WCG-TPU Cruncher's Hardware/Tech Support Discussion Thread

[PolRoger]

Edit: Humbled. I know I should always go back to the basics, but I didn't. I hadn't tightened a backplate screw down well enough, causing the the block to be a touch loose on the processor. Fixed that, new paste and temps are now in the 53c full load in eco mode.

I'm glad that you got it sorted out.

Depending on your use case for that setup you may find that a lower per CCX type of overclock with a lower voltage setting might give you better efficiency and thermals vs. native PBO and its higher single core boosting. I've put together a custom water cooling setup with a thin 240 rad that I've been testing and I'm planning on installing it in my Daily/SFF NCASE box at some point. Hopefully I'll be able to bring down the temps some as well as increase the clocks from where it is now currently running.

 

[phill]

So, I tried to add a 3900XT cruncher yesterday, but the temps were completely out of control, even with PBO. I've never owned a hotter chip at stock! I'll be working on getting the temps down over the weekend.

When I had mine on a 14D cooler, I was hitting 80 to 85C and wondering what the heck I was doing wrong. I was using Ryzen master to control the volts but then didn't even think to try and run it via the bios, so I cut the voltage down from auto or 1.35v to 0.9v and I think I ended up on 1.0v. Temps plummeted. I was able to run it about the 50 to 55C but granted what was being reported was not the correct clock speed. Thinking I was still hitting 4.20GHz, according to Task Manager I was sat about the 3.40GHz range which again, wasn't a bad thing for 247 WCG Crunching I didn't think with air.

Since then I put it under water, increased the vcore a little and updated my bios on my board, I now run 4.20GHz, at around 1.15vcore I think and temps are still in the 50 to 55C range depending on how warm a day it was... This is with a custom loop and a 560mm rad mind, but still. I did find out rather amusingly that I could run WCG with the same vcore etc. but with no fans turning and temps would increase only about 10C or so... Thought that was pretty impressive

It's amazing what simple and sometimes overlooked things can do when you try them....

 

[stinger608]

So, I've got a fairly new cruncher, thanks to our own @Solaris17 's generosity that's running a Ryzen 1700X chip. I also cut a deal on a 3700X chip but am having a hard time finding a B4xx or B5xx motherboard at a reasonable price.

My questions is this; how much better would the 3700X crunch over the current 1700X chip?

The motherboard is a Gigabyte X370 and supports the 3xxx processors, so until I find another board and grab some more DDR4, would the 3700X give me better crunching? If so, I'll pull the 1700 out until I find another motherboard at a reasonable price.

 

[thebluebumblebee]

how much better would the 3700X crunch over the current 1700X chip

AMD Ryzen 7 3700X Review

AMD's $330 Ryzen 7 3700X is an 8-core, 16-thread CPU that's clocked high enough to compete with Intel's offerings. Actually, its application performance matches even the more expensive Intel Core i9-9900K. Gaming performance has been increased significantly, too, thanks to the improved...

www.techpowerup.com

 

[stinger608]

Right @thebluebumblebee , I am just wondering if the crunching numbers are much better.

I guess with the much better performance than the 1700X it should result in better crunching numbers.

 

[thebluebumblebee]

Right @thebluebumblebee , I am just wondering if the crunching numbers are much better.

I guess with the much better performance than the 1700X it should result in better crunching numbers.

I would expect a 20+% increase. The only person that I know of who has both, who might be able to shed some light on this is @Norton .

 

[phill]

For a rough guesstimate @stinger608 mate, I'd probably saying nearly 20k a day wouldn't be too far off a good total Depending on a little towards the RAM spec/setup etc. (timings I mean) you might gain or get a few hundred or 1000 points possibly from that? I've never really been into testing that deep into it.

I think with the newer instructions and such, I'd say @thebluebumblebee is about right with his 20%+. Trouble is unless you run both identical systems together for a few days doing just one of the work units we can get, I guess we'll never know 100%. Some work units can really make a difference and even more so if one rig gets some GPU based units to do as well, that'll really boost the scores along...

