I wanted to start a new thread to gather information about environments that are seeing problems when running the Rosetta graphics. I've personally not had a problem, but here are some of the facts I think we should gather for each report.

1)Type of Operating System (Windows, Mac, Linux)

2)Type of graphics adapter (can someone please post instructions for how to view the pertenant details on the graphics card?)

3)Was the problem in viewing the graphics at all? Or only when using Rosetta as the screen saver?

4)Approximate date when you started seeing problems.

5)Go to your Rosetta preferences, and report your settings for "Percentage of CPU time used for graphics" and "Number of frames per second for graphics".

6)Have you noticed any specific WU names that seem to be more likely to have problems? Any LESS likely to have problems?

7)What errors were reported when the problem occured?

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

For Win2k and WinXP:
Right click on the desktop and choose Properties. click on the Settings tab. Underneath the pic of the monitor, it'll read something like, "Display: Plug and Play monitor on nVidia Geoforce FX 5600XT"
In my case, the video card chipset is an nVidia FX 5600XT. That's the hardware side.

For driver info..
choose the advanced tab in the bottom right corner of the Settings tab.
The General tab should have the same hardware information you just made a note of. Click on the drivers tab. Provide the Provider, Date, and version..
in my case, Provider is nVidia; date is 11.03.2003; version is 5.3.0.3

I run Boinc & Rosetta as a service, so I don't have any problems with the screen saver. I.e. no screensaver.. :)

---

A better way is to
START -> Run -> type dxdiag
Let DirectX Diagnostics collect the information.

Take note of these
Operating System
Language
System Manufaturer
System Model
BIOS
Processor
Memory
DirectX Version (all of it)

Then go to the display tab
Note everything in the Device section and everything in the Drivers section.

If you have webspace to host a file, use the save all information button at the bottom and upload it and create a link.

I would not recommend posting all the info here as it would just creat an unmanagable mess.

For Win2k and WinXP:
Right click on the desktop and choose Properties. click on the Settings tab. Underneath the pic of the monitor, it'll read something like, "Display: Plug and Play monitor on nVidia Geoforce FX 5600XT"
In my case, the video card chipset is an nVidia FX 5600XT. That's the hardware side.

For driver info..
choose the advanced tab in the bottom right corner of the Settings tab.
The General tab should have the same hardware information you just made a note of. Click on the drivers tab. Provide the Provider, Date, and version..
in my case, Provider is nVidia; date is 11.03.2003; version is 5.3.0.3

I run Boinc & Rosetta as a service, so I don't have any problems with the screen saver. I.e. no screensaver.. :)

---

Team mauisun.org

We should probably also ask folks to note the screen resolution and color quality they have set. In Windows, just right click out on the desktop, select Properties, and then the settings tab shows the screen resolution you have set. For example, I've got one here set to 1280x960 pixels with 32bit color.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I just found this .

Which may solve some of the errors.

We should probably also ask folks to note the screen resolution and color quality they have set. In Windows, just right click out on the desktop, select Properties, and then the settings tab shows the screen resolution you have set. For example, I've got one here set to 1280x960 pixels with 32bit color.

Feet1st, my DXDIAG suggestion covers all that ;-) (in windows)



Feet1st, I have a suggestion for you. Would it be possible if you would be a Rosetta@Home Live Help contact (via skype either text chat or voice chat) as part of the BOINC live help system. It's a volunteer basis and you help out at R@H a lot. http://boinc.berkeley.edu/help_volunteer.php

---

Team mauisun.org

Sorry Fluffy, I hadn't tried out the dxdiag to realize that.

Yes, I'll have to review this skype thing. Indeed I was proposing a similar chat system, so I realize the difficulty of staffing such a thing. ...on the other hand, I've only got so much spare time, and from the looks of the new host additions each day, most of the questions are not going to be Rosetta related. If anyone already has signed up as a question taker for the new BOINC skype system, please post your experience in this thread.

BTW, anyone ever hunt skype? I hear that if cooked just right over an open flame it tastes like chicken :)

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

OK so Chu confirms these problems seem related to when they changed things to add the sidechains to the graphic. So... how might that be impacting things as they run?

