Please REMOVE RX560

[Joe_H]

Why on earth would it be harder for AMD than Nvidia?

Didn't say it would be harder, just that it didn't happen. There was a third category for example, it was for AMD cards hat didn't have DP support. I think those have since been blacklisted as no current projects run without DP.

Everything is unstable. Turns out one card objected to Folding. Works on Milkyway though. But not primegrid. SP part broken and DP ok?

Your hardware or its setup is unstable, statement of fact. You can argue all you like, but that does not make it stable. As for everything being unstable, another false statement from you. F@h has thousand of machines contributing that have processed near 100% of the WUs assigned to them without error or crashing. They are not going to spend much time on working around a single person's hardware or setup issues, the time would be better spent on improving other aspects of the software.

[Lamberto Vitali]

Why on earth would it be harder for AMD than Nvidia?

Didn't say it would be harder, just that it didn't happen. There was a third category for example, it was for AMD cards hat didn't have DP support. I think those have since been blacklisted as no current projects run without DP.

If you managed it for Nvidia, why stop there? Don't tell me you have that racism against AMD I keep seeing?

Your hardware or its setup is unstable, statement of fact. You can argue all you like, but that does not make it stable. As for everything being unstable, another false statement from you.

Your arrogant rude post changes nothing. Again, hardware screws up. People change things around, things go wrong. Accept that fact. If I want to alter what I'm running or something crashes, then your software needs to cater for that or YOU lose out on a lost work unit. Fix it. I have given up caring if I throw WUs away. I came here to make a FRIENDLY HELPFUL SUGGESTION as to how you could improve things for YOUR BENEFIT, but you can't be bothered. Forget it.

As for your total and utter contempt for volunteers using their hard earned cash to donate to your program you shouldn't be an admin. Go ask your boss to read this.

[aetch]

4 GPUs.

4 GPUs.
Assuming for a moment they are all double slot RX560s, that's a lot of bandwidth and real estate for a motherboard to provide. My initial guess would be you're running a high-end gaming/workstation motherboard, except the GPUs need a lot of breathing room exhaust the heat. The only other type of motherboard I can think of with that many PCIe slots are mining motherboards.

My guess is you're running a mining rig with the GPU on PCIe risers, very likely with a dual core processor running the show.

[Lamberto Vitali]

4 GPUs.

4 GPUs.
Assuming for a moment they are all double slot RX560s, that's a lot of bandwidth and real estate for a motherboard to provide. My initial guess would be you're running a high-end gaming/workstation motherboard, except the GPUs need a lot of breathing room exhaust the heat. The only other type of motherboard I can think of with that many PCIe slots are mining motherboards.

My guess is you're running a mining rig with the GPU on PCIe risers, very likely with a dual core processor running the show.

280X. Antique DDR2 motherboard. Riser cables. Only need PCI-E v2 x1 lane to run a GPU.

[aetch]

Bad news.
FAH actually has much higher requirements.
The CPU is constantly moving data to/from the GPU while it runs.
Each GPU needs a dedicated CPU thread/core and a bare minimum bandwidth of x4, we actually recommend x8.
That means you need a quad core CPU as a bare minimum and the GPUs preferably plugged directly into the motherboard.

Your setup is killing your GPUs.

[Neil-B]

For fah any gpu running on a such a lane will be significantly inhibited for anything recent at least a v3 x8 is needed to ensure full performance and for very latest x16 will help ... mining and possibly boinc may perform well with such minimal attendance but fah needs better

[Joe_H]

If you managed it for Nvidia, why stop there? Don't tell me you have that racism against AMD I keep seeing?

That is another of those "reasons". There were volunteers working on the same approach for AMD GPUs, F@h Consortium members decided to use another approach for all performance related factors to be used in assigning WUs. That is the project that is currently on hold and not going forward after the person doing it was needed for other tasks.

As for AMD, their cards are not used as much for F@h. Part of it has been continuing problems with the drivers provided by AMD, especially on the Linux side. The majority of folders are running on Linux. At this point AMD GPUs being used are less than 10% of the number nVidia GPUs. https://stats.foldingathome.org/os

You are the one arguing for massive change is how the client works to fix your issues. Your suggestion so far has not been friendly or helpful. What we can give you here is suggestions how to use the client within its capabilities to have the least problems.

As for being for my benefit, I am a volunteer also. None of your suggestions really benefit me one way or another. I have been spending my money on the electricity to run folding for 17 years. I accept if I doing something outside the norm for the client, I may not get any points for it, and someone else may end up doing the WU. My completion rate is in excess of 99.8%.

[Lamberto Vitali]

Bad news.
FAH actually has much higher requirements.
The CPU is constantly moving data to/from the GPU while it runs.
Each GPU needs a dedicated CPU thread/core and a bare minimum bandwidth of x4, we actually recommend x8.
That means you need a quad core CPU as a bare minimum and the GPUs preferably plugged directly into the motherboard.

