TPU's WCG/BOINC Team

[phill]

Thank you @phill and @Norton ! I edited my post because I figured it out

Already finished a 2h testrun @3.9GHz to try and bother the VRM (it was not impressed - see TMPIN2):
View attachment 107289

Very nice work there @Dinnercore Please feel free to share if you'd like what the 1950X is pulling for wattage and such like, I'd definitely be interested to find out I look very much forward to seeing you in the leaderboard!

There are 74 people on this list. It is a list of people that started Crunching WCG, on the same day I did. There are only 4, including me, that are still Crunching...

Might take a while to get to the top spot but, 2nd place is coming into view!!


Although @phill passed me up, with his farm of servers, I am closing in at 69th (probably sometime tomorrow). @Sasqui mentioned, some time ago, he was coming for me!! That was before I bought the T 3500.

For the record, I bought another T 3500 and have a winning bid on another HP 8300 Ultra Slim. But this one has an i5 3470 (I think), and will be even more of a little monster. My Opti, w/ an i5 2400, is doing a 1000 more points a day than the i3 2120's. I expect the 3470 to be a little better!

Hopefully I can win the bids. I will know at @ 9:30 am CST.

I've noticed the same thing @Arjai that there's only 65 to 70 people crunching out of our entire team of 560+? Rather sad in some ways but if there's a way of getting a few older users or newer users crunching away, it would be awesome We'd be on the heels of those people over at XtremeSystems

Unfortunately the solar doesn't work at night otherwise I'd have the rigs on 24/7!

Please report back @Arjai about the power usage on your rigs and if you'd like to share with us in the thread here it would be amazing I try to leave things on for 2 days straight when I do do a power test or points test just so I hopefully get a fairly decent average.. That said, sometimes the luck of the draw with what points you can muster...
I'm very much excited to be looking to get the two Ryzen 1700X's up and running, I'm hoping that @Norton @thebluebumblebee and @blindfitter can help me there as well as anyone else can who might have a Ryzen I'd like to get them crunching as well... I do wonder if I could get the motherboards and coolers for Christmas, but I'm not so sure if that would be possible or a pie dream... We will see

Thanks to everyone who is supporting our team, it's such an awesome community I love it here

 

[Dinnercore]

Very nice work there @Dinnercore Please feel free to share if you'd like what the 1950X is pulling for wattage and such like, I'd definitely be interested to find out

Thank you! Well average measurement will take some time and I´m in the early stages of setting things up. I currently use the quick CPU-Z bench to estimate performance changes by clockspeed, then I monitor temps and power draw as measured by the die. This is faster to setup then crunch for some days straight and look at the point values.
My whole system will pull more power then needed, since the 20+ fans and 2 pumps for 2 water loops need juice too. In the end that might be a bit negated since the efficiency sweetspot of my PSU is between 300W and 400W, and CPU propably works better at 200W or lower. It also is more coolable at that point, tho my cooling system has 2x 420 rads in push pull dedicated to the CPU so I´m more bound by thermal transfer rate. Water temp stabilized at 3°C above ambient on the first test.

I also try to find the usage-% that allows me to still use the system for other tasks, right now I´m at 75% CPU time for 90% of threads and this allows me to continue playing games and watching netflix. I love the 1950x.

I can give you the following data right now:
1st test @ 3,9 GHz, 1.325 Vcore and medium LLC, 400kHz VRM switching freq. 60% CPU-time in BOINC = 186 W average draw CPU only. 9560 points CPU-Z, 65°C peak temp.

2nd test @ 3,725 GHz, 1.300 Vcore and medium LLC, 450kHz VRM switching freq. 75% CPU-time in BOINC = 200 W average draw CPU only. 9200 points CPU-Z, 56°C peak temp.

So between the two tests I lost ~5% compute performance, more then made up for it with 15% more compute cycles, increased power draw by 14W (7%) and dropped peak temps by nearly 10°C.
Next steps I will try to lower the Vcore further and ease off the LLC (needed that to stabilize the 3.9GHz on 1.325V, which was a decent result looking at the tomshardware-reviewsample with 1.35V for the same freq.).

