The recent posts are all rather peculiar to me.
I thought I must have missed these tasks going through, but there are none at all in my recent history.
I had a couple of errors, but I caused them myself after crashing one PC

There's not many faulty tasks, just a few people getting upset over nothing. I think we must have some journalists in here, that's their way of thinking. One little thing goes wrong and they think the world is about to end.

The biggest threat to my various systems and task completion is always me.
Having blown a motherboard in late March on my home machine I managed to corrupt Windows irreparably badly a few weeks ago during the heatwave we had and lost 6 days of running before it got repaired. Though it's running better since as I re-establish my overclock slowly over the last week. Possibly running faster and cooler now.
Overlapping that period, I decided to update a very old BIOS and all the other drivers on my 2nd PC to resolve some instabilities and again taken a few weeks to re-establish the overclock. Not running quite as fast but again much cooler.
And on a laptop I had some software hogging a few cores so 28k second tasks were taking between 60-90k seconds to complete, now solved.

I've probably dropped 50-100k credits over the last 2 weeks through local issues.
Slightly different to the occasional task crashing out after 2 or 3 minutes, which has never bothered me in the slightest.
By its nature, Rosetta is a kind of experimental project, looking for things that have never existed before, or trying to replicate things that do exist but in a way that's never been found before. So if there are some coding slip-ups or investigations of space that's out of bounds, so be it.

I also note that the queue of submitted tasks is 10-20x bigger than the number issued each day, so there's going to be a long lag between issues arising and working through tasks already submitted. No doubt they'll get aborted at some point according to the priority of the researcher. The occasional lost 2-3mins is a non-issue for users

---

Which is unlikely if the percentage is low! Let me explain this in simpler terms....

Since you're missing the point entirely (deliberately or otherwise), there's no point discussing it further.

I'm simply pointing out to you that your point is incorrect. If a bug wastes 50GB of data, but the whole project is using 50,000GB of data, then it's not worth worrying about. It's a 0.1% wastage. But if a bug wastes 50GB of data on a smaller project using 100GB of data, that's a 50% wastage and is doubling their costs. If I buy a car for 50 grand and there's a $50 charge to fill the tank with gas before I buy it, I'm not going to complain. $50 may be a lot, but it's nothing compared to the purchase of the car!

---

The biggest threat to my various systems and task completion is always me.
Having blown a motherboard in late March on my home machine I managed to corrupt Windows irreparably badly a few weeks ago during the heatwave we had and lost 6 days of running before it got repaired. Though it's running better since as I re-establish my overclock slowly over the last week. Possibly running faster and cooler now.

I used to overclock. Never again. You gain 15% speed, and wear the thing out twice as fast. Not worth it. They were designed a certain way, and presumably you don't know better than the manufacturer.... Same applies to cars. I bought a Honda CRV and converted it (professionally) to run on propane. Bad idea. That engine was designed by Honda to run on petrol. Changing it was a bad idea and melted the engine.

I've probably dropped 50-100k credits over the last 2 weeks through local issues.
Slightly different to the occasional task crashing out after 2 or 3 minutes, which has never bothered me in the slightest.

By its nature, Rosetta is a kind of experimental project, looking for things that have never existed before, or trying to replicate things that do exist but in a way that's never been found before. So if there are some coding slip-ups or investigations of space that's out of bounds, so be it.

I also note that the queue of submitted tasks is 10-20x bigger than the number issued each day, so there's going to be a long lag between issues arising and working through tasks already submitted. No doubt they'll get aborted at some point according to the priority of the researcher. The occasional lost 2-3mins is a non-issue for users

Indeed, as long as things are running 99% of the time, why worry?

---

The biggest threat to my various systems and task completion is always me.

I used to overclock. Never again. You gain 15% speed, and wear the thing out twice as fast. Not worth it.

15% (actually one at 11% and another at 18,5%) 24/7 x365 days x all the cores I actually do find worth it as my PCs are actually pretty old and still going strong after a surprising amount of time (Trigger's broom accounted for).
By the time they pack in I'll be long overdue an upgrade - everyone will make their own judgements on that. I'm slightly itching for one tbh

Edit - forgot my point. Manufacturers are encouraging overclockers nowadays by unlocking cores and providing advice on how to boost machines. It's a very different environment nowadays and I'm finally comfortable joining in

---

The biggest threat to my various systems and task completion is always me.

I used to overclock. Never again. You gain 15% speed, and wear the thing out twice as fast. Not worth it.

