Folding Community Board

[Juggy]

@LAR_SystemsI cannot seem to find your Chrome plugin in the web store anymore. I get the below error when clicking through from your website

 

[TVwazhere]

On 12/3/2020 at 8:19 AM, TVwazhere said:

 

SHIT

Back on track

[Kilrah]

Same, apart from my own excursions it's perfectly flat...

 

[LAR_Systems]

3 hours ago, Juggy said:

@LAR_SystemsI cannot seem to find your Chrome plugin in the web store anymore. I get the below error when clicking through from your website

 

Google has changed the way they review the extension sometime around 12AM EST today and therefore I have to resubmit it updated.

 

It has to do with the CPU tracking feature, they are suggesting the app does not use the CPU permission, and you can't have an app that asks for permissions it does not use... meanwhile it showing CPU names, logical processors and their PPD suggests otherwise.

I assume it's a matter of review automation not being able to tell how it's used, and it's hopefully resolved quickly.

Kinds feeling like Luc with floatplane app fun right now.  

[LAR_Systems]

5 minutes ago, TVwazhere said:

Back on track

I know there was some talk that F@H was trying to improve on the work servers giving specific GPUs WUs they are best suited for.

Wonder if this is that given the consistency. 

[chaozbandit]

8 minutes ago, TVwazhere said:

Back on track

[Juggy]

3 hours ago, LAR_Systems said:

Google has changed the way they review the extension sometime around 12AM EST today and therefore I have to resubmit it updated.

 

It has to do with the CPU tracking feature, they are suggesting the app does not use the CPU permission, and you can't have an app that asks for permissions it does not use... meanwhile it showing CPU names, logical processors and their PPD suggests otherwise.

I assume it's a matter of review automation not being able to tell how it's used, and it's hopefully resolved quickly.

Kinds feeling like Luc with floatplane app fun right now.  

Thank you and thanks for the great work with the plugin

[Happiness_is_Key]

On 12/7/2020 at 12:51 AM, LazyDev said:

GPU's come with 16 lanes by factory. However, they will only use the lanes that have been given, either by software (Bios) or by the connector/adapter that they're plugged into.

 

Ah, okay. That's news to me. I suppose that limits how much performance you can get out of them. Good to know! Thank you!

On 12/7/2020 at 1:18 AM, piratemonkey said:

I think (not too sure) fah helped create the spike model for covid19. Relatively speaking, folding is new (created around the start of 2010s maybe) so the results that may have come from it are probably not fully realized yet. I'm not aware of any other big results that have come folding.

 

That's neat. Yes, understood. I just want to make sure that the efforts are going toward something that benefits something rather than consuming a lot of electricity and computer resources for just points (no offense Bitcoin people ). While that is fun, I'd rather help people with my resources. That's the only reason why I ask.

Quote

But, it is worth noting that the network was faster than the fastest supercomputer at its peak this year (around March; when the call to fold was put out by LTT, Nvidia etc.). It has since fallen quite a bit, but it's still really fast. Like one of the fastest computing projects. I'm sure further understanding of diseases and ailments bring researched will be realized. Several universities afaik use the network after all. And don't worry about offending anyone. Even if you do, it's for a good reason. 

 

I heard about that! That was pretty cool and a great accomplishment. Once again, me just wanting to make sure part of my paycheck (electricity bill) and computer resources are being put to good use.... that's all. Overall, it's an amazing accomplishment that this is something that helps and all you need is a computer as every little bit helps as long as you get it done before the WU expires. I appreciate your response!

On 12/7/2020 at 1:37 AM, rkv_2401 said:

Welcome! It's super impressive that you've managed to score ~2 million points via just CPU folding. I was in a similar range when I resumed folding last year - my laptop's GPU only made 35k PPD running 24/7, and it seemed like even a million points would be practically undoable(I found it hard to sleep at night with the fan whine). But here I am now, having contributed just over 100 million points to the cause.

