Folding Community Board

[Bitter]

On 3/8/2022 at 9:16 PM, LazyDev said:

Well, my computer has developed an intermittent fault that I cannot identify.

 

The system would crash to a black screen. The CPU Ez led would be lit, and the debug code would be 00. After a minute or so, the system would boot as normal automatically. Sometimes, it would throw a BSOD, but it would be so fast, that the screen would go black as it was in a gradient to said blue screen. I've never been able to identify the error code, as windows doesn't have time to log it.

 

Other than that, it works fine. Just "fine".

 

Edit

 

I'm going to be looking for a more suitable chassis. It'll need to have space for an ATX motherboard, two power supplies, and at least 9 GPU's.

 

4 PSU's if I choose to go get some 4xxx GPU's.

How bad are risers gimping the card performance? Those look like x1 mining risers.

[IkeaGnome]

1 hour ago, Pezui said:

Anyone else been getting some sweet work units recently? I am pushing past 18M ppd right now which is fantastic!

I got shut down for the Windows update and won't be home for almost 2 more weeks. Enjoy the more points. 

Jealous.

[LazyDev]

2 hours ago, Bitter said:

How bad are risers gimping the card performance? Those look like x1 mining risers.

The risers are electronically wired as 4x 4x 4x 4x. Meaning, each GPU has 4 CPU lanes. You can run a 3080 on gen 4 with just 4 lanes, as it provides sufficient bandwidth for daily tasks, even Folding at Home (I have not tried my 3080 on these risers).  This does require of bifurcation of the primary PCIe port, of which not all motherboards support this feature. I've always had success with MSI motherboards.

 

https://riser.maxcloudon.com/en/

 

 

[LazyDev]

Update to my rig. I suspect the motherboard is dead. I've managed to record the fault on video. The system never successfully booted.

 

Seen at around 1:46, the BSOD that flashes, is "Critical Process Died". My only diagnosis, is a failure of the PCIe sub system on the motherboard, as Windows is running perfectly fine on the same drive, but on another motherboard.

 

My "New" Personal Rig

I had taken out the CPU, Ram and Nvme boot drive, and threw it into a brand new motherboard. It booted first time without issues. No modification or repair was done to the Nvme between the video above, and the installation of the drive on my new motherboard.

 

[Bitter]

11 hours ago, LazyDev said:

The risers are electronically wired as 4x 4x 4x 4x. Meaning, each GPU has 4 CPU lanes. You can run a 3080 on gen 4 with just 4 lanes, as it provides sufficient bandwidth for daily tasks, even Folding at Home (I have not tried my 3080 on these risers).  This does require of bifurcation of the primary PCIe port, of which not all motherboards support this feature. I've always had success with MSI motherboards.

 

https://riser.maxcloudon.com/en/

 

 

Oh neat!

[RollinLower]

@LazyDevdo you have any systems that, well, fit entirely in their case? 

[Captainmarino]

Hi, everyone!

 

I've been incommunicado for a while but not without good reason... my wife and I welcomed our third (and final!) Future Folder to the family a week before Christmas. Her name is Clara Noelle and she's doing great-- as indicated by her weight of 14.5 lbs at 2 months old! We, and her two older sisters (yes, three girls ) love her already.

 

On the folding front, I'm just about 60 days from joining the (not Diamond!!!) Platinum Club! Thanks for the support, everybody.

Edited March 17, 2022 by Captainmarino

[IkeaGnome]

1 minute ago, Captainmarino said:

Hi, everyone!

 

I've been incommunicado for a while but not without good reason... my wife and I welcomed our third (and final!) Future Folder to the family a week before Christmas. Her name is Clara Noelle and she's doing great-- as indicated by her weight of 14.5 lbs at 2 months old! We, and her two older sisters (yes, three girls ) love her already.

 

On the folding front, I'm just about 60 days from joining the Diamond Club! Thanks for the support, everybody.

Congratulations! And good luck in the future with three girls...

[Captainmarino]

7 minutes ago, IkeaGnome said:

Congratulations! And good luck in the future with three girls...

Much appreciated! I need it

 

I mean... they're great, it's great, I'm great.