 

[Arjai]

Well, I am the new owner of, another HP 8300 Ultra slim. Except this time, it is an i5 3rd gen. I am thinking it will be a 3450, that is the common CPU for HP's little monsters. Got it for under a hundred w/ keyboard and mouse.

If I remember correctly, this will be only the second EBAY auction I have won. The other was the 3770, that I have no home for, now.

Looks like Sunday, I will be shutting down for a spell. I need to re-arrange my farm rack. It is becoming a mess, with all the recent additions, and 2 more on the way! I have a plan, I have some WiFi dongles on the way and I have been eyeing some non PoE switch's. I need a cheap 10 output setup and I will be set. Meanwhile, the WiFi dongles will do the job and make the rearrangement a little simpler.

/rant



EDIT: After I rearrange the rack, I will use my new Kill-A-Watt to check things out. I have the K-A-W but, I am waiting on some 1 foot cords to show up. I find them uber useful with the Kill-A-Watt. So, when I get those I will start my data runs on power use. @phill

Edit ver.2: Oh, I am expecting the 2 new boxes to put out a minimum of 10 grand daily. The T3500 avg's North of 7 thou and the i5 2400 is just up above 4 thou. The new 22nm i5 should be better and the 2nd T3500 should not be any different.
Except, I am getting an MSI R7 240, on Monday, for the new T3500. I found one for a little less than the MSI GT 710 I have in the 1st one. I will also be getting those power cords, I mentioned, Monday.

 

[thebluebumblebee]

The other was the 3770, that I have no home for, now.

You *should* be able to swap it for the 3470. Same wattage and generation: https://ark.intel.com/compare/65719,68316

 

[Arjai]

You *should* be able to swap it for the 3470. Same wattage and generation: https://ark.intel.com/compare/65719,68316

You could be correct!! I looked at the HP site found a PDF that lists the 3470 and the 3770"s". But, since the 3470 and the 3770 are both 77W CPU's, I think it was just HP used the s's and not the full speed chip, because they could, not because they wouldn't work. I will DEFINITELY, tear into it, when I get it. If that works, AWESOME SAUCE!! If it don't, I still have a decent i5 working for me. Win-Win, in my book! Thanks for the info, I have been soooo tired I probably would not have even thought to check that out!! Thanks @thebluebumblebee



EDIT: I found out today, at work, I will be working Sunday and will have Wednesday off, instead. All because of some meetings my boss needs to attend, on Tuesday. I know, it doesn't make immediate sense but, it does in the convoluted work place I enjoy.

So, my Sunday plan, to re-arrange my shelves holding my Farm, is now moved to Wed. Unless I decide to come straight home from work and dig in...I may as well wait until Wednesday, since my GPU doesn't arrive until Monday...I can assemble the new T3500, 3 channel memory and the new gpu, by Wed. and then have everything set up at once. The new 8300, will have a place with the others. They will be stacked on the top shelf, with the monitor they all share.

I may pickup another cheap monitor for them, at some point. There's a pawn shop down the street that always has some discounted monitor's (I saw a Dell 24" 1080p there for $30! About a month, or two, ago!).

 

[phill]

@Arjai - Can we get some pictures when your done ??

 

[Dinnercore]

I may pickup another cheap monitor for them, at some point.

For cheap monitors, if you want to go really cheap you can find 4:3 LCD monitors basically thrown away in the dumpster nowadays. I got one from ebay actually for free, they gifted me 10€ coupon with no min. order limit, so I took that and bought some monitors that were up for instant buy at 2-3€. Yes those are power hungry, but if it´s for crunchers they don´t have to be on very often.

 

[stinger608]

Haven't been keeping up on the forums much lately, so I'm wondering if there has been any talk about an upcoming fall challenge?

 

[Norton]

Haven't been keeping up on the forums much lately, so I'm wondering if there has been any talk about an upcoming fall challenge?

