LTT Official Folding Month VI

[CT854]

Oh. I'm back.

 

I *think* I signed up properly. But I'll use this as a chance to heat my apartment anyway since my super, like usual, doesn't love turning on the heat consistently with the fluctuating temperatures.

[Alex Atkin UK]

Upscaler box locked up again, so this is going to be fun.   I have no idea what's broken here.

 

I suppose I could try disabling each CCD in case a CPU core has failed, but its odd as I did a multi-core test in CPU-Z and that ran fine.

 

I did have this problem in the old motherboard, but only when reducing the voltage in the CPU curve.  I've played 60 hours of Starfield on this CPU, just not on this motherboard.  Plus it ran 3 hours of MemTest ON THIS motherboard fine.  Ideas?

[justpoet]

42 minutes ago, Alex Atkin UK said:

Upscaler box locked up again, so this is going to be fun.   I have no idea what's broken here.

 

I suppose I could try disabling each CCD in case a CPU core has failed, but its odd as I did a multi-core test in CPU-Z and that ran fine.

 

I did have this problem in the old motherboard, but only when reducing the voltage in the CPU curve.  I've played 60 hours of Starfield on this CPU, just not on this motherboard.  Plus it ran 3 hours of MemTest ON THIS motherboard fine.  Ideas?

Most recent AGESA on the Motherboard?  Maybe do a manual set within bounds for voltage and clock multipliers?  I hate to say it, but...check for a hair or similar in the CPU socket?

[Alex Atkin UK]

2 hours ago, justpoet said:

Most recent AGESA on the Motherboard?  Maybe do a manual set within bounds for voltage and clock multipliers?  I hate to say it, but...check for a hair or similar in the CPU socket?

Yeah I only got this board about a week back so its the latest BIOS as it was the first thing I did.  Working perfectly since then, until today.

 

Windows Event Log doesn't seem to have anything interesting either, just the "unexpected power loss" from when I have to force the power off.

Its just one of those really annoying issues where it works absolutely perfectly, done several AI upscales while Folding which hammer the CPU due to software PRORES encoding.

 

If anything what I was guessing when I was under-volting on the old board, it was almost like when the CPU is idle it crashes, not under load, or in the transition between boost and idle. 

 

Still not used to Ryzen, it seems to run at 1.48v with just the NVIDIA driver thread on CCD 1 and F@H on CCD 0, yet drop to 1.18v under multi-core load, according to Ryzen Master.  Average core voltage is 0.55v (presumably all the sleeping cores).

 

Think I should leave Ryzen Master open so I can see what it says when it freezes?

 

Hang on, why does Curve Optimiser keep flashing up saying there is a negative offset when I reset the offset to 0 and disabled PBO?  Oh Ryzen Master says there is still an all-core -10mv offset?  Lets turn that off.

[Needfuldoer]

After running overnight, the R7910 settled in at around 420 to 450 watts from the wall. I forced the GPU fans to run harder with MSI Afterburner, so now the hot spot temperature isn't above 87 degrees.

 

Spoiler

 

 

But holy mackerel this thing's loud! I think it's just running its wall of blowymatrons at 100% regardless of the offset I put into iDRAC. I might actually have to see if the tower 7910 survived last month's basement flood. (That will take three of these cards, but two of them end up right next to each other so I'm concerned about one not getting enough fresh air.)

 

EDIT: If I disable the third-party PCIe fan response again, the volume drops to tolerable levels again. However, that's at the expense of thermals. This is after just a couple minutes:

 

Spoiler

Hopefully it's fine?

Edited October 29, 2023 by Needfuldoer

[TylerD321]

1 hour ago, Needfuldoer said:

After running overnight, the R7910 settled in at around 420 to 450 watts from the wall. I forced the GPU fans to run harder with MSI Afterburner, so now the hot spot temperature isn't above 87 degrees.

 

  Hide contents

 

 

But holy mackerel this thing's loud! I think it's just running its wall of blowymatrons at 100% regardless of the offset I put into iDRAC. I might actually have to see if the tower 7910 survived last month's basement flood. (That will take three of these cards, but two of them end up right next to each other so I'm concerned about one not getting enough fresh air.)

 

EDIT: If I disable the third-party PCIe fan response again, the volume drops to tolerable levels again. However, that's at the expense of thermals. This is after just a couple minutes:

 

  Hide contents

Hopefully it's fine?

I wanna see some pics of the setup

[CerberusEXHW]

Got linked here From ExtremeHW.net, I'd be glad to donate some power to the cause.

 

https://extremehw.net/topic/3280-ltt-official-folding-month-vi

[GOTSpectrum]

4 minutes ago, CerberusEXHW said:

Got linked here From ExtremeHW.net, I'd be glad to donate some power to the cause.