[Gorgon]

15 hours ago, LazyDev said:

Update to my rig. I suspect the motherboard is dead. I've managed to record the fault on video. The system never successfully booted.

 

Seen at around 1:46, the BSOD that flashes, is "Critical Process Died". My only diagnosis, is a failure of the PCIe sub system on the motherboard, as Windows is running perfectly fine on the same drive, but on another motherboard.

 

My "New" Personal Rig

I had taken out the CPU, Ram and Nvme boot drive, and threw it into a brand new motherboard. It booted first time without issues. No modification or repair was done to the Nvme between the video above, and the installation of the drive on my new motherboard.

Glad to hear it's likely only the motherboard. I believe you were just running one power supply - correct?

 

When I was running 6 GPUs in my mining frame I had second memory channel die on two separate motherboards when they were installed in the mining frame with 2 power supplies. I was never able to conclusively prove it but I suspect that there was a ground loop or current loop because of the two power supplies that took out the Memory Channels.

[LazyDev]

2 hours ago, Gorgon said:

Glad to hear it's likely only the motherboard. I believe you were just running one power supply - correct?

 

When I was running 6 GPUs in my mining frame I had second memory channel die on two separate motherboards when they were installed in the mining frame with 2 power supplies. I was never able to conclusively prove it but I suspect that there was a ground loop or current loop because of the two power supplies that took out the Memory Channels.

I did recently install a second, 1200 PSU. I used only a single 24 pin extension just to power it on. 

 

I'm going to leave the system for a few days, to see how my existing hardware that I pulled out (CPU, Ram, Nvme, 2060) functions, and see if the fault replicates. I do believe the motherboard is dead, as when the fault occurs, Windows doesn't have time to log the blue screen "Critical_Process_Died". As if the Nvme drive was unplugged from its port. This is what leads me to think the PCIe sub system has failed, with an intermittent fault. 

 

 

I have noticed as well, that the ethernet port would randomly drop, and would have to manually replug the cable. This never happened on the Wifi.

 

After a few days, if no faults are present with the current hardware, then I'll have a look at RMA'ing the motherboard.

[Gorgon]

33 minutes ago, LazyDev said:

I did recently install a second, 1200 PSU. I used only a single 24 pin extension just to power it on. 

 

I'm going to leave the system for a few days, to see how my existing hardware that I pulled out (CPU, Ram, Nvme, 2060) functions, and see if the fault replicates. I do believe the motherboard is dead, as when the fault occurs, Windows doesn't have time to log the blue screen "Critical_Process_Died". As if the Nvme drive was unplugged from its port. This is what leads me to think the PCIe sub system has failed, with an intermittent fault. 

 

 

I have noticed as well, that the ethernet port would randomly drop, and would have to manually replug the cable. This never happened on the Wifi.

 

After a few days, if no faults are present with the current hardware, then I'll have a look at RMA'ing the motherboard.

That was how I had the PSUs interconnected as well. Just a Thermaltake 24-pin adadtor to send just the "power on" signal to the second supply.

 

Generally, if using powered PCIe extensions, it is best practice to ensure the PCIe bus power for an extension comes from the same PSU as the PCIe Power connectors on the GPU. Even then weird things can happen due to the shared ground via the PCIe extension.

[LazyDev]

7 minutes ago, Gorgon said:

That was how I had the PSUs interconnected as well. Just a Thermaltake 24-pin adadtor to send just the "power on" signal to the second supply.

 

Generally, if using powered PCIe extensions, it is best practice to ensure the PCIe bus power for an extension comes from the same PSU as the PCIe Power connectors on the GPU. Even then weird things can happen due to the shared ground via the PCIe extension.

For now, the rig is decommissioned. I have a spare B550 Tomahawk motherboard that I need to test, and I'll end up rebuilding the system into another mining chassis with a better suited operating system. However, with the cost of electricity in this country is still rising (United Kingdom), this won't be a priority for now. I may have it built by the next folding month.

 

I certainly would like to see as well, the future of power supplies. For example, PCIe gen 5 connectors and the new ATX standard for 12 volts to the motherboard.

