Do you think that's in response to this specific situation, or just a general enhancement that they've arrived at for other reasons?

Do you think that's in response to this specific situation, or just a general enhancement that they've arrived at for other reasons?

Let's just say we know people with large rolling pins, lol

---

"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.

Do you think that's in response to this specific situation, or just a general enhancement that they've arrived at for other reasons?

Let's just say we know people with large rolling pins, lol

LOL...good to know!

Just noticed that the wingman on one of my error tasks from Monday evening, host 7440776, also got the same error. In fact, he currently appears to have generated at least 170 of them over the last 4 days. That host is also Windows 7 running BOINC 7.6.9. One major difference I see, however, is that he's also getting the error on Cuda tasks, something I haven't experienced (yet).

I'm getting the sense that if we really looked for them, we'd find that this is becoming an increasingly common error and something worth keeping an eye on to see if it's possible to nail down the configurations where it occurs.

That wingman is using a driver that no longer properly supports that card (it does so in legacy mode only) and needs to go back to a pre-340.xx driver.

Cheers.

The trick is not the size of the rolling pin, the trick is to know how and where to deploy.

What I don't quite understand is why we are suddenly seeing these errors.

Those who experienced them, can you pinpoint when it started? And while we are there, can you check if we are looking at a problem when a task first starts, when it restarts or even something that happens to running tasks.

Doing a controlled shutdown experiment and examining the entrails might help.

What kind of shutdown was it anyway?

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

The general gist of the way it usually goes, is some or another small change happened for security/safety or other reasons, that exposes a latent weakness in the way things were done before (that seemed OK to whoever implemented it, but may or may not have been totally in line with published best practices -- Who cares right ?). Add to that decades of cruft from code that was made to work on Win95/98 or earlier in the first place, which were fundamentally different operating systems to XP onwards (completely different team).

Analagously, probably a large proportion of Volkswagen owners in the US were completely happy with their purchases until someone exposed discrepancies (iirc from Cambridge or similar). --> New regulations + Old design = patch or replace it, just what happens when requirements change.

---

"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.

Using the search link that Janneseti posted in the second post in this thread, I've found cases dating back to January 2014 (sztaki) and August 2013 (CPDN), both from machines running Windows 7. It's just that the BOINC community at large doesn't take a robust attitude to identifying the meaning of error/exit codes, and tracking them back to source.

Aided and abetted by the poor reporting format used by BOINC for NTSTATUS codes. David's now checked in what looks like a full cleanup in that area, so that will slowly propagate to all BOINC projects as they update their webservers. Anticipated timescale? About a decade...

Anticipated timescale? About a decade...

Ever the optimist. Some projects run server code from 2008...

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

I found the trick to speeding up propagation of updated software while reducing costs, isn't the goose liver pate force-feeding (constant trivial update) approach like m$ have been attempting lately.

Make something twice as good, or half the cost(mostly time investment), or some mix of metrics that amount to twice as good, then people will update in droves (e.g. reduce project maintenance overhead -> projects will adopt)

GPU companies (AMD and nVidia) are doing that regularly. older institutionalised businesses not so much.

---

"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.

GPU companies (AMD and nVidia) are doing that regularly. older institutionalised businesses not so much.

That's not a good approach, if the updated code goes out with substantial breakages. Jacob Klein has got three NVidia bugs corrected in drivers released as mainstream - since they abandoned Beta testing.

GPU companies (AMD and nVidia) are doing that regularly. older institutionalised businesses not so much.

That's not a good approach, if the updated code goes out with substantial breakages. Jacob Klein has got three NVidia bugs corrected in drivers released as mainstream - since they abandoned Beta testing.

Had two or three corrected ourselves. Windows (or whatever OS) drivers aren't AMD's or nVidia's products, per se, but support material, and are specified by the OS Vendor & changing software needs.

Maybe reasonably they should be considered part of the hardware (and maybe are by many users). That's where AMD/ATI's 'Close to Metal' evolved through Mantle now into Vulkan ---> remove decades of M$ controlled driver cruft (with a scalpel if necessary)

---

"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.

Microsoft pushing feature-limited drivers (no OpenCL) through Windows Update is bad Microsoft practice.

NVidia posting 'recommended' drivers on their own website which fail to complete their own SDK sample applications is bad NVidia practice.

It's all just getting sloppy out there.

Microsoft pushing feature-limited drivers (no OpenCL) through Windows Update is bad Microsoft practice.

NVidia posting 'recommended' drivers on their own website which fail to complete their own SDK sample applications is bad NVidia practice.

It's all just getting sloppy out there.

Agreed on that point. Comprehensive regression testing is cheap insurance. I'm convinced modern business practices are focussed more on being first to market than the best, and in Agile development parlance things not working on X isn't a blocker when the target market is Y.

The nVidia products targeted at HPC applications are the Tesla range, and special Tesla Compute Cluster drivers. So with GeForce (Gaming) line we sadly play second (or 151st ) fiddle to the new Doom release.

---

"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.

Which is beginning to make me feel that when David jumped into bed with NVidia back in 2008, we got VHS instead of Betamax.

Which is beginning to make me feel that when David jumped into bed with NVidia back in 2008, we got VHS instead of Betamax.

Could well be the case, however you're starting to see AMD turn their failing business around through similar methods (with their own slant of course).

Remember of the GPU market, who has the largest share ? Do you know already ?

---

"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.

Remember of the GPU market, who has the largest share ? Do you know already ?

Intel?

Edit - and their driver management is crap, too.

Remember of the GPU market, who has the largest share ? Do you know already ?

Intel?

Edit - and their driver management is crap, too.

Yes Intel


source [edit:] fixed initially wrong link

Point being the fragile/broken driver infrastructure (wrt gpgpu on desktops) doesn't appear to be driving that (much).

Looks more like Intel turning the screws by integrating graphics, and the others left to fight over the scraps.

---

"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.

It would be interesting to see that expressed as actual numbers, rather than share - given that desktop machines are going out of fashion, in favour of laptops, tablets and smartphones.

I don't think you'd see the current generation of Intel GPUs advertised as 'for the gamers' - but they're plenty good enough for business use. Unless your business happens to be rendering CGI movies for IMAX.

"Total discrete GPUs (desktop and notebook) shipments decreased -11.2% from the last quarter"

It would be interesting to see that expressed as actual numbers, rather than share - given that desktop machines are going out of fashion, in favour of laptops, tablets and smartphones.

I don't think you'd see the current generation of Intel GPUs advertised as 'for the gamers' - but they're plenty good enough for business use. Unless your business happens to be rendering CGI movies for IMAX.

Yes I also figure that AMD's drop there has more to do with Bitcoin miners switching mostly to ASICs, while their game console market might have stayed more stable (I imagine).

In either case, both NV's and AMD's launch presentations made big about wanting DX12/Vulkan & VR gaming markets in particular, both things Intel Integrated graphics cannot achieve without years long development cycles.

You might get your wishes of better infrastructure and support for gpgpu compute by proxy, just because VR/Physics and AI require it integrated along with triangles.

---

"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.

Microsoft pushing feature-limited drivers (no OpenCL) through Windows Update is bad Microsoft practice.

NVidia posting 'recommended' drivers on their own website which fail to complete their own SDK sample applications is bad NVidia practice.

It's all just getting sloppy out there.

Cause it doesn't cost them anything actually. Reputational losses? LoL, if they ever were then M$ with Win10 upgrade policy would be in hell more than half year ago... Reputation hardly to monetize now... And in world that ruled by money non-monetizible things slowly or not so slowly get exterminated...