I've returned (also)

[Nicolas_orleans]

Hello everyone,

I used to be pretty active here in 2011-2016 (I think), then I had to buy a home, two kids joined the family etc. so my equipment got decommissioned and I stopped distributed computing. I was pretty involved in anything biomedical, from WCG/MCM to FAH, GPUGrid.net, and also Einstein@home since I liked their pipeline of scientific publications.

I learned a few weeks ago one of my kids had a pediatric cancer, so on top of supporting him in anyway I can, I thought it would be a good idea to support biomedical research again, especially when I read things like this.

So I am back with below priorities:
1/ Anything operational (search for markers on actual biomedical data etc.) that is cancer or Alzheimer related, like WCG/MCM or Smash Childhood Cancer, if new work is available.
2/ Anything more fundamental that is cancer or Alzheimer related, FAH & GPUGrid.net
3/ Einstein@home

My laptop #2 is still waiting for any work from WCG, but I still see 0 tasks available. I have used its GTX 1650 to do some GPUGrid.net but the current app yields a 50% failure rate, including on some tasks that have run for a few hours, so I have not pursued after a few days. The GPU was at 45-50°C (it's a gaming laptop) so I don't think it is hardware related. My Nicolas_orleans account was deleted by WCG due to inactivity and cannot be recovered so I have created a similar alias, not credited for the moment.

My folding/crunching rig #1 (that became my wife's computer with a CoolMasters case, and some spare parts from my old folding/crunching rig) was upgraded from my old GTX 980 Ti (accepted by FAH, but not current GPUGrid apps) to an RTX 4080 Super that folds nicely and is warming my office. I may need to add a triple fan on the case because the GTX 980 Ti was an Hybrid model from EVGA running at max 52-54°C, where I am at 67-71°C with the aircooled 4080 Super prior to messing with fan speeds.

I still have a damaged Asus Z9PE-D8 WS that used to run with two spicy Xeon E5-2660 @ 2.3 GHz and a spare PSU so I could probably run it in 1P (only the connector to PSU cable for CPU 2 burned) to feed the 980Ti again for FAH or Einstein@home. Will need to run some wattage simulations to see if I can bear the cost or not.

That's all for today, thanks for reading and keep in touch.

Nicolas

 

[pututu]

Sorry to hear about your kid. I hope with the advancement of AI and the exponentially increasing computational power will help to find cures for various cancers.

Anyway, welcome back to the DC world!

 

[Pocatello]

Welcome back Nicolas_orleans !

 

[skylaar24]

Welcome back! Sorry to hear about the recent news of your kid. That is tough to take but great you are searching for a cure.

While there is a Prime Grid challenge coming up in a couple days that is not biomedical related we do have the Boinc Games marathon running all year with plenty of biomed projects qualifying. That site takes an extra step to get your CPID verified but instructions are in the forums and it is fairly easy.

 

[Nicolas_orleans]

Some news.

I have finally updated the laptop from Ubuntu 20.04 / 8.0.2 BOINC client to Ubuntu 22.04. MCM tasks kicked in immediately after the update... I have been waiting for them for 2-3 weeks, but with WCG server outage and profile propagation issues documented in the WCG forum was (wrongly) thinking there was an issue going on on WCG's server side. I was wrong, it appears that, like for F@H Core 24, BOINC has a dependency to something that is in Ubuntu 22.04 and not in Ubuntu 20.04 with HWE enabled. Glad to map cancer markers again with the 4800H...

In order to keep the 980 Ti folding / crunching I launched a cheap AM4 build... and made a mistake : I purchased an MSI A520M-A PRO that was a perfect match with AM4 socket supporting Ryzen 5900X and PCIE 3.0 16x for my GTX 980 Ti... but it does not have an off / on button (sigh) so it needs a case connected to the jumpers to work... So the missing case is on the way, my plan to run a cheap AM4 build naked are dead.

Regarding the Z9PE-D8 WS, I tried 1 P mode but it does not even POST, it appears either there is more damage than the one I can see, or that the 2P connector should be connected no matter if the board is running 1P or 2P... dead for the moment... I had a look at used / new mobos but since I have spicy chips, I need to control BIOS versions so they are supported. Not sure I will find a solution to fold/crunch again on the Xeons...

