Low PPD on some WUs

[Bryman]

I think I will add a second fan to my GPU if it fits...

I overclocked my GPU to 1020mhz and the core only got up to about 90 but the VRM was hitting 110 which was causing the GPU to throttle...

I don't even know why it throttles at 110... it doesn't shut off until 135

Or I could just.. not overclock my GPU

Or I could turn the fan speed up... I have it running at 600 rpm right now

I guess re-applying the thermal paste would only help the core and not the VRM temp

I still had the voltage about 3% below stock... but I guess with my cooling it didn't work very well

I guess the speed was overclocked about 10% from stock... 7% from my factory overclock

So that would still be very roughly like 4% more heat than stock settings


I'm just going to try max speed on the fan for the heck of it... it's a little loud... well it says 22dBA but I'm really sensitive to noise lol... so it sounds loud to me... it's about the same as 60% fan speed on my video card


Started up a game and turned v-sync off with my video card clocked at 1060mhz with stock voltage(stock is 920mhz), core got up to about 83°C while the VRM temp 1 got up to about 104°C and VRM temp 2 got up to about 108°C

That's with the case fan at full speed


Is there a reason why with about 200 watts on the 12 volt rail the 12 volt and 3.3 volt report almost no voltage drop on my psu but my motherboard reports the 3.3 and 12 volt rail dropping a lot...

In fact the 3.3 actually increased to 3.313 at load and 3.297 at idle.... but my motherboard went down about 0.02 on the 3.3 volt rail... and about 0.15 on the 12 volt rail (it was reporting about 12.1 at idle) while my psu only went down about 0.02 on the 12 volt rail



If I run a stress test VRM temp 1 is 20-30°C higher than VRM temp 2... but if I leave the video card at a temperature for a while then VRM temp 2 is 3-4 degrees hotter than VRM temp 1...



I switched camera angles in the game and it raised the temp a couple degrees and caused my GPU to throttle...

Looks like if I want to overclock I'll need to add a second fan

I don't know why the heck they set the gpu to throttle if the VRM temp hits 110... they set the temp for it to shut off the video card at 135...

At 1020mhz the core temp drops 5 and the VRM temp drops 10... so I think that's what I'll keep it at, I'll still do 950 for folding

So really I can only get about 10% more speed with the fan being much louder... not sure it's worth it... even with headphones on

Even level 2 is a little annoying... level 9 is the highest and level 1 is the lowest

I'm just reaaaaaally sensitive to fan noise.. even just the smallest amount of noise bothers me

Also coil whine... coil whine is very painful to listen to... my video card makes it sometimes when starting up games...

OH GOD, it's like loud, VERY LOUD, when starting up Alan Wake

Like reaaaaally loud.. ear piercing.. okay I might be over exaggerating a little... but it was loud

https://www.youtube.com/watch?v=dlltomNrBrg

Huh.. that's weird.. it's not nearly as high pitched if I start up the game in windowed mode, which only cuts off a very small amount of the screen

I recorded it with a mic and normalized it so it's easier to hear

https://drive.google.com/file/d/0BwSqPT ... FFSk0/view

Honostly the version I recorded with my mic sounds less like it does irl than the iPhone version... for some reason the one I recorded with my iPhone sounds more like it does irl

If I normalize the audio in the video I recorded...

https://drive.google.com/file/d/0BwSqPT ... JsbUE/view

Sounds extremely painful

[Bryman]

The temps were going up pretty fast but not insanely fast near the beginning and middle... but near the end OCCT said something about my gpu hitting 95 and I had to refocus the window, then the temps started skyrocketing

https://www.youtube.com/watch?v=8VUL0sKSEto

The VRM temp especially exploded


I guess my video card was reporting about 400 watts going through it near the end...

The wattage increases as the temp increases

VRM temp would of EASILY hit 200°C... probably closer to 250°C

Maybe even higher than that...


