![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
DAN A4-SFX: The smallest gaming case in the world
- Thread starter dondan
- Start date
SpartanM07
n00b
- Joined
- Apr 9, 2017
- Messages
- 11
Looks great! Where did you get the foam gasket?
I need to do more low RPM testing, but signs point to 4.6-4.7 GHz being ideal for quiet sound levels around ~70 °C.
Actually I didn't get the adafruit ones, I used extras left over from a GPU mod from a couple years ago. They look identical though?
I think it was this one, again recycled from past modding projects, I don't remember exactly.
As an Amazon Associate, HardForum may earn from qualifying purchases.
Did you have to do any modding to the FTW cooler to get it to fit on your Titan? I'm wondering if I could get this cooler to work on a 1080 Ti FE.
I reapplied the Liquid Thermal Ultra & the Thermal Grizzly but temps are still out of control. 3:40 into Aida 64 it his the thermal limits. This is with the sides of the case off. The Noctua fan is only spinning at 2000rpm max?. I also have it set with the sicker facing the LP53.
I have no idea how some of you are temping in the 60-70. Thermal paste applications are not THAT hard to perform. Im confident its making solid and even contact.
I have no idea how some of you are temping in the 60-70. Thermal paste applications are not THAT hard to perform. Im confident its making solid and even contact.
Last edited:
dondan
[H]ard|Gawd
- Joined
- Apr 15, 2013
- Messages
- 1,751
Chapeau
Gawd
- Joined
- Jul 17, 2016
- Messages
- 753
Full disclosure - I don't have an A4, instead a Ncase M1 WC build.
I just don't see how your thin fan can effectively drive the radiator. A rule of thumb is that a fan needs 1 fan diameter for intake and for exhaust to function correctly.
Not many people have this kind of room, so it should really be taken with a grain of salt so to speak, but with your case just mm off the floor, a thin fan and a thick radiator, things don't look too good.
Have to tried with a simple (even stock) air cooler to wee whether it really is a thermal compound issue?
I experienced something similar with my bottom radiator - thin fans coupled with a dense 240mm radiator meant my temps were worse than on air initially.
As promised:
The FLIR is out of calibration and only has ± 5 °C accuracy at best, but that's okay because it's still useful for identifying hot spots. The MOSFETs in the exhaust path of the CPU heatsink are the hottest.
A quick check with an IR thermometer shows around 64 °C at the same hot spot, but these devices have limitations too.
Another way to measure the VRMs is using a thermocouople (type T) with a bit of polyimide tape to secure it in place. It's not perfect, but now the side panel can be put back on.
Thermocouple results after 20 minutes of Prime95 (Ambient: 22 °C):
illram
[H]ard|Gawd
- Joined
- Sep 19, 2011
- Messages
- 1,473
Nice. Your IR thermometer temps are the same as mine. The exhaust area is (predictably) the hottest part of the board. I also have similar looking heatsinks on my mosfets. (From adafruit). Tiny but they get the job done!
Sebastian @PCperspective reviewed the case, have a look at: https://www.pcper.com/reviews/Cases-and-Cooling/DAN-Cases-A4-SFX-Crowdfunded-Mini-ITX-Case-Review
SpartanM07
n00b
- Joined
- Apr 9, 2017
- Messages
- 11
Yes, I cut down the base plate probably more than I needed too but I did not want to cause a short by contacting any components (cuts could have been cleaner but I did not have my jigsaw at the moment so I used my dremel):
I used a dremel to cut down some fins to make room for components and the power plugs:
I used some M2.5-0.45 threaded hex nuts and some non-conductive washers for 7 of the 16 mounting holes (3 pictured) so I could install the backplate:
I first installed the 3 hex nuts/washers that would be covered by the base plate (seen in picture above), then the base plate and the other 4 hex nuts/washers, then the backplate, and lastly the fan/heatsink assembly:
The only thing that doesn't work is the LED lighting as the stock PCB does not have a second 4-pin plug. The mod was fun but if a aftermarket 1080 ti fan/heatsink would have popped up on eBay I would have gone with that since no mods would have had to been made. I went with this cooler mainly because it was a good price and aftermarket coolers rarely appear on eBay.
