Audio UPD.3: REVIEW of Mini GTX 1070 Cards (NFC S4 case)

[Broxin]

Update 1: Review finished
Update 2: Supersampling resolutions added
Update 3: Audio Supercut of Loudness added



FINALLY THE REVIEW IS HERE

READ BELOW

Hey there,

soon (this Weekend maybe) i will test three (four) mini 1070 cards in a S4 mini case.

The Cards:
ZOTAC GeForce GTX 1070 Mini 8GB (ZT-P10700G)
GIGABYTE GeForce GTX 1070 Mini ITX OC 8GB (GV-N1070IXOC-8GD)
KFA2 (Galax) GeForce GTX 1070 OC Mini (70NSH6DVO5MK)
MSI GeForce GTX 1070 Aero ITX OC

i want to measure in idle and under Load

the GPU temperature
the power draw
the noise
coil whine

Any other things i should test for as i have the cards for testing ?

any additional stuff i should look for ?

OK, Here it goes:

first of all, NONE of these cards is going to fit with the HDPLEX DELL Power connector interface originaly
as delivered by HDPLEX.
The only card that will work temporaly with turned around and cutted Power conector is the msi.
All can be seen on Photos below.

System:

• Case: NFC - S4 mini (SN:144)

• MB: Asus ROG Strix Z270I Gaming Mini ITX
• CPU: i7-6700K
• SSD: 250GB Samsung 960 Evo M.2
• RAM: Patriot D4 16GB 3000-16 Viper4 XMP2.0 1.35 K2 PAT | PVE416G300C6KBL
• PSU: HDPLEX 300 (Coils glued with epoxy to succesfuly eliminate coil whine)
• POWER BRICK: Dell XM3C3 340 Watt

Power draw:
Idle intel GPU, idle CPU: 25.5W
Max Furmark load @ intel GPU: 60W
Max Furmark CPU Burner load @ intel CPU: 127W
Max Furmark load @ intel GPU + CPU: 160W





I measured the following:

Looking @ any coilwhine that may occur

• POWER DRAW: idle system with GPU / without GPU (iGPU)
• POWER DRAW: CPU (Furmark CPU burner)
• POWER DRAW: GPU (Furmark GPU burner)
  
• POWER DRAW: CPU + GPU (Furmark CPU+GPU burner (+GPU-Z interface load))

Looking @ the Boostclock that the cards are boosting to in the Benchmarks

• BENCHMARK: Furmark FPS (min. avg. max.)
• BENCHMARK: Unigine Valley
• BENCHMARK: Unigine Heaven
• BENCHMARK: 3D Mark Timespy

Furmark full load:

• GPU TEMP/FAN LOAD/FAN RPM = NOISE: Lying
• GPU TEMP/FAN LOAD/FAN RPM = NOISE: Lying without bezel
• GPU TEMP/FAN LOAD/FAN RPM = NOISE: Standing
• GPU TEMP/FAN LOAD/FAN RPM = NOISE: Standing without bezel

Allways looking at any signs of throttling in GPU-Z Sensor view


Sound Level Meter

CHAUVIN ARNOUX C.A 832
http://de.rs-online.com/web/p/schallpegelmessgerate/4683895/
​

Measured in a Noisefree room (28dB on the SoundLevelMeter) approx. 50cm away from GPU
Standing/Lying positions marked on floor for precise repetition of test

Also included but not ready yet:

Recordings of the spinup of the Fan, from idle to working RPM, under heavy Futuremark stressing.
​

Recording Mic

Calibrated Behringer ECM8000 microphone
http://www.cross-spectrum.com/

------------------------------------------------------------------------------------------------------------------------------------------





On the Third Place:
--- Zotac

On the second comes:
-- Gigabyte

And the winner is...

!!! msi !!!

------------------------------------------------------------------------------------------------------------------------------------------



We begin with


ZOTAC GeForce GTX 1070 Mini 8GB (ZT-P10700G)



Pros:
+Cools down very fast ~30sec
+More quiet than gigabyte but: -bad resonance problems
+Least coil whine

Cons:
-Because of two fans, there are resonance problems on different RPMs (More in the Audio recordings to come)
-Makes weird noises (like "Breathing") at low rpms (a disqualifier for me)
-Has "ripple" rpm's (also at start), eg, you can hear the PWM controlling the fan
-Has to be modified /cutted to fit in the S4 mini with stock bezel
or has to use the 3D bezel
-Does not fit with the HDPLEX Power Connector



