TPU's WCG/BOINC Team

[librin.so.1]

Freedc is down so... somethings up.

Odd. It works fine for me.

 

[Jstn7477]

I will look for some other wattage numbers, bit-tech was the first I found.

I poked around through about 6 different 3770k reviews. Load wattage ranged from 112-222 with most having ~140. If I changed the wattage to the numbers you guys give me:

Using the ppd above, and these wattage numbers:
5.3k ppd/ 100 watts = 53 ppd/watt
5.75k ppd/ 150 watts = 38 ppd/watt

Roughly going by amazon prices I would say $300 for 3770k and $180 for 8350.
5.3k ppd/ $300 = 17 ppd/$
5.75k ppd/ $180 = 31 ppd/$

5.3k ppd/ 100 watts/$300 = 0.17 ppd/watt/$
5.75k ppd/ 150 watts/$180 = 0.21 ppd/watt/$

Okay, I decided to strip down my 3770K mATX rig temporarily to test with a Kill-A-Watt. At 4.3GHz/1.2v, the machine consumed 122w under 8 BOINC threads with the following hardware:

-ASRock Z77 Pro4-M
-Xigmatek Loki CPU cooler with stock fan
-80GB Seagate SATA HDD rated for +5v 0.72A and +12v 0.38A
-Sony 24x DVD burner (likely negligible power consumption at idle)
-Rosewill RG530 80% efficiency PSU
-No case fans aside from the PSU fan and the 92mm CPU fan were plugged in. If they were, they added 6 watts to the power consumption.

On Windows, this machine has a 5.65K PPD average on Windows 7 x64 using 87.5% of the CPU for BOINC. The remaining 12.5% is allocated for the GPUs (which were uninstalled for testing) to use for F@H. I momentarily allowed 8 threads to be used for testing. This processor is good for around 7K PPD average +/-500 PPD in Ubuntu (got that on my 2600K 4.4GHz last year when it wasn't running GPUs), but I cannot run it on my main machines because they house GPUs that require Windows to use for F@H purposes.

There may also be some test errata concerning what projects the machines are running and the point value of those projects. I have opted into all projects on my machines, which may not be the case for some other machines. Apparently from some posts not too long ago, there can be a significant PPD difference between projects.

Overall, the results are a toss up. AMD's inefficiencies are made up for the quite low initial cost of the equipment. Intel has superior efficiency and single-thread performance, but the latest i7 chips are $350 which is nearly double that of an FX chip. Also add into the fact that some people's energy costs per KWHR are double that of others, and comparing platforms further becomes troublesome. Just my two cents.

I've yet to see what a 4th generation i7 does with WCG, but I will gladly share my results with you guys next week when my equipment comes in and has been running long enough to generate a suitable average PPD.

 

[[Ion]]

Interesting power results--even with your system clocked 7.5% higher than mine that's a good 10-15w higher than I would have expected. Would you mind re-testing with a more efficient PSU?

 

[NastyHabits]

Freedc is down so... somethings up.

Down for me too. Can't ping it either.

 

[Norton]

Down for me too. Can't ping it either.

Their site has been dropping out randomly over the past few hours....

 

[xvi]

People over in the Bitcoin mining sites seem to worry a bit more about cost/gain. Their wiki has a pretty good hardware comparison, but results fluctuate depending on what program is used.

Of course, implementation can vary and there's no guarantee results will translate well from Bitcoin mining to WCG, but it might give a little insight as to how various cards stack up. For example, inside each generation (eg, cost/performance of a 7770 vs 7850).

("Mhash/J" being performance per watt, "Mhash/s/$" being, well, yeah.)

Also, those jerks are the ones buying up all the 5870s.

 

[Jstn7477]

Interesting power results--even with your system clocked 7.5% higher than mine that's a good 10-15w higher than I would have expected. Would you mind re-testing with a more efficient PSU?

I might be able to. Another thing to note is that the motherboard I tested with uses D2PAK power MOSFETs, which are the cheapest and least efficient variety. I'm pretty sure my other two LGA 1155 boards (Z77 Extreme6 and Z68 Extreme4) use better, low Rds(on) MOSFETs along with more robust circuitry, so maybe I will get around to testing my Z77 Extreme6 when it is ready to be pulled, since it also runs an 87% efficiency PSU. Just a better PSU alone should shave off 10-15 watts or so.

 

[librin.so.1]

If it slipped pass through Your radars - it appears GFAM entered Intermittent state today.

 

[Nordic]

If anyone was curious, a 5800k does about 2k ppd just cpu. The entire system using less than 200 watts which is cool.

 

[Arjai]

If it slipped pass through Your radars - it appears GFAM entered Intermittent state today.

I have five queued up and one ready to report.

 

[librin.so.1]

I have five queued up and one ready to report.

That's what? WUs? Pages of WUs? Tens of WUs?

 

[Peter1986C]

Two options: either the BOINC benchmark on Linux scores higher (this is why x64 is better than x86, mostly, as the benchmark scores higher) or there are things under the hood that are more efficient. I am surprised, however, as with a modern system idle load is basically 1% or less--which wouldn't be responsible for the 25%+ delta seen.

Do not forget that at the lower levels, the GNU C Compiler (a.k.a. GCC) is handling things a bit differently than whatever compiler has been used for the Windows BOINC clients. I am not someone with programming skills except some basics, so I cannot provide more details.