 

[Shoganai]

Welcome back. I'm really sorry to hear about your kid. There's a lot of crazy research going on right now. My dad was diagnosed with lung and brain cancer a couple of years ago, but he qualified for a new immunotherapy treatment that won a Nobel Prize in 2019. Instead of traditional therapy, the treatment makes the body's own immune system think the cancer is like a virus, and it fights it off itself. 6 months into the treatment, his cancer vanished, and two years later, he is still cancer free. I don't know anything about the research into childhood cancer, but after seeing the kind of things they're coming up with these days, I remain hopeful. I'm sorry you have to deal with that.

 

[Gilthanis]

Great to see you back. DC'ing is in a terrible world of hurt right now for good solid projects to support. Kremble is doing a terrible job with WCG. They should just scrap the whole thing and rebuild new with the latest BOINC server software. They do NOT have the support nor the infrastructure to keep pounding away at IBM's Frankensteining of everything.

GPUGrid as you are finding is a headache. You will find cards like a P100 to be really good right now. And obviously Linux...

Einstein is still pretty solid choice but yeah...not medical.

FAH is still FAH. I cannot really say anything is really changing much there.

Rosetta@home it hit and miss. Not like the solid project it was years ago.

Pretty much EVERYTHING else is either too small or not the right science or both. The only solid projects for competitions IMHO right now are FAH, Einstein, and PrimeGrid.

 

[Nicolas_orleans]

Welcome back. I'm really sorry to hear about your kid. There's a lot of crazy research going on right now. My dad was diagnosed with lung and brain cancer a couple of years ago, but he qualified for a new immunotherapy treatment that won a Nobel Prize in 2019. Instead of traditional therapy, the treatment makes the body's own immune system think the cancer is like a virus, and it fights it off itself. 6 months into the treatment, his cancer vanished, and two years later, he is still cancer free. I don't know anything about the research into childhood cancer, but after seeing the kind of things they're coming up with these days, I remain hopeful. I'm sorry you have to deal with that.

Thanks Shoganai for sharing this inspiring story. I hope your dad will remain cancer free...
Ability to treat cancer remains very dependent on the type of cancer itself, and sometimes also on the presence of certain markers. In the case of my son there is a 4-steps protocol, including immunotherapy in stage 4. He is in the middle of step 1. It's terrible to see some kids in the same hospital actually have some other forms of child cancer for which no protocol even exists...
Fold/crunch on

 

[Nicolas_orleans]

Great to see you back. DC'ing is in a terrible world of hurt right now for good solid projects to support. Kremble is doing a terrible job with WCG. They should just scrap the whole thing and rebuild new with the latest BOINC server software. They do NOT have the support nor the infrastructure to keep pounding away at IBM's Frankensteining of everything.

GPUGrid as you are finding is a headache. You will find cards like a P100 to be really good right now. And obviously Linux...

Einstein is still pretty solid choice but yeah...not medical.

FAH is still FAH. I cannot really say anything is really changing much there.

Rosetta@home it hit and miss. Not like the solid project it was years ago.

Pretty much EVERYTHING else is either too small or not the right science or both. The only solid projects for competitions IMHO right now are FAH, Einstein, and PrimeGrid.

Thanks for your warm welcome and for this overview that makes total sense. WCG is a nonsense with constant http transient error, and sometimes, when I hit refresh on the laptop, I manage to get a few tasks finalizing the stuck download. But well it's on my daily laptop and I have a personal interest in MCM. But I would no set a dedicated cruncher for MCM until they rebuild the thing as you suggest.

On a more philosophical standpoint, I was not expecting to see AI making prediction of 3D protein structures... yet - scientists would probably need a combo of quantum computers and GPUs, because from what I understand quantum computing should eventually be superior for quantum / molecular stuff. But I was expecting to see some big names in the tech space having built too many data centers for AI and willing to donate computing time to respected DC projects, a little like when Nvidia corp was making some tests with FAH a decade ago ?

I had issues with BIOS and AM4 socket / cooling but rig #3 is finally up and running with my old GM200, 1.5 M PPD not great, should help to be above 20 M PPD, will see what I can afford later.