The 3900X seems to do pretty well, I'm guesstimating between 30k and 35k depending on which way the wind is going I'm trying to work out what CPUs to put in to my two X370 boards, since I believe the 5 series CPUs aren't supported in the first gen Ryzen boards... It'll be lovely to replace the lot with 5950X's but as crazy as I am, I'm probably not going to do that Considered it mind!!

Also with the better crunching will be the better efficiency as well, I think the 1700X is a 105w CPU and the 3700X is a 65w CPU...

 

[80-watt Hamster]

How much does BOINC care about RAM speed? My memory divider reset itself, and I gained almost 10C dropping from 1400 to 1000, presumably since the memory controller isn't working so hard. The extra thermal headroom would potentially allow me another few hundred MHz. But will slower memory make that pointless?

 

[80-watt Hamster]

How much does BOINC care about RAM speed? My memory divider reset itself, and I gained almost 10C dropping from 1400 to 1000, presumably since the memory controller isn't working so hard. The extra thermal headroom would potentially allow me another few hundred MHz. But will slower memory make that pointless?

Anyone? Bueller...?

 

[PaulieG]

My default profile setting preferences won't change. I switched over to "minimum impact for a few days to see the difference in points from custom and standard. I've tried for 2 days in a row to switch back to standard, but every time I check, it's still at minimum. I understand it takes a little time, but 2 days?

 

[stinger608]

I don't think that BOINC or World Community Grid has much to do with memory speed while crunching bro. I think it's mainly the amount of ram and the processor more that the speed of the ram.

 

[phill]

Anyone? Bueller...?

I don't think that BOINC or World Community Grid has much to do with memory speed while crunching bro. I think it's mainly the amount of ram and the processor more that the speed of the ram.

Like with anything, if you have faster RAM it will help but nothing massive. I just set the standard RAM speeds for the setup and leave it be. If I could do a direct comparison I wouldn't say that there's masses in between them but if it's going to help with performance, then every little helps so to speak No point leaving RAM at 2133 speeds if it can run at 2667 or 3200 for example I'd say you'd notice some improvement but I wouldn't expect masses.

My default profile setting preferences won't change. I switched over to "minimum impact for a few days to see the difference in points from custom and standard. I've tried for 2 days in a row to switch back to standard, but every time I check, it's still at minimum. I understand it takes a little time, but 2 days?

From what little I know @PaulieG these changes should just work as soon as any of the clients connect to WCG for work upload/download... I'd check to make sure that you've not set up another profile and the PCs your using are on the other for example...
Is there something in particular that you are changing in the preferences that isn't working correctly?

 

[PaulieG]

From what little I know @PaulieG these changes should just work as soon as any of the clients connect to WCG for work upload/download... I'd check to make sure that you've not set up another profile and the PCs your using are on the other for example...
Is there something in particular that you are changing in the preferences that isn't working correctly?

All of my systems are linked as far as I can tell. I have now tried to change the default device profile from minimum impact to literally any other preference, standard, custom etc., and it just stays locked at minimum. WTF? Anyone with any idea what's going on or how to fix it?

 

[80-watt Hamster]

Like with anything, if you have faster RAM it will help but nothing massive. I just set the standard RAM speeds for the setup and leave it be. If I could do a direct comparison I wouldn't say that there's masses in between them but if it's going to help with performance, then every little helps so to speak No point leaving RAM at 2133 speeds if it can run at 2667 or 3200 for example I'd say you'd notice some improvement but I wouldn't expect masses.

That wasn't really my question. All else being equal, I'd up the memory speed and call it a day. But the processor is running cooler at the lower divider, so I might be able to run a higher overclock. I don't know a good way to benchmark WCG performance (like my n00b hat?), thus the question here. Put a different way, what's hypothetically more productive:

CPU@RAM
4.3@1400
4.5@1000

The 4.5 number is arbitrary; I don't know if I'd be able to hit that anyway. Probably should have tried first, but didn't want to put in the time if it's not worth it.

