Page 1 of 1

GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Jan 21, 2017 12:06 pm
by JT3rd
Hello

I'm new to the forum and to folding, and am looking for some advise on troubleshooting PPD performance. I'm still working on tweaking these cards, and the GPU slot in my FAH are named incorrectly (reversed).

ASUS GTX 1050, averages 110-130K PPD with 64 watts at 100% power

ASUS Strix Gaming GTX 1050 Ti is averaging 140-170K PPD with 80 watts at 100% power (with 6-pin power connector).

The FAH does not appear to be utilizing the full power of the cards, with a 77% power limit when compared to a GPU stress test.

I still need to run more test, before I can post solid numbers.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Jan 21, 2017 6:36 pm
by foldy
Hi JT3rd, PPD looks good for these cards.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Jan 21, 2017 10:40 pm
by davidcoton
I hope that's kPPD (ie, thousands). Otherwise you have something seriously wrong.
With two different cards you may want to see if the reversal issue is fixed in the beta client (7.4.16).

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sun Jan 22, 2017 6:51 am
by JT3rd
foldy wrote:Hi JT3rd, PPD looks good for these cards.
davidcoton wrote:I hope that's kPPD (ie, thousands). Otherwise you have something seriously wrong.
With two different cards you may want to see if the reversal issue is fixed in the beta client (7.4.16).
Thank you for catching that, I've added the K.

I was able to get the slots correctly named with the corresponding card :)

The FAH is using 100% of the 1050, but only about 77% of the 1050 Ti, and no adjustment I make seems to change this. Between the two cards, I appear to peak at about 350K on Core 21.

Also using Driver 376.48, since I'm new to this, I'm not sure if there's some unrealized potential yet to be utilized.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sun Jan 22, 2017 9:27 am
by rwh202
JT3rd wrote:
foldy wrote:The FAH is using 100% of the 1050, but only about 77% of the 1050 Ti, and no adjustment I make seems to change this. Between the two cards, I appear to peak at about 350K on Core 21.
That doesn't seem unreasonable - Folding doesn't use all of the hardware present on the card (mostly just the 'shaders' and some use of the memory subsystem). Other benchmarks push all of the decoders, renderers etc. too. Folding can therefore hit other limits (temp, voltage, frequency) before power draw hits its limit.
The power limit is fairly arbitrary and is what the power supply can deliver and the cooling dissipate - a GTX 1050 with monster VRMs and cooling could be given a 150 W power limit but would likely still draw 70 W when folding having maxed out on voltage and frequency first.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sun Jan 22, 2017 3:55 pm
by JT3rd
rwh202 wrote:
JT3rd wrote:
foldy wrote:The FAH is using 100% of the 1050, but only about 77% of the 1050 Ti, and no adjustment I make seems to change this. Between the two cards, I appear to peak at about 350K on Core 21.
That doesn't seem unreasonable - Folding doesn't use all of the hardware present on the card (mostly just the 'shaders' and some use of the memory subsystem). Other benchmarks push all of the decoders, renderers etc. too. Folding can therefore hit other limits (temp, voltage, frequency) before power draw hits its limit.
The power limit is fairly arbitrary and is what the power supply can deliver and the cooling dissipate - a GTX 1050 with monster VRMs and cooling could be given a 150 W power limit but would likely still draw 70 W when folding having maxed out on voltage and frequency first.
Thanks for the explanation, that makes me happy knowing I don't have to go chasing my tail, and can spend more time on other issues.

Thus far under FAH load, the GTX 1050 draws 70 watts and the GTX 1050 Ti draws 72 watts. Based on the numbers I'm seeing, the Ti's points per watt will be about 12% higher.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sun Jan 22, 2017 9:17 pm
by JT3rd
Here's today's project stats.

Both GPUs are conservately overclocked

GTX 1050 / 1050 Ti
Watts 70 / 72 (3% higher)
Temp 75 / 55 (36% lower)
Core 21 / 21
Proj 9180 / 13500
Credit 26005 / 48472

Estimated PPD at % completion of project

10% 157856 / 186311
20% 160738 / 186266
30% 159139 / 184803
40% 155766 / 180911
50% 145303 / 181874
60% 144803 / 181840
70% 144861 / 180904
80% 139016 / 182635
90% 145809 / 180908
99% 145794 / 180116
Avg. 149909 / 182657 (18% higher)

PPD Watt 2142 vs. 2537 (16% higher)

Driver 276.48
OS Win 7 64
CPU FX-8350 3.8Ghz, 1.225 Volts, 88 watts at idle :(
Mem 2x 8Gb @ 1866

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 1:04 am
by v00d00
I bought and installed a Zotac GeForce GTX 1050 Ti OC today to replace an aging GTX 760. Very impressed so far. Estimates are about 190k per day for a card that is running at 1700MHz using ~68W. I only have 97-98% GPU Usage though.

Driver: 375.20
OS: Slackware 14.1 x64
CPU: FX-8320e @ Stock (4 Cores enabled, 1x FAHClient, 1x Xorg, 2 for everything else, bound via taskset, and the FAHClient core is isolated isolcpus=3)
Mem 4x 8GB @ 2133 (HyperX)

GPU stats - pulled from nvidia-smi

Code: Select all

 GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 105...  Off  | 0000:01:00.0     Off |                  N/A |
| 50%   52C    P0    68W /  75W |     69MiB /  4069MiB |     97%      Default |
GPU Clock: 1700MHz (which is far beyond what its supposed to be capped at in boost mode and OC mode, but im letting the driver deal with it, no coolbits enabled, no manual overclock, for now)

I do have a second card in here that is for general use, ie, I run day to day stuff on it. The 1050 Ti is dedicated just for folding. Will probably replace the 450 im using for everything else with another one of these Zotacs down the line.