15% (actually one at 11% and another at 18,5%) 24/7 x365 days x all the cores I actually do find worth it as my PCs are actually pretty old and still going strong after a surprising amount of time (Trigger's broom accounted for).
By the time they pack in I'll be long overdue an upgrade - everyone will make their own judgements on that. I'm slightly itching for one tbh

Edit - forgot my point. Manufacturers are encouraging overclockers nowadays by unlocking cores and providing advice on how to boost machines. It's a very different environment nowadays and I'm finally comfortable joining in

If the chip was supposed to go that fast, it would already be set at that speed so they could sell it for more. Do you also put nitros in your car then wonder why the engine fails?

---

The biggest threat to my various systems and task completion is always me.

I used to overclock. Never again. You gain 15% speed, and wear the thing out twice as fast. Not worth it.

Manufacturers are encouraging overclockers nowadays by unlocking cores and providing advice on how to boost machines. It's a very different environment nowadays and I'm finally comfortable joining in

If the chip was supposed to go that fast, it would already be set at that speed so they could sell it for more. Do you also put nitros in your car then wonder why the engine fails?

It's very simple.
There's all the difference in the world between selling a product at a certain spec that's guaranteed to last a certain lifetime <at the manufacturer's liability> and selling essentially the same product in an unlocked form that the customer can overclock at their own risk, especially given the room for maneuver manufacturers allow themselves.
Both my overclocked CPUs have long outlasted their warranty already while running 24/7 without even a reboot for entire months at a time.
Obviously sometimes I get crashes, but it's usually my own fault and is often solved with a quick clean and a slight easing back on the overclock at the height of summer.
The amount of time my CPUs have run at their standard clock speed over the last decade must be less than a week combined.

---

There's all the difference in the world between selling a product at a certain spec that's guaranteed to last a certain lifetime <at the manufacturer's liability> and selling essentially the same product in an unlocked form that the customer can overclock at their own risk, especially given the room for maneuver manufacturers allow themselves.
Both my overclocked CPUs have long outlasted their warranty already while running 24/7 without even a reboot for entire months at a time.
Obviously sometimes I get crashes, but it's usually my own fault and is often solved with a quick clean and a slight easing back on the overclock at the height of summer.
The amount of time my CPUs have run at their standard clock speed over the last decade must be less than a week combined.

So if you go too far and break it, how does the manufacturer know you overclocked it? Does it log the clock speeds internally?

I gave up overclocking because with GPUs they break. With CPUs you get corrupt disks when they crash during a disk write.

I'll stick to the "guaranteed to last a certain lifetime"....

---

If the chip was supposed to go that fast, it would already be set at that speed so they could sell it for more. Do you also put nitros in your car then wonder why the engine fails?

Not true. The quoted speed is the lowest speed that chips from that die can be guaranteed to run - 90+ % of the chips will run faster than that and some will run very much faster than that but it cannot be guaranteed which it would have to be if the manufacturer increased the claimed speed.

If the chip was supposed to go that fast, it would already be set at that speed so they could sell it for more. Do you also put nitros in your car then wonder why the engine fails?

Not true. The quoted speed is the lowest speed that chips from that die can be guaranteed to run - 90+ % of the chips will run faster than that and some will run very much faster than that but it cannot be guaranteed which it would have to be if the manufacturer increased the claimed speed.

Well I never got anything significant out of overclocking, and almost always ended up with broken chips or corrupt disks.

I wonder if it would be economical for the manufacturer to test each one individually and rate them accordingly, selling at a higher price?

---

There's all the difference in the world between selling a product at a certain spec that's guaranteed to last a certain lifetime <at the manufacturer's liability> and selling essentially the same product in an unlocked form that the customer can overclock at their own risk, especially given the room for maneuver manufacturers allow themselves.
Both my overclocked CPUs have long outlasted their warranty already while running 24/7 without even a reboot for entire months at a time.
Obviously sometimes I get crashes, but it's usually my own fault and is often solved with a quick clean and a slight easing back on the overclock at the height of summer.
The amount of time my CPUs have run at their standard clock speed over the last decade must be less than a week combined.

So if you go too far and break it, how does the manufacturer know you overclocked it? Does it log the clock speeds internally?

Fair question. I don't know, but I don't think so. If I force it too far there seems to be some kind of failsafe that switches individual cores off or the PC freezes or shows a BSOD and I can go again.
As a generalisation, when I push the CPUs too far for too long it's something else that's eventually broken that's actually running within spec - motherboard and hard drive most recently.

I gave up overclocking because with GPUs they break. With CPUs you get corrupt disks when they crash during a disk write.