Not yet, but will likely take a look at the second half of October and then the WCG Birthday challenge in November

 

[Dinnercore]

I can give you the following data right now:
1st test @ 3,9 GHz, 1.325 Vcore and medium LLC, 400kHz VRM switching freq. 60% CPU-time in BOINC = 186 W average draw CPU only. 9560 points CPU-Z, 65°C peak temp.

2nd test @ 3,725 GHz, 1.300 Vcore and medium LLC, 450kHz VRM switching freq. 75% CPU-time in BOINC = 200 W average draw CPU only. 9200 points CPU-Z, 56°C peak temp.

I´ve got my settings for the first long term testrun now.

Made a 3rd test with level 3 LLC instead of level 5, lowering the temp by 2°C to 54°C.

And final settings are:
1950X @ 3.725 GHz, 1.26875 Vcore, low LLC (level 3), 450kHz VRM freq. increased to 80% CPU-time = 193 W power draw (average over time and CPU only), and I opened my windows to lower ambient temps. Peak-temperature is now 47°C (that is cooler then my 1800x under a 280mm aio).

I will try and get a power meter for measurement of the full system @ load.

 

[stinger608]

Not yet, but will likely take a look at the second half of October and then the WCG Birthday challenge in November

That would work perfect cap'n

 

[hat]

Colder weather seems to have secured a decent hold over my area. In response to this, my i5-2400 is now running BOINC. The Athlon II x4 will likely soon join, but I gotta reinstall Windows on that computer first... it hasn't even been able to successfully do Windows Update for a while due to crappy partitioning and low disk space.

 

[phill]

Thank you! Well average measurement will take some time and I´m in the early stages of setting things up. I currently use the quick CPU-Z bench to estimate performance changes by clockspeed, then I monitor temps and power draw as measured by the die. This is faster to setup then crunch for some days straight and look at the point values.
My whole system will pull more power then needed, since the 20+ fans and 2 pumps for 2 water loops need juice too. In the end that might be a bit negated since the efficiency sweetspot of my PSU is between 300W and 400W, and CPU propably works better at 200W or lower. It also is more coolable at that point, tho my cooling system has 2x 420 rads in push pull dedicated to the CPU so I´m more bound by thermal transfer rate. Water temp stabilized at 3°C above ambient on the first test.

I also try to find the usage-% that allows me to still use the system for other tasks, right now I´m at 75% CPU time for 90% of threads and this allows me to continue playing games and watching netflix. I love the 1950x.

I can give you the following data right now:
1st test @ 3,9 GHz, 1.325 Vcore and medium LLC, 400kHz VRM switching freq. 60% CPU-time in BOINC = 186 W average draw CPU only. 9560 points CPU-Z, 65°C peak temp.

2nd test @ 3,725 GHz, 1.300 Vcore and medium LLC, 450kHz VRM switching freq. 75% CPU-time in BOINC = 200 W average draw CPU only. 9200 points CPU-Z, 56°C peak temp.

So between the two tests I lost ~5% compute performance, more then made up for it with 15% more compute cycles, increased power draw by 14W (7%) and dropped peak temps by nearly 10°C.
Next steps I will try to lower the Vcore further and ease off the LLC (needed that to stabilize the 3.9GHz on 1.325V, which was a decent result looking at the tomshardware-reviewsample with 1.35V for the same freq.).

I´ve got my settings for the first long term testrun now.

Made a 3rd test with level 3 LLC instead of level 5, lowering the temp by 2°C to 54°C.

And final settings are:
1950X @ 3.725 GHz, 1.26875 Vcore, low LLC (level 3), 450kHz VRM freq. increased to 80% CPU-time = 193 W power draw (average over time and CPU only), and I opened my windows to lower ambient temps. Peak-temperature is now 47°C (that is cooler then my 1800x under a 280mm aio).

I will try and get a power meter for measurement of the full system @ load.