 

[Nicolas_orleans]

To share with you some news:
- after 8 cycles of chemotherapy, 4 of combined chemotherapy / immunotherapy, and two surgeries, step 1 will be finished soon and the disease should have decreased enough to enter step 2, that is the toughest one in the protocol. My son is still full of energy and joy, notably when he is home. There is more hope now compared to end 2024...
- I have been ramping up "a little" on the folding front, sometimes in times of uncertainty (good to have something you can control), sometimes to celebrate good news and support research for other kids. Thanks to firedfly and his magic vast.ai template, I can rent vast.ai RTX 4090 instances to do more when I want, without the heavy investing. Will still need to do something for my own riggery when the time is right.

 

[Red Falcon]

To share with you some news:
- after 8 cycles of chemotherapy, 4 of combined chemotherapy / immunotherapy, and two surgeries, step 1 will be finished soon and the disease should have decreased enough to enter step 2, that is the toughest one in the protocol. My son is still full of energy and joy, notably when he is home. There is more hope now compared to end 2024...
- I have been ramping up "a little" on the folding front, sometimes in times of uncertainty (good to have something to control), sometimes to celebrate good news and support research for other kids. Thanks to firedfly and his magic vast.ai template, I can rent vast.ai RTX 4090 instances to do more when I want, without the heavy investing. Will still need to do something for my own riggery when the time is right.

We are all rooting for you, your son, and your family.
With everything you have been going through, and yet are continuing to fight, you are an inspiration to us all.

This is why we do this.

 

[Holdolin]

To share with you some news:
- after 8 cycles of chemotherapy, 4 of combined chemotherapy / immunotherapy, and two surgeries, step 1 will be finished soon and the disease should have decreased enough to enter step 2, that is the toughest one in the protocol. My son is still full of energy and joy, notably when he is home. There is more hope now compared to end 2024...
- I have been ramping up "a little" on the folding front, sometimes in times of uncertainty (good to have something to control), sometimes to celebrate good news and support research for other kids. Thanks to firedfly and his magic vast.ai template, I can rent vast.ai RTX 4090 instances to do more when I want, without the heavy investing. Will still need to do something for my own riggery when the time is right.

Sending all my best to you and your family. Yes, in times like this it's great to work on something you CAN control.

 

[Gilthanis]

To share with you some news:
- after 8 cycles of chemotherapy, 4 of combined chemotherapy / immunotherapy, and two surgeries, step 1 will be finished soon and the disease should have decreased enough to enter step 2, that is the toughest one in the protocol. My son is still full of energy and joy, notably when he is home. There is more hope now compared to end 2024...
- I have been ramping up "a little" on the folding front, sometimes in times of uncertainty (good to have something to control), sometimes to celebrate good news and support research for other kids. Thanks to firedfly and his magic vast.ai template, I can rent vast.ai RTX 4090 instances to do more when I want, without the heavy investing. Will still need to do something for my own riggery when the time is right.

I love that you point out focusing on the things that we CAN control. Still praying for you guys and please keep us updated on the progress.

 

[pututu]

To share with you some news:
- after 8 cycles of chemotherapy, 4 of combined chemotherapy / immunotherapy, and two surgeries, step 1 will be finished soon and the disease should have decreased enough to enter step 2, that is the toughest one in the protocol. My son is still full of energy and joy, notably when he is home. There is more hope now compared to end 2024...
- I have been ramping up "a little" on the folding front, sometimes in times of uncertainty (good to have something you can control), sometimes to celebrate good news and support research for other kids. Thanks to firedfly and his magic vast.ai template, I can rent vast.ai RTX 4090 instances to do more when I want, without the heavy investing. Will still need to do something for my own riggery when the time is right.

Wishing and praying all the best to your son. Take care.

 

[jfb9301]

In order to keep the 980 Ti folding / crunching I launched a cheap AM4 build... and made a mistake : I purchased an MSI A520M-A PRO that was a perfect match with AM4 socket supporting Ryzen 5900X and PCIE 3.0 16x for my GTX 980 Ti... but it does not have an off / on button (sigh) so it needs a case connected to the jumpers to work... So the missing case is on the way, my plan to run a cheap AM4 build naked are dead.

Back in the day of Pizza Boxen, if I didn't have a power button, I would just short the power switch pins with a screwdriver. I did finally wire up a couple of switches to jumpers so I had essentially a loose power button that I cold connect to any mainboard installed in a random cardboard case. I couldn't run naked, my cat might stick her nose somewhere she shouldn't.

 

[jfb9301]

Sorry to hear about your kid.

