Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
if anyone was curious i'll be compiling a list of completed projects and its PPD on the 1070 SC. EVGA GTX 1070 SC 9160: 548,798 9162: 603,458 9205: 719,087 9205: 731,320 9212: 542,418 9442: 667,843 9661: 585,755 9704: 636,693 9704: 644,790 9704: 645,873 9704: 638,934 9704: 648,603 10494: 641,117 10494: 645,885 10496: 567,901 11401: 700,339 11418: 643,776 11419: 628,293 11419: 631,978 11423: 762,127 13131: 554,719 13131: 560,008 AVG PPD: 634,077 OK, first off my initial testing was off by nearly 10%, 569,252 PPD, because i found out that even though all my hardware was pcie 3.0 compliant the gpu was actually running pcie 2.0 until i forced it to run at 3.0 speeds. if you want a direct unit to unit comparison of pcie 2.0 vs pcie 3.0 look below. PCIE 3.0 9704: 636,693 9704: 644,790 9704: 645,873 9704: 638,934 9704: 648,603 Avg PPD: 642,978 PCIE 2.0 9704: 638,585 9704: 572,532 9704: 575,073 9704: 572,382 Avg PPD: 589,643 8.295% improvement with the 1070 going form pcie 2.0 to pcie 3.0. Now for comparison purposes the EVGA GTX 1080 (i assume FE) gets ~796,132 PPD AVG, Data here. With the 1070 avg being 634,077 PPD it is 26% lower than that of the 1080, link to data above. some other useful stats on the 1070 is that its PPD per Watt at 150W full load is 4,227 and its PPD per USD spent is 1,441. On the other hand the 1080 has 4,423 PPD/W and 1,244 PPD/$. with the 1080 achieving 26% higher PPD while only increasing its TDP by 17% to 180 Watts the 1080 wins in efficiency by 5%. BUT it looses in the PPD over cost department! With an increased cost of 48%, $650 vs $440, the 1070 is actually 18% higher in PPD per USD spent. so in conclusion, if efficiency is your goal and cost doesn't matter then the 1080 is the go to folding card right now. but if cost is of concern the 1070's are there to help you out and not break the bank.
post edited by Chris21010 - 2016/07/16 15:12:30
|
olddanit
New Member
- Total Posts : 78
- Reward points : 0
- Joined: 2012/05/22 07:52:14
- Location: Italy
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/16 06:54:10
(permalink)
Thank you, i am curious to compare your results with my 980ti's score (I'm not sure if there's a a reason to upgrade now to a 1080 or wait for the 1080TI), sadly I don't have any 9205 in my log so I'll wait for some other results.
post edited by olddanit - 2016/06/16 07:00:36
|
nathan_P
iCX Member
- Total Posts : 359
- Reward points : 0
- Joined: 2010/07/18 03:23:15
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/16 08:16:54
(permalink)
I've only got a result for a 970 on 9205 - tpf 6:15 - 333k ppd, looking forward to seeing more results - I'm not in the market for a 1080 at those prices
|
Re: GTX 1070 PPD
2016/06/16 12:36:52
(permalink)
Looks good. I've got my eye out for a 1070 FTW to add to my lone 970.
I wouldn't see much reason to upgrade a 980ti this go round. You can see a comparison on the anandtech bench results.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/16 14:43:32
(permalink)
alas, that 720K unit was a special one and well above the avg i have so far. i got unit 10475 twice so only 4 completed different units thus far. still though, ~600K PPD avg so far is quite nice. i'll keep this list updated for a couple days just so i can get a good avg set.
post edited by Chris21010 - 2016/06/16 15:34:08
|
Re: GTX 1070 PPD
2016/06/16 19:53:45
(permalink)
Hey, kids! Don't drink and post! You could end up posting in the entirely WRONG thread and not realize it until LONG MINUTES have passed! Don't be me! *laughs* Moving the original post to the right topic! d
post edited by drougnor - 2016/06/16 20:24:42
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/17 06:00:55
(permalink)
ok, woke up this morning to see i got unit 9704 three times, so the number posted is its avg like with 10475. also on a side note i noticed that my 1070 is only working at ~60% power and 70C but i am hitting the power/voltage limit consistently... this doesn't make any sense to me and i'll google around or post a topic about it after work. i dont see why im not getting near my power or temp thresholds of 103% and 81C respectively and still have the card throttle itself. so, these PPD's could be on the lower side if i can fix this and get the card to run like i think it should.