Last edited:
Ca11idus
Gawd
- Joined
- Jan 9, 2012
- Messages
- 519
Can you make another one of these for me? I will give you the monees. Also what fan gasket did you use?
I would take the .STL file that I posted to your favorite proto shop and get it printed in a color other than barf creme. Shapeways is okay (free shipping on first order).
Attachments
Yeah - as I mentioned, 65 was the absolute max with the case side off and a Cryorig C7 (default fan) in Aida64 after about 20 minutes in. I ran an encode under Handbrake for a bit and peaked around 80 or so. I'm curious how I would fare with the L9i (I'd rather not swap fans, etc. even though I know that's optimal). I have one so I may try it unless I feel too lazy to take the motherboard out again.
Hope you get yours sorted out soon!
thanks for the answer, can you even say when do you expect the first shippings to start?
ahmedelamin_
n00b
- Joined
- Dec 27, 2016
- Messages
- 31
Try swap your fan on the c7 for a Noctua a9x14 or ty100.
Best cooler seems to be the Cooltek LP53 but difficult to get your hands on it. I've emailed them and they have said they are currently doing a production run that should be wrapping up this month so should be more units hitting the market over the next few weeks.
Note: I have the Noctua NH-L9i and it is silent at idle and pretty quiet under load. I had to check it was actually spinning a few times!
Hello my dears!
After a lot of trying and testing I've managed to setup my 22 Cores and 44 Threads Xeon E5-2696 V4 properly.
(and the $%&**$%& Turbo Boost 3.0 Driver that was continuosly crashing).
Et voila!!!
It has broken the 3.000cb mark on cinebench R15, without any overclocking or tweaking, reaching 3.105cb .
This may be the fastest rendering machine per volume on earth, thanks to the A4-SFX.
It is:
· 17% faster than the 2646cb mark that an E5-2699 V4 did in this video link,
more or less the same that Linus got with these two E5-2699 v4 (5.214cb / 2 = 2.607cb) link
· 49% faster than the 2.114cb 4,2ghz oveclocked i7 6950x link
(this configuration can't be managed inside A4-SFX because of the heat)
· 74% faster than the 1.778cb stock i7 6950x link
· 97% faster than the 1.570cb 12 cores top end mac pro link
I don't know how much that mac pro costs but the E5-2696 V4 costs 1300€ on ebay from china, less than the 6950x.
So yes, the E5-2696 V4 is 17% faster than the E5-2699 V4.
(and that explains the question of why the first has a tdp of 150W and the second a tdp of 145W).
-------------------------------------------------------------------
I was continuosly making 2800cb because I wasn't realizing that RealTempGT, for whatever reason, crashed turbo boost 3.0 driver till I rebooted the compuer, stucking 8 cores on 2,2 ghz. all the time.
Since I had it all the time enabled to check the temps, I didn't notice the problem till I switched to HWiNFO64 (I thought it was the normal behavior of turbo boost, and was very happy with 2.800cb).
· Here you can see what happened while rendering with the RealtempGT- TurboBoost problem:
· After fixing it, this is how it looks while rendering (some times one core suddenly jumps to 3,7ghz, but thats so quick and random that i can't make the printscreen):
· And this is how it looks while using regular apps. It can stuck one core to 3,7ghz and the others in several clock speeds.
------------------------------------------------
I want to thank Dan for allowing the build of this dream "portable" workstation.
After a lot of trying and testing I've managed to setup my 22 Cores and 44 Threads Xeon E5-2696 V4 properly.
(and the $%&**$%& Turbo Boost 3.0 Driver that was continuosly crashing).
Et voila!!!
It has broken the 3.000cb mark on cinebench R15, without any overclocking or tweaking, reaching 3.105cb .
This may be the fastest rendering machine per volume on earth, thanks to the A4-SFX.
It is:
· 17% faster than the 2646cb mark that an E5-2699 V4 did in this video link,
more or less the same that Linus got with these two E5-2699 v4 (5.214cb / 2 = 2.607cb) link
· 49% faster than the 2.114cb 4,2ghz oveclocked i7 6950x link
(this configuration can't be managed inside A4-SFX because of the heat)
· 74% faster than the 1.778cb stock i7 6950x link
· 97% faster than the 1.570cb 12 cores top end mac pro link
I don't know how much that mac pro costs but the E5-2696 V4 costs 1300€ on ebay from china, less than the 6950x.