Thank you! Your GPU was doing pretty good compared to my 10-11k PPD. Haha. I too can relate to the fan whine, but I was able to optimize it where the fan didn't have to spin so fast which was taking it out of its casing. Congratulations to you on over 100 million points! That's amazing! I would run F@H on my MBP, but I can't get it to work in Linux. I can install the F@H client.deb, but not the control.deb for whatever reason. I have tried it through the Software Center and through the terminal using the F@H instructions on their website, but no luck. My MBP 8-core CPU is the most powerful one I have right now. I did get the F@H client to work on it earlier (hence the big spike on my record - see the attached image), but I can't control it when I'm doing work as it just fires up and goes to town, but I need to limit it to 6-cores so I can get my work done otherwise my main computer is too slow. Anyone know what's up with that on macOS Big Sur (Yes, I do know it's a new release, but I had to do it for work)? I suppose I could try to download Windows on it, but I honestly don't like Windows' phoning home/privacy issues. That's why I haven't gotten a desktop PC yet if I'm being honest as I'd love to be able to customize my computer, but that doesn't happen with a Mac anymore due to soldering stuff directly to the board. Ugh. I don't know enough about Linux to get that to work as every time I've tried to get something to work on it, it doesn't work for whatever reason (besides, I need the Adobe Creative Cloud suite for work).

Quote

I'm a bit rusty on the topic, but I believe all the projects turn into something as the scientists only create projects based on compounds they need to predict and filter prior to experimental validation. I heard from a friend who works in the field that waiting for computers to finish processing is one of the bigger delays in such research, so they have a clear incentive to send WUs of the highest priority first.

 

No worries, but that's good to hear and it makes sense. As per the waiting for the computers to finish the projects is interesting as the processing power is pretty monstrous, but I understand it. That must be why they're trying to optimize the projects based on the build, so it completes it more efficiently and time consciously.

Quote

For a single graphics card(like the 2060 Super you're considering), the no. of PCI-e lanes discussion isn't really relevant. Higher lanes means faster speeds/more bandwidth. One lane of PCI-e 3.0 can transfer data at upto 8 GT/s, so 4x can do 32GT/s, 16x can do 128GT/s, etc. It needs to be higher than 4x, yes, but there's no reason for you to set it to anything lower than the factory-default 16x. Lowering that may handicap your performance in other applications, such as gaming, anyway. These details are important when you're trying to create a machine with more GPUs than there are (high-speed) PCI-e slots, such as @LazyDev's F@h build. 

 

Your reply and the other replies have made me understand this better, so thank you. While looking at my CPU specs (here), I see it has a bus speed of 8 GT/s which seems like it's only for the memory controllers according to the help window (but that doesn't quite make sense as the PCI-e lanes have to end up somewhere), but it does say in the 'Expansion Options' section that there are x16 PCI-e lanes that operate at 3.0 speeds or you can have "up to 1x16, 2x8, 1x8+2x4" which does make sense as my Mac says it uses x8 lanes for the GPU and x4 lanes for the NVMe SSD, but I'm not sure what the other x4 lanes are for. Though, while on the NVMExpress page, it says the link width is x4 lanes, but the link speed is 8.0 GT/s, but your reply says it should be 32GT/s per your reply. Hmm.... Anyways, that's good to know! As per @LazyDev's build, wouldn't having 1 full x16 enabled slot be better than what they have? I see the motherboard will support 2 x16 lanes and x1 lane, but the CPU appears to only support 1 x16 slot (however, this is PCI-e 4.0 so I'm not sure what difference that makes other than probably speed). They currently have x25 lanes worth of GPUs in there which I don't know how that works other than having multiple CPUs, but it looks like they only have one in that case (the motherboard doesn't support multiple CPUs).

Quote

CPUs that have hyper-threading enabled have 2 threads for every CPU core, but a 2C/4T CPU isn't the same as having 4 actual cores. I believe the difference is that the extra threads can't all do floating point calculations at the same time, but the end difference is that the performance gain with HT on is around 30-40% compared to that with HT off for a workload where a CPU with 2x the cores would be roughly 2x faster. 

 

That's very interesting! That's basically the conclusion I came to, but it's good to have it in words. Looks like I need to read into that as well. I do know the Hyper-Threading is an Intel technology, and AMD uses something similar called Simultaneous Muti-Threading. I've been a long time Intel user, so I know more about their CPUs. I'm still curious about AMD though, don't get me wrong. It's truly amazing how far they've come.

 

 

Anyways, thank you all! I really appreciate the support and learning about this. It's super interesting to me, and something I hope to be able to pass on to others as I gain knowledge. I hope you have a wonderful day!

Edited December 8, 2020 by Happiness_is_Key

[GOTSpectrum]

@Happiness_is_KeyI'm not sure if I have said it yet, but as folding team leader,  welcome to the team! It's great to have you on board! 

 

We'll fold you into our ranks! 

[piratemonkey]

6 minutes ago, GOTSpectrum said:

We'll fold you into our ranks! 

Times like these are when a disappointed reaction is needed the most

[GOTSpectrum]

2 minutes ago, piratemonkey said:

Times like these are when a disappointed reaction is needed the most

I'm pretty sure I could hear the painful groans from you as I clicked submit 

[LazyDev]

22 minutes ago, Happiness_is_Key said:

As per @LazyDev's build, wouldn't having 1 full x16 enabled slot be better than what they have? I see the motherboard will support 2 x16 lanes and x1 lane, but the CPU appears to only support 1 x16 slot (however, this is PCI-e 4.0 so I'm not sure what difference that makes other than probably speed). They currently have x25 lanes worth of GPUs in there which I don't know how that works other than having multiple CPUs, but it looks like they only have one in that case (the motherboard doesn't support multiple CPUs).

The 3000 series Ryzen CPU's come with 24 PCIe lanes, compared to the 16 that Intel currently ship with. The primary X16 slot is bifurcated into 4x4x4x4, so with the correct adapter, four GPU's can plug into that single slot. The secondary x16 slot runs off of the chipset, which only gets four lanes, and is also subject to increased latency. The 6th (Or where my primary display GPU sits) is plugged into the top most NVME slot, which uses 4 CPU lanes.

 

In my system, 

Primary X16 slot = CPU = 16 lanes

Primary M.2 = CPU = 4 lanes

 

Secondary x16 slot = X4 chipset

Seconday M.2 = x4 chipset

- Shared x4 lanes, hence the increased latency.

 

Also, More speed you have, the less lanes you'll require. For example, my 3080 is on PCI gen 4, and the most utilization is around 4%. In theory, the card will run without issues on just 1 gen four lane.

 

Quote

• Version 1.x: 2.5 GT/s

  • x1: 250 MB/s
  • x16: 4 GB/s

• Version 2.x: 5 GT/s

  • x1: 500 MB/s
  • x16: 8 GB/s

• Version 3.x: 8 GT/s

  • x1: 985 MB/s
  • x16: 15.75 GB/s

• Version 4.0: 16 GT/s

  • x1: 1.97 GB/s
  • x16: 31.5 GB/s

• Version 5.0: 32 GT/s

  • x1: 3.94 GB/s
  • x16: 63 GB/s

https://en.wikipedia.org/wiki/PCI_Express#PCI_Express_4.0

[Favebook]

42 minutes ago, Happiness_is_Key said:

That's neat. Yes, understood. I just want to make sure that the efforts are going toward something that benefits something rather than consuming a lot of electricity and computer resources for just points (no offense Bitcoin people ). While that is fun, I'd rather help people with my resources. That's the only reason why I ask

Not sure I understand this reference. Bitcoin is not "just point(s)". Bitcoin has value.

[LAR_Systems]

4 hours ago, Juggy said:

Thank you and thanks for the great work with the plugin

Thanks for your support, makes a big difference on days like this when things are not going as planned.

[Kilrah]

Blah my flat graph will be spoiled again, just came to find my PC powered off, it just shut down in the middle of a WU and I only noticed 2 hours later. 

[Happiness_is_Key]

5 hours ago, GOTSpectrum said:

@Happiness_is_KeyI'm not sure if I have said it yet, but as folding team leader,  welcome to the team! It's great to have you on board! 

 

We'll fold you into our ranks! 

I appreciate that! Haha, nice. 

[Happiness_is_Key]

5 hours ago, LazyDev said:

The 3000 series Ryzen CPU's come with 24 PCIe lanes, compared to the 16 that Intel currently ship with. The primary X16 slot is bifurcated into 4x4x4x4, so with the correct adapter, four GPU's can plug into that single slot. The secondary x16 slot runs off of the chipset, which only gets four lanes, and is also subject to increased latency. The 6th (Or where my primary display GPU sits) is plugged into the top most NVME slot, which uses 4 CPU lanes.

 

In my system, 

Primary X16 slot = CPU = 16 lanes

Primary M.2 = CPU = 4 lanes

 

Secondary x16 slot = X4 chipset

Seconday M.2 = x4 chipset

- Shared x4 lanes, hence the increased latency.

 

Also, More speed you have, the less lanes you'll require. For example, my 3080 is on PCI gen 4, and the most utilization is around 4%. In theory, the card will run without issues on just 1 gen four lane.