Some Data:
Stock Clock / Boost Clock / Mem Clock: 1519 / 1709 / 8008MHz

• POWER DRAW: 39W idle system with GPU
• POWER DRAW: 239W GPU (Furmark GPU burner)

Looking @ the Boostclock that the cards are boosting to in the Benchmarks

• BENCHMARK: 114. 128. 132. (min. avg. max. FPS) - 1518-1645MHz Boost - Furmark
  
• BENCHMARK: 3856 Points - Unigine Valley
• BENCHMARK: 4741 Points - 1759-1822MHz Boost - Unigine Heaven
• BENCHMARK: 5648 Points - 3D Mark Timespy

Furmark full load:

Forced 100% FAN LOAD=FAN RPM=NOISE​

• 100%=Missing Data - Standing
• 100%=Missing Data - Lying

GPU TEMP=FAN LOAD=FAN RPM=NOISE:​

• 81*C=71%=2648rpm=41.5db - Lying
  
• 77*C=68%=2550rpm=39.8db - Lying without bezel
  
• 75*C=60%=2250rpm=37.4db - Standing
• 73*C=51%=1889rpm=35.5db - Standing without bezel

GIGABYTE GeForce GTX 1070 Mini ITX OC 8GB (GV-N1070IXOC-8GD)

Pros:

Cons:
-Very high PCB Layout, bad to cram in the S4 mini.
-Does not fit with the HDPLEX Power Connector



Some Data:
Stock Clock / Boost Clock / Mem Clock: 1531 / 1721 / 8008MHz

• POWER DRAW: 38W idle system with GPU
• POWER DRAW: 250W GPU (Furmark GPU burner)

Looking @ the Boostclock that the cards are boosting to in the Benchmarks

• BENCHMARK: 109. 128. 132. (min. avg. max. FPS) - 1582-1620MHz Boost - Furmark
  
• BENCHMARK: 3847 Points - 1809-1835MHz Boost - Unigine Valley
• BENCHMARK: 4710 Points - Unigine Heaven
• BENCHMARK: 5581 Points - 3D Mark Timespy

Furmark full load:

Forced 100% FAN LOAD=FAN RPM=NOISE​

• 100%=3400rpm=53.3db - Standing
• 100%=3400rpm=50.0db - Lying

GPU TEMP=FAN LOAD=FAN RPM=NOISE:​

• 81*C=71%=2426rpm=40.6db - Lying
  
• 79*C=69%=2360rpm=40.0db - Lying without bezel
  
• 74*C=63%=2153rpm=40.0db - Standing
• 72*C=62%=2102rpm=40.8db - Standing without bezel

Silent Fan mode:

• 88*C=51%=1960 rpm=35.0db - Lying (temp throtling @82*C)
  
• - Lying without bezel
• - Standing
• - Standing without bezel

MSI GeForce GTX 1070 Aero ITX OC

Pros:
+Most quiet card of the three
+Lowest Max Speed of Fan
+Lowest Temperatures on full load
+No Coilwhine

Cons:
-Does not fit with the HDPLEX Power Connector


Some Data:
Stock Clock / Boost Clock / Mem Clock: 1531 / 1721 / 8008MHz

• POWER DRAW: 38W idle system with GPU
• POWER DRAW: 238W GPU (Furmark GPU burner)

Looking @ the Boostclock that the cards are boosting to in the Benchmarks

• BENCHMARK: 123. 129. 132. (min. avg. max. FPS) - 1594-1620MHz Boost - Furmark
  
• BENCHMARK: 3865 Points - 1835-1885MHz Boost - Unigine Valley
• BENCHMARK: 4775 Points - Unigine Heaven
• BENCHMARK: 5673 Points - 3D Mark Timespy

Furmark full load:

FAN LOAD=FAN RPM=NOISE​

• 100%=2280rpm=44.8db - Standing
• 100%=2280rpm=42.9db - Lying

Furmark full load:

GPU TEMP=FAN LOAD=FAN RPM=NOISE:​

• 80*C=85%=1940rpm=37.8db - Lying
  
• 73*C=76%=1730rpm=35.0db - Lying without bezel
  
• 70*C=72%=1650rpm=35.5db - Standing
• 66*C=66%=1502rpm=33.5db - Standing without bezel (also the core/boost clock is higher 1632MHz than other cards due to low temperature)

Conclusion

I dont know how msi has achieved this great Performance.
Its the most recent 1070 Mini card (03.2017), maybe they learned from the pitfalls of other manufacturers, i dont know.