So yes, the E5-2696 V4 is 17% faster than the E5-2699 V4.
(and that explains the question of why the first has a tdp of 150W and the second a tdp of 145W).
-------------------------------------------------------------------
I was continuosly making 2800cb because I wasn't realizing that RealTempGT, for whatever reason, crashed turbo boost 3.0 driver till I rebooted the compuer, stucking 8 cores on 2,2 ghz. all the time.
Since I had it all the time enabled to check the temps, I didn't notice the problem till I switched to HWiNFO64 (I thought it was the normal behavior of turbo boost, and was very happy with 2.800cb).
· Here you can see what happened while rendering with the RealtempGT- TurboBoost problem:
· After fixing it, this is how it looks while rendering (some times one core suddenly jumps to 3,7ghz, but thats so quick and random that i can't make the printscreen):
· And this is how it looks while using regular apps. It can stuck one core to 3,7ghz and the others in several clock speeds.
------------------------------------------------
I want to thank Dan for allowing the build of this dream "portable" workstation.
Last edited:
The C7 heatsink is as good as any, so just lower the fan speed through bios/software/adapter!
kingtron
Limp Gawd
- Joined
- Feb 2, 2017
- Messages
- 254
pcpartpicker
DAN Cases A4-SFX Mini ITX Build (April 20)
https://player.twitch.tv/?volume=0.5&video=v136909963&time=09m40s
DAN Cases A4-SFX Mini ITX Build (April 20)
https://player.twitch.tv/?volume=0.5&video=v136909963&time=09m40s
Last edited:
Check out my post here: https://hardforum.com/threads/dan-a4-sfx-the-smallest-gaming-case-in-the-world.1799326/page-264
Shortened down version:
Please note that these temps can't represent "Real World Performance", I mostly just wanted to get an overview of the temperatures quickly - all these benchmarks are taken in the following way:
- 5 min
- Open Test Bench
- Idle temp in bios
- Bios reset, Ram upped to 2400mhz
Summary
- If you ask me today if LP53 is worth it? I really do not know. My results indicate that the C7 + Noctua fan might be a better option (more cost effective)
- I would recommend you try a C7 + a NF A9x14 fan on it
- The NF A9x14 is way quieter than the the C7 stock fan
Sound Levels
Some notes about the noise during the bios temp benchmarks:
- LP53 + NF A9x14 (quietest)
- C7 (slightly higher)
- Stock Intel (loudest)
Want to save money?
- Get C7 + NF A9x14
Got money to spare and want a kick ass full copper CPU cooler?
- Get LP53 + NF A9 x14
Last edited:
Check out my post here: https://hardforum.com/threads/dan-a4-sfx-the-smallest-gaming-case-in-the-world.1799326/page-264
Shortened down version:
Please note that these temps can't represent "Real World Performance", I mostly just wanted to get an overview of the temperatures quickly - all these benchmarks are taken in the following way:
- 5 min
- Open Test Bench
- Idle temp in bios
- Bios reset, Ram upped to 2400mhz
Summary
- If you ask me today if LP53 is worth it? I really do not know. My results indicate that the C7 + Noctua fan might be a better option (more cost effective)
- I would recommend you try a C7 + a NF A9x14 fan on it
- The NF A9x14 is way quieter than the the C7 stock fan
Sound Levels
Some notes about the noise during the bios temp benchmarks:
- LP53 + NF A9x14 (quietest)
- C7 (slightly higher)
- Stock Intel (loudest)
Want to save money?
- Get C7 + NF A9x14
Got money to spare and want a kick ass full copper CPU cooler?
- Get LP53 + NF A9 x14
Thank you! Question though, doesn't the Noctua push less air?
HoodedHusky
n00b
- Joined
- Mar 21, 2016
- Messages
- 42
Got money to spare and want a kick ass full copper CPU cooler and with 4mm standoffs?
- Get LP53 + NF A9 PWM (over 35% greater air flow over the thinner Noctua)
Last edited:
Does that combination actually fit in the case?
HoodedHusky
n00b
- Joined
- Mar 21, 2016
- Messages
- 42
It is exactly what I am using. Is tight, but fits with no bowing of the side case panel.
Last edited:
Alright, thank you. Now I just need to find it cheaper.
