TPU's WCG/BOINC Team

[jboydgolfer]

Reading through previous posts , it seems to be the case, but maybe it'll be different?

+1, my Ryzen 1700 doesn't like MIPS at all. gives about 45% output when compared to OET or OZ.

It must be an AMD thing though ,because earlier today I noticed one of these previous posts ,so I looked at my WCG manager ,and noticed that several of the MIPS work units were about to start ,but when I looked back it already completed 30 minutes later

 

[Arjai]

I have been hearing this, for a while now. MIPS runs more efficiently on INTEL chips and not so, rather slow, on AMD chips. I am assuming that since INTEL is using a similar design, my i3's are 2120's i5 is a 3917u and the future i7 is a 3770. But I digress, I am unsure whether older AMD cpu's have the same results. Could just be the new architecture in the Ryzen chips.

Sorry, "chip's" equals CPU in my archaic language!

 

[thebluebumblebee]

I think that the credit given for OET may have changed. One of my systems has had its PPD cut in half. 2600K, Linux, only running OET.

 

[jboydgolfer]

I think that the credit given for OET may have changed. One of my systems has had its PPD cut in half. 2600K, Linux, only running OET.

i dont run full 24 hour days, but i have noticed recently that my PPD has dropped considerably as well.

 

[mstenholm]

Heads up for this one - https://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,40734_offset,0
There is new very big OET WUs around. I deleted two in progress that had run 200 % longer than normal before I checked the forum.

I think that the credit given for OET may have changed. One of my systems has had its PPD cut in half. 2600K, Linux, only running OET.

My little 2520M gets around three times more than normal....Linux, mostly OET.

 

[Norton]

Heads up for this one - https://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,40734_offset,0
There is new very big OET WUs around. I deleted two in progress that had run 200 % longer than normal before I checked the forum.

Looks like the point totals are starting to get close to matching the effort now.

@HammerON - these wu's may have been what was affecting your ppd when you first started up your EPYC rig.. going to stick with them on my rig for a little while longer/re-evaluate over the weekend

 

[HammerON]

Might have been. I am wishing that I had went with a Threadripper CPU instead as they are clocked quite a bit higher.

One thing I noticed when I ran the CPUZ benchmark is the low single thread score:

I may end up dumping this build and starting fresh with a Threadripper....

 

[Norton]

I may end up dumping this build and starting fresh with a Threadripper....

iirc @XZero450 has a TR build underway, would be interesting to compare yours to his

I'll be sticking to my 1/2 TR builds for now- getting decent output from 1600/1700 chips atm

 

[T-Bob]

Looks like the point totals are starting to get close to matching the effort now.

@HammerON - these wu's may have been what was affecting your ppd when you first started up your EPYC rig.. going to stick with them on my rig for a little while longer/re-evaluate over the weekend

I noticed the same thing on both of my xeon 2P rigs, in sight, recently. All of a sudden they we're loaded with OETs that required 3 to 4x the normal time to compute.

 

[mstenholm]

I noticed the same thing on both of my xeon 2P rigs, in sight, recently. All of a sudden they we're loaded with OETs that required 3 to 4x the normal time to compute.

After 9 hours and 60 % completion (normally they takes 2 and some) they went 10 % completed and 3d remaining. They are now terminated. Have a look at my points -

 

[jboydgolfer]

I have OET's running into 7+ hours! Thats a long time Considering the HP the cores on this chip put out. Wow, that explains why my results dropped SO much

 

[HammerON]

I will check my two rigs when I get home....

One thing I noticed to day is that since I put switched the EPYC to Windows from Ubuntu, Free-DC is now showing the rig in place of the E5-2683 v3 build. Totally just took over all of the points accumulated by the Xeon cruncher. I did use that SSD with Windows 10 for a bit in the Xeon rig, but switched to Ubuntu quite a while ago.

 

[XZero450]

@Norton @HammerON I'm coming I'm coming... I have a handful of parts that have recently arrived and will soon arrive that should help me get much closer to completing it.

 

[HammerON]

Wish I could check the temps of the EPYC cruncher,,, Might have to go old school and buy a thermal probe. I remember installing one in of my first water cooling builds, probably around 2005...
Anyways, I am happy to see that the 6900K is running real cool with no overclock. For the last 1.5 years I have had it at 4.4 GHz or 4.2GHz while crunching.

I may boost her back up soon/\

 

[manofthem]

Wish I could check the temps of the EPYC cruncher,,, Might have to go old school and buy a thermal probe. I remember installing one in of my first water cooling builds, probably around 2005...
Anyways, I am happy to see that the 6900K is running real cool with no overclock. For the last 1.5 years I have had it at 4.4 GHz or 4.2GHz while crunching.

I may boost her back up soon/\

Incredible, those temps look very very low, crunching 100% in mid 30s?