The big C took my dad in '19, and when I was back home (on the wrong side of the country) my wife was diagnosed with a tumor. The big C took her in '21.

 

[Red Falcon]

In order to keep the 980 Ti folding / crunching I launched a cheap AM4 build... and made a mistake : I purchased an MSI A520M-A PRO that was a perfect match with AM4 socket supporting Ryzen 5900X and PCIE 3.0 16x for my GTX 980 Ti... but it does not have an off / on button (sigh) so it needs a case connected to the jumpers to work... So the missing case is on the way, my plan to run a cheap AM4 build naked are dead.

I know it has been a minute since you posted this, but would something like this work to allow it to be standalone?

 

[Gilthanis]

I know it has been a minute since you posted this, but would something like this work to allow it to be standalone?
View attachment 721504

I just used to pull them out of old cheapo cases that I would scrap over the years. lol

 

[Nicolas_orleans]

Sorry to hear about your kid.

The big C took my dad in '19, and when I was back home (on the wrong side of the country) my wife was diagnosed with a tumor. The big C took her in '21.

I am very sorry to hear that. I hope you had and have people and family around you to support...

 

[Nicolas_orleans]

I know it has been a minute since you posted this, but would something like this work to allow it to be standalone?
View attachment 721504

I actually sent back the board to the retailer and got a slightly higher end one, that has been working perfectly since. Thanks for sharing anyway!

 

[Red Falcon]

Sorry to hear about your kid.

The big C took my dad in '19, and when I was back home (on the wrong side of the country) my wife was diagnosed with a tumor. The big C took her in '21.

Damn, I'm sorry that happened to you and your family, I hope things get better for you.

 

[Nicolas_orleans]

Today I would like to share my Blackwell adventures on Linux, which were incredibly painful compared to Kepler, Maxwell & Ada.

My setup of the last weeks was: 1x980 Ti on (legacy) rig 1, 1x 4080S on (gaming) rig 2, and 2x4090 on vast.ai. I wanted to:
- internalize a part of my production since vast.ai is nice, flexible, without hardware issues or heat to manage, but also comes with a premium in terms of PPD per $ spend
- do it on rig 1 without other hardware purchases. For the record, the 4080S on rig 2 takes 3 slots so not an option to add a second card there... also putting the 5080 on rig 2 sounded risky since it's a greatly working Proton gaming setup
- keep folding on the 980Ti because it's doing smaller work units bigger cards don't, and I want to support all research, not only the one associated to higher cards species

So I did a shortlist of some 5080 cards taking 2 slots or 2,5 slots (SFF) to fit the board with the 980 Ti and waited for a reasonable pricing, which happened this week. I happily purchased an Inno3D x3 5080 model and by the way, compared to the overpricing we have been having in Europe, all my vast.ai spending since February has been covered by the difference between the inflated launch prices and current price that is much closer to founder's edition MSRP and below this specific model MSRP (1269 €, got it at 1208€).

Then the pain starts. The rig was already on driver 570.124.04 which was Blackwell-compatible, so I was hoping it would be the usual linux plug and play without driver reinstall. It was not. My Ubuntu install was completely broken, either in dual card or single card setup. I purged & reinstalled several time the driver but nothing good was coming out of it. So I installed Plucky (Ubuntu 25.04) that ships with 570.133.07, without the need for a PPA. Also, I abandoned for the moment the idea of the dual GPU setup given what I saw in my driver install attempts, not at all the plug and play of my triple GPU Kepler setup that used to be on this board.

On Plucky the card is recognized with minimal driver support in nvidia-settings but, after install, gets the infamous black screen bug at boot that has been widely reported. The workaround I found was to use the Ivy Bridge iGPU of my CPU for the operating system and to use the 5080 as a computing card. Plucky performs incredibly well on the iGPU by the way, as long as you don't want a super high screen resolution.

Now I thought I was ready to fold but... the fah .deb installer has unmet dependencies on 25.04. I had to mess a little with the tarball file, it's not installed as a service, but... IT NOW FOLDS !!!

Regarding the performance, on such an old CPU, with an SATA III HDD (no SSD), it's not that bad, even though an hardware update will be required at some point. Currently pulling 23,5 M PPD when the 4080S does 21 M PPD with a much more recent CPU. This setup is even quite educational to see the difference of behaviours in projects, for example P18227 takes 5 minutes to initialize, starts at 5 M PPD after 5%, and finishes around 22,3 M PPD. Good reminder FAH initialization is single threaded when checkpoints are not.

