Sanity check for a workstation/home studio/light gaming PC

New to PC silencing? Read & post your questions here. Dedicated to rosy_toes.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Sanity check for a workstation/home studio/light gaming PC

Post by Xinef » Wed Jun 14, 2017 1:59 pm

First of all, hi everyone. I'm new here.

I'm trying to build my first quiet PC, and after days of research into the topic, I've come up with a rough idea of what I'd want. However, before I spend a big chunk of my budget on hardware, I though it might be a good idea to have some experts look at my shopping list, and see if I've made any wrong assumptions or other mistakes. Or if anyone is interested in hearing my review once I buy and test it.

1. Requirements/Use Cases:

- Programming (CUDA, OpenCL), rendering, 3D modeling, video and audio editing. I'm a PhD student in Computer Science (virtual reality), and I'm writing my own game engines, graphics engines, physics engines... and occasionally doing other gamedev related things like 3D modeling in Blender.
- Home studio - I'm recording gaming sessions (online tabletop roleplaying games), and thinking of recording video tutorials and maybe let's play series. I don't want the noise from the PC to be recorder by my microphone, even when the system is under heavy load. Sure, I can edit the noise out, but it lowers audio quality.
- Light gaming. Games get better as they get older ;)
My favorite games include: Factorio, Kerbal Space Program, Minecraft (all heavily modded, I've run with 100+ mods sometimes), Heroes of Might and Magic 1-5, Civilization 1-5 + freeciv, the X-Com series, various city builders... yes, including Dwarf Fortress...
As far as MOBA/MMO games go, the only ones that caught my interest are Heroes of the Storm and Dungeons and Dragons Online.
Most of the titles I play seem to be CPU/RAM heavy, but light on the GPU. I don't mind playing on lower details or with lower FPS (<30 FPS doesn't hurt in turn based strategies), but CPU-limited games tend to slow down when simulating things, and that's worse.
- Linux. I'm buying the computer for myself, not for Microsoft's nefarious world domination plans, thank you :)
Currently my distro of choice is Linux Mint, with the MATE environment.

I'm living in rather quiet suburbs, far away from main roads etc. so the ambient noise is probably quite low. When the PC is off, I can hear birds chirping, wind blowing outside, and often I can't really hear anything. As such, any noise the PC makes stands out easily even when idling.

Worst case scenario, I might want to leave the PC overnight, making some CUDA/OpenCL computations on both the CPU and GPU... while I'm trying to sleep 1.5 meter away. So yeah, if I can make it practically silent even under heavy load, that would be great.

Ok, with that out of the way... let's get to the good part:

2. System:

CPU: i7 7700k
Cooler: be quiet! Dark Rock 3 (single tower, not the pro version which is dual tower)
GPU: Palit GTX 1050 Ti KalmX
PSU: be quiet! Dark Power Pro 11, 550W
Case: Nanoxia Deep Silence 5 rev. B
Fans: 3 x 140mm Nanoxia fans (included with the case), 3 x 140 mm Nanoxia PWM fans (purchased separately), 2 x 140mm be quiet! SILENT WINGS 3 on the CPU cooler - I'm going with the assumption that many, big, slow fans are much more quiet than few, small, fast fans would be. So having 8 (9 if counting the PSU) 140mm fans, all set to never exceed 800 RPM should do the trick.
Mobo: MSI Z270 SLI Plus
RAM: 2x8 GB DDR4, any brand, will probably upgrade in the future. I could buy 2133 MHz to save money for future upgrades, although there are some 3000 MHz at just a 10% premium, so maybe I should pick those? And I guess I should go for low profile sticks without heatsinks, so they don't conflict with the CPU cooler?
Drives: system, home directory = ADATA SP900, big files = TOSHIBA DT01ACA300 - that's actually what I'm using right now. I may upgrade in the future, but it's not a priority. The HDD is rarely in use, so I can spin it down when recording/sleeping. In fact, I'm mostly using external HDDs for backups, but since my old mobo has only USB 2.0 support... yeah, I think it's time for an upgrade.

Peripherals (for anyone interested, since I'm just taking them from my old system):
Monitors: Dell 1703FP (my older monitor) and Dell P2416D (my newer monitor)
External sound card: Sound Blaster X-Fi HD
Headphones: AKG K271 MkII
Microphone: Samson C01U Pro
Keyboard: a quiet one :P I don't know why someone would want a loud keyboard

3. A few other important tidbits:

- I'm not planning on overclocking just yet. It sounds fun, but since it may void the warranty, I'll likely wait until the warranty is over before attempting anything of that sort. I might in fact decrease the voltage, since I've heard 7700k works stable at lower voltages, which should reduce the heat and prolong the processor's life I suppose?
- I think building PCs that look good inside is an interesting challenge, so that's part of the plan. I'm not going to sacrifice performance or silence for the looks, but I might be willing to spend a bit extra on parts that are sufficient, and look good. (as such, Noctua fans are no go, they're ugly :/ ) That's my reason for the Dark Rock 3 cooler - should be sufficient, and looks good.
- My previous system lasted 7 years. And the main reason I'm upgrading is noise, otherwise I could probably use if for a few more years. As such, I'd be happy if the new build lasted 10 years or longer, with probably some upgrades along the line, but I'm not one to buy a new PC every year or two ;)
- Still, having room for upgrades is nice... what if I decide I need dual graphics cards in SLI at some point? Or if I decide to go with a custom watercooling loop to keep the thing quiet when I do start overclocking?

I've never done overclocking or watercooling, but it sounds fun, so I might try it at some point in distant future... and that's why I'm buying some parts with the future in mind. Such as the case which I won't have to replace even if I purchase a larger motherbaord, or go with a custom watercooling loop. And the mobo and PSU, which shouldn't need upgrades if I decide to go dual GPU at some point.

So... does anything seem out of place? Will the cooling be enough? Will I run into RAM clearance issues? Is anything not future-proofed? Is anyone interested in a review of the results? Anything I should keep in mind if I'm thinking of overclocking and building a custom watercooling loop in the future? Or any other comments?

Any suggestions for Linux software to test/control temperatures, fan speeds, voltages, clock speeds etc? My old motherboard doesn't support most of those things, so I never looked for good software. Now things will change :)

My friends suggested I should look at the Ryzen processors. My impression is that they make good server/mainframe CPUs because of all the cores, but for a desktop/gaming CPU they are likely worse than Intel's. Considering my "requirements" I suppose Ryzen could be better for some of them (like rendering, compiling, etc.) but the large majority of things I do will not utilize that many threads. Also, high end Ryzen processors (1800x) according to some reviews draw considerably more power than 7700k does at full load. While it's true 7700k (unless delided) with it's poor heat dissipation can overheat easily... it doesn't affect cooling much. The cooler still has the same amount of heat to dissipate, whether the TIM is good or bad. It just means the CPU will be hot, while the radiator will be much colder. However if Ryzen generates more heat, it means more heat for the cooler to dissipate. As such, I'd expect Ryzen to be louder in practice. Or am I missing something? Is there any point in setting the CPU fans to higher RPMs to blow at that cold radiator as much as possible, just because the CPU is poorly transferring heat to the radiator?

... if I have to ditch Intel for a Ryzen I'll have to redo a lot of my shopping research... :c

Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Re: Sanity check for a workstation/home studio/light gaming

Post by Xinef » Thu Jun 15, 2017 5:51 am

Also, I realized that because the three intake case fans are somewhat obstructed (dust filters, the indirect air tunnel at the front, and the bottom fan being not that far away from the floor), while the three exhaust fans have very few obstructions, this might in fact lead to negative pressure despite the 3 to 3 ratio...

Which from what I've heard means dust will be sucked into the case through the various non-filtered openings.

Any suggestions? Should I just run the intake fans at higher RPM than the exhaust fans?
Replace intake fans with something with higher static pressure?
Go with a 4 intake to 2 exhaust ratio?

I know some people would suggest removing dust filters... well... my room has mediocre levels of dust (my old computer was fine with a yearly maintenance), but it also had rather poor airflow, with just two small fans (side intake and rear exhaust).

Now this build would probably suck much more dust considering a bottom intake sitting right above the dusty floor, with probably stronger airflow than in my previous build... so dust could become a bigger issue.

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Sanity check for a workstation/home studio/light gaming

Post by Abula » Thu Jun 15, 2017 6:26 am

Since this is going to be a Linux build, you have two options as i see it, you either try to get a good motherboard with good bios fan control or you resort into an external fan controller.

This are my impression on each of the 4 mobo manufactuers on bios fan control.

ASUS
The best on software, but on bios, a little buggy, their QTunning is an option on the bios to test fans and establish min max values, it doesn't work well with all type of fans, thus in some cases not recognizing the min of operation on the test, this makes you not able to go below what the motherboard establishes as minimum. This issues i encounter with some (not all) noctua fans and some scythe. I have open a petition to asus to crosscheck this, but i feel they are going to fix it on newer motherboards, as they have done in the past. The good about Asus is that they have Switchable headers, meaning you can plug a 3pin or 4pin in any fan header, you just need to enter the bios and select the type of fan. The bad, is that if the bios can establish the min max correctly of the fans you buy, you might not be able to drop them low enough for your liking. I personally don't recommend Asus for bios fan control.

AsRock
My favorite atm, it has no restriction on the bios (like asus and msi), you can go as low as your fans can go, its pretty simple, you have prestablish presets (i dont use them) or you can set your own graph, you have 4 breakpoints that help you design the ramping of the fan based on the cpu temperature. The bad is that their fan headers are not switchable as Asus, you gotta understand how they design them, usually CPU_FAN and CHA_FAN1 are autodetect, meaning that bios senses if you connect a 3pin or 4pin and switches the control automatically, but the rest of the headers aka CHA_FAN2, 3, 4, etc are only controlled by PWM. So in a pure AsRock build i would plan on full pwm or at least most of them (max of 2 none pwm although i recommend going full pwm). PWM splitters (specially ones powered by PSU) are a good asset toward helping you control large amount of fans with a single header.

MSI
Their bios fan control my favorite until Z87/97, very easy to setup, as long as you understood how the headers worked. Usually CPU_FAN1 and CPU_FAN2 were PWM and CHA_FAN1, 2, 3 were DC controlled. So with MSI you simple had to chose carefully the fans on which header and the rest was pretty easy. They have limitation on bios on PWM of 12.5% increments, meaning 12.5, 25, 37.5, 50, 67.5, 75, 87.5, 100% (asrock is better here as they allow 1% increments on 1C temp variations, practically full range of operation). Now the bad is that on Z170 MSI change some stuff on the bios fan control, the CHA_FAN1 and CHA_FAN2 remain PWM, and work great, the problem was with the CHA_FANs where i couldn't undervolt dynamically with the bios, i could still go for a steady undervolting meaning for example 50% that was around 6.5V but it wont move if the load of the CPU changes, thus making it a steady set and forget, instead of planning to increase the cooling as the load goes. That said, i still think MSI can work great, even newer, but with only using PWM header aka CPU_FAN1 and CPU_FAN2 and PWM fan splitter and going full PWM fans. I really wished i had a MSI Z270 to test to know if MSI change something from Z170.

Gigabyte
I cant say much, i haven't own a gigabyte in 10 years, in the past they were terrible on bios fan control, but i have read they have made it a lot better in newer motherbaords, but i cant say much as i don't own one to know how it really work out.

About the amount of fans, you should read
Bit Tech The Big Air Cooling Investigation
LinesTechTip Case Fans - How many should you have?

Personally i lie to myself and about to do a 10 fan setup, and ill keep saying to me that its beneficial, but the reality is that above.
Last edited by Abula on Thu Jun 15, 2017 9:05 am, edited 1 time in total.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Sanity check for a workstation/home studio/light gaming

Post by CA_Steve » Thu Jun 15, 2017 7:44 am

Welcome to SPCR.

First up: Ryzen is useful if you have applications (like Blender) that can make effective use of high core/thread counts. For gaming, it's a toss-up. High thread count games can do better under Ryzen if the game's been optimized. Not a lot of games are optimized. Over time, this should change...maybe not for old games...then again old games probably won't stress the CPU. Under stress load, AMD will use more power than Intel. Under likely load, not a big difference. AMD uses solder for thermal interface between CPU and heatspreader, Intel uses thermal paste. Net effect, AMD can run cooler even if it uses more power. This is for stock clock. BIOS updates continue at a quick pace...what I don't know is Linux compatibility/stability for Ryzen/specific motherboards - you might check Phoronix.com for that. Conclusion: Ryzen is a new platform with a lot of potential upside for your uses...the big unknown is whether it's settled down enough yet for your specific applications.

Your system's stress power use with stock clocks will be 95W (CPU) + 75W (GPU) + call it 50W for everything else...unless you stick with 9 fans...that's 220W. A typical high load on CPU/GPU will be around 85% of stress load for ~190W. For this amount of heat, you should try using just the included three case fans as mounted (2 front, 1 rear). Use the motherboard's BIOS level fan controls to set a profile you like. Remove the front drive trays for unobstructed airflow to the GPU.

Case: ok. If you don't have an optical drive, consider the Fractal Define C.

GPU: Note the lower end GTX cards don't allow for SLI. KalmX gets good reviews. Requires some airflow to maintain std clocks. If front case fans don't provide enough air at low enough rpm for inaudibility, strap a low rpm fan to the KalmX.

RAM: Prices are pretty flat until you get to the higher speeds. Note that std clock speed for Intel Kaby Lake RAM is 2400. Some slight benefit for 2666. AMD Ryzen does benefit from faster RAM speeds, but dependant on BIOS updates for RAM tuning - in progress.

SSD: I'm not an Adata fan (warranty return service issue). If it's what you've got and it works, fine.

Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Re: Sanity check for a workstation/home studio/light gaming

Post by Xinef » Fri Jun 16, 2017 8:49 am

Abula wrote:Since this is going to be a Linux build, you have two options as i see it, you either try to get a good motherboard with good bios fan control or you resort into an external fan controller.
The case comes with a fan controller (2 sliders, each controlling up to 3 fans), so this might come in handy.
The PSU also has 4 fan connectors, supposedly varying fan speed based on power draw?

However, none of these would give me control from BIOS/OS level, or based on temperatures... but that can be fixed with a PWM splitter (using CPU_FAN). Or I could purchase a more capable fan controller, they don't cost that much, right?

One interesting thing I found just now is this NZXT GRID+ V2 controller, which while not officially supported on Linux, has this:
https://github.com/CapitalF/gridfan
unofficial script which could give me all the control needed.

Sadly, it uses voltage control only, even with PWM fans. Are there any controllers that actually use PWM with PWM fans?
(as I understand, PWM fans can work at much lower speed when powered by PWM than when controlled by voltage? Also, some sources claim that PWM vs voltage affects noise and lifetime of fans)
Abula wrote:This are my impression on each of the 4 mobo manufactuers on bios fan control.
My current board is AsRock N68C-GS UCC, and I only used it with stock fans, never really playing much with it, so I don't even know what that old board or it's bios is capable of. I'm just beginning to learn about quiet PC building and fan control (over the past few weeks), because I decided it's time to upgrade, so I thought why bother figuring out if I can do anything with the old board, I should just buy a new one and see from there.

However, that MSI board (Z270 SLI Plus) seems to have great features (plenty of USB 3.1 ports, type A and C connectors, 2x M.2 ports, 3x PCI-e 16, 3x PCI-e 1, 6x SATA III, and SLI support), at a reasonable price, meaning I have plenty of room to add features (most likely: a WiFi card, more SSD drives, and maybe more graphics cards), so this board caught my interest.

So I think I'll go with MSI, and if things don't go well, I'll see what external fan controllers can do (or if my case/PSU controllers are enough).

Btw. I looked at motherboards with integrated WiFi, but:
a) They seem to be a step up in price (understandably), so it wouldn't save me money.
b) It means if/when I change the board, the WiFi would go to waste. If I buy a card, I can move it to the new board.

So that sort of makes buying the most expensive boards a poor investment, it would seem :P
Abula wrote:About the amount of fans, you should read
Bit Tech The Big Air Cooling Investigation
LinesTechTip Case Fans - How many should you have?

Personally i lie to myself and about to do a 10 fan setup, and ill keep saying to me that its beneficial, but the reality is that above.
The above two experiments seem to completely ignore fan control / noise. It seems they run everything at full speed too.

My plan to go with 6 case fans is because:
(let's go with Nanoxia 140 mm non-PWM 1400 RPM fans)
3 fans going at full speed move 3 * 76,5 CFM = 229,5 CFM, making a noise of 3 * 16.2 dB = 21 dB (I used https://www.noisemeters.com/apps/db-calculator.asp ) :P
6 fans going at half speed move 6 * 38,25 CFM = 229,5 CFM (from what I read, halving speed halves airflow... I dunno if it works only for unobstructed airflow, or if it holds for obstucted fans too?), while the noise will be 6 * 10.5 dB = 18.3 dB (I'm taking the listed noise for 7V, should be close enough).

In other words - same airflow, less noise because now we're operating at half speed. None of the experiments you linked attempted anything of that sort. I've seen some fan reviews however, suggesting that indeed two fans at half speed are quieter than one at full speed.
CA_Steve wrote:First up: Ryzen is useful if you have applications (like Blender) that can make effective use of high core/thread counts. For gaming, it's a toss-up. High thread count games can do better under Ryzen if the game's been optimized. Not a lot of games are optimized. Over time, this should change...maybe not for old games...then again old games probably won't stress the CPU. Under stress load, AMD will use more power than Intel. Under likely load, not a big difference. AMD uses solder for thermal interface between CPU and heatspreader, Intel uses thermal paste. Net effect, AMD can run cooler even if it uses more power. This is for stock clock. BIOS updates continue at a quick pace...what I don't know is Linux compatibility/stability for Ryzen/specific motherboards - you might check Phoronix.com for that. Conclusion: Ryzen is a new platform with a lot of potential upside for your uses...the big unknown is whether it's settled down enough yet for your specific applications.
I know plenty of old games that can stress the CPU. Strategy games when running AI (all those "wait for the AI to make a move" moments), Minecraft when generating new terrain, Factorio when doing this: https://www.youtube.com/watch?v=Kry8lbrHjeY

However, I doubt many of them can benefit from more than 4 threads.

Anyway, even if Ryzen could work for a few of my uses, I think Intel will work better for a majority of tasks I'm doing, and computations that can be parallelized well on many threads are probably better run on the GPU anyway. I mean, if I want to render things in Blender, I can use CUDA and then the CPU will matter less, since GPU will do most of the work. If I want more computational power in my own engines? I'll also try hardware acceleration.
(to be honest, my code for the moment runs on CPU only, apart from some shaders, but I'm getting to it! Learning CUDA and OpenCL is on my TODO list)

Not to mention I don't really need my own engines to work super fast on my machine. I want them to run fast on other people's computers :P For programming, all I need is modern hardware with all the APIs etc.
CA_Steve wrote:Your system's stress power use with stock clocks will be 95W (CPU) + 75W (GPU) + call it 50W for everything else...unless you stick with 9 fans...that's 220W. A typical high load on CPU/GPU will be around 85% of stress load for ~190W. For this amount of heat, you should try using just the included three case fans as mounted (2 front, 1 rear). Use the motherboard's BIOS level fan controls to set a profile you like. Remove the front drive trays for unobstructed airflow to the GPU.
I already addressed my reasoning for more fans. With more fans, I can run them at lower speeds getting the same amount of airflow. Is this a flawed way of thinking?
Also, this is a big case, with a passively cooled GPU = two reasons why it might need more airflow than the average case.

However, the suggestion to remove drive trays sounds good. I can probably fit everything (the HDD and SSD) in the top 5.25 inch bay.
CA_Steve wrote:Case: ok. If you don't have an optical drive, consider the Fractal Define C.

GPU: Note the lower end GTX cards don't allow for SLI. KalmX gets good reviews. Requires some airflow to maintain std clocks. If front case fans don't provide enough air at low enough rpm for inaudibility, strap a low rpm fan to the KalmX.

RAM: Prices are pretty flat until you get to the higher speeds. Note that std clock speed for Intel Kaby Lake RAM is 2400. Some slight benefit for 2666. AMD Ryzen does benefit from faster RAM speeds, but dependant on BIOS updates for RAM tuning - in progress.

SSD: I'm not an Adata fan (warranty return service issue). If it's what you've got and it works, fine.
I do have an optical drive, although as of lately I only use it to rip music CDs... it's nice to have physical CDs if only so that friends can see your music collection :P With the popularity of digital-only music sales, I can't even tell what music most of my friends have, so I figured physical music CDs still have their use in that regard ;)

I know if I ever need SLI, I'll need higher end cards. However, there are a few reasons I'd want the KalmX anyway:
a) I can switch between 2 loud GPUs or 1 quiet GPU as needed.
b) It IS possible to use multiple GPUs even without SLI. There are limitations (only one GPU will be connected to the monitor, and even some non-graphical uses require SLI to share data between cards), however I've heard of people using one GPU to do physics while another one does graphics, without SLI. And I think it shouldn't be too hard to split tasks for my own engines between multiple GPUs even without SLI. However, it's good to have SLI support on the board just in case ;)

As for RAM - indeed. I guess there's not much point investing in high speed RAM if so many sources indicate it either doesn't work as advertised, or just doesn't give much advantage even when it works.

So far I've never had Adata drives fail, but I'll keep your warning in mind. Buying some new SSDs will likely be one of my next upgrades anyway.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Sanity check for a workstation/home studio/light gaming

Post by CA_Steve » Fri Jun 16, 2017 9:33 am

I already addressed my reasoning for more fans. With more fans, I can run them at lower speeds getting the same amount of airflow. Is this a flawed way of thinking?
I think it's a faulty generalization. The primary purpose of case fans is to replace the accumulated hot air with cooler ambient air (with the secondary purpose to provide direct cooling to any components along the way). If you have a lot of accumulated hot air (ie: high wattage components), then more airflow is beneficial. Fans at higher rpm tend to be disproportionately noisier, so if more fans can provide additional airflow at much reduced rpm, there can be a sweet spot for noise and temps.

With your build, there won't be a lot of accumulated hot air because there isn't a lot of wattage involved. It may well be that just the three included fans will be able to provide decent temps while running at such low speeds to be inaudible. Adding more fans may lower temps, but may also rise above your noise floor.

Also, try the BIOS fan controller. It will provide more flexibility than what's provided with the case, and any add-on controller is just not worth it (more $'s, little value add over BIOS).

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Sanity check for a workstation/home studio/light gaming

Post by Abula » Fri Jun 16, 2017 1:12 pm

Xinef wrote:However, none of these would give me control from BIOS/OS level, or based on
the PWM fan splitters are cheap imo, for what they bring, that said you do need to semi match as you wont have control over them, just one, the rest will work at the PWM signal, so similar or identical would be best.

Xinef wrote:Sadly, it uses voltage control only, even with PWM fans. Are there any controllers that actually use PWM with PWM fans?
This is not recommended, not saying it wont work, as you already know they can be undervolted, but to what i have read, PWM fans their motors are design to operate always at 12V, and not to stand different voltages, weather this had an impact or not on the longevity... idk, but i would certanly recommend you to use 3pin with voltage control and 4pin with pwm control, its pretty easy as long as you see how your headers from your motherboard work out.
Xinef wrote: The above two experiments seem to completely ignore fan control / noise. It seems they run everything at full speed too.

My plan to go with 6 case fans is because:
(let's go with Nanoxia 140 mm non-PWM 1400 RPM fans)
3 fans going at full speed move 3 * 76,5 CFM = 229,5 CFM, making a noise of 3 * 16.2 dB = 21 dB (I used https://www.noisemeters.com/apps/db-calculator.asp ) :P
6 fans going at half speed move 6 * 38,25 CFM = 229,5 CFM (from what I read, halving speed halves airflow... I dunno if it works only for unobstructed airflow, or if it holds for obstucted fans too?), while the noise will be 6 * 10.5 dB = 18.3 dB (I'm taking the listed noise for 7V, should be close enough).

In other words - same airflow, less noise because now we're operating at half speed. None of the experiments you linked attempted anything of that sort. I've seen some fan reviews however, suggesting that indeed two fans at half speed are quieter than one at full speed.
More case fans will have an impact as long as your outside and inside temps are different, but once you get close, they have little impact, with the tests they were getting that with 2 fans, the 3rd you gain very little, and less each extra fan added. Would be interesting if you could test it at very low rpms, and see if you have diminishing returns past 2 fans.

Fans behind fan filters diminish a lot what they can move of air, depends a lot on the fan design and how restrictive is the filter, I usually take out 1/3, then again i never really pay attention to manufacturers CFM and noise, as all measure different, so no point on using those specs.

Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Re: Sanity check for a workstation/home studio/light gaming

Post by Xinef » Fri Jun 16, 2017 2:21 pm

Abula wrote:Would be interesting if you could test it at very low rpms, and see if you have diminishing returns past 2 fans.
Ok, thanks everyone for the input so far.

I think I'll just do some testing and report my findings (probably in a few weeks).

lego45
Posts: 6
Joined: Tue Jun 13, 2017 11:55 pm

Re: Sanity check for a workstation/home studio/light gaming

Post by lego45 » Tue Jun 20, 2017 7:27 pm

That will be an interesting build. Subscribed for any update.

Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Re: Sanity check for a workstation/home studio/light gaming

Post by Xinef » Fri Jun 23, 2017 10:53 am

Alrighty, system assembled and I will soon try to find some time to do various tests and experiments.

In the meantime, I've taken the time to learn the basics of the Phoronix Test Suite. Quite an interesting tool, to say the least, although it's hard to find any decent "quick start guides", so I had to read the entire manual, and do some trial-and-error, to finally get some reasonable configuration.

Nevertheless, the temperature graphs and other monitoring data are quite interesting :)
http://openbenchmarking.org/result/1706 ... ONOTICTE27
http://openbenchmarking.org/result/1706 ... AYTRACIN27

It seems the "system temperature" sensor doesn't work... it always gives 27.80°C even when running CPU/GPU intensive tests for hours, even though it was listed as an available sensor. I'll have to look into that. I think in BIOS it showed different temps.

Anyway, first a quick note:
I decided to buy a 120mm Silent Wings 3 fan instead of the 140 mm I originally planned to buy. Quite a few reasons:
  1. Even though the Dark Rock 3 cooler comes with a 135 mm fan, it recommends attaching a 120 mm fan if you want to go dual. It made me look deeper into why wouldn't they recommend 140 mm fans.
  2. If I place the 120 mm fan on the RAM side and the 135 mm fan on the I/O port side, it gives more RAM clearance. The alternative would be to hang a 140 mm fan higher up, but then it would stick 2 cm above the top of the radiator, which would waste some airflow, and would look ugly too. Also, in smaller cases it could be impossible, although not a problem in the DS5.
  3. The main heatsink itself seems to be about 120 mm in height - even the provided 135 mm fan in the default configuration extends below the radiator, providing some airflow between the main heatsink and the small heatsink that is directly on the base plate. This is good for cooling, but would be problematic on the RAM side of the cooler. As such, having a larger fan on the I/O side and a smaller fan on the RAM side is perfect.
  4. While comparing noise/airflow/air pressure between different manufacturers just by the values given in the specs may be pointless, comparing products from the same manufacturer I guess makes more sense. And in this case there are interesting differences. Just look at the four main contenders from the Silent Wings 3 series:
    1. 120mm PWM - 1,450 rpm 16.4 dB(A) 1.79 mm(H2O)
    2. 140mm PWM - 1,000 rpm 15.5 dB(A) 1.08 mm(H2O)
    3. 120mm PWM high-speed - 2,200 rpm 28.6 dB(A) 3.37 mm(H2O)
    4. 140mm PWM high-speed - 1,600 rpm 28.1 dB(A) 2.16 mm(H2O)
    First thing we can notice is that the high-speed versions will likely be quite audible at 100% speeds, being a whole 12-13 dB(A) above their slower brothers, so if we want a system that is quiet under full load, we'll want to stick to slower speeds. Now, looking at the slower ones we can see a huge difference in air pressure in favor of the smaller fan, with only a 0.9 dB(A) difference. If this data is even remotely accurate, it gives the 120 mm slow PWM fan a huge lead in air-pressure to noise ratio. Keeping the logarithmic scale of decibels in mind, it wins at pressure/noise ratio even against the high-speed fans.
Image

Image

As for the BIOS fan control, things are looking good so far. While the default settings start with the CPU fans set to PWM mode and case fans set to DC mode, this can be switched around for each fan, and also each fan can be set to "smart fan mode" to enable variable speed based on temperatures with 4 control points.
Although, even when speed is set to 0%, the fan still spins at minimum speed. (it also displays a warning that fans set to below 40% speed can behave unpredictably or somesuch)
It also seems that all fans are looking at the CPU temp only (or perhaps at the higher of the two), so there seems no way to e.g. set case fans to just look at the "system" temperature sensor, or do other shenanigans.

In "smart fan mode" each point of the graph can be adjusted to 1 degree, 1% speed precision, and with smart fan mode disabled, the speed can still be adjusted 1% at a time.

I'll have to look into the BIOS some more, but that's what I noticed for now.

Image

===

As for the fans themselves, I'll have to do the experiments to see how different configurations and speeds affect noise and temperatures, but so far my observations are that:
- After setting all fans to the lowest speeds, the PC seems very quiet, although there is still some noise I was unable to identify. I think it might be the air just flowing through the case though, or maybe it's vibrations of the case, in which case that would likely be the lowest noise level I can get with this build. It's barely audible though, and I have to focus to even hear it. Any other sounds nearby easily overpower that faint sound, so overall I doubt it will be noticeable in daily use.
- Setting any of the fans to 100% speed (I have 3 PWM fans right now controllable through BIOS and 3 DC fans controllable through front panel) makes that fan the main source of noise. Which suggests fan control works, and also suggests none of the fans are truly inaudible at 100% speed - as expected, just making sure ;)

===

More to come after I make some heavier testing.

So far, even when under 100% load for half an hour, the CPU and GPU temperatures reached a maximum of 78/74°C, so I guess that's good enough. Not much room for overclocking, as I expected for this system (if I want it to stay quiet under load while overclocked, I'll need watercooling for the CPU at the least), but also not bad for a quiet system?
Although, those temperatures were with fans running at higher rpms, which while still reasonably quiet, were noticeably louder than idle. So my testing will mostly revolve around seeing where fans can be slowed down while keeping reasonable temperatures under 100% load.
Last edited by Xinef on Fri Jun 23, 2017 2:33 pm, edited 1 time in total.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Sanity check for a workstation/home studio/light gaming

Post by CA_Steve » Fri Jun 23, 2017 2:04 pm

Build looks nice.

Stress testing is a great way to understand the corner conditions of your build / to make sure you won't overheat/throttle. For noise optimization, the best way would be to just use your applications over the course of a day and then adjust fan rpm to fit them.

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Sanity check for a workstation/home studio/light gaming

Post by Abula » Fri Jun 23, 2017 3:27 pm

Very interesting picture of the msi bios fan control, specially since its a SYS_FAN header that has PWM/DC switching options, i have some questions,

1) What % does it allow, in the past it allowed 12.5% to 100% in 12.5% increments on PWM and on DC was 40% to 100% in 10% increments, so has this change?

2) Can 0% be reached on DC o PWM?

Im sure ill have some more questions later on.

Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Re: Sanity check for a workstation/home studio/light gaming

Post by Xinef » Fri Jun 23, 2017 3:54 pm

Abula wrote:Very interesting picture of the msi bios fan control, specially since its a SYS_FAN header that has PWM/DC switching options, i have some questions,

1) What % does it allow, in the past it allowed 12.5% to 100% in 12.5% increments on PWM and on DC was 40% to 100% in 10% increments, so has this change?

2) Can 0% be reached on DC o PWM?

Im sure ill have some more questions later on.
I answered those questions already xD

Sorry if I'm writing so much that details are flooded by my ramblings...

But yes, for 1) you can adjust in 1% increments, whether in PWM mode, DC mode, "smart fan mode" on or off, doesn't matter. The only visible change in the GUI is that in PWM mode it displays percentages and in DC mode it displays volts. And "smart fan mode" gives 4 control points instead of just one. The values you see here on the right are (defaults) for both modes, with the inactive mode grayed out.

Like here:
Image

and for 2) you can go down to 0%, but the fan would still turn at ~300 rpm, so I'm guessing it still has a hidden lower cap on voltage/pulse. Also, when holding the control point anywhere between 40% and 0% it displays a message that values below 40% can result in abnormal fan behavior and are not recommended, or somesuch. However my fans are spinning at ~25%-30% speeds without issues.

I'd have to check if the 1% changes are actually making a difference, or if it's only a GUI feature hiding old behavior in software, though :P

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Sanity check for a workstation/home studio/light gaming

Post by Abula » Fri Jun 23, 2017 4:06 pm

Thanks for the reply.

Well it seems that MSI is now the best in bios fan control, from what you said there are no limits or restrictions (so its same as asrock), it has 1% increments on 1C variations (same as asrock), 4 breakpoints for the graph (same as arock), but two things that has over asrock are

1) Switchable headers, like Asus.
2) All headers seems to be able to control PWM and DC (asrock only can do this on two headers, the rest are pwm).

overall seems like my next mobo its going to be MSI again, that said, i still think AsRock is offering pretty good mobos for the money.

Xinef
Posts: 7
Joined: Thu Jun 08, 2017 7:02 am
Location: Gdańsk, Poland

Re: Sanity check for a workstation/home studio/light gaming

Post by Xinef » Fri Jun 23, 2017 4:22 pm

I've just checked youtube for z270 MSI BIOS reviews, and one thing I noticed is that there appear to be some differences between the BIOS for the "gaming" mobos and the "pro" mobos.

"gaming z270"
https://www.youtube.com/watch?v=mcjH5dsteU8

"pro z270"
https://www.youtube.com/watch?v=lkt_jKaxz8s

Other than the aesthetics, I've noticed that gaming BIOS had a third fan mode "auto" which was not present in the "pro" BIOS. Can't really tell from those videos if there are other differences as well. Anyway, I'd just keep that in mind, since it means some features could be missing from one or the other.

Post Reply