Folding Community Board

[Gorgon]

3 minutes ago, Inkertus said:

The VRM bit is why I hate the heat-slab thing MB designers have done on some boards.  You can't OC them without either a lot of directed airflow or a board surgery to put a heatsink on instead (it is otherwise a great board from what I can tell). I posted a picture a few weeks back of me stacking a sink from another board on top of a slab in a bit of silliness and it did help more than expected. Not OC good though. I think Gamers Nexus did a bit on that a good while back.

My issue on my Aorus z370 Gaming 5 was a poor VRM design (4-phase with 2 drivers per phase) combined with a stupid heatsink and said heatsinks had plastic spacers that prevented good contact with the MOSFETs. I removed the plastic washers and tightened down the screws and it made a difference but I still need a -300MHz AVX offset to CPU fold and keep the VRM under 85C. It just barely passes prime95 small FFTs maxing out at 110C.

 

Yes, the CPU could probably go faster with an all-core over-clock but the MOSFETs would likely melt through the motherboard.

 

GigaByte to their credit has learned. My z570 Pro Wifi has a solid VRM design and even the z570 Gaming X I'd be willing to try a 3950x on.

 

I tend to nerd out watching BuildZoid's uTube videos and relive my youth when I studied electronics back when the lab computers at College were Commodore Pets.

[Xsilent(X)]

32 minutes ago, Gorgon said:

I know you've already spent the money but take a look at the build thread for the 6-GPU rig in my sig.

 

After a couple of years of folding I've come to these conclusions:

1. Two GPU rigs in a standard chassis are a bugger to manage the heat load on. If you must go this way buy a hybrid card for the upper card.
2. x299/399 just aren't worth it. The cost per slot is too high. A 6-GPU rig built in a mining chassis using a PCIe m.2 card and m.2 to PCIe3 x16 adapters is about the lowest cost-per-slot and you can run twin- or tri-axial cards more quietly in a mining frame than a chassis where you almost have to go to Delta level airflow to get decent thermals.
3. If your considering doing CPU as well as GPU compute spend the money up-front on fat processor. The 3900x is about the sweet spot. I recently did a cost analysis and the only thing that beats it are the 2000-series threadrippers if you can find them at a discount deep enough to offset the cost of a half-way decent x399 motherboard.

 

All of the GPU's i have in that mining rig are all AMD cards, I've looked at you're 6rig build and those all are nVidia cards, AMD cards do not run that heavy on the CPU but i will update once i have the CPU on my PPD development.

 

And the 3900x i ain't got the money for that.

[Gorgon]

25 minutes ago, Xsilent(X) said:

 

All of the GPU's i have in that mining rig are all AMD cards, I've looked at you're 6rig build and those all are nVidia cards, AMD cards do not run that heavy on the CPU but i will update once i have the CPU on my PPD development.

 

And the 3900x i ain't got the money for that.

Understandable but how about a lower-end x570 motherboard and a 1600AF? I still have my 1200 that I got for $99Cdn. It was a good deal at the time and was all I could afford then but enabled me to drop a 2700 in the rig later.

 

I'm looking forward to seeing your results and hope you do see a massive uplift in production. You'll be limited by the PCIe3x1 risers but with those lower end cards (RX480s?) it shouldn't be too bad. If it's still low and your running Windows then you might want to look at Linux. I've seen big differences in Windows 10's PCIe utilization compared to Linux. @Unilevers I believe had a similar setup and might be able to give you some advice.

[Inkertus]

1 minute ago, Gorgon said:

Understandable but how about a lower-end x570 motherboard and a 1600AF? I still have my 1200 that I got for $99Cdn. It was a good deal at the time and was all I could afford then but enabled me to drop a 2700 in the rig later.

 

I'm looking forward to seeing your results and hope you do see a massive uplift in production. You'll be limited by the PCIe3x1 risers but with those lower end cards (RX480s?) it shouldn't be too bad. If it's still low and your running Windows then you might want to look at Linux. I've seen big differences in Windows 10's PCIe utilization compared to Linux. @Unilevers I believe had a similar setup and might be able to give you some advice.

People like you are why I chime in with general advice I recall from earlier and then say "ask the pros for more" rather than make an ass out of myself with wrong specific advice. You know stuff. 

[Gorgon]

4 minutes ago, Inkertus said:

People like you are why I chime in with general advice I recall from earlier and then say "ask the pros for more" rather than make an ass out of myself with wrong specific advice. You know stuff. 

