Message boards :
Number crunching :
Intel HD Graphics 530 - Keeps "Postponing"...
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · Next
Author | Message |
---|---|
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
Well this is not new! Thanks for the input, Vyper. I think the "built-in" Intel GPUs just plain SUCK!!!!! :-( When you are using Lunatics in addition to it, a GPU should ALWAYS crunch faster than a STOCK CPU. This was NOT the case for Farragut, (HD530), here... MOST disappointing. I'll stick with EVGA for crunching on GPU. I have had EXCELLENT progress with Exeter, (GTX-760), and Prometheus, (GTX-750 TI SC), so EVGA is the GPU for me. :-) TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
BetelgeuseFive Send message Joined: 6 Jul 99 Posts: 158 Credit: 17,117,787 RAC: 19 |
I'm also having issues with my HD Graphics 530 (Core i5-6500). Over on beta I had lots of inconclusives and even an invalid with MB v8. There is a new Beta driver available (15.40.18.4380) on the Intel website, but there is no mention of OpenCL changes/fixes in the release notes so I haven't tried it yet. For now I only use the integrated GPU for my display and I have my EVGA GTX-750 decicated for crunching. I hope that Intel solves these issues as it seems a waste to have high end processors with integrated GPU's that cannot be used for scientific purposes. If there is anyone out there who is able to use the 530 GPU without inconclusives/errors, please post a message here. Tom |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
HP just Upgraded the Video Driver for Farragut, (HD530), to 20.19.15.4360; OpenCL 2.0, now taking 4.8+ GB of System RAM from the total available of 12 GB. So, I've chosen to Upgrade BOINC from 7.4.42 to 7.6.22, and installed Lunatics 0.44. I chose NOT to install support for AP Units; ONLY MB v8. I then told BOINC to Allow New Tasks, and Resume Project. So, tomorrow when the computer is booted and logged into, BOINC should have SETI DL new v8 Work. Hopefully, v8 will work better on this system than the APs did... (I know, Apples and Oranges...) I desperately want to see if the HD530 will perform better than other systems crunching CPU. (This was my chief complaint... The HD530 was just pathetically slow... Maybe under 7.6.22 and Lunatics 0.44, and with the Upgraded Video Driver this thing might now work... Who knows...) TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
MarkJ Send message Joined: 17 Feb 08 Posts: 1139 Credit: 80,854,192 RAC: 5 |
According to Downloadcenter.intel.com there is a 4380 beta driver available. No mention of 4360 on the Intel site so it's probably a Microsoft Win10 special version. The 4380 driver lists the following fixes KEY ISSUES FIXED: As you can see it's all game stuff, no mention of OpenCL so I wouldn't hold your breath on this one fixing anything, nor the 4360 that you've got. I've raised a bug with Intel about their OpenCL not working properly (too many validation errors) along with another user eta over at Einstein. Intel told me it's the OEM responsibility, so they don't want to know about it. My tasks run fine, just fail to validate, but then I'm running Win7 x64 on my machines. BOINC blog |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
According to Downloadcenter.intel.com there is a 4380 beta driver available. No mention of 4360 on the Intel site so it's probably a Microsoft Win10 special version. Yes, 20.19.15.4360; OpenCL 2.0 is for Win 10. Farragut, (HD530), is a Win 10 Pro x64 system by HP. It is an HP Envy Desktop System. It's now been running MB v8 since 9 AM this morning - Pacific Time. I think that the Units are crunching OK; haven't checked results yet. Estimated times for each WU are currently 30 Minutes. TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
OK - Looks like it's still playing the "Inconclusive" game with WUs... Work Unit 2062109641 Farragut is Computer ID: 7066606 and the third send out has not happened, yet. :( TL [EDIT] NOTE: However; at least the MB v8 units are NOT "Postponing" like the AP units were. TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
This wu will be valid. Just a few signals (best) above threshold probably. With each crime and every kindness we birth our future. |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
This wu will be valid. Thanks Mike. :-) Looks like Farragut, (HD530), is processing these v8 Units between 40.5 to 40.75 minutes each. I hope that's good... TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
You only had VHARs so far. Mid range tasks will take longer. With each crime and every kindness we birth our future. |
BetelgeuseFive Send message Joined: 6 Jul 99 Posts: 158 Credit: 17,117,787 RAC: 19 |
It certainly looks look the "Inconclusive" game again ... Your host now has 8 (out of 8) inconclusives. I hope that one of these days we get an Intel driver version that has better OpenCL support. Tom OK - Looks like it's still playing the "Inconclusive" game with WUs... |