This was already posted.
HopkinsTea
n00b
- Joined
- Nov 25, 2015
- Messages
- 35
For the life of me, I can't find the thread with all the pictures of people's builds for A4, can someone link me?
richiegore
Limp Gawd
- Joined
- Feb 1, 2017
- Messages
- 322
So the NH L9X65 uses the a9x14 fan, they could have made a shorter heatsink and used a 25mm fan. To me that means that the a9x14 plus an extra 11mm of heatsink is a better tradeoff than the increased airflow provided by a 25mm fan and a shorter heatsink. I trust noctua on this.
This leads me to think that the cryorig c7 with the a9x14 should be a great performer. I can't find much showing that in the excel document though. Does anyone have the two to compare?
This leads me to think that the cryorig c7 with the a9x14 should be a great performer. I can't find much showing that in the excel document though. Does anyone have the two to compare?
Hi Jack,
I have an similar setup to you, in the sense that I'm currently using the 1503 and 1303 GeChic portable monitors to work with my A4-SFX. They are both too big for any bag which fits the A4 snugly.
I would have suggested yet another bag but...it's pricey and would allow you both the Tenba AND the Messenger 250.
I need to figure out how to share images to better describe my set up. [ I travel across the world for work, and this allows me to bring a crazy 980/1080GTX setup with me to my hotel room ]
SpartanM07
n00b
- Joined
- Apr 9, 2017
- Messages
- 11
Credit to ceski for the initial idea! I don't have a 3D printer so I used some acrylic from a hardware store, my drill and jigsaw, double-sided sticky tape, and a Phobya 140mm foam gasket from Newegg to make myself a shroud. Dropped idle temps by 3-5C and load temps from 5-10C. That's on a de-lidded, undervolted 3770k at 4.0GHz under an LP53+Noctua a9x14.
For those who are using the shrouds, are you noticing an increase in sound reflection and thus louder fan noise for a given RPM? I noticed with my shroud the noise levels that temps are lower but despite the lower fan RPM, noise reflection meant that it was still louder. Maybe the use of foam in the recent shrouds mean this is less of an issue.
I don't notice additional noise due to reflections, maybe because it's not flared out like yours or maybe because the foam absorbs some of the high frequency components of the sound. However, the seal the duct creates forces the air to be pulled in through the ventilation holes rather than the partial recirculation from within the case that occurs without the duct. This means that at high fan speed (>2000 RPM), there is definitely a slight suction sound that is not present without the duct. At reasonable fan speed, there is no difference in sound.
richiegore
Limp Gawd
- Joined
- Feb 1, 2017
- Messages
- 322
Because this shroud is flat you could easily add some thin sound dampening foam that you can get for applying to computer case panels, I doubt it is worth it though, nice mod!
dondan
[H]ard|Gawd
- Joined
- Apr 15, 2013
- Messages
- 1,751
I think if you flip the LP53 to 180° you can lower the temps of another 5°C because the heatpipes will have a more efficent orientation. In you orientation the fluid inside have to work against gravity.
By the way if you don't have a 3D printer you can easily buy this low profile fan duct 92mm to 120mm to get the same effect:
Bitspower Ultimate 92mmm to 120mm (BP-FA1208090-BK)
That's another sweet mod. I'd be interested in seeing if changing the fan from push to pull changes the performance.
A quick video of the E5-2696 V4 cinebench.
After disabling turbo boost 3.0 driver it is performing better (turbo boost works anyway).
It made 3176 cb, so it is now:
· 20% faster than an E5-2699 V4.
· 50% faster than a 4,2ghz overclocked i7 6950x.
· 78% faster than a stock i7 6950x.
· 100% faster than a 12 cores top end mac pro.
Temps are 80 - 82 ºC on full load, with the Dynatron T-318, so it is perfectly suitable for the A4-SFX.
After disabling turbo boost 3.0 driver it is performing better (turbo boost works anyway).
It made 3176 cb, so it is now:
· 20% faster than an E5-2699 V4.
· 50% faster than a 4,2ghz overclocked i7 6950x.
· 78% faster than a stock i7 6950x.
· 100% faster than a 12 cores top end mac pro.
Temps are 80 - 82 ºC on full load, with the Dynatron T-318, so it is perfectly suitable for the A4-SFX.