S4 MINI Classic (S4M-C)

[Smallform Gaming]

http://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=879#overview
Biostar mini-itx board is official now. Only 1 m.2 slot

In the description....
"Support Dual Channel DDR4 3200(OC)/ 2933(OC)/ 2667/ 2400/ 2133/ 1866 MHz
Support Non-ECC & ECC Un-buffered DIMM Memory modules
2 x DDR4 DIMM Memory Slot"

Does both ECC and non-ECC support mean you can utilize the VLP ECC sticks from Kingston and Crucial?
*If so then I could use this with my 120mm Blue LED Gelid fans... and that's awesome.

https://www.amazon.com/dp/B01K1JVY22/?tag=theminutiae-20
https://www.amazon.com/dp/B01IBW74JI/?tag=theminutiae-20

 

[Nimikins]

In the description....
"Support Dual Channel DDR4 3200(OC)/ 2933(OC)/ 2667/ 2400/ 2133/ 1866 MHz
Support Non-ECC & ECC Un-buffered DIMM Memory modules
2 x DDR4 DIMM Memory Slot"

Does both ECC and non-ECC support mean you can utilize the VLP ECC sticks from Kingston and Crucial?
*If so then I could use this with my 120mm Blue LED Gelid fans... and that's awesome.

I'm not 100% sure, but that's the way I read it as well.

 

[zovc]

IIRC Ryzen supports ECC.

I don't understand why they would specify that it supports non-ECC RAM, as far as I understand most chips/chipsets that support ECC also support non-ECC... on my (Xeon) server board I'm using non-ECC RAM. (For shame!)

 

[Josh | NFC]

ECC ram should always be purchased if possible. Oh my goodness it takes care of so many instability problems! I love the stuff.

 

[|||]

http://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=879#overview
Biostar mini-itx board is official now. Only 1 m.2 slot

The single M.2 has been discussed in other threads. If there were two, you wouldn't be able to put them in RAID, anyhow. Just put applications and data you need for very fast loading on the M.2 and use SSDs for all other data...you won't notice that much of a detriment if you delineate the data fairly well, let you have more storage space, and it will be cheaper for you.

In the description....
"Support Dual Channel DDR4 3200(OC)/ 2933(OC)/ 2667/ 2400/ 2133/ 1866 MHz
Support Non-ECC & ECC Un-buffered DIMM Memory modules
2 x DDR4 DIMM Memory Slot"

Does both ECC and non-ECC support mean you can utilize the VLP ECC sticks from Kingston and Crucial?
*If so then I could use this with my 120mm Blue LED Gelid fans... and that's awesome.

https://www.amazon.com/dp/B01K1JVY22/?tag=theminutiae-20
https://www.amazon.com/dp/B01IBW74JI/?tag=theminutiae-20

Yes, so long as they are unbuffered (UDIMM).

IIRC Ryzen supports ECC.

I don't understand why they would specify that it supports non-ECC RAM, as far as I understand most chips/chipsets that support ECC also support non-ECC... on my (Xeon) server board I'm using non-ECC RAM. (For shame!)

Ryzen supports it so long as the motherboard and BIOS do, which is up to the motherboard manufacturer. I think they just included non-ECC to be explicit.

 

[Zero]

ECC ram should always be purchased if possible. Oh my goodness it takes care of so many instability problems! I love the stuff.

It has been suggested that ECC introduced additional latency since it runs a parity check each cycle, but when people actually tested it for gaming, it ran as well or better than normal ram: http://www.techspot.com/article/845-ddr3-ram-vs-ecc-memory/

It's a fucking shame DDR4 has been out for a year and no one will do non-ECC VLP... my ram sticks are blocking my fan from cooling the HDPlex more.

Speaking of, bonus shots of my case... I'll tidy it up and put the lid back on eventually but it looks so badass like this:

 

[Josh | NFC]

Actually that looks really awesome. I dig the assemtry. That's my jam.

I bet with some more doo-dads here and there it would be a world class mod.