The graphics card's going to be running hotter, and perhaps using more memory. The Rosetta program must keep track of more detail to build the graphic with, is there any heap space or temp storage that is related to the graphic?

The other thing that happened around that time was the addition of the Docking WUs. Which actually show all the sidechains for TWO proteins. These are also longer running models. Do they require more memory to run? Is there any correlation between the screensaver failures and the type of WU being run? Some people basically report that no matter what they do, they can't seem to keep it running. This might be because they got a batch of WUs that are all of the same type?

Some people don't seem to have any problems. Is this just because they don't run the screensaver? Or is there more to the pattern?

If I wanted to see if I can recreate the problem, what should I do? What should I look for to confirm that the problem has occured?

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I could probably create the problem with one of my machines, by using the graphics (though I'm now using BOINC ver. 5.7.5 to play with the skins).


I checking now... with a WU with about 5hr's on it.

Now running both WU's in graphics mode (both about 5hr's in) both with side chains.

~2 minuets in I've lost the ability to zoom/rotate on one of the WU's both still working.

~2.5 min I can now zoom/rotate in both WU's.

Seems to be when I switch to this (firefox tab) I lose the control on the second WU.

I can no longer zoom/rotate both WU's.

closing one/ both then opening them/ opening only one etc.. dosn't help (I didn't switch to firefox in this time).

BOINC is now using the wrong buttons to display the active task.

I've streched/shrunk, zoomed in /zoomed out, the windows/cells, no fatal errors yet.

Nope it's just not going to die on me.

My five minute test anyway.

I'll let it go over night and see what happens.

Does anybody know if the BOINC graphics has been updated 5.4.x and 5.7.5? maybe they squished a few bugs.

If they complete OK I'll try out the screensaver.

I speak too soon, WU 2 has died, I can still zoom/rotate on WU 1 but I can't access WU 2 graphics and I have dropped to ~55% CPU.

It is still on my bar, but I can't view/close it.
WIndows tells me that:

This program is not responding

etc...

6/12/2006 7:39:44 PM|rosetta@home|Unrecoverable error for result 2reb__ETABLE_TEST_ABRELAX_rhh13sm6atrrep__1416_5220_0 ( - exit code -1073741819 (0xc0000005))


Because it's an 0xc0000005, it may of been something other than graphics, they happen.

I will have another go tomorrow.

Yes I will get tell everybody all of the graphics drivers and stuff, this was just a quick test because I was passing through.

Edit: The bad WU is Here

The problem is I cannot recreate it.
I have
All WindowsXP

Pentium3-m, 512MB with Mobility Radeon (16MB) (laptop)
Pentium4-m, 512MB with S3 Supersavage (16MB) (laptop)
PentiumM 1.7 (o/c 1.9), 1GB with GeForceFX-Go 5200 (128MB) (laptop)
PentiumM 1.9, 1GB with Mobility Radeon 9200 128MB [I think!] (laptop)

Athlon 3200+, 512Mb with GeForceFX 5200 (128Mb)
Athlon64 3500+ (o/c to 2.4), 1Gb with GeForce 6800GS (256Mb)
Athlon 1600+ with GeForce 4200 (128Mb)

That I have tested on and it just doesn't crash for me.

I have tried normal install and showing graphics, the unofficial service install with graphics.
Screensaver launching on some of them.

---

Team mauisun.org

For the information.


Since the debugging seems to be working now (and every one should be using at least 5.4.11 for it to give the most information)

The graphics card & version can be seen from there (hence the devs should esily be able to compile a list and compare the crashing units)

For instance hug's last crashed task
https://boinc.bakerlab.org/rosetta/result.php?resultid=50173347
at little bit into it
ModLoad: 69000000 00678000 C:WINDOWSsystem32atioglxx.dll (6.14.10.5014) (-exported- Symbols Loaded)
File Version : 6.14.10.5014
Company Name : ATI Technologies Inc.
Product Name : ATI OpenGL driver
Product Version: 6.14.10.5014

So an ATI card, using driver version 6.14.10.5014 a old Catalyst ~3.4
I would give updating the drivers a go.

But the devs can have a look to guess what really went wrong.


Though the Access Violation (0xc0000005) is almost a catch all problem.
From hardware, memory, drivers or bad software even DEP problems...