Your setup is killing your GPUs.

Wrong, I do monitor how well they perform. It gives me about 85-90% power from the GPUs. A 280X isn't exactly fast.

[Lamberto Vitali]

That is another of those "reasons". There were volunteers working on the same approach for AMD GPUs, F@h Consortium members decided to use another approach for all performance related factors to be used in assigning WUs. That is the project that is currently on hold and not going forward after the person doing it was needed for other tasks.

As for AMD, their cards are not used as much for F@h. Part of it has been continuing problems with the drivers provided by AMD, especially on the Linux side. The majority of folders are running on Linux. At this point AMD GPUs being used are less than 10% of the number nVidia GPUs. https://stats.foldingathome.org/os

They work fine here, but I use a proper paid for OS, not the freeware one.

[Nice snippage....]

You are the one arguing for massive change is how the client works to fix your issues.

Not my problem if work is lost. I make my machines do something useful all the time. If work is downloaded and not done, that only hurts the project.

Your suggestion so far has not been friendly or helpful. What we can give you here is suggestions how to use the client within its capabilities to have the least problems.

And what I suggested is how to make the client be more efficient. But you fail to see that.

As for being for my benefit, I am a volunteer also. None of your suggestions really benefit me one way or another.

When I said "your" I was of course referring to FAH.

I have been spending my money on the electricity to run folding for 17 years. I accept if I doing something outside the norm for the client, I may not get any points for it, and someone else may end up doing the WU. My completion rate is in excess of 99.8%.

Depends what you mean by completion rate, but I make sure the time my computer spends is useful. If I download one then don't process it, because I didn't actually ask the client to download it, then that affects my completion rate?

[aetch]

Wrong, I do monitor how well they perform. It gives me about 85-90% power from the GPUs. A 280X isn't exactly fast.

Sadly, I'm not wrong. Too many times people with mining rigs, just like yours, have came here for help because their system was unstable and underperforming. Just like yours they had a low power processor and bandwidth bottlenecks galore. They eventually got stable running folding rigs but had to essentially rebuild and reconfigure them, in many cases they pared back on the GPUs as well.

[Lamberto Vitali]

Sadly, I'm not wrong. Too many times people with mining rigs, just like yours, have came here for help because their system was unstable and underperforming. Just like yours they had a low power processor and bandwidth bottlenecks galore. They eventually got stable running folding rigs but had to essentially rebuild and reconfigure them, in many cases they pared back on the GPUs as well.

I know what my own system is doing, I know how much calculations need to occur on the CPU, I know how much bandwidth is required between them. The machine is sat right in front of me. If it was underperforming I could tell and I'd put it on another project.

And it's not a mining rig, mining is a pointless utter waste of electricity, but I do use their riser hardware.

If you don't believe me, how many PPD do you think a 280X should get? Because I get up to 450K.

Just checked with folding.lar.systems stats and I'm getting what I should out of that card.

[aetch]

You're actually making a really good argument that you don't know what you're doing.
FAH is not BOINC.
FAH is not bitcoin or one of its variants.
None of the "suggestions" you have made will change the fact that your system is not setup for FAH. Please put your system onto another project.

[aetch]

And it's not a mining rig, mining is a pointless utter waste of electricity, but I do use their riser hardware.

1). Your system has the topology of a mining rig, therefore it is a mining rig. The proprietary mining motherboards came along later.
2). Arguably FAH and BOINC are an even bigger waste of electricity as you don't even get virtual currency in exchange for your work, just points and the satisfaction you have progressed science.

If you don't believe me, how many PPD do you think a 280X should get? Because I get up to 450K.

Just checked with folding.lar.systems stats and I'm getting what I should out of that card.

Your system is no better than 1st Gen Core2, maybe 2nd Gen Core2, and it has a crippled PCIe bus.
This is not a credible claim i.e. it's a lie.

[Lamberto Vitali]

I can see what I'm achieving and I get the stated number of PPD which is similar to other GPUs of that model of mine on faster machines and other people's. Stop mouthing off, because you have not seen my machine, you do not know what it's doing. If it was giving less points than expected for that card, or the usage % in MSI Afterburner was rather low, I would know it wasn't being used to full potential. But I am getting a decent amount of points, there is hot air blasting off the cards, and the usage is reported as high.

As for point 2, how is the progression of science a waste of electricity? Why are you even here? Anyway, you do get curecoin and gridcoin, so you haven't researched that either.

[aetch]

My system doesn't crash and burn every few days.
Also, I'm not demanding changes to the client to work around an unstable system.

I tried to help, you're too arrogant to accept it. I'm out.