Anyhow, the fact that it keeps the Temperature allways lower that the other two Competitors, gives the card more headroom for higher Boostclock, so the performance in small cases is better (No thermal throtling->Higher Boostclocks)

I have to say, that right now i realized that i should have tested the cards more with high FPS content for coil whine test, but i tested the msi now in my S4 case in and out for days and have still not heard any coilwhine so far.

Part of coilwhine came also from the HDPLEX PSU, where i instantly used the epoxy glue trick on the coils to get rid of the whine.

at the end, some pictures of it all:

Be sure to check in a few days to hear the uploaded audiofiles of the fannoise if your interest is still up.

Sorry for any Spelling errors. english is not my native language.


------------------------------------------------------------------------------------------------------------------------------------------

ZOTAC
--------------------

with Power Connector?
No, wont fit...

Tight...

It fits (without Power Connector)

Length of the card...

Cables

Shroud to cut:

Bonus: Comparsion: 340W Brick <--> Zotac :-)

Gigabyte
-------------------

Same Same...
Wont fit:

Due to gigantomous Height of card.. have to make room...
every mm counts

Still strugling..

until...

Voila

msi !!!
---------------

Will it fit ?

Not quite :-/

Temporal fixing on the way (for the shroud to fit)

Flipped it. sits not straight but it is very firmly fixed (for the next weeks at last)

i have to remove 1-2 mm of the case so the pcb can slide in.
Next week. Enough tinkering and testing for today!

Finally:

This case has been arround the block...
Cheers

Hope this helped some of you guys as it took quite some Days.

Update 2:

I think it'd be nice to, in your round up, list the ports on the cards and also mention what resolutions/refresh rates you're able to run your display at by default. Nvidia's Control Panel (in my experience) will let you drive a monitor at whatever it's capable of, but sometimes you have to go in there and test/enable certain resolutions/refresh rates yourself.

Yeah, i did.

All of the cards support the following resolutions above 1080p :

2560x1440
2880x1620
3200x1800
3600x2025
3840x2160
and just for the lulz:
5120x2880
:-D


Update 3:

Audio Loudness Supercut:

Recorded from 0 RPM (idle GPU) -> Started Furmark -> waited for the Fan RPM (this is the interesting part) to stabilize at 100% GPU usage -> Quick spinup to 100% Fan RPM to measure the noise.

Audio is: Zotac -> Gigabyte -> msi

Zotac and msi recorded with a "max RPM" noise test at the end
(Gigabyte max RPM recording didnt work out)

Every Spinup took roundabout 5 minutes and was speed up to 30 seconds to compare better (the speed was altered without altering the pitch of the recording)

It was recorded 5cm away from the case to overcome the basenoise of the microphone and amp.

you have to disable any soundcard compression/normalization to hear the real loundness difference between each Card.

http://www38.zippyshare.com/v/yCbAkNkC/file.html

boy is the msi is pushing some air. Just realized it now after hearing the air allmost clipping the mic in the recording :-D

 

[K888D]

Could you measure the height of the power connectors on the top of the card from the bottom edge of the PCIe slot? Thanks!

 

[zovc]

I think it'd be nice to, in your round up, list the ports on the cards and also mention what resolutions/refresh rates you're able to run your display at by default. Nvidia's Control Panel (in my experience) will let you drive a monitor at whatever it's capable of, but sometimes you have to go in there and test/enable certain resolutions/refresh rates yourself.

 

[spintje]

I would love if you could rate the overall compatibility of these cards with S4 and the HDPLEX ie. which cards requires modification to fit? And what kind of mods are necessary?

 

[fminus]

@spintje I have an S4 + HDPlex300w + MSI 1070 Aero ITX. You need to follow Josh's guide if you want the power plug to fit. Similarly to the Gigabyte card.

 

[chx]

Yes, a measurement of the cards would be great for people who want to build custom cases

 

[Broxin]

is there any other 1070 mini card that fits the S4?
i want to test all the possible cards at once.

I think it'd be nice to, in your round up, list the ports on the cards and also mention what resolutions/refresh rates you're able to run your display at by default. Nvidia's Control Panel (in my experience) will let you drive a monitor at whatever it's capable of, but sometimes you have to go in there and test/enable certain resolutions/refresh rates yourself.

could you explain furrther ?
i just have a 1080 monitor so i can only test up to 1920x1080 60Hz

Could you measure the height of the power connectors on the top of the card from the bottom edge of the PCIe slot? Thanks!

I would love if you could rate the overall compatibility of these cards with S4 and the HDPLEX ie. which cards requires modification to fit? And what kind of mods are necessary?

will do.

 

[spintje]

@spintje I have an S4 + HDPlex300w + MSI 1070 Aero ITX. You need to follow Josh's guide if you want the power plug to fit. Similarly to the Gigabyte card.