ExtraTerrestrial Apes Send message Joined: 5 Jan 02 Posts: 13 Credit: 1,370,403 RAC: 0 |
i3 6100 with HD530, Win 10 64, BOINC 7.6.22 MB 7/8: works with acceptable invalid rate, about 1 per day AP: same issue with suspended tasks as described by others. Issue triggered after the 1st check point shall be loaded. Driver: currently x.4352, same with 1 or 2 earlier versions (forgot which ones). Am going to try x.4380 soon. Hardware: Asus Z170 or MSI Z170 mainboard, same issue with prior i7 6700 Collatz: works fine. Einstein: ~50% invalids, very wrong results from reference WU. I've been chasing this Einstein issue since September 2015 without any luck :/ (I'm the one MarkJ mentioned) MrS Scanning for our furry friends since Jan 2002 |
ExtraTerrestrial Apes Send message Joined: 5 Jan 02 Posts: 13 Credit: 1,370,403 RAC: 0 |
Success! My 1st Astropulse crunched on a Skylake GPU. Changes since the last post: - Intel GPU driver x.4380 (beta) - installed the MS C++ 2012 runtimes, 32 and 64 bit (fixed Collatz errors) The performance was about a factor of 10 better than a core of a Xeon E31225, basically a Sandy Bridge Core i5 at 3.1+ GHz. The yield was 219 cr/h, whereas for MB tasks I get about 150 cr/h (running 2 tasks simultaneously). The power consumption of the GPu and main memory was a bit less than running MB tasks, although this difference is practicaly negligible. MrS Scanning for our furry friends since Jan 2002 |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
HI TL I have a similar configuration as you. Core i5 with HD530 GPU, and I am experiencing the same problem with AP's. For me the AP can run anywhere from an hour or so up to over 2 hours before crashing out. But I find that when the AP pauses another GPU task starts and will run through uninterrupted before the AP can resume. And I have to agree that running AP's under these conditions is pointless, 4 out of 5 AP's all crashed and the 5th I aborted to save time. But since the HD530, while slower than other GPU's such as nVidia and Radeon, is still a lot faster than running CPU only I think it is worth persevering with. Though most GPU tasks come up inconclusive and require extra validation, with about 5 to 10 % failing as invalid, the end result is a much higher productivity than not using it. But I am looking for a nice cheap nVidia card to pop in and tweak things a little bit. Stephen |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
I would recommend to use OpenCL MB on iGPU instead. I have lot of inconclusives/invalids with AstroPulse on my IvyBridge but very stable processing with MultiBeam on the same iGPU. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
I have about 80% of all GPU tasks on my i5/HD530 setup come back as inconclusive, and after they are compared to a 3rd host about 10% of those come back as invalid. But, I am getting much faster crunching times from the GPU than from CPU alone so I am sticking with it for now. Anyone have a cheap GTX750 TI for sale LOL Stephen |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
For what it is worth that is faster than mine .... mine takes about 1 hour to 1.5 hours for WU Stephen |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
It is on MB tasks that I have that "inconclusive" problem. AP's all crash out. Stephen |
TimeLord04 Send message Joined: 9 Mar 06 Posts: 21140 Credit: 33,933,039 RAC: 23 |
Well, I'm finding that with Lunatics 0.44 that v8 MBs are crunching faster than what I experienced with v7. Mike was right, I started getting more mid range v8 Units on Farragut, (HD530), and the system is taking 1 Hour ++ to complete those Units. I'm actually finding many CPU paired systems that are running faster than the HD530. More of them were with v7 MBs than v8; but, there are some v8s that I've been paired against CPU runners and they are faster than the HD530 GPU. Like others here have mentioned, I HIGHLY recommend an EVGA GeForce GTX-750 TI SC GPU. REALLY AWESOME BANG FOR THE BUCK!!!!! Good luck, TL TimeLord04 Have TARDIS, will travel... Come along K-9! Join Calm Chaos |
ExtraTerrestrial Apes Send message Joined: 5 Jan 02 Posts: 13 Credit: 1,370,403 RAC: 0 |
2 more valid APs on HD530, no problems any more. Is anyone else seeing this? Was it the new driver or the MS C++ runtimes? MrS Scanning for our furry friends since Jan 2002 |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14653 Credit: 200,643,578 RAC: 874 |
2 more valid APs on HD530, no problems any more. Is anyone else seeing this? Was it the new driver or the MS C++ runtimes? If the MS C++ 2012 runtimes were needed (but missing), I think the app would fail to start - crash immediately (I've had that with self-built BOINC clients). I've seen Raistmer build with VS2008 and VS2010, but nothing newer than that. I don't think the Intel app will need 2012/13. |
©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.