That's interesting... HWiNFO reported the PCIe #1 at 36 amps... which is A LOT more accurate compared to what Corsair Link was reporting (it was reporting about 40 amps)

That would be about 430 watts, and there's about a 20 watt drop or so due to voltage drop to the video card (which was reporting about 400 watts)

Or maybe it's because I was running folding@home and stress testing at the same time... I'll test it later

[Bryman]

Ah, I see the problem... it caps the wattage for the core at about 370

Even as the amperage of the PCIe #1 was going up, the wattage was still not going above 370

So really when the amperage reports 40 on the psu, the wattage on the video card is about 440 (410 watts for the core + 30 watts for the memory)

https://www.youtube.com/watch?v=nwESnY2d21o

PCIe #1 got up to 38 amps, under optimal conditions I can get it up to 40, which is right at the OCP limit

[Bryman]

So.... good news... I can now underclock with folding@home

Bad news: for some reason my video card is only underclocking to 500mhz core 1425mhz at idle

I tried shutting off my computer and turning it back on... did not fix the problem


I guess this is good though... I can underclock with folding@home now



I'm not exactly sure how this happened... I ran folding@home and a game at the same time for a few minutes to test something and then I noticed my GPU was throttling when the VRM temp hit 110 (which folding@home normally prevents that)

So then I stopped the game and folding@home and my video card only underclocked down to 500mhz core 1425mhz memory...

[Mstenholm]

Time to change the title to something like - My folding experience - a monolog

[Bryman]

Lol, sorry

I'll stop posting stuff on this topic that isn't related to the PPD after this post

Ughhhhh, my video card is... and the drivers... so glitchy

I even added a second fan to cool the VRMs and it lowered the core temp about 5 and the VRM temp like just 5 too...

*sigh*


VRMs easily get up to 100°C if I downvolt my video card 10% from stock... core only gets up to about 80

I have the fans on a low rpm, but still.. I think I about matched the second fan speed with the first fan

And if I downvolt my video card by 5% and raise the clock speed about 8% from stock, then the core gets up to about 85 while the VRM gets up to 110... which... is when it throttles... it doesn't shut off the video card until 135 but it throttles at 110, apparently... don't know why they'd set the thermal shut off so high... I mean, 135 for VRM temp is almost like 110 for core temp

My core is a lot less gap... at 97 it forces fans to 100%, at 98 it throttles, at 104 it shuts off the video card


Oh.. and my video card doesn't have a power limit... I was able to get my 7970 to use 450 watts.... *sigh*

Video card only lets me underclock the memory 150 below the core clock... otherwise it'll set it to 1425 ... Facebook makes the video card go to factory clock settings... folding@home disables underclocking (at least it used to)

[Bryman]

Code: Select all

20:03:32:WU01:FS00:0xa4:Completed 12500 out of 1250000 steps  (1%)
20:07:53:WU01:FS00:0xa4:Completed 25000 out of 1250000 steps  (2%)
20:09:14:WU02:FS01:0x21:Completed 150000 out of 5000000 steps (3%)
20:09:57:WU01:FS00:0xa4:mdrun returned 255
20:09:57:WU01:FS00:0xa4:Going to send back what have done -- stepsTotalG=1250000
20:09:57:WU01:FS00:0xa4:Work fraction=0.0247 steps=1250000.
20:10:01:WU01:FS00:0xa4:logfile size=0 infoLength=0 edr=0 trr=25
20:10:01:WU01:FS00:0xa4:logfile size: 0 info=0 bed=0 hdr=25
20:10:01:WU01:FS00:0xa4:- Writing 641 bytes of core data to disk...
20:10:01:WARNING:WU01:FS00:FahCore returned: UNSTABLE_MACHINE (122 = 0x7a)
20:10:01:WU01:FS00:Sending unit results: id:01 state:SEND error:FAULTY project:11625 run:1 clone:121 gen:7 core:0xa4 unit:0x000000094c71bbb0569e7db0cdf02a49
20:10:01:WU01:FS00:Uploading 659B to 128.252.203.2
20:10:01:WU01:FS00:Connecting to 128.252.203.2:8080
20:10:01:WU01:FS00:Upload complete
20:10:01:WU01:FS00:Server responded WORK_ACK (400)
20:10:01:WU01:FS00:Cleaning up

:/

Took it about 40 WUs before an error happened, though

[bruce]

project:11625 run:1 clone:121 gen:7 has been returned once (yours) and has been reassigned. We can't prove one-way-or-another whether it's a problem with your hardware or with the WU itself until others have the opportunity to complete the same WU.

