Almost. The Teslas cost 1000 to 2000$ more, whereas I'm talking about 10 to 20$ more. I suppose what makes the Teslas really expensive is the extensive testing and "guaranted" functionality (if there is something like that for chips at all). They wouldn't neccessarily need that for "heavy duty GP-GPUs".

MrS

---

Scanning for our furry friends since Jan 2002

Radiation
GPU’s, and everything else, suffer from continuous Radiation bombardment. As well as Ionising radiation, Neutrons, Protons, Muons and Pions we are even hit by Cosmic Radiation. These all cause random system Errors. Most impacts do not cause permanent damage, but sometimes RAM has to be replaced or a BIOS reset. They might cause the most damage to hard disk drives, and it is why your CD’s of precious memories won’t be readable in 15 or 20 years! You can’t hide from a particle that can go through 10’ of lead. It’s also why aircraft are grounded when there is an increase in solar flares. So the next time your system restarts you have something to blame – it could always be solar radiation!

High energy particles cause transient errors by impact ionization. They don't generally cause permanent erros: the mass/energy difference between these particles and the atoms of your chip are too large for significant momentum transfers. Therefore they can (temporarly) kick electrons out of their bindings, but they can hardly move atoms.

MrS

---

Scanning for our furry friends since Jan 2002

I agree that most (radiation) impacts do not cause permanent damage - a restart and you are up and running again.
Theoretically speaking however (and only in very rare circumstances) Neutrons can cause Permanent damage (and yes they can move atoms, and sometimes not just one). Unfortunately however, radiation does not have to move atoms to cause permanent damage, just the odd atomic bond; causing material degradation.
Humans have a protein called telomerase that repairs such damage when it occurs in DNA, but Computers don’t have an equivalent just yet.
We might all live for much longer if the Telomerase gene was not stuck at the end of a Chromosome (which shrinks with age)! Cancers would probably not be such the problem either. It’s such a pity nobody is studying this Cure for All Cancers Solution.

Well get to work then.

Theoretically speaking however (and only in very rare circumstances) Neutrons can cause Permanent damage (and yes they can move atoms, and sometimes not just one). Unfortunately however, radiation does not have to move atoms to cause permanent damage, just the odd atomic bond; causing material degradation.

You're right, neutrons can move atoms. good that there aren't too many of them in the cosmic ray mix :)

And you wouldn't have to reboot on every transient error: the fault might not lead to any disturbing consequences. Furthermore I heard a talk about 3 years ago where the Prof. said Intels core logic is entirely "radiation hardened" to the point where they can detect 2-bit errors and correct 1-bit errors. Don't quote me on the these numbers, though.. it's been quite some time.

Interesting that you mention breaking bonds. This is actually what causes the slow degradation of chips over time, it's just not mainly caused by cosmic radiation. The defects at the Si-SiO2 interface (or now Si-HfOx) are passivated by hydrogen atoms. Over time the occasional highly energetic electrons (the boltzmann tail, or from the substrate) kick these light hydrogen atoms out and a dangling bond is created. This is a "trap state" for charge carriers. Once such a trap contains charge the transistor operation is influenced (the threshold voltage shifts), which can only be bad in either direction.

MrS

---

Scanning for our furry friends since Jan 2002

I was not implying that neutrons were part of the cosmic radiation! Most sources are quite terrestrial. They are also very rare, and are usually produced by other rare particle bombardments. The biggest single radiation concern for humans is Radon, as it is in the stone of many buildings, work benches, ornaments and the rocks beneath us. I’m sure the ionising radiation that directly or indirectly results from Radon causes computer problems too.

My bad, I read too much into that! Regarding computer problems due to radioactivity: fortunately we don't need to worry about the alphas here, as they can not even penetrate paper. Betas and Gammas, on the other hand, can cause transient errors by ionization if the design is not radiation hardened or there are too many of them (but you likely wouldn't care much as you'd sit in the middle of a fission reactor :D). But at a few MeV they can probably create dangling bonds and thus lead to component decay.

MrS

---

Scanning for our furry friends since Jan 2002

Humans have a protein called telomerase that repairs such damage when it occurs in DNA, but Computers don’t have an equivalent just yet.
We might all live for much longer if the Telomerase gene was not stuck at the end of a Chromosome (which shrinks with age)! Cancers would probably not be such the problem either. It’s such a pity nobody is studying this Cure for All Cancers Solution.

The "shrinking with age" regarding telomerase probably has little effect on how long we live currently given upper population life expectancies of around 85 years for Japanese women. Essentially, this repair process and shrinking is related to the "Hayflick Limit" in cell division, which places a finite limit on natural human life span at around 250 years (when "shrinking" results in lengths too short for the division to occur properly). Research on telomerase (and related issues) has been ongoing for three or more decades, including some work on cell division in some cancers.