 

[Curiosity]

@||| you have valid points, but for me personally the speed factor of M.2 is just a bonus you get, with the main benefit being no wires and smaller size.
I'm sure there's at least one person that agrees with me.

 

[CubanLegend]

@||| you have valid points, but for me personally the speed factor of M.2 is just a bonus you get, with the main benefit being no wires and smaller size.
I'm sure there's at least one person that agrees with me.

I agree with you. dual m.2 will be amazing once we have 1 and 2tb m.w drives on the cheap. Oh and having DUAL m.2 is even better because i can have one m.2 boot drive and one Intel Optane drive, you guys see the intel optane stuff? Kinda acts like a huge cache, its like having a nonvolatile 16 or 32GB cache thats faster than a hybrid SSD/HD's cache, but a little slower than RAM. Linus did a video on it, albeit sponsored by Intel.

 

[Nimikins]

I agree with you. dual m.2 will be amazing once we have 1 and 2tb m.w drives on the cheap. Oh and having DUAL m.2 is even better because i can have one m.2 boot drive and one Intel Optane drive, you guys see the intel optane stuff? Kinda acts like a huge cache, its like having a nonvolatile 16 or 32GB cache thats faster than a hybrid SSD/HD's cache, but a little slower than RAM. Linus did a video on it, albeit sponsored by Intel.

I did see that yesterday. My understanding is it's acts kind of like a third level cache, and people with high end ssds would get less benefit from it.
The only reason I care about dual m.2 is that a m.2 drive means less cables for me to run in the case, which means better airflow and a neater looking system.
Additionally, many of the CAD products that I use want to be installed on the same drive as the OS, meaning my 256gb Intel ssd is crammed to the max, and then I want to put my most commonly played games (such as ARK, which is over 100GB) on my fastest drive.

 

[zovc]

Holy cow. I never knew ARK took up that much space.

 

[Zero]

Yeah I'm not regretting dropping $480 for the 1TB 950 Evo. Boots in 4 seconds, keeps the S4 (considerably) clean since I don't have to muck about with 2.5 drives and cabling, and having a bunch of VR games on my drive adds a ton of GB on top of multiple 50+ GB installs like GTAV, Mankind Divided, and so on. I'm sure Optane will be insane-o priced and I don't really see the point if you have something as fast as the 960-series SSD.

 

[Curiosity]

Yeah I'm not regretting dropping $480 for the 1TB 950 Evo. Boots in 4 seconds, keeps the S4 (considerably) clean since I don't have to muck about with 2.5 drives and cabling, and having a bunch of VR games on my drive adds a ton of GB on top of multiple 50+ GB installs like GTAV, Mankind Divided, and so on. I'm sure Optane will be insane-o priced and I don't really see the point if you have something as fast as the 960-series SSD.

in the video Linus did he said the optane drives should be starting at 50$.
I think the idea isn't that people like us use it, more that it would be an easy upgrade to get ssd-like performance on a system with a hdd as the main/ only drive.

 

[|||]

in the video Linus did he said the optane drives should be starting at 50$.
I think the idea isn't that people like us use it, more that it would be an easy upgrade to get ssd-like performance on a system with a hdd as the main/ only drive.

That might be in Canadian Dollars. On Anandtech is says $44 for the 16GB model and $77 for the 32GB model.

 

[McTeags]

@||| you have valid points, but for me personally the speed factor of M.2 is just a bonus you get, with the main benefit being no wires and smaller size.
I'm sure there's at least one person that agrees with me.

I'm right there with you. I actually purchased a 480 GB SATA M.2 SSD for the form factor. It was less expensive (less than $120) than NVMe, lower thermals, got rid of some wires and the M.2 slot on my motherboard only has 2 PCIe lanes. Overall great purchase and recommend an M.2 for anyone doing SFF.

 

[CubanLegend]

guys...my Thermolab LP53 came in yesterday afternoon.

And I started work on temp benchmarking my current NHL9i this afternoon...

Here's a preview of what I tested with my NT-H1 thermal compound on my 7700K at 4.5GHz turboboost @ stock vcore 1.24v:

• IDLE temp
• LOAD temp (after 10 minutes) or LOAD temp once I hit 99c/thermal throttle with a time-stamp
• AMBIENT temp.

