The problem with Diesel engines is that they emit small powders (PM10). Milano is awash in small powders since Milanesi buy many Diesel cars. Also all cars with catalytic converters (most cars now have one) produce N2O, which sunlight transforms in ozone, and smog. AQUA@home has started using CUDA. This is ironic, since the aim of D-wave is to build a quantum computer and all AQUA programs simulate one. But I don't think nVidia has supported them as it has done with SETI.
Tullio
My Diesel has a particle filter so it doesn't emit more particles than a petrol car.
It's good to see more and more projects going CUDA, more choice can't be bad.
It's good to see more and more projects going CUDA, more choice can't be bad.
To true ...
It is also true that we are still seeing significant problems on almost all GPU projects of one sort or another.
SaH & SaH Beta has real troubles running VLAR tasks on CUDA and sometimes crashes the GPU eating all subsequent tasks until reboot.
GPU Grid off and on has a set of tasks that won't run... they will all crash (slowly becoming more rare)
Aqua - from the initial reports, Ugh!
Ramsey - See Aqua ...
Milky Way - check is in the mail
The Lattice Project - same error rate as CPU tasks, sadly that is 10% or more (my experience) and though the project seems to think that that is acceptable, well, I for one, don't ... especially after all these years ...
Still, it will be nice when we have more than one or two choices ... personally I was really hoping for Einstein to be in the lead ...
I am running AQUA@home units because the subject interests me. Both 72-qubits units and 96-qubits take about 69 hours on my CPU. A 128-qubit may take up to 150 hours. Checkpointing is now working, contrarily to the past. At least my work is getting done. Ohne haste aber ohne raste (correct?). Ernest Rutherford.
Tullio
It's intresting to hear that the BOINC favorite NVIDIA apps have so many problem while Cluster Physik ATI gpu application runs absolutely flawlessly. :)
Well, I think there were some problems as well with the ATI app (e.g. because ATI/AMD changed some DLL names, APIs, whatever in their drivers), so flawless is a bit of an exaggeration. Also AFAIK the ATI app doesn't interact with the bOINC co-processor scheduling, right? That removes a whole layer of problems but introduces other issues I guess. Still, it's an amazing app, that's for sure.
Well, I think there were some problems as well with the ATI app (e.g. because ATI/AMD changed some DLL names, APIs, whatever in their drivers), so flawless is a bit of an exaggeration. Also AFAIK the ATI app doesn't interact with the bOINC co-processor scheduling, right? That removes a whole layer of problems but introduces other issues I guess. Still, it's an amazing app, that's for sure.
Though I was able to get it to work somewhat I was one of those that did have issues with the ATI application. It worked "well enough" for long enough that it was worth the aggrivation to me ... but, not what I would call flawless either...
And no, it was not integrated into the base of the BOINC Client and we are still waiting for recognition of the ATI cards and the OpenCL that is forthcoming ... sadly, the way CUDA was implemented means that the integration of those two and other GPU systems will be just as painful because the use of a virtual class or some other abstraction for the specific processing class was not used.
So, you have the current three classes, NCI, CPU and CUDA all kinda smushed together with no clear distinctions in how things are handled. Theory says it should not matter, but, practice says that the code leaves a lot to be desired. Worse, without the abstraction and with the code intermixed it is hard to tell if one thing is stepping on another ... I have been told I do to much analysis and draw the wrong conclusions, but I demonstrated a clear case where a suspended CPDN task, resident in memory prevented the running of two tasks on the GPUs ... stopping and restarting BOINC to bump the CPDN task out of memory cleared that up ...
No matter, they are now adding more useless features for AM systems that hardly anyone uses and likely will not fix any of the severe issues that have plagued BOINC like since forever ...
It's a question of tempus guys ;)
English is not my first language but i think
that 'runs absolutely flawlessly' indicates that I referee
to present time.
In fact for me it has working flawlessly for more that 2 months which
is the time I have bin using the application.
It's a question of tempus guys ;)
English is not my first language but i think
that 'runs absolutely flawlessly' indicates that I referee
to present time.
In fact for me it has working flawlessly for more that 2 months which
is the time I have bin using the application.
Which project do you use your ATI card to crunch for?
Read the readmed file that is included in the 0.19e zip.file
If you are using Catalyst 9.2 or higher you have to install the...
amdcalrt.dll or amdcalrt64.dll
amdcalcl.dll or amdcalcl64.dll
amdcaldd.dll or amdcaldd64.dll
manually in the Windows\system32 folder.