Attached Image(s)
|
mdk7777777
Superclocked Member
- Total Posts : 220
- Reward points : 0
- Joined: 2014/01/03 12:01:05
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/17 06:56:20
(permalink)
interesting.
smaller circuits mean higher density and faster heat up.
While the clocks that can be achieved intermittently / for short duration during gaming are obviously what is hyped...what can be achieved continuously for FOLDING may be entirely different.
I would start the experiment by setting the fan profile manually to 87% continuous to see the affect...
But it may well be a heat transfer problem (rate limited by scale and interface) rather than heat sink capacity(max BTU).
been expecting to see this problem crop up for some time. It will be interesting to see if water cooling is still sufficient, or if other solutions are required.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/17 09:30:57
(permalink)
But Temps are well below any thermal threshold and as the chart also shows it never triggers the thermal limit on the card. Also i have a manual fan curve on the card and its currently @84% when temps are ~70C. Something else is amiss here and i'll get to the bottom of it when I can research it properly later tonight.
post edited by Chris21010 - 2016/06/17 09:34:54
|
mdk7777777
Superclocked Member
- Total Posts : 220
- Reward points : 0
- Joined: 2014/01/03 12:01:05
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/17 09:40:00
(permalink)
Yup, many other possibilities, 1. limited because designed for limited continuous Voltage load(don't want to overheat VRM, exceed the 6 pin capacity ....etc. etc. etc.) 2. inefficiency, lack of optimization of open cl on new architecture(pipeline fills and empties instead of maintaining continuous load)memory latency etc. etc. 3. simply artificially limited to avoid too closely matching 1080 performance. Might not find a solution in one night.  But everyone will be interested in you research. 
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/17 15:35:45
(permalink)
OK, i got a video for you to watch and i think you'll be as perplexed as i currently am about this power issue. https://youtu.be/t8Iyp3BFF30 in short, if you want 100% GPU utilization you'll have to open a youtube video, pause it, and leave it there. otherwise it will drop right back down to ~60%. i did this test after doing the clean driver install again when i got home, just to double make sure no issue there, and i also removed my second monitor, just to eliminate another variable. EDIT: i think i'll run a couple units tonight after doing that workaround thing and see if and by how much the PPD may change.
post edited by Chris21010 - 2016/06/17 15:43:19
|
mdk7777777
Superclocked Member
- Total Posts : 220
- Reward points : 0
- Joined: 2014/01/03 12:01:05
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/17 15:53:53
(permalink)
OK,
You have an excellent demonstration of why many people are using LINUX for GPU FOLDING.
I'm not up on the details, but the problem seems to be with how windows allocates resources.
The more layers you go up from "bare metal" as they say, the greater your chance of loosing efficiency.
Gotta go now, see If I can find you a thread on the FF or slack later if I can find it.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/17 15:58:31
(permalink)
ya know if i wasn't a gamer i would try it on linux, but alas i must have my games. so i can not personally test this card on a lynix system unless i setup a dual boot for over night testing... not a bad idea, i'll give that a whirl and see if i can not get that to work tonight instead. any easy guide to setup F@H on linux? also which version of linux should i test with, Ubuntu? EDIT: nm, stanford has a good guide on their website, looks simple enough on that side of things. now i just need to remember how to setup dual boots... Edit2: this isn't as easy as I hoped... I got fahcontrol to run but it isn't seeing the GPU, so I look at what it takes to install a graphics driver and I face palm... why is nothing Linux touches simple or easy...
post edited by Chris21010 - 2016/06/17 18:11:39
|
mdk7777777
Superclocked Member
- Total Posts : 220
- Reward points : 0
- Joined: 2014/01/03 12:01:05
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/17 19:07:59
(permalink)
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/17 19:14:59
(permalink)
welp, it failed... dont know why or how but linux will not happen on my machine. 4 hours dealing with code to get stuff to work and failing isn't something i am keen to. So, i'll simply force the GPU to 100% and test a few units overnight and see if that workaround lasts for extended periods of time or not.
also that link is to another thread i made about a completely different issue. that walk through just shows how to add multiple GPU's to the FAHControl.
|
mdk7777777
Superclocked Member
- Total Posts : 220
- Reward points : 0
- Joined: 2014/01/03 12:01:05
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/17 19:20:11
(permalink)
Sorry, Like I said, I gave up on Linux years ago...just thought it might help with the drivers. It will be interesting to see if the work around actually speed up folding, or if the drivers are just in a loop...increasing the load for the shut down vid. time will tell.