[Bryman]

My processor was at 4.4ghz, at 4.6ghz Gromacs couldn't even start

I'm guessing it's my hardware

I'll just turn my clock speed down to 4.3... will still be plenty fast for games

Also my processor is running about 85°C at 80% cpu usage because I don't like fan noise so I have my fans turned pretty much all the way down... which is fine, but at 4.3ghz it would be 3-4°C cooler

Pretty sure my voltage is high enough.. haven't gotten any WHEA-loggers within about a week



Just realized I restarted my computer yesterday I think... and I think I might of changed the voltage while the computer was booted so it didn't permanently change it... so my voltage might of been changed when I restarted my computer



I'll just keep it at 4.2ghz while folding and if I want to game I'll overclock it to 4.5 or 4.6ghz

At 4.7 I get memory reference errors though, so 4.6 is probably the highest I should go

It's really easy, I can change the clock speed and voltages without having to restart the computer



Temps dropped 8°C just by going from 4.4 to 4.2ghz... heh

So instead of 80-85 it's about 75

Which is MUCH better



Tried 4.9ghz at 1.575 volts... ran for a few min and nothing crashed, before when I tried I got a memory reference error within seconds... maybe I just didn't have enough voltage

Then I tried 5ghz at 1.6 volts and my computer restarted itself after about 15 seconds

Only ran at about 20-30% cpu usage though... it was like 80-90°C even at that amount lol... even had the fans turned up



I guess if it's stable with games it's not necessarily stable with folding@home though... I'm kind wondering if Gromacs couldn't start at 4.6ghz because of the clock speed, or because the voltage wasn't high enough...

[bcavnaugh]

Well... PPD is back down to ~15,000

The one I'm folding right now only has a base credit of 2,700... the one I did a few days ago had a base credit of 10,400

@Bryman
I would recommended that you change your CPU to use only 6 Threads an Even Number and not Fold on Odd Number of Threads.
I also noted that HFM.NET 0.9.5.478 has download a lot of changes over the last few weeks and Point changes are some that I remember seeing.

[bruce]

The one I'm folding right now only has a base credit of 2,700... the one I did a few days ago had a base credit of 10,400

That's what I was trying to tell you earlier. WUs from different projects are absolutely NOT the same.

Suppose the first WU can be completed in 2.7 hours (or days) and the second WU can be completed on identical hardware in 10.4 hours (or days). The base credits SHOULD be different in order to approximate a constant PPD.

It's not possible to have a constant PPD since when you change hardware configuration to one which is, say 10% faster, that new hardware might be 5% faster on one type of computation and 15% faster on another, which eventually average out to being 10% faster.

Stanford does attempt to establish base credits which are fair for all types of hardware, but it's an impossible task so they have to settle for something approaching a reasonable approximation.

[Bryman]

Well... PPD is back down to ~15,000

The one I'm folding right now only has a base credit of 2,700... the one I did a few days ago had a base credit of 10,400

@Bryman
I would recommended that you change your CPU to use only 6 Threads an Even Number and not Fold on Odd Number of Threads.
I also noted that HFM.NET 0.9.5.478 has download a lot of changes over the last few weeks and Point changes are some that I remember seeing.

Yes, I've been folding with 6 threads for a while now... works much better and there's not a big loss in PPD

[Bryman]

15:54:48:WU00:FS00:0xa4:Gromacs cannot continue further.
15:54:48:WU00:FS00:0xa4:Going to send back what have done -- stepsTotalG=500000
15:54:48:WU00:FS00:0xa4:Work fraction=0.3967 steps=500000.
15:54:55:WU00:FS00:0xa4:logfile size=13744 infoLength=13744 edr=0 trr=23
15:54:55:WU00:FS00:0xa4:logfile size: 13744 info=13744 bed=0 hdr=23
15:54:55:WU00:FS00:0xa4:- Writing 14280 bytes of core data to disk...
15:54:55:WARNING:WU00:FS00:FahCore returned an unknown error code which probably indicates that it crashed
15:54:55:WARNING:WU00:FS00:FahCore returned: UNKNOWN_ENUM (-1073741819 = 0xc0000005)

Processor was only clocked to 4.2ghz and voltage should of been high enough

Last time this happened my processor was clocked to 4.4ghz and even then it took about 40 WUs for it to happen, so 4.2ghz should be perfectly fine

[foldy]

You don't want to run your CPU unstable, this will cause errors in any APP once in a while.
And can even corrupt your file system if wrong values are saved.

You need to test your CPU overclock if it is stable.
IntelBurnTest worked best for me. Enter 10000 times and stop after 1 hour. Your CPU is stable if no error occured.

[Bryman]

You don't want to run your CPU unstable, this will cause errors in any APP once in a while.
And can even corrupt your file system if wrong values are saved.

You need to test your CPU overclock if it is stable.
IntelBurnTest worked best for me. Enter 10000 times and stop after 1 hour. Your CPU is stable if no error occured.

I'll try that later



I might clock to 4ghz anyway.. I'm not sure why I had it clocked to 4.2.. I can raise the speed pretty easily without having to restart my computer if I want to play a game or something


I'm gonna do the test just for the heck of it though