Thanks but I just try to learn from my mistakes. I have learnt a LOT

[Inkertus]

Just now, Gorgon said:

Thanks but I just try to learn from my mistakes. I have learnt a LOT

As have I from my tinkering and honed Google-Fu. So much random troubleshooting has made me more capable of fixing a PC than my car, and I had a beater Kia that broke a lot. Two things I've learned from hardware: Don't OC unless absolutely necessary and make sure your components are fully seated. Software is a massive pain to troubleshoot and 50/50 you can do a workaround fix. Windows updates occasionally blow but completely current drivers, BIOS, and chipset stuffs tends to prevent most issues IME. Linux is a pain in my ass to work with due to 1) nothing being where or how my habits are set up and 2) it's generally less friendly to do stuff with it (ex: command line to install non-store-found software vs just double-clicking a exe file and you're off, don't even get me started on GUI or not) though devs are finally figuring out how to make it idiot friendly, though preferably not fisher price friendly like Mac feels to me with how much they hide stuff (even Mac knows how to make it easy to use and install software vs many Linus distros and its UNIX-based).

[cbigfoot]

1 hour ago, Zberg said:

Man these fail to send WU's are killing me.  6 very chonky boys (250-300k each) in the last 2 days sitting in a send queue, down to 10% their original value, if they every go through.  

all my stuck units just went after cycling them try that...the 3.x.x.x uploads took 3-5 mins a peice

[Gorgon]

2 minutes ago, cbigfoot said:

all my stuck units just went after cycling them try that...the 3.x.x.x uploads took 3-5 mins a peice

Yeah, it seems the servers are out of Beta WUs or those servers are down as I've been stuck in pause/un-pause hell. The joy of those stalled downloads is if it is trying to send and gets a connection then stall then the client refuses to download any new work.

 

I had 10 or 12 WUs that had aged out and were sitting at the 28-30k points value I had to delete to get my slots un-stuck

[cbigfoot]

5 minutes ago, Gorgon said:

Yeah, it seems the servers are out of Beta WUs or those servers are down as I've been stuck in pause/un-pause hell. The joy of those stalled downloads is if it is trying to send and gets a connection then stall then the client refuses to download any new work.

 

I had 10 or 12 WUs that had aged out and were sitting at the 28-30k points value I had to delete to get my slots un-stuck

yeah the pentathlon is looking more and more necessary to allow F@H to catch up...i plan to move my nvidia and some CPU work(at Night) to the Pent...wont run the CPUs during the day due to heat...at least i say that now...things can change in the HEAT of the moment...

 

@jctappel67

You officially on my radar

[Zberg]

15 minutes ago, cbigfoot said:

all my stuck units just went after cycling them try that...the 3.x.x.x uploads took 3-5 mins a peice

by cycling do you mean just closing and reopening client or pause unpause?  Those are my only moves and have tried both (plus PC restart) a bunch but to no avail  

 

[cbigfoot]

1 minute ago, Zberg said:

by cycling do you mean just closing and reopening client or pause unpause?  Those are my only moves and have tried both (plus PC restart) a bunch but to no avail  

 

Pause unpause

[Unilevers]

39 minutes ago, Gorgon said:

Understandable but how about a lower-end x570 motherboard and a 1600AF? I still have my 1200 that I got for $99Cdn. It was a good deal at the time and was all I could afford then but enabled me to drop a 2700 in the rig later.

 

I'm looking forward to seeing your results and hope you do see a massive uplift in production. You'll be limited by the PCIe3x1 risers but with those lower end cards (RX480s?) it shouldn't be too bad. If it's still low and your running Windows then you might want to look at Linux. I've seen big differences in Windows 10's PCIe utilization compared to Linux. @Unilevers I believe had a similar setup and might be able to give you some advice.

rx580's on pcie1x risers are hit and miss depending on the WU.  One day with easy WU you might get 1.5-1.8milppd, then the next day all 6 cards full tilt all they get is 500kppd.  The 1x risers really hold back in windows, i dont know with the new core if linux is really going to be that much better but there are some PCI lane optimization that may help you out.  a rx580 8gb on a pci4x slot or higher hold a consistant 350kppd.

 

You can get a 3600 ryzen and a cheap end x570 motherboard and order the PCI splitters for cheap if you already have the rig with the PSU's.  Then in the future if you ever decide to switch it up to better cards your ok.  Really you gotta be 2080 Super or higher to max out a pci4x slot and at that point your pushing 2 mill PPD per card anyways and you could have 6 of those in a rig.

 

My sweet spot currently is a el cheapo x570 board that supports PCI bifurcation with a ryzen 3600, two 850 psu in a open air mining rig, 6 2070 supers.  currently 2070 supers are realistically one of the best buys in my opinions for price to PPD.

 

my opinion - sell the 580's because they are still good for 1080 gaming, use the funds to finance a ryzen 3600 and a x570 board and build from there.

[Inkertus]

If you need bandwidth for the cards on a riser, a surprisingly brief eBay search yields no x2 cables (must be a very uncommon config) but a few x4 to x16 cables or x16 to x16 cables no issue.  Even found a m.2 NVMe to x16 cable.  Neat!  Worth a look if you need the throughput.

[Gorgon]

3 hours ago, Inkertus said:

... Even found a m.2 NVMe to x16 cable.  Neat!  Worth a look if you need the throughput.

Combining 4 of those with a 4 NVMe PCIe3 x16 Card is what I used for my 6 GPU rig.

[Favebook]

18 hours ago, Xsilent(X) said:

No unfortunately i don't have a motherboard with at least 7 PCIE slot in to it to test. But i really think that I5 CPU what i bought will will do the trick of Tripling my PPD.

Do not expect triple PPD, especially if you are using x1 risers. 

[cbigfoot]

Well my 2060 KO came in...waiting on the asus to finish its WU before i rip it out...

[Bitter]

Well that was weird, FAH just quit working on my Fold1 Linux box. Seems to have happened around midnight as that was the last log entry I could pull through FAHControl or that was when the systme locked up, it was stuck at 99% done on the GPU WU, very weird. The PC was sitting with GPU fans running hard, had to use the hard reset button on the front of the case. Once I did that it booted right up as normal but FAHclient wasn't running and I couldn't get it to run. So I opened up software and it didn't even show it was installed...but if I opened the .deb I had downloaded it would let me uninstall it lol. So I uninstalled the deb, rebooted, saw it was gone, tried to reinstall from the deb and it wouldn't run, uninstalled, rebooted, installed it from the snap in software, and now it's all working again I think, but of course it grabbed a CPU WU IMMEDIATELY before I could do any configuration so I'll have to let the poor little Celeron chug away at that and finish, let the GPU finish, then reboot so I can get FAHControl working to view it remote again.

 

SO WEIRD. I wonder if there was some oddity with the month date changing hanging FAH or if it was a system hang issue?

[cbigfoot]

1 hour ago, cbigfoot said:

Well my 2060 KO came in...waiting on the asus to finish its WU before i rip it out...

Bad News Every one...the 2060 KO folds like a 2060 Vanilla...but that is ok its a tad faster than the asus

[Bitter]

48 minutes ago, cbigfoot said:

Bad News Every one...the 2060 KO folds like a 2060 Vanilla...but that is ok its a tad faster than the asus

Truly a bummer!

[Bitter]

3 hours ago, Bitter said:

Well that was weird, FAH just quit working on my Fold1 Linux box. Seems to have happened around midnight as that was the last log entry I could pull through FAHControl or that was when the systme locked up, it was stuck at 99% done on the GPU WU, very weird. The PC was sitting with GPU fans running hard, had to use the hard reset button on the front of the case. Once I did that it booted right up as normal but FAHclient wasn't running and I couldn't get it to run. So I opened up software and it didn't even show it was installed...but if I opened the .deb I had downloaded it would let me uninstall it lol. So I uninstalled the deb, rebooted, saw it was gone, tried to reinstall from the deb and it wouldn't run, uninstalled, rebooted, installed it from the snap in software, and now it's all working again I think, but of course it grabbed a CPU WU IMMEDIATELY before I could do any configuration so I'll have to let the poor little Celeron chug away at that and finish, let the GPU finish, then reboot so I can get FAHControl working to view it remote again.

 

SO WEIRD. I wonder if there was some oddity with the month date changing hanging FAH or if it was a system hang issue?

It did it again. This is so weird, it'd been running perfectly fine for a couple weeks at least, and then suddenly it's locking up and fahclient won't work right after I restart it. fahcontrol won't connect to fahclient and throws an error about the gpu slot or configuration, then I try to see if the fahclient service is running it says it is but cpu useage and gpu useage says it's not doing anything.

[cbigfoot]

2 hours ago, Bitter said:

Truly a bummer!

well so far its average has been stable so far 

it is bender slot 1 it has been averaging 112-116xxxx the whole time so far where the others fluctuate quite often 90-112 but tine will tell. lol hopefully it will get a job all my others have done so i can truly see.

well i say that and this happens

So @jctappel67 Watch out lol i think i got you next update then i need to figure out how to catch @LFreeman...

[jctappel67]

19 minutes ago, cbigfoot said:

So @jctappel67 Watch out lol i think i got you next update then i need to figure out how to catch @LFreeman...

Don't be so sure about 6 pm. Maybe 9 pm will be the one. 

 

It's unfortunate that I won't crack the top 300 first, but c'est la vie

[Xsilent(X)]

Update: I have paused 3 of the 7 clients and the PPD are stable at 1.6 Million ppd with the Celeron processor.

[cbigfoot]

9 minutes ago, Xsilent(X) said:

Update: I have paused 3 of the 7 clients and the PPD are stable at 1.6 Million ppd with the Celeron processor.

so all of those are RX580s? That seems pretty hight for Folding Rig slot 00 and 01 weird im saying that but would like my rx580s to be that good lol

[Xsilent(X)]

5 minutes ago, cbigfoot said:

so all of those are RX580s? That seems pretty hight for Folding Rig slot 00 and 01 weird im saying that but would like my rx580s to be that good lol

 

No the folding Rig contains 6 RX 470's and 1 RX 580 for video output.