---

Team mauisun.org

> I am glad that some people out there do not have a problem with the graphics as I did not have such a happy experience.
I posted a number of times over the last 2 months that the graphics were a problem but nothing was done about it. At one stage I was only getting about 1 in every 4 or so workunits to actually process to the end and give me some cobblestones.
In the end I said "stuff this" and turned off the Boinc screensaver altogether (I now run basic Windows screensavers) and all my workunits started to process without any problems, so I have left it off for the last 3 weeks and had no more trouble.
This also applies to Ralph.
I have an Intel P4 2.53 with an NVIDIA Geforce4 Ti4600 (128MB) graphics card.
Also an AMD 4800+ with dual Winfast PX6600 GT cards in SLI (128 MB ea)
Both have had the problem but more so with the AMD.
All my other computers do not have graphics (either service install or Linux) and have not had the problem.
So at this stage I will not be turning the graphics back on as it was causing too many problems and the note from feet1st saying chu has acknowledged a problem is the first I have heard that the project admits there is something wrong.

---

Just for the infor we are not the only project with the
Access Violation ..... I posted above.

just put that and BOINC into google.

Though ROM has a write up on how he thinks he fixed it (and why it was happening) when he was helping Rosetta@Home get rid of it in the first place
Google cached link

(cahced because I cannot get the actual article at the moment)

---

Team mauisun.org

OK, so last night I enabled the screensaver on one of my machines. Overnight it has botched several WUs. This is noteable because I have a 24hr runtime preference, and my machine is well versed in crunching them that long.

They all seem to be ended by the watchdog and I seem to get 20 credits granted, regardless of how much time was spent on crunching the WU.

It is this host. The messages tab doesn't even show anything out of the ordinary, other then the time it took to complete.

I will download the PDB version now. ...actuall... I thought Chu had said he put it out there. I'm expecting to see a Windows copy for v5.41 with a .pdb suffice, just as you see there for v5.16. So either there are still some permissions problems with the file, or it's already incorporated into everyone's download... or I am not understanding what I should do to run with additional debugging capabilities.

...screensaver... back to NONE!

Conan, I'm glad you are here posting. I wouldn't say your posts were ignored, although with no results from them I certainly understand why you would say that. It's gotta be tough when people post about things you don't see in your lab and have no idea on how to fix. In fact it may be a BOINC bug rather then Rosetta.

That was my purpose for opening this thread. I'm hopeful that something will turn up, and when it does, we'll all know where to go to look.

I think the pdb is built in since they moved to getting the symbols remotely.
It just hadn't been enabled. Some people seem to be getting debug info.

---

Team mauisun.org

OK, so last night I enabled the screensaver on one of my machines. Overnight it has botched several WUs. This is noteable because I have a 24hr runtime preference, and my machine is well versed in crunching them that long.

They all seem to be ended by the watchdog and I seem to get 20 credits granted, regardless of how much time was spent on crunching the WU.

It is this host. The messages tab doesn't even show anything out of the ordinary, other then the time it took to complete.

I will download the PDB version now. ...actually... I thought Chu had said he put it out there. I'm expecting to see a Windows copy for v5.41 with a .pdb suffix, just as you see there for v5.16. So either there are still some permissions problems with the file, or it's already incorporated into everyone's download... or I am not understanding what I should do to run with additional debugging capabilities.

...screensaver... back to NONE!

Conan, I'm glad you are here posting. I wouldn't say your posts were ignored, although with no results from them I certainly understand why you would say that. It's gotta be tough when people post about things you don't see in your lab and have no idea on how to fix. In fact it may be a BOINC bug rather then Rosetta.

That was my purpose for opening this thread. I'm hopeful that something will turn up, and when it does, we'll all know where to go to look.