View attachment 107359

I find with crunching @Dinnercore much like mining, it's not the point to have everything running flat out as possible but to have something ticking over is so much better not just for your pocket but for the hardware as well I do not believe that there's any reason to have it screaming away to try and grab those extra few points each day to have it dying a few months or so later.. That to me is just daft!
That said, most of my crunchers if not all but one runs at stock speeds, simply because the power requirements for some of the hardware I use is through the roof, so I don't believe there's any point in running it flat out and with 8 systems that could possibly run with a possibility of another 2 coming at some point, that's a big no no for my electric bill

Well, I am the new owner of, another HP 8300 Ultra slim. Except this time, it is an i5 3rd gen. I am thinking it will be a 3450, that is the common CPU for HP's little monsters. Got it for under a hundred w/ keyboard and mouse.

If I remember correctly, this will be only the second EBAY auction I have won. The other was the 3770, that I have no home for, now.

Looks like Sunday, I will be shutting down for a spell. I need to re-arrange my farm rack. It is becoming a mess, with all the recent additions, and 2 more on the way! I have a plan, I have some WiFi dongles on the way and I have been eyeing some non PoE switch's. I need a cheap 10 output setup and I will be set. Meanwhile, the WiFi dongles will do the job and make the rearrangement a little simpler.

/rant



EDIT: After I rearrange the rack, I will use my new Kill-A-Watt to check things out. I have the K-A-W but, I am waiting on some 1 foot cords to show up. I find them uber useful with the Kill-A-Watt. So, when I get those I will start my data runs on power use. @phill

Edit ver.2: Oh, I am expecting the 2 new boxes to put out a minimum of 10 grand daily. The T3500 avg's North of 7 thou and the i5 2400 is just up above 4 thou. The new 22nm i5 should be better and the 2nd T3500 should not be any different.
Except, I am getting an MSI R7 240, on Monday, for the new T3500. I found one for a little less than the MSI GT 710 I have in the 1st one. I will also be getting those power cords, I mentioned, Monday.

Great work @Arjai I look forward to seeing all about it when your ready and done with the setup and project I really do need to do some work with my gaming PC as that needs to get in a case which is sat on the floor at the moment... Might have to go spend some money on some tubing and get it setup and sorted out!!

 

[Boatvan]

Hi guys, been a minute. I have most of my crunchers shut down currently because I overdrafted my bank account on my last electric bill
Servers may not have been the best choice for efficiency as they chug along on 1200 Watt psu's (x2!). I'm hoping that as the weather cools off I can fire them back up as my A/C has been running a lot with this heat we've been having here in Ohio. This week is already in the 60's-70's F so it is looking promising. I hope to be back crunching along side you gents soon!

 

[phill]

Hi guys, been a minute. I have most of my crunchers shut down currently because I overdrafted my bank account on my last electric bill
Servers may not have been the best choice for efficiency as they chug along on 1200 Watt psu's (x2!). I'm hoping that as the weather cools off I can fire them back up as my A/C has been running a lot with this heat we've been having here in Ohio. This week is already in the 60's-70's F so it is looking promising. I hope to be back crunching along side you gents soon!

I feel your pain @Boatvan !! The R710 I have I find if I use both PSUs it actually uses more power than just one PSU but I can understand that down to the redundancy I do like them, but they are no very efficient when compared to the likes of Ryzen CPUs both in crunching power and wattage.. The difference is kinda scary!!

I hope it all works out soon

 

[Radical_Edward]

Sneaking back into crunching again. Been a while. Changed my line of work and had to take a break from some of my hobbies for a while.

 

[Dinnercore]

I find with crunching @Dinnercore much like mining, it's not the point to have everything running flat out as possible [...]

True! That´s why I stepped down from 3.9GHz. But I don´t really want to drop down to a perfect sweetspot at something like 3.2GHz or lower, because I still use it while crunching. And for fps in games like CS:GO it is helpful to have some more clocks per second. And I don´t think ~1.27V and 50°C will degrade the Threadripper any time soon. Atleast I hope so, don´t want to buy another now that prices went up again by 35%.