Did any of you have a better linux experience with Blackwell and if yes, which distro / driver ?

 

[Red Falcon]

Nope, Blackwell drivers are completely broken on Debian distros as far as I can tell.
Ironically enough, I am using a 5070 Ti on an Ivy Bridge platform, and gave up the ghost and decided to run it with Windows 10 until the Linux drivers mature to a stable state.

I'm sure others like yourself have had success with it (good job!), but I ran out of time and needed to get it going.
Ivy Bridge quad-cores still do an impressively good job considering their advanced age, and the initialization and checkpoints don't take overly long, depending on the project.

 

[Nicolas_orleans]

To share with you some news:
- the two high dose chemotherapies end of spring and during summer were very hard, but it could have been much worst in terms of side effects when you compare to other kids who undergone the same treatments. Some "luck" for a very unlucky kid in a way. After that, my son got "good" results with no measurable tumoral activity since August. He started a new treatment phase (immunotherapy) aiming to prevent relapses. Three months after, still no measurable tumoral activity which is good. He's a fighter and is now allowed to go to school again which has a been a major victory for all of us.
- I replaced the old Ivy bridge with a 9950X that is performing very well. My 5080 presents signs of hardware malfunction that I cannot solve, I hope it can be RMAed since the card is only 6-7 months old. If not I will try to see if at least it is Einstein-stable to do some good with it. If it has become useless for DC I will eventually replace it when I can afford it. I don't want F@H to suck all my "philanthropic/cancer-related money" since I participate in the funding of some groundbreaking research on two child cancer for which no combination of treatment works for all kids, including the very cancer my son has (they search cures for relapses).

 

[Red Falcon]

To share with you some news:
- the two high dose chemotherapies end of spring and during summer were very hard, but it could have been much worst in terms of side effects when you compare to other kids who undergone the same treatments. Some "luck" for a very unlucky kid in a way. After that, my son got "good" results with no measurable tumoral activity since August. He started a new treatment phase (immunotherapy) aiming to prevent relapses. Three months after, still no measurable tumoral activity which is good. He's a fighter and is now allowed to go to school again which has a been a major victory for all of us.

Congratulations to your son and your family, that is major progress!

- I replaced the old Ivy bridge with a 9950X that is performing very well. My 5080 presents signs of hardware malfunction that I cannot solve, I hope it can be RMAed since the card is only 6-7 months old. If not I will try to see if at least it is Einstein-stable to do some good with it. If it has become useless for DC I will eventually replace it when I can afford it. I don't want F@H to suck all my "philanthropic/cancer-related money" since I participate in the funding of some groundbreaking research on two child cancer for which no combination of treatment works for all kids, including the very cancer my son has (they search cures for relapses).

Check the 12VHPWR power cable on your 5080.
The 5070 Ti I was folding with was becoming more and more unstable and would just go to a black screen and reboot/lock the system.

Tried 3 motherboards, 3 PSUs, and 3 different OSes, same behavior across the board.
I was about to RMA it and looked at the 12VHPWR adapter that came with it, and the positive connectors were all darkened and/or scorched.

The 5070 Ti itself only used about 150 watts with F@H, so it was probably a faulty cable.
The GPU itself was totally fine with a new adapter and has been good since, so it might be worth checking the 12VHPWR.

Also, the 50 series works good now with the 580 drivers on Debian distros since my last post.

 

[Nicolas_orleans]

After investigating, it was a PSU issue. Both my old PSUs are AX860 Platinum. They are stable with 5080 stock and 9950X stock all cores @4,9 GHz. They are not with 5080 stock and 9950X PBO auto all cores @5,3 GHz.

For reference:
- GPU power starvation did not produce system crash, only FAH errors like "Error invoking kernel: CUDA_ERROR_LAUCH_FAILED (719)"
- with latest agesa PBO auto is superior to any PBO Enhanced tutorial I have seen: far less heat, super stable frequency, even vs CO -30 per CCD
- SMP 9950X PBO auto all cores @5,3 GHz alone (without GPU) is insane on Linux and can punctually reward with > 3 M PPD on certain WUs. > 1,9 M PPD average on 30 cores (not tested 31)