Here's the dxdiag info. from that machine (there's a nice little "save all information" button that writes it all to a text file, just delete out what you don't need)

------------------
System Information
------------------
Time of this report: 12/6/2006, 21:23:03
Machine name: Feet1st
Operating System: Windows XP Home Edition (5.1, Build 2600) Service Pack 2

(2600.xpsp_sp2_gdr.050301-1519)
Language: English (Regional Setting: English)
System Manufacturer: IBM
System Model: 819533U
BIOS: Phoenix FirstBios(tm) Desktop Pro Version 2.0 for IBM

ThinkCentre.
Processor: Intel(R) Pentium(R) 4 CPU 2.80GHz (2 CPUs)
Memory: 1278MB RAM
Page File: 532MB used, 2521MB available
Windows Dir: C:WINDOWS
DirectX Version: DirectX 9.0c (4.09.0000.0904)
DX Setup Parameters: Not found
DxDiag Version: 5.03.2600.2180 32bit Unicode


---------------
Display Devices
---------------
Card name: NVIDIA GeForce4 MX 440 with AGP8X
Manufacturer: NVIDIA
Chip type: GeForce4 MX 440 with AGP8X
DAC type: Integrated RAMDAC
Device Key: EnumPCIVEN_10DE&DEV_0181&SUBSYS_90301462&REV_A2
Display Memory: 64.0 MB
Current Mode: 1024 x 768 (32 bit) (60Hz)
Monitor: NOKIA 445Xi
Monitor Max Res: 1600,1200
Driver Name: nv4_disp.dll
Driver Version: 6.14.0010.5303 (English)
DDI Version: 9 (or higher)
Driver Attributes: Final Retail
Driver Date/Size: 11/17/2003 11:33:00, 4323968 bytes
WHQL Logo'd: n/a
WHQL Date Stamp: n/a
VDD: n/a
Mini VDD: nv4_mini.sys
Mini VDD Date: 11/17/2003 11:33:00, 1618939 bytes
Device Identifier: {D7B71E3E-42C1-11CF-6563-3AB003C2CB35}
Vendor ID: 0x10DE
Device ID: 0x0181
SubSys ID: 0x90301462
Revision ID: 0x00A2
Revision ID: 0x00A2
Video Accel: ModeMPEG2_A ModeMPEG2_B ModeMPEG2_C ModeMPEG2_D
Deinterlace Caps: {212DC723-3235-44A4-BD29-E1652BBCC71C}:

Format(In/Out)=(YUY2,0x3231564e) Frames(Prev/Fwd/Back)=(0,0,1)

Caps=VideoProcess_YUV2RGB VideoProcess_StretchX VideoProcess_StretchY

DeinterlaceTech_MedianFiltering
{335AA36E-7884-43A4-9C91-7F87FAF3E37E}:

Format(In/Out)=(YUY2,0x3231564e) Frames(Prev/Fwd/Back)=(0,0,0)

Caps=VideoProcess_YUV2RGB VideoProcess_StretchX VideoProcess_StretchY

DeinterlaceTech_BOBVerticalStretch
{212DC723-3235-44A4-BD29-E1652BBCC71C}:

Format(In/Out)=(NV12,0x3231564e) Frames(Prev/Fwd/Back)=(0,0,1)

Caps=VideoProcess_YUV2RGB VideoProcess_StretchX VideoProcess_StretchY

DeinterlaceTech_MedianFiltering
{335AA36E-7884-43A4-9C91-7F87FAF3E37E}:

Format(In/Out)=(NV12,0x3231564e) Frames(Prev/Fwd/Back)=(0,0,0)

Caps=VideoProcess_YUV2RGB VideoProcess_StretchX VideoProcess_StretchY

DeinterlaceTech_BOBVerticalStretch
Registry: OK
DDraw Status: Enabled
D3D Status: Enabled
AGP Status: Enabled
DDraw Test Result: Not run
D3D7 Test Result: Not run
D3D8 Test Result: Not run
D3D9 Test Result: Not run

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I've also been having problems with the screensaver graphics, but I've been reporting them in the 5.41 problems thread (just noticed this one). I do have a machine which runs the graphics and has no problems, it happens to have a single-core, non-HT processor in it. Perhaps this is significant. I'll try to get more of these on Rosetta to see if that works as a rule.

I was just thinking, doesn't BOINC use OpenGL for the graphics? Why would the DirectX version you have installed make any difference?

---

I've been wondering... in order to draw the graphic, the thread doing so must access that data that the crunching thread is working on. (or at least I'm led to believe that graphic for the screensaver is done in a second thread) What mechanism is in place to assure there are no conflicts between the continued crunching, and the thread drawing the graphic? Is crunching halted while a frame is drawn? I believe the answer is going to be that a semaphore or some other locking mechanism is in place to serialize access to the required data, if perhaps only long enough to stash away a seperate copy for the graphic thread to use...