 

[Norton]

Here are some stats from a couple of members at XS who are running Ryzen/Zen based rigs (1950X, 1700/1700X/1800X, and even the 2990WX)

https://boincstats.com/en/stats/15/host/list/12/0/125242

https://boincstats.com/en/stats/15/host/list/12/0/342585

Pretty impressive ppd totals!

 

[hat]

Colder weather seems to have secured a decent hold over my area. In response to this, my i5-2400 is now running BOINC. The Athlon II x4 will likely soon join, but I gotta reinstall Windows on that computer first... it hasn't even been able to successfully do Windows Update for a while due to crappy partitioning and low disk space.

Made good on this today. Cleaned out both machines with my trusty shop-vac, and reinstalled Windows on the machine with the Athlon II x4. Due to lack of foresight when partitioning the small SSD, Windows was left without enough room to even do Windows Update...

There may be a dangerous experiment in store for that Athlon, though... I may try to delid it and run it direct die one day. It currently runs at 44c under 100% WCG load with a known good OC profile (3.5GHz) created long ago in the days of old... if I can experience a significant enough temp drop from running direct die, maybe it can take a bit of a higher clockspeed. It would be fun to see it hit that magical 4GHz mark.

I guess now I'll muck around with ePSXe a bit more. It used to work like a charm, but now it lags and stutters like crazy. Not sure what they did to it in recent updates...

-ed so the Athlon system likes to turn itself off now. Not sure why. Maybe the OC is suddenly unstable for some reason. I never had a problem before when it was just running Plex, but the system had virtually no load on it then, either...

-ed again bluescreenview shows nothing... so it's just powering off for some reason. What/where can I look for in Event Viewer? Of course, suspecting faulty power supply at this point, since unstable OC doesn't usually cause the system to just shut off...

-ed again, again... seems the system just shut down, again. Gonna try running it without BOINC for now and see if it stays up for a while.

Unless it's fucking going to sleep/hibernating. D'oh!

 

[phill]

Here are some stats from a couple of members at XS who are running Ryzen/Zen based rigs (1950X, 1700/1700X/1800X, and even the 2990WX)

https://boincstats.com/en/stats/15/host/list/12/0/125242

https://boincstats.com/en/stats/15/host/list/12/0/342585

Pretty impressive ppd totals!

The one thing I ask myself is, would the price of a 2990WX or any high core CPU be out done by lower core CPUs and more of them? A little like Mining in a way since if you have a system with 8 cards and they all stop, you're loosing out on 8 cards mining, but if you have a system of 4 or 6 and multiple systems, would that actually be more efficient for you to run?

I'm curious with running a few Ryzen 1700/2700 for example and one 2990WX CPU, I wonder if the cost to buy and keep it running, would it be out weighed by having two or three 1700/2700 systems? Would it perform as well, would it use as much electric and so on...

Penny for anyone's thoughts??

 

[Arjai]

Output wise, a 1700,2700 looks to be almost exactly half of the 16core TR. Power wise, I suppose would be close unless overclocking one or the other. I am not sure TR's are all that great at Crunching...someone here bought one and sold it, due to being underwhelmed by the crunching. So, I would guess, that the 8 core Ryzens would do better in the long run. A 1950x is a 180W chip. 1700 65W 1700x 95W 2700&2700x are both 95W.

I think 2 2700 would more than double the 1950x at the same Wattage, simply due to core speeds. Granted not by much of a margin but over time I am sure the separation would be significant.

Speculation on my part, but that is how I see the printed numbers.

 

[hat]

Looks like the Athlon II x4 system was indeed just going to sleep on me. Whoops... silly oversight, but it's not as bad as some sudden hardware fault at least.

 

[thebluebumblebee]

Penny for anyone's thoughts??

I'd better get a penny!