 

[laptop-hpc]

If it slipped pass through Your radars - it appears GFAM entered Intermittent state today.

I noticed it was getting close a few days ago.
Luckily I increased my cache in time, so I have about 3 pages in progress, and 7 WUs pending.

HPF2 is almost done too, maybe a week left to go.

 

[[Ion]]

I might be able to. Another thing to note is that the motherboard I tested with uses D2PAK power MOSFETs, which are the cheapest and least efficient variety. I'm pretty sure my other two LGA 1155 boards (Z77 Extreme6 and Z68 Extreme4) use better, low Rds(on) MOSFETs along with more robust circuitry, so maybe I will get around to testing my Z77 Extreme6 when it is ready to be pulled, since it also runs an 87% efficiency PSU. Just a better PSU alone should shave off 10-15 watts or so.

That would be great! I think that my Z77 UD4P has better MOSFETs--so that may explain it as well. And the different PSU is also certainly important.

Do not forget that at the lower levels, the GNU C Compiler (a.k.a. GCC) is handling things a bit differently than whatever compiler has been used for the Windows BOINC clients. I am not someone with programming skills except some basics, so I cannot provide more details.

You're right, another very realistic possibility. Either way, Linux is definitely doing something right

 

[Norton]

FreeDC may go dark???

It looks like FreeDC is having trouble again and the guy running the site is considering shutting it down

Server failed again
I'm looking,again it's the main drive that takes a beating with a ton of writes, but it really shouldn't have failed this quick. I'm getting very tired of this I have a spare drive, but this one was an Intel SSD and only lasted 5 months which is crazy. (Unless it's the mobo)

This can't really go on, I think I have 3 options: 1. Reduce the update frequency to around every 4 hours 2. Rewrite the scripts to use mysql master-slave replication which would require extra hardware 3. Give it all up

Please no donations whilst I consider all of this. If anyone has any ideas please post them at Free-DC forum

http://stats.free-dc.org/stats.php?page=team&proj=bwcg&team=22175&sort=today

That site is a great resource for all crunchers- things won't be the same without it....

 

[AnnCore]

It looks like FreeDC is having trouble again and the guy running the site is considering shutting it down


http://stats.free-dc.org/stats.php?page=team&proj=bwcg&team=22175&sort=today

That site is a great resource for all crunchers- things won't be the same without it....

I saw earlier, but if it's his SSD that is failing, I wonder why he doesn't use "normal" hard drives in some sort of raid configuration. Granted it isn't as fast, but they are proven.

 

[Norton]

I saw earlier, but if it's his SSD that is failing, I wonder why he doesn't use "normal" hard drives in some sort of raid configuration. Granted it isn't as fast, but they are proven.

I agree, but without knowing what type of system he's using or resources he has, it's hard to determine why he's having so many issues.

 

[AnnCore]

I agree, but without knowing what type of system he's using or resources he has, it's hard to determine why he's having so many issues.

From his forum:

Good ideas of course - this is all on linux and I certainly could use Ramdisks (or mysql memory tables). I would love to fit the stats tables into Ram but without removing host data and potentially freeing up some other types of data I would need at least 64Gb Ram and more likely 128Gb which would require some high end hardware.

I currently have 32Gb on the DB server which is the max for that particular motherboard.

Here is what I've posted in our private thread.

I have TWO separate databases on two distinct SSD drives.

All updates and calculations are done in the dcfree database to tables name like temp_boinc_user, temp_boinc_team etc etc.

Second database (stats) contains the web facing data with tables named boinc_user, boinc_team etc etc.

Once all updates are done in a single iteration, I copy the database 'files' from dcfree to stats (outside of mysql), so stats then contains both boinc_user and temp_boinc_user. I then drop boinc_user and rename temp_boinc_user-> boinc_user.

This operation is pretty much instantaneous so anyone doing an update on the webpages never tries to access a table that is locked for update.

Problem being though that I write 30Gb of data to the 'stats' drive 20 times per day. That on top of the reads just wears them down.

I could lower the frequency considerably.

I'd love to have the whole 'stats' database in memory personally, but would need 128Gb Ram likely be able to do that (totally new hardware) and it would take some rewrite. (Or remove the hosts part)

Probably other options, but I'm open to suggestions.

*EDIT* this one was an Intel SSD as well which are supposed to be the most robust. Lasted only 5 months.



So I guess he wants to keep it so the site doesn't have any downtime...

 

[Chicken Patty]

Really hope Free-DC doesn't shut down. Its by far my favorite stats website.

 

[NastyHabits]

Really hope Free-DC doesn't shut down. Its by far my favorite stats website.

It looks like he won't shut down. It is the THE place for checking on stats. Those of who can afford it should give a donation to Bok once he has a plan in place.

 

[TRWOV]

I think that keeping the db on RAM and committing to (hard) disk twice a day is the way to go. I'd be happy to donate the $25 I got in the last challenge.

 

[Jstn7477]

I agree, the DB commits are what is killing the drives. He either needs to get some SLC drives or go with a large HDD array for commits, and try to run as much as he can in RAM to prevent latency issues. If your SSD is pretty full and you write a lit to it, it will easily fail within 6-12 months, and newer flash with less write cycles isn't helping one bit.

 

[librin.so.1]

it makes me wanna cry seeing Free-DC at a state like this

 

[Bow]

 

[TRWOV]

I think I'll move an X6 to the office CCTV PC after all...

EDIT: