Stereodude
Not really a
Definitely the former.
question Where'd the X79 motherboards go?
- Thread starter Stereodude
- Start date
mubs
Storage? I am Storage!
How can a well-known brand ship such crappily finished product? Sheesh.
Stereodude
Not really a
I'm not sure what you mean. It was perfectly shiny and smooth when I got it. It just wasn't flat. It's now very flat, but a bit ugly looking.
Howell
Storage? I am Storage!
So it that supposed to be copper?
Stereodude
Not really a
That depends if you're asking if the exposed material is copper, or if you're asking if the entire base is supposed to be copper. It was shiny chrome looking originally. They put a layer of copper down first during the chrome plating process. When I sanded down the high spots I cut through the chrome and nickel layers and into the copper layer. Hence the exposed copper.
Howell
Storage? I am Storage!
I understand what I'm seeing now. I had thought the bottom was convex not concave. What percentage of the base is the cpu heat spreader?
Stereodude
Not really a
Almost all of it. It only overhangs the CPU by a few mm on each side.
CPU issues, motherboard issues, RAM issues, heat-sink issues, noisy fan issues, ebay issues...what's next? This system just does NOT want to go smoothly. I wonder if this is karma payback for something in your computer building past? If not, you are going to have a full karma bank and the next 4 projects should go uber smoothly.
<fingers crossed> Hopefully the rest of the build will be issue free.
<fingers crossed> Hopefully the rest of the build will be issue free.
I got all my remaining parts in this morning and got time this evening to put one of the two systems together. The Intel case and components are well-engineered. Everything just seemed to fit together nicely. After removing the included low profile heat sinks that shipped with the board I cleaned out their thermal past and removed each CPU from their socket to inspect them. Both were in good health with no bent pins or any dust/crap left in the socket area. One small thing I noticed is that possibly during their extraction someone left a screw wedged on top of the motherboard near a RAM slot clip. I'm happy I noticed it before powering everything on but it didn't seem to cause any issue.
The install went well and the biggest complaint I might have was trying to get the fans to slow down. After spending some time updating the firmware, BMC, FRU, etc, I didn't realize two things contributed to the fan speed noise. One is the case is well-engineered enough to know when the side lid is left open thereby causing the fans to run at 100%. I wasn't aware it has an intrusion detection switch. Second, in order for the acoustic thermal changes to be applied I had to power cycle the entire box meaning, discount the power cable for 30 seconds to cause the BMC to power cycle. So far the noise level isn't that bad when the server is idle. I haven't been able to get an OS loaded in order to run some load on the CPUs so I'll be curious to see how the fan noise with that. I tried installing ESXi 6.0 but it died when trying to partition my HDD. I then tried to install Xubuntu on it and it wouldn't even load the installer. Now I've booted into memtest86 and I'm going to let that run through the night to see if there is any issues with the RAM.
The install went well and the biggest complaint I might have was trying to get the fans to slow down. After spending some time updating the firmware, BMC, FRU, etc, I didn't realize two things contributed to the fan speed noise. One is the case is well-engineered enough to know when the side lid is left open thereby causing the fans to run at 100%. I wasn't aware it has an intrusion detection switch. Second, in order for the acoustic thermal changes to be applied I had to power cycle the entire box meaning, discount the power cable for 30 seconds to cause the BMC to power cycle. So far the noise level isn't that bad when the server is idle. I haven't been able to get an OS loaded in order to run some load on the CPUs so I'll be curious to see how the fan noise with that. I tried installing ESXi 6.0 but it died when trying to partition my HDD. I then tried to install Xubuntu on it and it wouldn't even load the installer. Now I've booted into memtest86 and I'm going to let that run through the night to see if there is any issues with the RAM.
Stereodude
Not really a
I figured that out when I opened up the case and it got super loud. I put tape over the switch for the sake of testing.
I didn't have to do that. Simply saving the settings in the BIOS and letting it go through reset worked for me.
I used a bootable USB with the Ultimate Boot CD written to it to run mprime (small FFT) and I was unable to get the fan speeds to budge from their idle RPMs. Even after 15+ minutes of running. Ambient temp is about 65F. Because it runs from the console with no GUI I can't monitor the temps while it's running. Well, I might be able to use IPMI over LAN to see the temps, but I don't have LAN connected and haven't tried that. Exiting mprime, and using the utility on the bootcd to check temps (takes a few seconds to switch) showed CPU temps were still only in the 60's using the Acoustic fan profile. Using the Performance fan profile the temps were in the upper 50's after almost 1 hour of mprime. Next step is to get Windows on it so I can watch temps while Prime95 is running.