 

[phill]

Looks like a decent CPU there @HammerON I would like to get my 5960X up and running.. 4.5Ghz @ 1.12vcore It's been a while since I've had it running
@manofthem, the temps are rather decent on these CPUs at stock under decent water cooling Also just depends on the load used as some 100% loads are more intense than others

 

[jboydgolfer]

these 8600 i5's are pretty decent Crunchers for 6 core no HT'ing. Im getting around 7,000ppd, or 56,000 (if i account for the 8 hours each day my PC isnt crunching) currently getting 5,200ppd at 15-18 hours /day.

 

[m&m's]

@HammerON your EPYC scores are intriguing me so I did a little of research and I found that only 3 hosts are reported with an EPYC 7401P CPU. Two of which are yours (Windows 10 and Linux) and the other one is getting over 30K PPD.

I looked at other EPYC hosts on stats.free-dc and their integer and floating points are on par with yours.
There is a 7551P getting 40K+ PPD and a 2x7281 getting 50K+ PPD. Seeing this reassures me that it isn't a bug with EPYC and BOINC/WCG.
There isn't much info to be had. Kernel numbers don't match, some are using older ones others aren't. Same goes for the BOINC version.

Your CPU-Z score seems normal too. If I'm not mistaken, when BOINC test the performance of the host it only load 1 thread. Could your CPU be throttling when under full load for a prolonged time? Or just a bad install?

Also I saw you mentioned you couldn't check your temperatures, Linux kernel 4.15 (and 4.16 released 2 days ago) officially supports temperature reporting for EPYC CPUs.

I don't know if it'll help you find the issue but I thought I'd still tell you what I found.

 

[HammerON]

Real Temp does not support my CPU so I tried Core Temp and this is what it shows. Seems about right to me...
I have not seen any throttling when checking it at random times. I noticed that the OET tasks were taking about 17 hours to complete versus the roughly 10 hours on the 6900k. So I changed the Device Profile again to exclude OET tasks and aborted most of the others.
Let us see what happens next. If this does not solve the low points output then I will do a fresh install of Windows 10. I also need to update the BIOS on the motherboard - keep forgetting to do that

 

[blindfitter]

i report on my Ryzen 1700x, @4ghz, 1.3volts, 205 watt with cpu temp 50C water cooling H100 crunching 24/7

 

[manofthem]

Just noticed that I got some HST wu's again, feels like forever. Noticed that they're longer, estimated about 7hrs run time on the 1600x. Does that sound right for HST? It's been so I've forgotten about them.

 

[jboydgolfer]

Just noticed that I got some HST wu's again, feels like forever. Noticed that they're longer, estimated about 7hrs run time on the 1600x. Does that sound right for HST? It's been so I've forgotten about them.

I just checked my results status ,and there's actually no help stop tuberculosis , seems I haven't seen any in quite sometime. I don't recall them being particularly long

 

[thebluebumblebee]

On my 3770K, W7:

 

[manofthem]

So the 1600x is down right now. I shut it down because I got my AM4 bracket for the Raystorm block, and I wanted to try to setup a loop with the leftover parts I had around. So far I've had several major issues, including a huge leak between the pump and pump top.

I complain about it below.

Spoiler

Now time for to complain about myself... Several years ago I first bought my d5 pump used off the forums, and was stoked when it came in to make my loop. Got everything running and about 10 minutes later I noticed water leaking out of the bottom of the case. I realized after braking it back down that I never checked if the o-ring was in the pump, and if course it wasn't thus creating a huge issue and down time.

Since then, I said I'd never have that issue again... But I lied. I top the stock top off this small pump (bought here used years ago from a member that used to sell these small mcp35x, hasn't been tested in years) and put on my old acrylic nn . But stupid me, I didn't make sure the o-ring was still there, and of course it wasn't. After filling up as much as I could I noticed leaking, and all due to the o-ring omission.

So has to break it down, drain some, take the pump apart to dry it out, and then reassemble. Feel like banging my head, like I'm at the DMV

Just sent the wife to grab some distilled water so hoping to get it running again tonight. Then again I think she has plans for us

On my 3770K, W7:

Sounds about right then, thank you!

I just checked my results status ,and there's actually no help stop tuberculosis , seems I haven't seen any in quite sometime. I don't recall them being particularly long

They have proven to be rare and difficult to get. I only have a few of them

 

[jboydgolfer]

While I'm awaiting my nephew to increase his school grades I've been intending to use my old set up 4790 to run BOINC, sadly it was overheating to tjmax, into the 90's on stock intel HSF I ordered a replacement but in the meantime I tried a different route and the results have been extremely impressive.

You can see how bad the TIM had dried under the IHS, replaced with LM, and dropped near 30c!!! .Now the proper heat sink that's on its way to my house will easily handle whatever my nephew throws at this computer, as well as whatever WCG can put it through.