 

 

[Gorgon]

3 hours ago, LazyDev said:

For now, the rig is decommissioned. I have a spare B550 Tomahawk motherboard that I need to test, and I'll end up rebuilding the system into another mining chassis with a better suited operating system. However, with the cost of electricity in this country is still rising (United Kingdom), this won't be a priority for now. I may have it built by the next folding month.

 

I certainly would like to see as well, the future of power supplies. For example, PCIe gen 5 connectors and the new ATX standard for 12 volts to the motherboard.

Here in Canada (Eastern Ontario) we have Time of Use rates which I believe are also available in the UK. What I’ve been doing is only running my rigs during the less expensive ToU periods using cron scripts.

 

Its nice that you have the option to procure renewables on your supply and they will actually pay you when they have excess wind production.

 

I’m writing up a Howto on ToU folding where you can also just raise and lower CPU and GPU power limits depending on the current ToU periods which I’ll post in the next couple of weeks.

 

I’m also working on updating my Linux HOWTO to use Ubuntu 20.04 LTS rather than 18.04 LTS which will be end-of-life shortly but I’ll have to wait a bit for that as I’ll have to take some of the rigs off-line for an extended period which I’d rather do after the cold season here ends.

[justpoet]

36 minutes ago, Gorgon said:

I’m also working on updating my Linux HOWTO to use Ubuntu 20.04 LTS rather than 18.04 LTS which will be end-of-life shortly but I’ll have to wait a bit for that as I’ll have to take some of the rigs off-line for an extended period which I’d rather do after the cold season here ends.

Looking forward to this for the GPU side of things, as it seems to be a bit hit or mostly miss to get it working right.  

[dogwitch]

i planned on getting another gpu. but had a large raid set up fail... so replacing drives etc... i simple cant atm for a gpu

[RollinLower]

2 hours ago, justpoet said:

Looking forward to this for the GPU side of things, as it seems to be a bit hit or mostly miss to get it working right.  

same here, currently folding on 18.04 because 20.04 was a damn nightmare to get going.

[marknd59]

Damn I have not been paying atention I  passed 3B points and didn't notice. 

 

[Gorgon]

2 hours ago, RollinLower said:

same here, currently folding on 18.04 because 20.04 was a damn nightmare to get going.

18.04 LTS is end of support later this year and 22.04 will be out very soon so I figure it's time to get going on 20.04 LTS. Besides, I want the newer kernels available in 20.04 which have better support for Zen3.

 

The biggest drawback for many with 20.04 LTS is the lack of Python 2 support and the attendant issues with getting the F@H GUI working but I believe there is a Python 3 FAHClient in beta and, if not, there are work-arounds. Doesn't really bother me as I run all my Linux Rigs headless but I'll get that working and document it.

 

My main compaint is the crappy system.d support. It would be really nice if

service FAHClient start|stop|restart

actually worked properly so I'll be spending some time on that as others have gotten it working with custom init.d scripts and making the client a member of the "video" group like what's needed for AMD GPU drivers.

 

First I have to upgrade my monitoring system to 20.04 which means moving from Zabbix 3.4 to 4.x then when I'm upgrading the rigs to 20.04 the native Zabbix 4.x client package will actually work and I'll maintain my telemetry.

[Gorgon]

11 hours ago, dogwitch said:

i planned on getting another gpu. but had a large raid set up fail... so replacing drives etc... i simple cant atm for a gpu

I feel you.

 

My TrueNAS box has one disk starting to drop off-line occasionally. The disks are 7 years old at this point and though the SMART data still shows them as OK it's time to think about replacing them and the RAID 10 array on my Daily Driver which is the same vintage.

 

But funding for 10 new 4-6TB NAS disks, UPS upgrades, a replacement for my 10 year-old Daily Driver and memory upgrades for my Distributed Computing Rigs will mean most of my budget that could have gone towards GPUs is gone.

 

I've so far refused to pay the exorbitant prices for GPUs and over the course of COVID have instead been upgrading CPUs as I can.

[dogwitch]

3 hours ago, Gorgon said:

I feel you.