I'll stick to the "guaranteed to last a certain lifetime"....

Fair enough - it's hardly compulsory and everyone can go as far as they're comfortable with.
For me, I haven't started a fire yet. Melted things, yes, fire no...

---

I wonder if it would be economical for the manufacturer to test each one individually and rate them accordingly, selling at a higher price?

They do, it's called Chip Binning.

---

Grant
Darwin NT

The recent posts are all rather peculiar to me.
I thought I must have missed these tasks going through, but there are none at all in my recent history.

And today is the first day i don't have any in my current Task list either, after having them continuously for over a week and a half.
Even the number of errored/Invalid Tasks for the top systems has dropped off a lot over the last few hours.

To get back into the subject of the thread (Rosetta 4.2+), I rec'd 5 more foldit1 validation errors for one of my Android devices. However, these were all resent tasks for not starting by deadline, first sent 8/15. This is forgivable, as would be difficult for system to screen for problematic tasks in this case. These had all originally been sent to Windows or Linux systems. It appears that these are slowly getting cleaned out of the system thankfully.

---

I wonder if it would be economical for the manufacturer to test each one individually and rate them accordingly, selling at a higher price?

They do, it's called Chip Binning.

There are even companies that bin the manufacturer binned chips and sell the highest performing cpus for even higher cost.
https://siliconlottery.com/

Fair enough - it's hardly compulsory and everyone can go as far as they're comfortable with.
For me, I haven't started a fire yet. Melted things, yes, fire no...

You should see what happened when I tried water cooling....

---

I wonder if it would be economical for the manufacturer to test each one individually and rate them accordingly, selling at a higher price?

They do, it's called Chip Binning.

Then why isn't this always done? Why sell a 3.5GHz chip and say "you might be able to overclock this to 4GHz", instead of testing it then selling it for a guaranteed speed for more money?

---

I wonder if it would be economical for the manufacturer to test each one individually and rate them accordingly, selling at a higher price?

They do, it's called Chip Binning.

Then why isn't this always done? Why sell a 3.5GHz chip and say "you might be able to overclock this to 4GHz", instead of testing it then selling it for a guaranteed speed for more money?

They're quantising a continuous spectrum, there will always be room between the values they guarantee for the enthusiasts to eek out an advantage.

It's not something I'd do myself - in fact I'm seriously considering underclocking my 3600 to reduce the power consumption if I can do it without compromising performance too much. It's running at 1.448v which sound high from what I've heard.

They're quantising a continuous spectrum, there will always be room between the values they guarantee for the enthusiasts to eek out an advantage.

It's not something I'd do myself - in fact I'm seriously considering underclocking my 3600 to reduce the power consumption if I can do it without compromising performance too much.

I once underclocked a previously overheated AMD CPU which kept crashing. It made it work!

It's running at 1.448v which sound high from what I've heard.

Is that an old chip? I thought everything was around a volt nowadays.

And AFAIK, clock speed up = crash, but voltage up = death.

---

It's not something I'd do myself - in fact I'm seriously considering underclocking my 3600 to reduce the power consumption if I can do it without compromising performance too much.
It's running at 1.448v which sound high from what I've heard.

A quick google shows people oc'ing at under 1.2v - have a look round for tips

---

Why sell a 3.5GHz chip and say "you might be able to overclock this to 4GHz", instead of testing it then selling it for a guaranteed speed for more money?

They do both- those that are guaranteed to hit a minimum speed above the basic specifications, people pay an arm and a leg for (well, those with more money than sense anyway).
But those chips that can reach the highest of overclocked speeds require all sorts of tricks & tweaks to actually achieve that. The time and effort needed to find those & label them accordingly during manufacturing would cost way more than even the craziest of over clockers would be prepared to spend.

---

Grant
Darwin NT

Why sell a 3.5GHz chip and say "you might be able to overclock this to 4GHz", instead of testing it then selling it for a guaranteed speed for more money?

They do both- those that are guaranteed to hit a minimum speed above the basic specifications, people pay an arm and a leg for (well, those with more money than sense anyway).
But those chips that can reach the highest of overclocked speeds require all sorts of tricks & tweaks to actually achieve that. The time and effort needed to find those & label them accordingly during manufacturing would cost way more than even the craziest of over clockers would be prepared to spend.

The coolest computer I ever saw was oil cooled. Someone had immersed the entire machine in (I think cooking) oil. Non-electrically conductive, but very thermally conductive. No fans or pumps required. Not sure why it isn't done more often. Just take all the fans off, immerse the case in a large container of oil, and away you go!

---