 

https://extremehw.net/topic/3280-ltt-official-folding-month-vi

Welcome aboard, remember you must fold for team 223518 to be counted for the event 

[Windows7ge]

Even as fall slowly sets in and the weather gets colder. The house needs heating and folding being a good excuse. My landlord is still immediately on my butt about what I'm doing with them, what it's going to cost - ignoring that I offered him additional compensation for the month and we haven't even started yet.

[Brakta]

it shuts down my pc a few sec after restart, showed me some amd error only the first time, had to uninstall , gpu amd 6800 xt, cpu 7950x

[Imakuni]

1 minute ago, Brakta said:

it shuts down my pc a few sec after restart, showed me some amd error only the first time, had to uninstall , gpu amd 6800 xt, cpu 7950x

In other words, your computer is unstable and you need to find the source of it. If I had to bet, that would be the CPU's PBO exagerating on the overclock, or possibly the memory EXPO not being as stable as they'd like you to believe.

[GOTSpectrum]

4 minutes ago, Imakuni said:

In other words, your computer is unstable and you need to find the source of it. If I had to bet, that would be the CPU's PBO exagerating on the overclock, or possibly the memory EXPO not being as stable as they'd like you to believe.

There's nothing like folding to tell you if your system is unstable 

[GOTSpectrum]

I'll be firing up Foldatron 3.0 this evening, it has been rebuilt since the last event as I have said in previous replies here, with a new case for better airflow, a 22 core, and 2x 4070ti... 

 

She's only had a day or so of testing so let's hope she holds up 

[GOTSpectrum]

The first blog post is up! 

 

 

[Blufire]

Haven't had the chance to participate before but I finally got setup this year! Don't have a dedicated folding machine, just my main PC, but at least it will double as a space heater as things get colder here

[cbigfoot]

Friendly reminder going through the points there are a few people actively folding on default team or another team You must fold for LinusTechTips_Team  # 223518 for it to count when the event starts. 

[Fluxmaven]

Another ExtremeHW degenerate here. I'll switch a rig or two over for the event. Finally starting to cool down where I'm at so I've been wanting to be able to fold more. 

[Alex Atkin UK]

Touch wood, 14 hours and counting since I disabled the Curve Optimiser and were still folding......

[Shadowbert]

Is the passkey sign up working for anyone? It never loads more than the background on my side.

In the dev console, it mentions that Vue is not defined. Maybe somebody over there stuffed up a website deployment?

[GOTSpectrum]

1 minute ago, Shadowbert said:

Is the passkey sign up working for anyone? It never loads more than the background on my side.

In the dev console, it mentions that Vue is not defined. Maybe somebody over there stuffed up a website deployment?

I've had a couple of people ask me the same thing, maybe we hugged it to death? 

[GOTSpectrum]

Can I fold on this /smaybe /s

 

https://www.currys.co.uk/products/lenovo-tab-p11-11.5-tablet-128-gb-storm-grey-10244078.html

 

It comes tomorrow so maybe I'll have a play 

[Imakuni]

How do I access the client without using a browser? Opening advanced view just keeps "connecting" endlessly, without ever actually doing so.

[Alex Atkin UK]

13 minutes ago, Schnoz said:

Definitely! Folding@Home is interesting where (at least on Nvidia GPUs), it doesn't actually peg the GPU fully like a graphically-intensive game or Furmark would. The usual quick-and-dirty overclock method (run Furmark, crank up core and memory speeds separately until it crashes) doesn't work, as it results in failed WUs. Since the GPU isn't being driven at 100%, it doesn't run into its power limit as easily and operates at a frequency higher than the maximum under-load boost clock, which may be unstable. Fortunately, dialing back the overclock fixes this issue.

Yeah its kinda confusing as it will show 100% utilisation, which I presume refers to the GPU pipeline, as its basically impossible to 100% utilise the silicon in ANY scenario, game or compute.

 

I've already shown the real-world power consumption between a 3080, 4070 Ti and 4090 previous on the 4090 is crazy efficient at Folding.  I was actually looking into including it in my real-time stats, but I need to figure out how to reliably feed that information back to my server.

 

You can basically probe this on Windows or Linux with nvidia-smi though:

nvidia-smi --query-gpu=index,timestamp,power.draw --format=csv -f /tmp/www/gpu.log

So I may just feed it to a network share.

[BiotechBen]

1 hour ago, Imakuni said:

How do I access the client without using a browser? Opening advanced view just keeps "connecting" endlessly, without ever actually doing so.

I usually just open F@H control and hit "start" it's not the prettiest interface, but it gets the job done.

[Alex Atkin UK]

Just now, BiotechBen said:

I usually just open F@H control and hit "start" it's not the prettiest interface, but it gets the job done.

There's an unofficial Android app but drawback is it doesn't let you control individual slots.