Loaded power draw was not as high as I was expecting at ~291W with the Acoustic fan profile and mprime (Small FFT) with my SSD and 2 7200 RPM HDs connected and powered. The same test with the Performance fan profile gave me ~293W. I don't have my graphics card (GeForce GT720) in the system yet since it hasn't arrived yet. I guess I could have safely gotten one of the 550W chassis though I don't know that they would have been any cheaper and they only silver rated, not platinum.
Edit: I need to figure out the fan situation. I haven't decided if I want 3000 or 2000 RPM Notuas. As of now I don't see having more than 2 PCIe cards. A rather low power graphics card, and maybe a 10Gbe ethernet card. I don't think they'll need much airflow. So, I would think a 2000RPM should suffice there though with a PWM in the low 20's it will only spin like 500RPM. I don't know if that will trigger an alarm with the BMC like the problem I'm having with my Supermicro or not. The CPU fan is a little more tricky IMHO. It seems the system will either run it at ~28.5% PWM or ~46% PWM depending on whether you use Acoustic or Performance. 28.5% on the 2000RPM is likely to be pretty slow in terms of RPM and airflow. I could use the Performance profile to spin the 2000RPM fan faster, but using acoustic with the 3000RPM fan would probably achieve the same net effect and give it more cooling headroom.
Last edited:
I'm going to use the included fans for a while and see how things go for me Given these will live in my basement and I'll be using the RMM4lite for management I don't care about a little noise so long as it isn't the craziness I heard when it first ramped up. Good points about the fan RPMs. I was hesitant to link to or suggest the 2000RPM version of that Noctua because of similar reasons you described but mostly that at 100% it may not be enough to keep the system cool if it needed to. One thing though you might want to consider is the heat generation on whatever 10Gb NIC you decide to get. Those can run warm sometimes. I'll also be looking into a couple 10Gb adapters down the road once I make progress with my projects.
I've got about 8 hours of memtest under the belt so far on one of the systems but it hasn't finished a single round yet. Hopefully by this evening it has completed at least one so I can move on to prime95 stress testing while I build the other system. I missed out on the SSD sale at newegg. They sold out of the Samsung 850 pros in the size I wanted so I'm using my 1.5TB samsung eco green drives temporarily to get me going until I can decide on an SSD to use.
Stereodude
Not really a
If I keep using the included fans it's going in my basement. However, I'd like to put it in my library. PWM fan controls are supposed to be linear from 20% to 100%. I need to plot the two Noctuas so I can estimate RPM against PWM duty to see about how fast each will spin with the two profiles based on my measurements.
I ran memtest86+ & Goldmemory on both of my E5 / LGA2011 systems. However, I don't have the final RAM for either since each set of RAM I bought had a bad stick. I borrowed from each other to test. I tested the dual E5-2670 Intel board with 32gB per processor. I've tested the Supermicro X9SRA with 64gB with an e5-2670 (mostly to make sure all 8 slots were fine), and then with 32gB with an E5-2689. I haven't tried Prime95/mprime on the Supermicro.
I'm looking at how I'm going to work around the FAN PWM nonsense on the X9SRA this morning. Right now I plan to use a level shifter to take the 12V RPM signal from each of the fans drop it to 5V for a microcontroller to measure, the micro will multiply the frequency by some amount, right now I'm thinking 3x, and output that new signal to a different level shifter that will turn the 5V output from the micro back into 12V. I connect those 12V signals to the motherboard. Then the motherboard will think the fans are running 3x as fast and I won't be dealing with the stupid fan pumping anymore while using the fans I already have and I don't lose PWM control from the motherboard.
Stereodude
Not really a
So I did the math. I think the 3000 RPM makes more sense. However, since I haven't seen the PWM's budge yet I'm not quite ready to pull the trigger. I will see about installing Windows 7 Pro and digging a little more into what it does under load. I also get my RMM4lite iKVM module today so I will see how well that works. If it works well I may just stick it in the basement and not worry about the fans.
Did you enable the quiet mode in the BIOS for the fans? Also what altitude did you select? I enabled it and left the default altitude which I believe was the lowest.
I'm unclear how the RMM4lite would work with another graphics card installed. I assumed it redirected the output of the built-in graphics chip which is pretty lousy for anything other than basic admin tasks. I like it so I can mount iso remotely and install the OS, troubleshoot, etc. Once I'm done I don't use it for using the OS in any real way.
