Someone from DPC over here: we've notified the guy running the Nifhack account of this thread and asked if he wants, and is able to, clarify this. He's know for having access to huge amounts of computational power (at work, I believe) but can't deploy all of it all the time. He's also known to rarely part with specifics. My guess is as well those machines are indeed some sort of hosts to the computers behind.

Something similar to using Amazon cloud to fire up Rosetta instances on a grand scale, but routing it through a single BONIC client? Except instead of renting time on Amazons system, he has the ability to do this all on a private cloud?

My question about credits is, what is up with this guy? Within 3 days, he has the top three "fastest" computers by nearly a factor of 6.

They are returning a lot of Tasks for such a small number of core/threads.
0.72 day turn around. 8 hour runtime. 4,600 Tasks in progress on one system, over 6000 Valid.
0.72 day turn around. 8 hour runtime. 1,300 Tasks in progress on the others, roughly 1,650 each Valid on the others.

Number of times client has contacted the server, 3 for one system. 0 for the others?

Some sort of CPU compute cluster feeding it's results through those host IDs?

Is the work really done? If it is, great. But the "credit" only goes through one host? If the work done is real, who cares?
Neat trick. If he's got authority to do it, great. If not, it's nothing to do with me...

---

Someone from DPC over here: we've notified the guy running the Nifhack account of this thread and asked if he wants, and is able to, clarify this. He's know for having access to huge amounts of computational power (at work, I believe) but can't deploy all of it all the time. He's also known to rarely part with specifics. My guess is as well those machines are indeed some sort of hosts to the computers behind.

Thanks for the insight! I'm just getting into edge computing and would love to know some of the details to play around with a similar setup.

Yup, nothing wrong in what he is doing, it all seems good, valid, crunching. It's just funny seeing a single host with that RAC

---

Our DPC-member is working for Nikhef - our National Institute for Subatomic Physics.
It's not the first time Nikhef is testing new toys on Rosetta :)

The Dutch Power Cows have their yearly "stampede" on Rosetta this year. The stampede will end on april 30 so expect a slow down in our production.

If you want to know a little bit more about the computing power of Nifhack parse this https://www.nikhef.nl/news/nieuwe-nikhef-rekenclusters-gaan-eerst-aan-het-coronavirus-rekenen/ through DeepL or some other translator. Original article is in Dutch.

In short:

64x Lenovo SR655 systems, most of them with AMD EPYC 7702P and 512GB ram.
It is housed in 4x 19" cabinets.
Each system has a Mellanox Connect LX4 that connects with 2x 25Gbit/s to the network.
Subset of the systems are equipped with GPUs.

Kuuke, moderator DPC forum

---

too late ;

tks Kuuke

The WU time initial is 8 hours. While processing the WU time rise up to 48 hours or more.
(Sorry my english).

Greetings

You seem to have specified a runtime preference of the highest value allowed (36 hours). It will take BOINC Manager a few days to get used to tasks taking 36 hours to complete. The estimates are not very accurate when you first start out.

---

Rosetta Moderator: Mod.Sense

I think if you deploy a vast amount of nodes from fixed OS images that have the BOINC folder hard coded then a clashing host ID could result something like this.

Yup, nothing wrong in what he is doing, it all seems good, valid, crunching. It's just funny seeing a single host with that RAC

I'm sure I've seen a news report in the past where the user didn't have the authority to take over all the machines
Now that was funny (to read)

---

WTH is this?! Took my overclocked Ryzen 3900x over SIX hours to get SIX........SIX credits?! Is this a credit error, or the norm? What a waste of electricity..? SMDH...https://photos.app.goo.gl/rGjmPdzUNLErS5CeA



Did I get a batch of bad WU's?! Can someone please look into the following?

TASK: 1165850138 WORK UNIT: 1045787673 SENT: 30 Apr 2020, 22:27:22 UTC REPORTED: 3 May 2020, 19:02:13 UTC Completed and validated TOTAL TIME: 23,504.78 CPU TIME: 22,564.69 CREDIT= 7.06 Rosetta v4.15
windows_x86_64
1165798670 1048233647 30 Apr 2020, 21:57:15 UTC 3 May 2020, 19:02:51 UTC Completed and validated 24,268.53 23,202.55 CREDIT= 6.46 Rosetta v4.15
windows_x86_64
1165774723 1048213014 30 Apr 2020, 21:23:53 UTC 3 May 2020, 18:07:52 UTC Completed and validated 22,475.43 21,490.02 CREDIT= 7.65 Rosetta v4.15
windows_x86_64
1165751409 1048192909 30 Apr 2020, 20:50:56 UTC 3 May 2020, 16:42:29 UTC Completed and validated 22,156.37 21,163.57 CREDIT= 6.22 Rosetta v4.15
windows_x86_64
1165779858 1048149788 30 Apr 2020, 20:42:40 UTC 3 May 2020, 16:41:20 UTC Completed and validated 23,246.45 22,230.59 CREDIT= 6.32 Rosetta v4.15
windows_x86_64
1165742379 1048185110 30 Apr 2020, 20:37:25 UTC 3 May 2020, 22:31:38 UTC Completed and validated 31,895.94 30,596.40 CREDIT= 45.84 Rosetta v4.15
windows_x86_64
1165733818 1048177553 30 Apr 2020, 20:25:06 UTC 3 May 2020, 22:32:28 UTC Completed and validated 34,077.06 32,814.71 CREDIT= 39.39 Rosetta v4.15
windows_x86_64

Those were all crunched by computer 4246752. Measured floating point speed: 4958.25 million ops/sec
Measured integer speed: 19723.67 million ops/sec

WTH is this?! Took my overclocked Ryzen 3900x over SIX hours to get SIX........SIX credits?! Is this a credit error, or the norm? What a waste of electricity..? SMDH...

With your computers hidden it's difficult to help, but maybe if you post this in the "Is the amount of credits I'm getting normal?" thread and make your systems visible it would be a start.

---

Grant
Darwin NT

Thank you. I will post there.

I took a look and it appears your host was successfully returning results but then became unstable. Might there be an issue with the host?

---

There was, indeed, a hiccup today! The CPU over-temp'd, and I had to dial it back... I guess 4.2GHz on 12 cores/24 threads @ 1.35v was too much. Does it look ok now? She's been running a lot cooler since the "incident". Lol. Came down from 96°C to ~72°C. Thanks for looking into my conundrum.... You guys and gals are awesome! You can delete my original post, if you want...

Now, back to the topic at hand... Bring on the monster WU's!

if the BOINC Manager sees requests for memory that exceed that 4GB the task is actually aborted

If aborted tasks still return results, it's possible to assign credits for computed models.

---

if the BOINC Manager sees requests for memory that exceed that 4GB the task is actually aborted

If aborted tasks still return results, it's possible to assign credits for computed models.

I believe it does now - according to the CPU time the task reports back

---

We need to define terms carefully. With "Aborted" work units, the BOINC Manager does not return the results files. If a work unit ends abnormally, or it is ended by the watchdog, then results are returned and credit is granted based on the number of completed models.

---

Rosetta Moderator: Mod.Sense

I can't wait to get to 1,000,000 points so I can get my reward.

Well I made my million points and made me some pretty good tacos.

Would like to see GPU work units in this project is where my desktop has its most efficient throughput.