From @Norton 's links above, it appears that the 2990WX scales perfectly. Although we don't know what the setup is that is being used, averaging 100K is 4 times what I'm getting with my 1700, running only OET. The other 2990WX listed is at ~75K, which is what I think you'd get if you allowed all projects. The 250 watt TDP sounds huge until you realize its less than 4 x 65 watts of a 1700/2700, and it's running closer to the speeds of the 2700X. Guru3D showed the power usage of a 2700X system at 200 watts while the 2990WX was at ~365 watts running CineBench. So it looks like the 2700X builds would actually use more electricity, or ~800 watts for 4 of them.

On with the builds:

I'm not accounting for GPU's for either of these builds.

PCPartPicker part list: https://pcpartpicker.com/list/tQFkbX
Price breakdown by merchant: https://pcpartpicker.com/list/tQFkbX/by_merchant/

CPU: AMD - Threadripper 2990WX 3GHz 32-Core Processor ($1748.98 @ SuperBiiz)
CPU Cooler: Cooler Master ML360 RGB TR4 Edition ($199.99 @ Newegg) (had to manually insert that)
Motherboard: ASRock - X399 Taichi ATX TR4 Motherboard ($293.98 @ Newegg)
Memory: Corsair - Vengeance LPX 8GB (2 x 4GB) DDR4-3000 Memory ($93.99 @ Amazon)
Memory: Corsair - Vengeance LPX 8GB (2 x 4GB) DDR4-3000 Memory ($93.99 @ Amazon)
Storage: Kingston - A400 120GB 2.5" Solid State Drive ($24.99 @ Amazon)
Case: Fractal Design - Meshify C ATX Mid Tower Case ($90.98 @ Newegg)
Power Supply: EVGA - SuperNOVA G3 550W 80+ Gold Certified Fully-Modular ATX Power Supply ($69.88 @ OutletPC)
Total: $2568.78
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2018-09-24 13:00 EDT-0400

PCPartPicker part list: https://pcpartpicker.com/list/3j8ByX
Price breakdown by merchant: https://pcpartpicker.com/list/3j8ByX/by_merchant/

CPU: AMD - Ryzen 7 2700X 3.7GHz 8-Core Processor ($318.89 @ OutletPC)
CPU Cooler: Scythe - FUMA Rev.B 79.0 CFM CPU Cooler ($47.00)
Motherboard: ASRock - X370 KILLER SLI/ac ATX AM4 Motherboard ($79.99 @ Newegg)
Memory: Corsair - Vengeance LPX 8GB (2 x 4GB) DDR4-3000 Memory ($93.99 @ Amazon)
Storage: Kingston - A400 120GB 2.5" Solid State Drive ($24.99 @ Amazon)
Case: Fractal Design - Meshify C ATX Mid Tower Case ($90.98 @ Newegg)
Power Supply: EVGA - SuperNOVA G3 550W 80+ Gold Certified Fully-Modular ATX Power Supply ($69.88 @ OutletPC)
Total: $725.72 (4 needed or $2902.88)
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2018-09-24 13:10 EDT-0400

TL;DR: The 2990WX would cost less and use substantially less electricity! 4 x 2700X systems would cost $334 more and would use better than twice the electricity!

I am not sure TR's are all that great at Crunching...someone here bought one and sold it, due to being underwhelmed by the crunching.

That was an EPYC 7401P 24-Core 2.0 GHz (3.0 GHz Turbo) based server system.

2700&2700x are both 95W.

Actually, the 2700 uses 65 watts while the 2700X uses 105 watts.

Looks like the Athlon II x4 system was indeed just going to sleep on me. Whoops... silly oversight, but it's not as bad as some sudden hardware fault at least.

Just another reason to use Linux.

 

[hat]

Why, Linux doesn't go to sleep?

I could probably get away with Linux on this system if I wanted to... but I don't really want to mess with it.

 

[Norton]

Why, Linux doesn't go to sleep?

I could probably get away with Linux on this system if I wanted to... but I don't really want to mess with it.

Takes about 15 minutes to load up a hard drive with Linux and put it to work crunching

Even nicer is that once that drive is loaded up you can move it to another rig without changing anything