 

https://en.wikipedia.org/wiki/PCI_Express#PCI_Express_4.0

 

Really? The AMD website says it only has 4.0 x16 unless I'm missing something. I suppose it could mean that the rest are not 4.0 and a different version, though I'm not sure if that's possible. Your description does make sense though. I appreciate the expanded explanation.

 

I didn't know that the different generations relate to how many lanes you need to have even though it makes sense. Very cool!

 

This is why I need to read about it more.

5 hours ago, Favebook said:

Not sure I understand this reference. Bitcoin is not "just point(s)". Bitcoin has value.

 

Looking back at that now, I'm not sure what I meant. You got me. My apologies. I guess I forgot about Bitcoin having money value while typing all of that. Hahahaha, just one of those days. I'll strikethrough that in an edit.

[GOTSpectrum]

It's looking like around the start of next year I could end up having a more healthy income. if it works out like I'm expecting it may be time to upgrade my 10 series cards to either 20 or 30 series! I'm trying not to get too excited as the money isn't guaranteed but theres that small part of my that is looking forward to it. 

[Kilrah]

16 hours ago, LazyDev said:

The 3000 series Ryzen CPU's come with 24 PCIe lanes, compared to the 16 that Intel currently ship with.

4 of those are the chipset link, so you can't really count that way.

Intel has 16 user lanes + DMI 3.0 (which is basically 4 PCIe3 lanes) to the chipset.

 

AMD has 20 user lanes + 4 to the chipset.

 

The 4 extra lanes on AMD go to an NVMe SSD on pretty much all consumer mobos.

[TVwazhere]

On 12/7/2020 at 8:38 AM, LAR_Systems said:

I know there was some talk that F@H was trying to improve on the work servers giving specific GPUs WUs they are best suited for.

Wonder if this is that given the consistency. 

On 12/7/2020 at 8:40 AM, chaozbandit said:

At this point it has to be. It's been a week and other than a small fluctuation every now and then, the points every 3 hours is remarkably consistent.

It should be noted ITNOS is accepting beta WU's currently, perhaps that's part of the reason? (meaning they haven't rolled out WU optimized server settings to everyone just yet)

[Favebook]

13 hours ago, Happiness_is_Key said:

Looking back at that now, I'm not sure what I meant. You got me. My apologies. I guess I forgot about Bitcoin having money value while typing all of that. Hahahaha, just one of those days. I'll strikethrough that in an edit.

No worries, you forgot about "only" 350 billion $, so, nothing serious.  

[RollinLower]

5 hours ago, Favebook said:

No worries, you forgot about "only" 350 billion $, so, nothing serious.  

eh, pocket change. in zimbabwan dollars that is.

[Juggy]

Woohoo, 500 million points eventually and top 140 in the team... Happy days

 

[Captainmarino]

On 12/6/2020 at 4:31 PM, Bitter said:

Re-paste it? Stick case fans on it?

 

I have 2 80x15mm fans on my P106-100 because it came without fans or shroud, they're held with screws jammed into the fins. I have a GTX 650 (not folding) that the original fan lost a blade on, so I shucked the whole fan/shroud and zip-tied a 92x15mm case fan to it. Both cards run cool and quiet this way, the GTX 650 runs cooler than it ever did before, the P106 I don't have a previous referenced for. What is nice though, on the P106 I used a video card to case fan adapter. If I wanted, I could connect the fans to the motherboard or an external fan controller for manual control to run them at a set speed or whatever since they're standard fans. The 650 had a weird connector so I just soldered/heat shrunk.

I took a couple of case fans and MacGruber-ed them onto the heatsink and it's... better. Temp still climbs quite fast so I undervolted it and brought down the power slider. That got the temps to stay in the 60s at least. I may have to check the paste. The heatpipes get hot so there's some kind of contact. Maybe the fluid in the pipes leaked or something.

[Bitter]

1 minute ago, Captainmarino said:

I took a couple of case fans and MacGruber-ed them onto the heatsink and it's... better. Temp still climbs quite fast so I undervolted it and brought down the power slider. That got the temps to stay in the 60s at least. I may have to check the paste. The heatpipes get hot so there's some kind of contact. Maybe the fluid in the pipes leaked or something.

Heat pipes rarely fail, if it's never had new paste it needs it badly. Arctic MX4 is 8 year stable, not the highest heat transfer but it's very long lived.