Stereodude
Not really a
I set it to the quietest mode the BIOS supports (Acoustic / lowest altitude / etc.) short of the mode that will stop the fans under light loads. The Nidec BETAV fan (upper fan) has a terrible noise characteristic, or at least mine does.
The GeForce GT720 is intended to be a hardware accelerator for the video decoding portion of transcoding / x264 encoding. I wasn't planning to hang a monitor on it. For my current server in the basement I pretty much entirely use Remote Desktop. There's a limitation that the hardware video decoding engine on the Nvidia cards can't be initialized when using remote desktop. I've worked around it before while using Remote Desktop, but it's inconvenient. Basically you get the job all setup and have the computer auto login and run the encode jobs as part of the startup. So, you initiate a reboot and it will start the job on it's own after the reboot.
Since encoding would be the primary function of this PC, that's not a good solution. I need to test the iKVM function against something like UltraVNC or similar. If they work well enough I could use them in place of remote desktop. If not, I can get everything set using remote desktop but emulate a local login using the iKVM or a VNC type program to start the process. My intent is to getting everything setup, tweaked, and adjusted on one of my other PCs, copy over the files to be worked on to the dual E5-2670 box and kick it off. Some paths may need to be adjusted when moving it from one machine to another. I could probably script the path adjustments to the few file with perl or Haskell. But I don't have a good solution to remotely opening a command prompt and executing a .bat file without doing it visually from VNC, iKVM, etc.
Mercutio
Fatwah on Western Digital
Does the S2600CP board you guys are buying have a rectangular or square LGA2011 socket?
Stereodude
Not really a
It does not have the narrow ILM. It has the square.
In other news I received an AXXRMM4R from OEM XS, not an AXXRMM4 Lite It has the separate board with the NIC. From my quick Googling it appears to be the direct dock version of the AXXRMM4. :scratch: It also has the little plug in "key".
In other news I received an AXXRMM4R from OEM XS, not an AXXRMM4 Lite It has the separate board with the NIC. From my quick Googling it appears to be the direct dock version of the AXXRMM4. :scratch: It also has the little plug in "key".
The lite version also comes with a separate board with a nic but no ribbon cable to connect it. All you need is to install the little plug (key) into the slot in the location circled below and after you restart the system you can access the BMC through the web. You'll need to activate an account in order to log in. I enabled the root account for my own config but I'm not sure what all the others restrict.
Stereodude
Not really a
I plugged in the little module to the board. How do I change the password or set up a password?
Edit: Nevermind, I got it all working.
Last edited:
Stereodude
Not really a
The system pulls a lot more power in Windows. Idle is around 125W and under load it's 380W or so with Balanced Prime95. Aside from running Windows, I plugged in an Ethernet cable, and added my GeForce GT720. That can't account for an extra 100W though. The GeForce GT should be idle. It has no display plugged into it.
Stereodude
Not really a
I went back and remeasured under Linux. The power draw numbers were basically the same. No idea why it pulls more power in Windows. The CPUs seem to get hotter in Windows too.
I don't get Intel's thermal strategy. Basically they leave the fan RPMs alone until the CPUs get to just over 80C then it will make small tweaks the the fan RPMs to hold the temps at 80C. I would prefer something more like desktop systems where they gradually ramp the fans as the CPU heats up. Acoustic and performance both do the same thing. Only the idle RPMs are different.
Alternatively, I wish I could adjust the temperature it is trying to hold the CPU at.
I don't get Intel's thermal strategy. Basically they leave the fan RPMs alone until the CPUs get to just over 80C then it will make small tweaks the the fan RPMs to hold the temps at 80C. I would prefer something more like desktop systems where they gradually ramp the fans as the CPU heats up. Acoustic and performance both do the same thing. Only the idle RPMs are different.
Alternatively, I wish I could adjust the temperature it is trying to hold the CPU at.
Good to know about windows usage causing increased power and heat. I could make up theories as to why but it won't help. 