Telomerase repairs damaged DNA, but the Telomerase gene resides close to the ends of chromosomes (the end region is called a telomere). So my point was that when chromosomes shrink overall Telomerase production is reduced in the body, and non-existent in some cells. Without Telomerase DNA stays damaged, so there is a greater risk of Cancer and other illnesses.

Just because someone is researching something does not mean they are looking for a cure. They might just be looking!

Anyway back to the topic – Video Card Longevity

It’s a good idea to use a fine mesh fan filter on your system inlet fans. To clean the system all you have to do is point the vacuum cleaner at it for about 2 seconds every other week. But they don’t just keep the dust off your components. I looked at a system about 2 years ago and was told it made a loud pop/bang noise and stopped suddenly. There was a bluebottle fly lying at the bottom of the case.

...Without Telomerase DNA stays damaged, so there is a greater risk of Cancer and other illnesses.

Just because someone is researching something does not mean they are looking for a cure. They might just be looking!

Just an FYI...see here and here for example.

Anyway back to the topic – Video Card Longevity

It’s a good idea to use a fine mesh fan filter...But they don’t just keep the dust off your components. I looked at a system about 2 years ago and was told it made a loud pop/bang noise and stopped suddenly. There was a bluebottle fly lying at the bottom of the case.

Dust is definitely not the only problem. I saw a system about 5 years ago that had the same "pop/bang" noise problem...opened the case only to find a nice colony of ants (some rather toasted)! I doubt even the fine mesh would have kept them out.

Both these research teams took the inverted smart approach. If something is essential for life, they want to kill it. They know they will take out a few cancer cells on the way, be able to publish in a few obscure journals, and further their career. If they get really lucky a drug company will develop some sort of anti-telomerase, to slowly kill people with, and they will get a bit of money out of it. Drug companies don’t do cures! Unfortunately this sort of research undermines science and interferes with the work of descent scientist that are really trying to do something positive.

Hear, Hear.

It's the Cholestorol arguement.

Your quack will tell you your cholestoral is to high and give you drugs to reduce it (Statins) lipitor etc.

What your quack never tells you is that most (up to 90%) of the Cholestorol in your body is MADE by your own Liver. From that you could assume your Liver is trying to kill you by making an excessive amount of Cholestorol. Actually Cholestorol is needed by every cell in your body for life and if your Liver is making more of it, it's doing it for a reason.

Atherosclerosis (hardening of the arteries) is caused by plaques made of Cholestorol (soft) and Calcium (hard) so drugs companies came up with 'lets make a drug which reduces the Livers' ability to produce this terrible substance' and they came up with Statins (Billions of $$$ are spent on Statins every year) despite the fact that Lipitor is known to cause memory loss and is in the dock for causing Cancers.

So why does your Liver begin producing excessive Cholestorol? To repair you, that's why. Your Arteries get damaged in use and your body would normally repair them using substances like Collagene, however, if you lack sufficient quantities of Collagene your body uses...YES Cholestorol which acts like a sticking plaster which the body places over the damaged area in your arteries. Your Liver is NOT trying to kill you it's trying to save your LIFE.

---

Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

Both these research teams took the inverted smart approach. If something is essential for life, they want to kill it.

Neither of these were research teams doing work on telomerase and cancer. Both were review articles (from 1996 and 2001) demonstrating that at least theraputic research in this area has been going on for quite some time and were provided by me to counter your statement that "It’s such a pity nobody is studying this Cure for All Cancers Solution".

...publish in a few obscure journals...

Though I probably wouldn't really call "Scientific American" a journal (the first review piece), it is hardly obscure. "Human Molecular Genetics" (the second review piece) is a prominent journal in the area.

Unfortunately this sort of research undermines science and interferes with the work of descent scientist that are really trying to do something positive.

I am really at a loss with this kind of statement. Are you really suggesting that the U.S. and Japanese researchers from the second article are not descent scientists?


Anyway, this has gone way off topic, so I apologize for Hijacking the thread.

For Scott Brown only, everyone else skip to the last paragraph!