Btw, and totally OT, did they do something to the DVI port on these cards? None of my VGA to DVI convertors will plug into the 1050's DVI port. Its made tonight a serious pain in the ass.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 2:11 am
by Joe_H
v00d00 wrote:Btw, and totally OT, did they do something to the DVI port on these cards? None of my VGA to DVI convertors will plug into the 1050's DVI port. Its made tonight a serious pain in the ass.
Looking at a picture of one, its port is a DL DVI-D type. That does not have the pin connections for an analog signal, so even if you could plug it in the adapter would not work. For a DVI to VGA adapter to work the port needs to be either DVI-I or the less common DVI-A.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 3:46 am
by v00d00
Cheers Joe.

Looks like I can work around it for now by using an older card for day to day, and keep the 1050 just for folding. Not going to buy a new screen just for that card, especially when my 970 runs happily on it. Might have to eventually go up to a 710 and run it off an x1, just something that doesnt generate a lot of heat.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 4:23 am
by Joe_H
If you still need to connect to a VGA display, you could replace your DVI to VGA adapter with either a Display to VGA or HDMI to VGA adapter.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 11:49 am
by JT3rd
v00d00 wrote:I bought and installed a Zotac GeForce GTX 1050 Ti OC today to replace an aging GTX 760. Very impressed so far. Estimates are about 190k per day for a card that is running at 1700MHz using ~68W. I only have 97-98% GPU Usage though.
Glad to see you are enjoying it, I'm still running the 1050 TI Strix OC version (with 6-pin power plug). I can get the core stable up to 1974, and the last 10 WUs average 190K. But in hindsite, the OC doesn't yield me much more than 5-15K at best and the watts needed for that push isn't cost effective. I'm currently running the 378.66 driver along with a standard GTX 1050, in a G4560 CPU, total watts 160 with 330PPD, so about 2000K per watt. The CPU only has 2 cores (with 4 threads) but is working very well for this dedicated rig.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 2:08 pm
by v00d00
Joe_H wrote:If you still need to connect to a VGA display, you could replace your DVI to VGA adapter with either a Display to VGA or HDMI to VGA adapter.
Looked at it, but those adapters cost almost as much as a 610 or 710. I still have 3-4 pcie x1 slots available on the motherboard, so it may be more effective to simply go up to a low end card with a vga port, which in turn frees up the second pcie x16 that the 450 is sat in, meaning another Zotac 1050 Ti. I reckon on this rig I can get away with a second one without to much issue, and the thing is generating very little heat compared to the 760.

I used to game on this machine, and use it for other things, but its rare now, since i have dedicated rig for those things that doesnt fold. This system is just web browsing, emails, office work and other serious things.
JT3rd wrote:Glad to see you are enjoying it, I'm still running the 1050 TI Strix OC version (with 6-pin power plug). I can get the core stable up to 1974, and the last 10 WUs average 190K. But in hindsite, the OC doesn't yield me much more than 5-15K at best and the watts needed for that push isn't cost effective. I'm currently running the 378.66 driver along with a standard GTX 1050, in a G4560 CPU, total watts 160 with 330PPD, so about 2000K per watt. The CPU only has 2 cores (with 4 threads) but is working very well for this dedicated rig.
I tried the 378.xx drivers, but something weird kept happening. I kept getting a core reload error and it appeared that fahclient was spawning a second core_21 causing a loop error of some sort. Tried a reload a fair few times and in the end downloaded every driver published that had 10xx support and supported older chipsets. The first one I tried was 375.20 and it worked brilliantly. PPD went from 75k to an estimated 200k (but that would even out to something more around 180-190k if I ran 24/7, which I dont). I can extrapolate PPD when ive run a few off.

According to the cards maximum clock numbers it will also go as high as your Strix, but I dont see it doing it on motherboard power alone. I think around 1700 is probably its limit. It would need the additional power to push beyond that point, unless its being limited by a thermal cutoff. At some point I will enable coolbits and see if it can go a little further, but for now Im happy at this level.

To think I almost bought one of those Palit StormX for £130, but couldnt live with something that used a single fan for cooling. This cost £18 more, has two fans but is rated annoying by some people because the minimum fan speed percentage is hardcoded at 45% (you cant adjust lower). But for someone who is going to use it at its limit, having a hardcoded fan speed is a useful feature. This card will never run at stock speed, it will always be maxed out.

I shall endeavour to put some numbers up when I have them.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sat Mar 11, 2017 4:09 pm
by Joe_H
I have seen listings for DisplayPort or HDMI to VGA adapters in the $10-20 range, so that is the only reason I suggested one. The expensive ones were DVI-dital to VGA converters.

Re: GTX 1050 vs. GTX 1050 Ti PPD performance

Posted: Sun Mar 12, 2017 1:02 am
by v00d00
Joe_H wrote:I have seen listings for DisplayPort or HDMI to VGA adapters in the $10-20 range, so that is the only reason I suggested one. The expensive ones were DVI-dital to VGA converters.
Im UK bound. But yes I could get one. Just dont see the point in adding another layer of emulation if you want to call it that.

Eventually I will upgrade my monitor to one of those 144Hz gaming ones. But I will need to replace the KVM when that occurs, as that uses vga.