|
kram36
Omnipotent Enthusiast
- Total Posts : 14519
- Reward points : 0
- Joined: 2009/10/27 19:00:58
- Location: United States
- Status: offline
- Ribbons : 46
Re: GTX 1070 PPD
2016/06/17 21:51:03
(permalink)
What's the fan noise like on this card? I have a EVGA GTX 970 ACX 2.0 and the fan noise is terrible. Have to fold it in a system in a closed room so I don't hear the fan noise. I have a MSI GTX 960 folding 24/7 in my front room and the fan noise isn't an issue. Wish I had not bought the EVGA GTX 970 ACX 2.0 and don't want to make the same mistake on another EVGA card.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/18 07:38:04
(permalink)
i can not answer that question as my PC sounds like a roaring dragon at all times and i have a box fan running as well.
secondly, the workaround did hold true all night though i only had time to finish the delayed unit and one other, that delayed unit i will not count but the other is listed with an asterix so we can see if it makes any difference.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/18 21:40:32
(permalink)
OK, ran another 9 units @100% power and the conclusion is at best a 10% improvement when directly comparing unit 9704 before and after. at worse only 5% improvement on the group avg if you remove the unusually high PPD unit 9205. SO, those numbers should be good enough to say that at best a 1070 may get you around 600K PPD, after a little OCing.
post edited by Chris21010 - 2016/06/19 08:26:23
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/19 09:02:57
(permalink)
ok, last update. went ahead and simplified the numbers on first post as the difference between was minuscule at best and posted the unit ID and PPD for every completed unit over the last 3 days. its final avg is 569,252 PPD which is 28% lower than that of the 1080 numbers linked to on the first post. some other useful stats on the 1070 is its PPD per Watt at 150W full load is 3,795 and its PPD per USD spent is 1,294. On the other hand the 1080 has 4,423 PPD/W and 1,244 PPD/$. achieving nearly 30% higher PPD while only increasing its TDP by 17% the 1080 wins in efficiency but its basically tied in PPD over cost department.
post edited by Chris21010 - 2016/06/19 09:12:58
|
zshadez
New Member
- Total Posts : 29
- Reward points : 0
- Joined: 2016/06/16 21:46:33
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/19 10:14:53
(permalink)
You seem to know what you're doing. I'm also running a 1070, and just installed last night. The log file in advanced control shows 4 work units completed over night and the point estimates after each one add up to around 290,000... web control shows 46,000 points earned for the team. Does the web control point total only update every so often, or is the point estimate on the advanced log that far off? And do you have any tips to maximize PPD other than OC?
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/19 10:28:31
(permalink)
the points estimated are the points granted for that one unit. you may complete 10 of these units in a day. so what you need to do to calculate your points per day (PPD) is know the time you started and finished that unit in hours. then take 24 and divide that by the time it took to finish that unit. this will tell you how many of those units you could do in a day. and finally multiply that number by the estimated points to get your PPD for that unit. example: log data 07:28:58:WU02:FS01:0x18:Completed 0 out of 5000000 steps (0%)
09:59:12:WU02:FS01:0x18:Completed 5000000 out of 5000000 steps (100%)
09:59:22:WU02:FS01:Sending unit results: id:02 state:SEND error:NO_ERROR project:10475 run:372 clone:0 gen:35 core:0x18 unit:0x00000024ab40415c567481a39d79b288
09:59:37:WU02:FS01:Final credit estimate, 56296.00 points
Start: 7:28:58
Stop: 9:59:12
Total run time: 2.504 Hours
Project: 10475
Credit: 32540
PPD = (24/2.504)*32540 = 539,602
also if you havent yet you must get yourself a passkey. if you dont your PPD will be gimped. you can do that by going to the configure->Identity tab and clicking link to the passkey request page.
post edited by Chris21010 - 2016/06/19 10:32:18
|
zshadez
New Member
- Total Posts : 29
- Reward points : 0
- Joined: 2016/06/16 21:46:33
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/19 11:23:43
(permalink)
Can you let me know if I'm doing this right. I'm using the advanced log to get start/end times and point estimates for each work unit:
10:25:41:WU00:FS01:0x21:Completed 5200 out of 520000 steps (1%) <- First entry for work unit
11:20:18:WU00:FS01:Cleaning up <- Last entry for work unit
11:20:18:WU00:FS01:Final credit estimate, 26150.00 points <- The final credit estimate for the unit.