 

[phill]

That wasn't really my question. All else being equal, I'd up the memory speed and call it a day. But the processor is running cooler at the lower divider, so I might be able to run a higher overclock. I don't know a good way to benchmark WCG performance (like my n00b hat?), thus the question here. Put a different way, what's hypothetically more productive:

CPU@RAM
4.3@1400
4.5@1000

The 4.5 number is arbitrary; I don't know if I'd be able to hit that anyway. Probably should have tried first, but didn't want to put in the time if it's not worth it.

It's bound to run cooler on a lower divider since it's not straining the memory controller or anything as such. Like I said, I'd just set the RAM at whatever its stock 'XMP' speeds are and leave it

As for the CPU, just make it run cool so if your hitting 70C with it at 4.5 and 55C at 4.3, I'd just run at 4.3, the difference will probably not be worth the extra strain on the CPU... But again that depends on your personal preference.. Other than doing the in built CPU test, there's nothing much else, we all know a higher CPU/RAM etc. clock will mean it's faster but from experience if you want something to last a long time and run cool and quiet sort of thing, running the CPU slight slower is going to give you a much better return on how long it lasts Making it scream isn't always the best answer

Do a few CPU based tests and see if there's much of a difference between the two. I'd hazard a bet that the lower CPU clock and higher RAM clocks would make up for the higher CPU and lower RAM timings etc.

 

[PaulieG]

Is there a way to rename devices on the World Community Grid website device manager page?

 

[thebluebumblebee]

I am unaware of any way to change the name without changing the system/device name locally.

 

[Arjai]

This.

 

[PaulieG]

I am unaware of any way to change the name without changing the system/device name locally.

This.

I was hoping there was another option for a multitude of reasons. Oh well, if I must.

 

[PaulieG]

Most of my rigs are on Wifi simply due to lack of space where I have hardwire LAN. Is my workload negatively impacted by this?

 

[mstenholm]

Most of my rigs are on Wifi simply due to lack of space where I have hardwire LAN. Is my workload negatively impacted by this?

No.
Edit: the only thing that I can think will be impacted is the QRB (quick return bonus) but that is a folding@home issue were wifi can be slower and hence reduce the bonus.

 

[thebluebumblebee]

No.

You took the word right out of my mouth!

 

[PaulieG]

No.
Edit: the only thing that I can think will be impacted is the QRB (quick return bonus) but that is a folding@home issue were wifi can be slower and hence reduce the bonus.

You took the word right out of my mouth!

I'm sure this is kind of a rookie question, but before the last couple of months I haven't crunched in over 5 years and pretty much forgot everything I knew about WCG. I suppose grad school while working full time will do that to you.

 

[thebluebumblebee]

I'm sure this is kind of a rookie question, but before the last couple of months I haven't crunched in over 5 years and pretty much forgot everything I knew about WCG. I suppose grad school while working full time will do that to you.

It's a good question, I was just laughing that @mstenholm and I had the same one word answer. Uploads and downloads are independent of the work being performed on the CPU. Sure, a work unit could arrive late for today's count, but so what. Also, wireless speed doesn't matter. I know that @Norton uses https://www.amazon.com/gp/product/B...etailpage_o04_s00?ie=UTF8&psc=1&tag=tec06d-20 At least that's what he told me in 2017!

 

[Lorec]

It's a good question, I was just laughing that @mstenholm and I had the same one word answer. Uploads and downloads are independent of the work being performed on the CPU. Sure, a work unit could arrive late for today's count, but so what. Also, wireless speed doesn't matter. I know that @Norton uses https://www.amazon.com/gp/product/B...etailpage_o04_s00?ie=UTF8&psc=1&tag=tec06d-20 At least that's what he told me in 2017!

i use prety much same thing on my xeon and on my r7 1700!
xeon doesnt have any problems whatsoever, but my 1700 (which is running ubuntu 18.04) keeps on disconnecting wifi after couple days
Anyone know anything?