I'll be testing the following software:

• Prime95 26.6 (custom, run FFTs in place, & max FFT size of 8kb)
• ASUS ROG Realbench v2.43 Stresstest (set to use 8GB of RAM)
• a custom x264 load/stability test, on all 8 threads (custom test from the Overclockers.net Kaby Lake OC'ing thread)

I'm testing both coolers with the fins parallel to the RAM in the following configurations (with the S4 Mini side panels CLOSED):

• NHL9i +NF-A9x14 - STOCK (with motherboard VRM heatsinks ON)
• NHL9i +NF-A9x14 - without mobo VRM heatsinks
• LP53 +NF-A9x14 - without mobo VRM heatsinks

Right now, as I've just completed the benchmarks and am typing up the results.... the results are IMPRESSIVE. 1st, after removing the VRM heatsinks, I came across a decent temp drop that I will share in the results. 2nd The LP53, it WAS INDEED worth the 1 month wait after purchase to arrive, for sure, it has replaced my NHL9i going forward into my delid to occur tomorrow.

... okay I typed up the results before posting, so I merged them into this post.

NHL9i vs LP53 S4 Mini RESULTS (AMBIENT TEMP was 74F or 23C at all times):

Prime95_v26.6____IDLE__________LOAD
NHL9i Stock______39c_________99c@5min
NHL9i no VRMs___39c____________97c
LP53 no VRMs____36c____________92c

Realbench Stress__IDLE__________LOAD
NHL9i Stock______39c_________99c@2min
NHL9i no VRMs___39c____________94c
LP53 no VRMs____36c____________90c

x264 Custom Load_IDLE__________LOAD
NHL9i Stock______39c__________99c @ 2min
NHL9i no VRMs___39c________(94c@2min) 97c
LP53 no VRMs____36c________(92c@2min) 95c

RESULTS ANALYSIS:
Removing the VRM heatsinks gave me anywhere from a 2-4c improvement in load temps when using the NHL9i. this also allowed my load temps not to fail before the 10 minutes, where they were hitting 99c otherwise.
The LP53 beat the NHL9i in the same configuration without the VRM heatsinks, it dropped my idle temps by 3c, and my load temps dropped between 2-5c.

So in here's the kicker! I went from... using the NHL9i with the VRM heatsinks (which I was using all this time), to the LP53 without the VRM heatstinks. This resulted in my CPU no longer throttling and failing my load tests, while dropping my load temps from reaching 99c in 2 & 5 minutes, to maintaining my load temps at a steady 90/92c after the full 10 minutes! That's an average of -9c on my load temps! on air! inside the S4 mini!

Pics and LP53 fitment story "1 Small compromise, for a no-compromise All-copper CPU cooling solution!" :

First step was removing the VRM heatsinks after I tested the NHL9i with them ON... the 2 screws for one of the VRM heatsinks was easy to access from the bottom of the mobo, the other two were hidden behind the S4 mini sidepanel, requiring I pull out the HDPLEX, disconnect SATA from the mobo, disconnect my PCI-Ex riser, to slide the mobo over to access the other two screws:

Screws:

how much I had to slide over the motherboard to reach the last 2 screws:

VRMs off with my screw pointing at the VRMs (little black squares)

VRM heatsinks off (I'll have mini heatsinks coming this Friday for all the exposed MOSFETS/VRMs, will install with Pics at that time)

My thermal paste application of NTH1 had great coverage (amount was the size of 1 uncooked rice grain):

LP53's ALL COPPER mounting plate/surface and fins :

LP53 vs NHL9i + NF-A9x14 size comparison NOTE, the LP53 actually overhangs the mounting screws more on the pointy end of the heatpipes, as the LP53 is offset a bit, so that the bent ends of the heatpipes actually protrude LESS than the other end:

Top: NHL9i mounting hardware, bottom: LP523 mounting hardware... truly amazing how minimalistic it is!