I'm guessing their strategy is to not really care because this board has a market aimed at lab usage rather than desktop. I agree that with some refinement of the thermo algorithm this could be addressed in a reasonable way. Which fan header connectors did you use on the motherboard? I used system 1 and system 2 for mine. I haven't had enough time to experiment with the thermos of the board. I spent 3-4 hours last night identifying the issue with my second system rather than being able to make progress with setting it up as an ESX server. I messaged the ebay seller for a return on the case because of the PSU but we're in discussion to see if there is an alternative. They're concerned it's the motherboard but the board posts and operates when I use the known-working case. I even updated all the firmware, BIOS, FRU, etc while it was connected to the working chassis and that all went fine. I'm not looking forward to having to ship the 45lb box back.
I'm guessing their strategy is to not really care because this board has a market aimed at lab usage rather than desktop. I agree that with some refinement of the thermo algorithm this could be addressed in a reasonable way. Which fan header connectors did you use on the motherboard? I used system 1 and system 2 for mine. I haven't had enough time to experiment with the thermos of the board. I spent 3-4 hours last night identifying the issue with my second system rather than being able to make progress with setting it up as an ESX server. I messaged the ebay seller for a return on the case because of the PSU but we're in discussion to see if there is an alternative. They're concerned it's the motherboard but the board posts and operates when I use the known-working case. I even updated all the firmware, BIOS, FRU, etc while it was connected to the working chassis and that all went fine. I'm not looking forward to having to ship the 45lb box back.
Stereodude
Not really a
I did the same and followed the color coding between the two.
Mercutio
Fatwah on Western Digital
Thanks. I'm sort-of mulling the feasibility of putting one of those in my office as a test rig, but I'm not going to mess with it unless I can get away with desktop cooling
Stereodude
Not really a
You're going to be in for a bit of a challenge. Standard LGA 2011 heatsinks will fit on the sockets. However, the motherboards may have odd cooling behavior compared to desktops like the Intel board I outlined above. They also tend to have minimum fan RPM thresholds. Supposedly boards with IPMI can usually work around the minimum fan speed. The Supermicro X9SRA I have can't work around it. I'm in the process of building a hardware solution to the problem.
Stereodude
Not really a
:idea: I figured out how to manipulate the Intel temperature thresholds by editing the SDR file and doing an FRUSDR update! It now sits around the 65-70C under 100% load instead of 75-80C.
Edit: I decided to push for another 5C reduction in temps and edited the files further. It worked and runs another 5C cooler under load. :mrgrn:
Edit: I decided to push for another 5C reduction in temps and edited the files further. It worked and runs another 5C cooler under load. :mrgrn:
Last edited:
Stereodude
Not really a
That sucks, but you can always Paypal dispute them if they don't pop up in a day or two. You will win for sure, but that would leave you with a 2nd not working system. You could also try contacting Intel and attempt to get a new power distribution module (or whatever it's called) under warranty.
I'd like to work something out with them as I want to use the case but if I can't then I'll dispute it through eBay. I don't know why they'd just ignore me after starting a dialog already.
I removed the working motherboard and mounted in the one that was beeping into the known good case and now I have a functional esxi server so the motherboard, CPUs, and memory seem OK so far. I'm running stress test vms right now. Esxi shows power consumption around 450W at 100%. I set the fans to performance and disabled the quiet mode. They don't seem to be very loud under load. Other than the case and PSU issue I'm very happy with this setup.
I removed the working motherboard and mounted in the one that was beeping into the known good case and now I have a functional esxi server so the motherboard, CPUs, and memory seem OK so far. I'm running stress test vms right now. Esxi shows power consumption around 450W at 100%. I set the fans to performance and disabled the quiet mode. They don't seem to be very loud under load. Other than the case and PSU issue I'm very happy with this setup.
The ebay seller got back to me this morning and again a few minutes ago. They said they shipped me a new power supply along with a fedex tracking number and asked me to close out the ticket. I figured I'll wait until I get it before closing the issue. I wrote back and asked what exactly they sent me as I'm concerned they just send the hot swappable PSU portion when I need the actual logic board inside the case.
Aside from that the motherboard that formerly had the beep codes has been stressed tested for many hours with no issues inside the functioning case and has since been up and running ESXi 6.0 U2 for a couple days now with my Linux server. I'm feeling fairly confident that both motherboards work well at this point. I'm excited to get this cluster of ESXi servers doing some work.
Aside from that the motherboard that formerly had the beep codes has been stressed tested for many hours with no issues inside the functioning case and has since been up and running ESXi 6.0 U2 for a couple days now with my Linux server. I'm feeling fairly confident that both motherboards work well at this point. I'm excited to get this cluster of ESXi servers doing some work.
Stereodude
Not really a
I'm pretty sure if you close it you can't reopen it or open another, but I'm not 100% sure. Either way you should end up with a nice collection of spare parts.
