byte I found that [ ] should not be used in those strings - in BOINC log info is OK but the web page started to act strange (weird things after the name) At first the new string was "ATI Radeon HD 6570 (NI TURKS) [ASUS EAH6570/DI/1GD3(LP)] " so the problem may be also the ending space I used HxD - Freeware Hex Editor http://mh-nexus.de/en/hxd/ Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
I myself wrote: If I would add a NV GPU card to the J1900 PC ... - - - - - - - - - - BilBg wrote: Code for -cpu_lock is probably the same in all Raistmer apps. - - - - - - - - - - BilBg wrote: Code for -cpu_lock is probably the same in all Raistmer apps. Raistmer wrote: Yep, they should to do so.... but was never really checked with few different types of GPU installed. And it's quite easy to test. - - - - - - - - - - Raistmer wrote: So, maybe -instances_per_device 2 will be needed in this case too. Check that. BilBg wrote: @Dirk - - - - - - - - - - OK, in ap_cmdline_win_x86_SSE2_OpenCL_Intel.txt: -cpu_lock -instances_per_device 1 in ap_cmdline_win_x86_SSE2_OpenCL_NV.txt: -cpu_lock -instances_per_device 2 (I let run 2 tasks simultaneously on GT730 (app_info.xml CUDA/0.5)) The result in Task-Manager: 1 Intel iGPU app at Core#0 1 NV GPU app at Core#0 and 1 NV GPU app at Core#1 What I need to change, that 1 NV GPU app (1st) use Core#1 and 1 NV GPU app (2nd) use Core#2? (And Intel iGPU app use still Core#0) Thanks. BTW. With the above mentioned settings, the Intel iGPU app/task calculation increased from 17hrs to 18.5hrs (since I added the GT730 VGA card). |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
try -instances_per_device 3 |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
In ap_cmdline_win_x86_SSE2_OpenCL_Intel.txt: -cpu_lock -instances_per_device 1 in ap_cmdline_win_x86_SSE2_OpenCL_NV.txt: -cpu_lock -instances_per_device 3 The same result in Task-Manager: 1 Intel iGPU app at Core#0 1 NV GPU app at Core#0 and 1 NV GPU app at Core#1 |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
In ap_cmdline_win_x86_SSE2_OpenCL_Intel.txt: -cpu_lock -instances_per_device 1 You could try -cpu_lock_fixed_cpu 3 (or 4) for the Intel iGPU. That should work for now. I expect Raistmer may modify the mutex implementation to make some future versions cooperate between different brand GPUs, but no promises... Joe |
©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.