It took 55 minutes, so .917 hours
PPD = (24/.917)*26150 = 684,405
Next one
11:23:43:WU01:FS01:0x21:Completed 50000 out of 5000000 steps (1%)
16:23:41:WU01:FS01:Cleaning up
16:23:41:WU01:FS01:Final credit estimate, 163188.00 points <- Cancer fighting karma
5 hours pretty much exactly.
PPD = (24/5)*163188 = 783,302
Is that how you're calculating, or am I using the wrong numbers? Sorry for the bother, I've just never tried to play with folding but interested in how the 1070 is doing.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/19 11:28:33
(permalink)
Looks good, though you need to pick the 0% start not 1% start. It shouldn't effect your numbers much but it's more accurate. Also I'm curious as to which unit gave you that 783k ppd. You can see it on the sending unit results line, just like I posted.
|
zshadez
New Member
- Total Posts : 29
- Reward points : 0
- Joined: 2016/06/16 21:46:33
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/19 11:34:38
(permalink)
Wow, hadn't even noticed the 0% lines since they're mixed in the the end of the previous WU. It didn't change the first one at all, but it dropped the second one. Thanks for helping me figure it out, had a few people asking about PPD results with the OC and had no clue.
10:25:08:WU00:FS01:0x21:Completed 0 out of 520000 steps (0%)
11:20:18:WU00:FS01:
Final credit estimate, 26150.00 points
It took 55 minutes, so .917 hours
PPD = (24/.917)*26150 = 684,405
Next one
11:20:15:WU01:FS01:0x21:Completed 0 out of 5000000 steps (0%)
16:23:41:WU01:FS01:Cleaning up
16:23:41:WU01:FS01:Final credit estimate, 163188.00 points <- Cancer fighting karma
5.05 hours
PPD = (24/5.05)*163188 = 775,546
|
zshadez
New Member
- Total Posts : 29
- Reward points : 0
- Joined: 2016/06/16 21:46:33
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/19 11:40:37
(permalink)
Chris21010
Also I'm curious as to which unit gave you that 783k ppd. You can see it on the sending unit results line, just like I posted.
16:23:12:WU01:FS01:Sending unit results: id:01 state:SEND error:NO_ERROR project:10493 run:2 clone:3 gen:116 core:0x21 unit:0x0000009d8ca304f555d6167046cf6660 No clue what it means, but that should be the one.
|
mdk7777777
Superclocked Member
- Total Posts : 220
- Reward points : 0
- Joined: 2014/01/03 12:01:05
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/19 11:46:22
(permalink)
Your ppd calculations are a tad high. Yes, that is the time to run the WU, but the download, startup, shutdown, and upload all take time which eats into your average.
Not trying to find any fault/critic... just letting you know that your long term actual daily average will be lower based on those logistics, and the WU variability that Chris21010 has listed.
thanks for reporting all.
|
Chris21010
FTW Member
- Total Posts : 1154
- Reward points : 0
- Joined: 2006/05/03 07:26:39
- Status: offline
- Ribbons : 6
Re: GTX 1070 PPD
2016/06/19 12:01:37
(permalink)
mdk7777777
Your ppd calculations are a tad high. Yes, that is the time to run the WU, but the download, startup, shutdown, and upload all take time which eats into your average.
This is true but not to the extent that you think as it downloads the next work unit before you finish it. And it starts that new unit once it starts to upload of the results of the old unit. Total down time between units isnt long, matter of seconds not minutes. Though the variance in work units is the biggest variable to your ppd.
|
Re: GTX 1070 PPD
2016/06/19 12:19:30
(permalink)
9158 x18 core = 471689 for a 980 ti (so its 30k points less then the 9157) 10489 = 493380 
post edited by toyking22 - 2016/06/19 12:22:15
Attached Image(s)
Cpu= I7-4790k 4.9Ghz Cpu cooler= Nzxt Kraken x61 GPU= Evga 980 Ti's (2way sli) 1450mhz all day Motherboard= Asus Maximums Vii Hero Z97 Ram= G.Skill TridentX Series 16GB (two 8 gig) Ssd/Hdd Samsung 850 pro 512 gig /Seagate 1T drive Monitor= ASUS PB287Q Black 28" 4K Power Supply= EVGA SuperNOVA 1200 P2
|
nathan_P
iCX Member
- Total Posts : 359
- Reward points : 0
- Joined: 2010/07/18 03:23:15
- Status: offline
- Ribbons : 0
Re: GTX 1070 PPD
2016/06/20 09:38:03
(permalink)
Are all these numbers under windows? Linux should add another 10-15% if its works properly
|