I've been going back and forth with Supermicro about my fan pumping issue on the LGA 2011 workstation . I do have the parts to build my hardware tachometer feedback signal hack, but I haven't done that yet. I've tried to stress test it and monitor the CPU temps. The CPU fan speed ramps up and down the whole time, but the CPU doesn't seem to get much over the very low 50C range with the Thermalright True Spirit 140 BW Rev. A. I even used cheap white thermal compound that came with the HSF since I figured I would be doing some CPU swapping during testing and didn't want to waste the good stuff.
On the dual processor Intel I was thinking to pull the CPU heatsinks, make sure they're very flat and smooth, replace the thermal compound with the top-tier non-metal stuff I bought not too long ago, edit the .sdr file again to set it to run another 5C cooler under load, and then do the FRUSDR update to load the settings. I'd also like to put Windows 10 on it. So far I haven't done anything with it but run CPU stress tests or memory tests.
I've been going back and forth with Supermicro about my fan pumping issue on the LGA 2011 workstation . I do have the parts to build my hardware tachometer feedback signal hack, but I haven't done that yet. I've tried to stress test it and monitor the CPU temps. The CPU fan speed ramps up and down the whole time, but the CPU doesn't seem to get much over the very low 50C range with the Thermalright True Spirit 140 BW Rev. A. I even used cheap white thermal compound that came with the HSF since I figured I would be doing some CPU swapping during testing and didn't want to waste the good stuff.
On the dual processor Intel I was thinking to pull the CPU heatsinks, make sure they're very flat and smooth, replace the thermal compound with the top-tier non-metal stuff I bought not too long ago, edit the .sdr file again to set it to run another 5C cooler under load, and then do the FRUSDR update to load the settings. I'd also like to put Windows 10 on it. So far I haven't done anything with it but run CPU stress tests or memory tests.
Stereodude
Not really a
Maybe you guys are right about eBay...
So I returned 1 DIMM each to each of the two sellers last week since I had one bad DIMM from each. USPS sat on them for 3 days, but eventually got them to their destinations. One seller said they would send me a out very very similar replacement (we discussed if it was acceptable first) out last Monday. I hadn't gotten it as of yesterday and saw USPS delivered the bad DIMM to them yesterday. So I sent a friendly note asking for the tracking number / status. They subsequently informed me that they don't have a replacement DIMM and offer me a refund. I looked at the listing of the DIMM they offered me as a replacement and saw it was now sold out. They had 12 of the DIMMs they offered me as a replacement and sold all of them after I had contacted them, after they had promised to send me one of them, and after I had shipped and provided them a tracking number for the bad DIMM I was returning.
What am I supposed to do with 3 matching DIMMs and a $15 refund? I can't buy a 4th matching DIMM for $15. There's not even one listed for sale on eBay regardless of price. I told them I was going to initiate a return through eBay for the other 3 DIMMs and they're paying the return shipping this time. I've never left someone negative feedback, but this seller has earned it.
So I returned 1 DIMM each to each of the two sellers last week since I had one bad DIMM from each. USPS sat on them for 3 days, but eventually got them to their destinations. One seller said they would send me a out very very similar replacement (we discussed if it was acceptable first) out last Monday. I hadn't gotten it as of yesterday and saw USPS delivered the bad DIMM to them yesterday. So I sent a friendly note asking for the tracking number / status. They subsequently informed me that they don't have a replacement DIMM and offer me a refund. I looked at the listing of the DIMM they offered me as a replacement and saw it was now sold out. They had 12 of the DIMMs they offered me as a replacement and sold all of them after I had contacted them, after they had promised to send me one of them, and after I had shipped and provided them a tracking number for the bad DIMM I was returning.
What am I supposed to do with 3 matching DIMMs and a $15 refund? I can't buy a 4th matching DIMM for $15. There's not even one listed for sale on eBay regardless of price. I told them I was going to initiate a return through eBay for the other 3 DIMMs and they're paying the return shipping this time. I've never left someone negative feedback, but this seller has earned it.
They sent my the swappable PSU exactly like I told them not to send me. I'll give it a try just in case this is the issue but I'm very doubtful.
Sorry to hear of your delightful experience with eBay. I'm equating the savings of using eBay to increased hassle. I know had I bought this stuff from newegg or Amazon I'd already be done with this project but at 4x the cost.