But what if there are little bits of code here and there that are not properly using the serialization mechanism? Well, then you might have fairly random, yet frequent conflicts where the crunching thread is writing to the storage that the graphic thread is reading.

If something like this were the cause, then I'd expect that the problem should be more frequent on a system with a faster frame rate. Unfortunately, the default value at 10 frames per second is actually the lowest value you can select.

...could I ask that anyone that feels they've got a system that is rock solid, regardless of running as screensaver, enable the BOINC screensaver, then go in to your Rosetta preferences (perhaps for a specific location, if you've got more then one machine) and set the frame rate up a notch per day? And see if it remains stable, or if you eventually hit a point that you see problems as well?

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

I was just thinking, doesn't BOINC use OpenGL for the graphics? Why would the DirectX version you have installed make any difference?

...I have no idea :) I was just dumping what it showed for the display info. in case it is significant.

Yes, since various Rosetta version seem to have screensaver problems, I felt a seperate thread might help pull the issue out, and help keep the version-specific thread stay focused on any other issues that arise.

I agree, perhaps the hypterthreading is an issue. I've now enabled the screensaver on the same PC again for tonight's overnight crunching, but I've set this location to run only on a single CPU (BOINC General Preferences).

Let's see if I can turn the 15hrs I've got in to these two WUs here into 20 credits :)

By the way, I meant to point out that review of all the WUs that seemed to end prematurely when I had screen saver active, shows 20 credits, a normal end, and that the watchdog ended the run due to a stuck Rosetta score.

Perhaps this explains some of the credit issues as well? It's really only due to my NORMAL long runtime that these stand out like such a sore thumb. If my normal were the 3hrs, these might blend in with other WUs.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Rosetta's graphics routines call DirectX (DDraw) also a just incase ;-) but the readout gived the computer information. e.g. driver version, base computer setup.


Interesting is yesterday (or maybe today for the US people)
a fix has been checkied in to the CVS, SO when 5.7.6 gets released you may want to give it a go.

David 6 Dec 2006
- maybe fixed crash bug in core client:
If an app was doing screensaver graphics
and didn't respond for 4 seconds to message to hide graphics,
we were calling kill_task() but not setting
ACTIVE_TASK::task_state = PROCESS_ABORTED.
As a result, the ACTIVE_TASK wasn't getting removed.
Further attempts to reference its shared-mem segment
or its handle would produce errors;
eventually the core client crashes.
Solution: set task_state to ABORTED in kill_task();
Note: actually we don't want to kill the task in this case.
I'll do this later.
- core client: print separate message when killing non-responding
screensaver app
- core client: prepend [file_xfer] to messages where needed
- server DB code: add ping_user, ping_time fields to TEAM
(not used in C++ code, but should be here for good form)

client/
app.C,h
app_control.C
app_graphics.C
pers_file_xfer.C
ss_logic.C
db/
boinc_db.C,h

It may ro may not be related to this bug.


So maybe increasing the graphics processor % may keep it responding quicker... and .. who knows ..

---

Team mauisun.org

I'm running CC 5.6.5 on all my machines. I tried 5.7.2 and 5.7.5 for a while, but had screensaver issues not just with Rosetta, but with all projects. Here's what would happen: normally, the graphics would change at 10 minute intervals, but at some point (before its 10 minute time slice was up), the graphics would disappear and be replaced with the default BOINC graphic saying "screensaver shutting down", or something like that. Then, at the end of what would have been the 10 minute period, the next graphic would appear, but only for a few seconds, to be pre-empted again by the same "shutting down" message. I don't know if this bug fix will help Rosetta specifically, but it may make the 5.7 series usable. I'll certainly give it a try.

I now have 3 single-threaded machines (2 desktops, 1 laptop) attached to Rosetta, we'll see if they have problems. All have NVidia cards with various driver versions except my one here at home with the ATI card I'm testing. I've seen a few graphics freezes with the ATI (not all of which ended the WU), so I guess that isn't likely to help solve it.

---