Cool, so dremel the connector and that's it?

 

[zovc]

could you explain furrther ?
i just have a 1080 monitor so i can only test up to 1920x1080 60Hz

Regarding the I/O, the Galax 1070 has two DV-I, one DisplayPort, and one HDMI. This is a shortcoming of the card and actually makes me regret choosing it, honestly. I much prefer using more modern cables (HDMI and DisplayPort, DP being my favorite) and if I want to use three monitors, I have to use a DV-I on one.

Regarding the resolutions, it's possible Nvidia's Control Panel will let you 'overclock' (resolution is technically "oversampling") your monitor to a higher refresh rate or a higher resolution. My monitors are sold as 1080p monitors that run at 144Hz, but I've overclocked them to display 1440p.

 

[Broxin]

They arrived :-D

Sadly i cant test the kfa2 (Galax) because i could not get a hold of this one.

Maybe one of you could write some data/photos of the galax 1070 if you own one @zovc ?
Loudness, temps e.t.c.

I want to put all the cards in the same ballpark when testing, meaning:
Putting them all on the same gpu and memory clock.

Is it possible to set both, base and boost clock to a manual value?

 

[zovc]

I actually managed to fry my Galax 1070 Mini, which is a long stupid story. (And it's totally my fault.) Galax is supposed to be shipping me a replacement--which I didn't expect--but I haven't gotten confirmation that it's on its way.

I'd be happy to do a review and run tests on the card, but I don't feel like I have good means of testing the card. I can try to replicate your methods as best as I can if I happen to get that replacement, though.

As far as setting base and boost clocks to minimal values, you should look into some sort of Pascal overclocking guide. I did some toying with undervolting and underclocking the 1070 before my silly accident, and I found that GPU Boost 3.0 seems to do strange things to clock speeds and voltages. Maybe I was 'missing a piece of the puzzle' as far as tuning things in MSI Afterburner is concerned, but I would try to lock the voltage to certain values and it would 'randomly' jump up or down from certain settings when I would click 'Save.'

I imagine there's got to be one of the bigger tech youtubers who put out a guide on overclocking Pascal, whether it's JayzTwoCents or LTT or Gamers Nexus or whoever else. Learning about overclocking will also help you figure out getting cards onto uniform 'stock' or 'base' settings.

 

[Broxin]

Oh, how did you fry it?
So if it is damaged, could you measure the dimensions of the card? Also how the connector is placed, like it was asked here thanks

 

[zovc]

I don't have the card, Galax does.

I was removing the card from the PCI-e slots in a case and it bumped the (plugged-in) AC-in port of my case. There was a visible arc from the PCB of the port to the PCI-e bracket on the video card. Somehow, I wasn't shocked, but the card did not survive.

 

[rchetal]

Do all of them fit in the S4 mini without modification? I'm also deciding between all of these options. I'm using a GX12 jack so I'm not sure if it'll have the same space issues the Gigabyte 1070 mini has.

 

[zovc]

Do all of them fit in the S4 mini without modification? I'm also deciding between all of these options. I'm using a GX12 jack so I'm not sure if it'll have the same space issues the Gigabyte 1070 mini has.

The Galax card fits into the S4 Mini without any complication. That's the only 1070 I have any experience with.

 

[Broxin]

ok, quick preview:

msi won in every category...
this card is just crazy. its cooler, smaller, much more quiet and has the best ports for me at least.

i update the thread with the full material and data i gathered sometime later. its 5am and i need some sleep

 

[kevlar]

ok, quick preview:

msi won in every category...
this card is just crazy. its cooler, smaller, much more quiet and has the best ports for me at least.

i update the thread with the full material and data i gathered sometime later. its 5am and i need some sleep

I was really eye-ing this card earlier this year because I was reading some crazy stuff about its temps and noise levels but it seems stock is super sparse. How did you get your hands on one?

 

[Davila]

msi won in every category...
this card is just crazy. its cooler, smaller, much more quiet and has the best ports for me at least.

So this card beats the Zotac GTX 1070 Mini? Why does everyone go for the Zotac then? I haven't seen a single build with this card, and I've seen tons with the Zotac. I also expected the Zotac card to be cooler but I guess not.

 

[Broxin]

It is not. I dont know, maybe its because the msi is fairly new (introduced 03.2017) and noone compared it to the others.

Its defenetly cooler, more quiet and fits better.

I am just tense on time to publish the results.
Will do as soon as possible

 

[Broxin]

Phew...

Updated with Review in first Post

Enjoy