 

My TrueNAS box has one disk starting to drop off-line occasionally. The disks are 7 years old at this point and though the SMART data still shows them as OK it's time to think about replacing them and the RAID 10 array on my Daily Driver which is the same vintage.

 

But funding for 10 new 4-6TB NAS disks, UPS upgrades, a replacement for my 10 year-old Daily Driver and memory upgrades for my Distributed Computing Rigs will mean most of my budget that could have gone towards GPUs is gone.

 

I've so far refused to pay the exorbitant prices for GPUs and over the course of COVID have instead been upgrading CPUs as I can.

yeah. currently running 8 bay set up and a 4 bay dense one. have on other but atm. it having issues.(one that failed on me os and drive)

[Bitter]

Each Nvidia card must have it's own CPU core because of Nvidia driver weirdness, AMD doesn't care. The G4290 is appropriate to drive one Nvidia GPU and run Windows with the other core, adding more cards than that to it will start to gimp down performance. That's why my OG folding rig had a 35W Haswell i5 with 4 cores, I could run 3 Nvidia cards no problem on it and turning turbo off lowered it to a 25W CPU.

[Gorgon]

52 minutes ago, Schnoz said:

I've been making a few upgrades lately to generate more F@H PPD. Up until November/December, I was folding on a Sapphire Nitro+ 5700 XT (1.1-1.3M PPD), which I replaced with a 1660 Ti (1.6M PPD).

 

In December, I set up another system with a Tesla P100 borrowed from my dad's workplace (3.4M PPD). The system is a veritable amalgamation of parts I've accumulated over time.

 

• CPU: Delidded Celeron G4920, courtesy of @TheCoder2019
• Motherboard: Asus Z370-I Gaming, courtesy of Den-Fi
• RAM: 2x4GB DDR4-2133 DIMMs I just had laying around
• SSD: A very-solidly-built PNY CS1200 120 GB SSD, courtesy of Den-Fi
• Case: Some cheap flimsy Rosewill thing I literally found laying on the side of the street
• CPU cooler: Cooler Master Hyper T4 from my first desktop, with no fan because the G4920 produces almost no heat
• PSU: Some Thermaltake 500W 80+ Bronze thing from my first desktop
• GPU fans: Noctua NF-F12 iPPC-3000, courtesy of Den-Fi, and Noctua NF-A12x25 chromax.black.swap
• Display: Some Sharp 720p thing I found when dumpster-diving back when I lived in Southern California

 

I've since replaced the 1660 Ti with a Titan Xp (3M PPD), and I'm also working on getting that 1660 Ti up and running again in a separate Core 2 Quad system.

I really want to maximize the PPD of the cards I'm using. Are there any tips I should keep in mind? I'm having each one fold Alzheimer's WUs, but are there any other WUs that yield a higher point value?

You could also replace the Celeron with a used Pentium G5400 or 5500 which are 2c/4t you should be able to drive 2 GPUs in one system but you'd need an m.2 to PCIe3 x16 adapter on that iTX motherboard to get a second slot and you'd likely need to Power limit the cards a bit to not exceed the 500W on the Power Supply.

[rkv_2401]

3 hours ago, Schnoz said:

I've been making a few upgrades lately to generate more F@H PPD.

I really want to maximize the PPD of the cards I'm using. Are there any tips I should keep in mind? I'm having each one fold Alzheimer's WUs, but are there any other WUs that yield a higher point value?

The new CUDA Folding WUs are quite CPU intensive, as me and a couple of forum members discussed during the last Folding month. Make sure you don't have anything else CPU intensive running if you're using Windows, but @Gorgon said that on Linux, one core is enough to drive one GPU (IIRC).

 

If you don't get free electricity, I would suggesting optimizing for PPD/W (within the limits of your hardware ofc) than outright max. PPD. Some undervolting (embedded down below) will go a long way in improving the efficiency, netting you higher core clocks (probably even an overclock) at a lower power draw, and be that much easier on the card and the fans (and your wallet!)

[rkv_2401]

1 hour ago, Schnoz said:

By the way, are Alzheimer's WUs the best in terms of PPD production, or should I switch to something else like Prions or COVID-19?

I'm not sure, but you can check LARsystem's database for your GPU/similar GPUs and find out.