If you update the Catalyst driver you have to reinstal
amdcal... files again after you have instaled the Catalyst driver.
Read the readmed file that is included in the 0.19e zip.file
If you are using Catalyst 9.2 or higher you have to install the...
amdcalrt.dll or amdcalrt64.dll
amdcalcl.dll or amdcalcl64.dll
amdcaldd.dll or amdcaldd64.dll
manually in the Windows\system32 folder.
If you update the Catalyst driver you have to reinstal
amdcal... files again after you have instaled the Catalyst driver.
RE: The problem with Diesel
)
My Diesel has a particle filter so it doesn't emit more particles than a petrol car.
It's good to see more and more projects going CUDA, more choice can't be bad.
Bikeman
RE: It's good to see more
)
To true ...
It is also true that we are still seeing significant problems on almost all GPU projects of one sort or another.
SaH & SaH Beta has real troubles running VLAR tasks on CUDA and sometimes crashes the GPU eating all subsequent tasks until reboot.
GPU Grid off and on has a set of tasks that won't run... they will all crash (slowly becoming more rare)
Aqua - from the initial reports, Ugh!
Ramsey - See Aqua ...
Milky Way - check is in the mail
The Lattice Project - same error rate as CPU tasks, sadly that is 10% or more (my experience) and though the project seems to think that that is acceptable, well, I for one, don't ... especially after all these years ...
Still, it will be nice when we have more than one or two choices ... personally I was really hoping for Einstein to be in the lead ...
I am running AQUA@home units
)
I am running AQUA@home units because the subject interests me. Both 72-qubits units and 96-qubits take about 69 hours on my CPU. A 128-qubit may take up to 150 hours. Checkpointing is now working, contrarily to the past. At least my work is getting done. Ohne haste aber ohne raste (correct?). Ernest Rutherford.
Tullio
It's intresting to hear that
)
It's intresting to hear that the BOINC favorite NVIDIA apps have so many problem while Cluster Physik ATI gpu application runs absolutely flawlessly. :)
Well, I think there were some
)
Well, I think there were some problems as well with the ATI app (e.g. because ATI/AMD changed some DLL names, APIs, whatever in their drivers), so flawless is a bit of an exaggeration. Also AFAIK the ATI app doesn't interact with the bOINC co-processor scheduling, right? That removes a whole layer of problems but introduces other issues I guess. Still, it's an amazing app, that's for sure.
CU
Bikeman
RE: Well, I think there
)
Though I was able to get it to work somewhat I was one of those that did have issues with the ATI application. It worked "well enough" for long enough that it was worth the aggrivation to me ... but, not what I would call flawless either...
And no, it was not integrated into the base of the BOINC Client and we are still waiting for recognition of the ATI cards and the OpenCL that is forthcoming ... sadly, the way CUDA was implemented means that the integration of those two and other GPU systems will be just as painful because the use of a virtual class or some other abstraction for the specific processing class was not used.
So, you have the current three classes, NCI, CPU and CUDA all kinda smushed together with no clear distinctions in how things are handled. Theory says it should not matter, but, practice says that the code leaves a lot to be desired. Worse, without the abstraction and with the code intermixed it is hard to tell if one thing is stepping on another ... I have been told I do to much analysis and draw the wrong conclusions, but I demonstrated a clear case where a suspended CPDN task, resident in memory prevented the running of two tasks on the GPUs ... stopping and restarting BOINC to bump the CPDN task out of memory cleared that up ...
No matter, they are now adding more useless features for AM systems that hardly anyone uses and likely will not fix any of the severe issues that have plagued BOINC like since forever ...
It's a question of tempus
)
It's a question of tempus guys ;)
English is not my first language but i think
that 'runs absolutely flawlessly' indicates that I referee
to present time.
In fact for me it has working flawlessly for more that 2 months which
is the time I have bin using the application.
RE: It's a question of
)
Which project do you use your ATI card to crunch for?
Milkyway@home And the GPU
)
Milkyway@home
And the GPU apps is here
0.19e is the latest version.
Read the readmed file that is included in the 0.19e zip.file
If you are using Catalyst 9.2 or higher you have to install the...
amdcalrt.dll or amdcalrt64.dll
amdcalcl.dll or amdcalcl64.dll
amdcaldd.dll or amdcaldd64.dll
manually in the Windows\system32 folder.
If you update the Catalyst driver you have to reinstal
amdcal... files again after you have instaled the Catalyst driver.
RE: Milkyway@home And the
)
Thanks