CPU cleaned and ready for LP53 mounting fitments tests:

LP53 mounting tests, without having to remove anything (and I even refitted the VRM heatsinks just to verify, although not pictured) you dont have to remove RAM or VRM heatsinks to fit the LP53 with the heatpipes over the M.2 slot, BUT this sadly forces the Heatsink's fins to be perpendicular to the RAM, limiting cooling potential:

But with the fins parallel to the RAM, the LP53 protrudes OVER the first RAM slot on my ASUS STRIX z270i motherboard. Forcing me to pray that it will fit with the heatpipe bends facing the RAM instead, so I don't have to sacrifice RAM...

Here you can see that the LP53 is offset from the motherboard mounting holes by quite a bit when pushed up against the first RAM stick, if trying to fit it with the fins parallel to the RAM...

AAAAND! IT FITS! But see if you spot the small sacrifice I made:

I removed the first RAM slots outer heatspreader, as it was BARELY making contact with the LP53's heatpipe bends... and I was NOT comfortable with that. So you may not even have to do this, if you don't mind the heatpipes touching your RAM.

So now comes the hard part, figuring out how to mount the NHL9i's fan onto the LP53!

I used black & thin plastic zip ties, I looped them over the pointy ends of the heatpipes snugly, but I had to loop them under the start of the bendy end of the other side of the heatpipe. The first pic is showing how much I had to slowly, notch by notch, tighten and refit the the zip-ties until they kept the fan SNUG against the LP53:

PROTIP: do the pointy ends first, at they can be lifted off and retightened easily, once you do the bendy ends, you cant move the fan off the heatsink to have enough room to loop the remaining zip-ties easily:
[

Then just cut your ends off:

Reinstall the side panel, and you'll have 5mm of space between the fan and the side panel, where the NHL9i had 9.5mm of space. Your fan might have a slightly more pronounces WOOSH sound, or hum at higher loads, but its not very noticeable to me. My S4 sits 3 feet away at ear level on my desk in a quiet room.

If there's any other temp tests you think I should do before my CPU delid in 2 days, PLEASE LET ME KNOW.

 

[zovc]

So a challenger approaches! Do you think low profile ram would have been short enough to avoid the clearance issues you had?

I don't see Realbench scores for the LP53 under load. Was that an error, did it overheat, or did you not get to that yet?

 

[CubanLegend]

So a challenger approaches! Do you think low profile ram would have been short enough to avoid the clearance issues you had?

I don't see Realbench scores for the LP53 under load. Was that an error, did it overheat, or did you not get to that yet?

Sorry @zovc I forgot to add that result. It's been edited in there now.

i also added a short "results analysis" and a "conclusion", after the results section.

EDIT: Also, about your mention of low-profile RAM... I dont think you'll need to worry about it. I ACTUALLY HAD clearance to keep the heatspreader, but it would be touching the heatpipes just BARELY, you see.

If your RAM has a heatspreader, i'd venture to say it's okay if the LP53's bendy end touches it because its touching a heatspreader, not the RAM itself, so your RAM should be safe.

Just to emphasize why I did remove that heatspreader: I just wanted to ensure there was ZERO contact with the RAM (even if just a little), as I dont want ANYTHING from the CPU to be in contact with my DDR4-4000MHz RAM modules, even a little bit. Those modules were expensive and I don't want any variables limiting their sound operation for years to come as a result of their contacts being even slightly bent. I kept the heatspreader in a plastic baggy though, in case i ever want to reinstall it. The thermal goop stayed perfectly intact to a reinstall will be a peice of cake.

 

[Curiosity]

Very interesting.
How does the height of the LP53 W/ noctua can compare to the L9i?

 

[CubanLegend]

Very interesting.
How does the height of the LP53 W/ noctua can compare to the L9i?

The LP53 seems to be 4.5mm taller than the NHL9i. I mention it in the last part of the most that the NHL9i gave me 9.5mm of space to the start of the side panel, where this LP53 now leaves me roughly 5mm (I eyeballed it lol) but I'll measure it for real once I get home again this afternoon and will let you know the REAL measurement.