Quote from your most recent abstract, 2001, link in your post.
“By reverse transcription, the telomerase RNP maintains telomere length stability in almost all cancer cells”.
Why, however did they work that out?
Perhaps it’s because telomerase maintains telomere length and stability in ALL cells – yeah, that’s the cells without cancer too! Without Telomerase you would age Very rapidly, your DNA would fall off the ends of your Telomeres, the cells would stop working and die. There are still no magic bullets. You can’t turn it off in one cell, without turning it off in another. So, fundamentally, researching how to stop telomerase working won’t find a cure for cancer. They might be able to squeeze a drug out of it that manipulates telomerase, and make some money when someone spends an extra few weeks at deaths door, but that’s about it. Just because it’s a Professor arguing doesn’t mean the black crow is white. The skewed reasoning behind their research was exposed and compared to the drug manufacturing industries Cholesterol con. I expect they were playing the game; wanted some new shiny microscopes, so they drew an improbable link to a cancer treatment to get the drug companies interested in their research, and who knows maybe a few post-docs to do some of that tedious teaching.
Scientific America is a sensationalistic rag. Human Molecular Genetics, well OK if that’s your thing, I don’t read it, it’s not exactly Nature or Cell. Perhaps some other researchers are trying to aid telomerase functionality; looking for cancer prevention by trying to find out what is interfering with telomerase to stop it repairing the DNA correctly in the first place (it’s DNA damage that results in cancer) - but none of this is relevant to this thread, unless they are using CUDA or at least some sort of processor modelling, and you never mentioned that they are. There was no need for you to draw the conversation away from the theme. So again, and in a less subtle attempt to get back to the subject, my point was that GPU’s don’t have a Telomerase, they don’t repair themselves, so their life expectancy is more limited!

Is anyone working on a processor (CPU or GPU) that can perform a self diagnostic test and do an instruction set work around, like a Bios patch? I’m sure space agencies and aircraft manufacturers would be very interested.
Anyone want to pick up on that (and not Cell Biology)?

Is anyone working on a processor (CPU or GPU) that can perform a self diagnostic test and do an instruction set work around, like a Bios patch? I’m sure space agencies and aircraft manufacturers would be very interested.
Anyone want to pick up on that (and not Cell Biology)?

There are transient and permanent errors. Transient ones are errors which happen due to whatever reason (e.g. ionizing radiation) and disappear shortly afterwards. As I stated before I believe Intels designs are radiation hardened to the point where they can detect 2 bit transient errors and correct 1 bit errors, in both, the core and the cache. For regular use this is quite good already.

The permanent errors are more challenging. If they appear in the cache you can blend the affected cache line out. The fat CPUs (IBM Power, Itanium, Sparc) can certainly do this, whereas for desktop chips I think it's a one time action - before the chip leaves the factory.

Permanent errors within the logic parts of the chip are currently unrepairable. One could think about disabling certain blocks after failures, but there's not much redundancy in cpus, so you can't take much away so that they still work. It's different for GPUs: disabling individual shader clusters should be possible by software / bios, maybe requiring little tweaks.

Another option is to use redundant hardware from the beginning on. This is fine for safety-critical markets (space, military, airplanes, cars etc.), but wouldn't work in the consumer sector. Who'd buy a dual core for the price of a quad, just so he can still have 2 working cores even if 2 of them fail? We'd want to go 4-3-2 instead.

An interesting option are FPGAs, reconfigurable logic. With this stuff you could build chips which can adapt to the situation and which could repair themselves. The problem is that you need 10 times the transistors and you can only run the design at about 1/10th the frequency. To put this into perspective: with 130 nm tech you could build a regular Athlon XP at 2 GHz. Or you could build a Pentium 1 at 200 MHz, something already available at the 350 nm node. It's a very interesting research area, but no option for the consumer market.

Otherwise.. IBM is researching such stuff, but I don't know how far they got by now. And you can be sure Intel's in the boat as well ;)

MrS

---

Scanning for our furry friends since Jan 2002

A question to you
I have a inprogress WU 19-KASHIF_HIVPR_dim_ba4-28-100-RND7953_0 running but not in my tasks. How do i get rid of it ?
It is putting strain on my 295 to the extent that I have had to go back to 1 WU at a time. tempertures have going over 65C
Thanks
Ross

---

Hi Ross

actually the WU you mentioned is in your list, but way down.

In your task list select the "Show: in progress" instead of "Show: all", just above the top row, and you will see it.

kind regards

Alain

Edit - and BTW 65 C is by now means (too) hot, but well within limits. My GTX260 runs at 77 C, which is still cool.

Hi
that WU is not showing up in my tasks list . it and other were the cause of all the problems early this week . has July 30th expire date on it.
normaly I can crunch a WU in 7.5 hrs but now takes 9hrs.
As it does not show up in tasks how do I abort or kill it?
Thanks
Ross
my 295 freaks at over 73c smartdoctor alarm goes off new fan etc
using good airflow
Ross

---

OK, let us try to get this right now.
Looking again in more detail at that WU, it indeed is cancelled and crunching it further serves no purpose and yields no credit.
So, are you saying that it is still on your machine?
In that case, select the WU in your BOINC task list and abort it. Not fun, I know, but part of the game I am afraid.

Hope this solves your issue now.

Kind regards

Alain.