Sorry to hear of your delightful experience with eBay. I'm equating the savings of using eBay to increased hassle. I know had I bought this stuff from newegg or Amazon I'd already be done with this project but at 4x the cost.
Stereodude
Not really a
The other seller says they shipped me out an identical DIMM to the one I returned and gave me a tracking number that doesn't work yet. I guess I will buy 4 of these from Natex for the other machine.
I did the Windows 10 upgrade to the S2600CP2J yesterday. I used the Server 2012 graphics driver from Intel's site. I don't know why there isn't a graphics driver for any other OS's. The Avago website has a Pilot2 and Pilot3 driver. According to the Intel Technical Product Spec, the board has the Pilot III controller. The Pilot 3 driver I used in Windows 10 reports the controller/board has 32MB of SDRAM, but the Intel spec only says it has 16MB. Got me.... :scratch:
I did the Windows 10 upgrade to the S2600CP2J yesterday. I used the Server 2012 graphics driver from Intel's site. I don't know why there isn't a graphics driver for any other OS's. The Avago website has a Pilot2 and Pilot3 driver. According to the Intel Technical Product Spec, the board has the Pilot III controller. The Pilot 3 driver I used in Windows 10 reports the controller/board has 32MB of SDRAM, but the Intel spec only says it has 16MB. Got me.... :scratch:
So...as I guessed the new PSU they sent me did not solve the issue. Both motherboards don't post with this case. I sent them a message again with a picture pointing to the part that's broken. I'm unclear what they'll offer me at this point.
Stereodude
Not really a
Sorry to hear about your continued troubles. Try to keep the power supply though. :mrgrn:
I did a clean Windows 10 install tonight on mine. I also messed around to see if I could stop the fan stupidity at power on. That didn't go anywhere.
It looks like the Emulex Pilot 3 drivers tell the OS it supports 32-bit color at higher resolutions when it really doesn't and you actually get 16-bit color even though you've picked 32-bit.
Also, for whatever reason a reboot is different with Windows 10. Reboots with Windows 7 didn't incur the fan ramp or any of the linux live images I messed around with. The fans ramp with Windows 10 and it's a different ramp. It takes at least a full minute for the fans to return to normal speed with a reboot from Windows 10. It seems to incur the fan hysteresis which doesn't let the PWM value settle very quickly.
I did a clean Windows 10 install tonight on mine. I also messed around to see if I could stop the fan stupidity at power on. That didn't go anywhere.
It looks like the Emulex Pilot 3 drivers tell the OS it supports 32-bit color at higher resolutions when it really doesn't and you actually get 16-bit color even though you've picked 32-bit.
Also, for whatever reason a reboot is different with Windows 10. Reboots with Windows 7 didn't incur the fan ramp or any of the linux live images I messed around with. The fans ramp with Windows 10 and it's a different ramp. It takes at least a full minute for the fans to return to normal speed with a reboot from Windows 10. It seems to incur the fan hysteresis which doesn't let the PWM value settle very quickly.
I took apart the power supply control board. I don't visually see anything wrong with it but this is the part that I'm pretty sure is keeping me from making progress. Now to see if I can find a replacement for it online somewhere if the eBay seller won't send me a new one.
I think I found one on eBay for $20 or I can make an offer if things don't go well. It's an Intel power board PBA G15585-401.
Stereodude
Not really a
I got a beta / test BIOS from Supermicro that lets the fans run down to 300RPM instead of freaking out at 700RPM. It requires a drop below 300RPM to go full throttle. Unfortunately, the fan on my CPU, a Thermalright TY-147A, runs around 300-320RPM at idle, and will intermittently drop below 300RPM which triggers a temporary run to 100% duty. It happened a few times while I was messing with the system this morning testing the BIOS.
The Standard fan setting in the BIOS doesn't behave any differently than Optimal. It should run the fans faster per the manual, but the PWM duty is exactly the same (32.71%) with either setting. The baseline PWM is supposed to be 50% with Standard and 30% with Optimal. I gave Supermicro my feedback.
I have a few Thermalright TY-147A fans between my various computers on their heatsinks so I will see if I have one that runs a little faster with the same PWM input. Otherwise I'm going to get a small DC/DC step up regulator and increase the voltage to the fan slightly so it runs just a little bit faster than it does now. I'd expect it might not even take an extra 0.5V to speed it up a few RPM.
I again noticed that the CPU doesn't seem to get very hot. With crappy white thermal compound CPU temps were barely cracking 50C on any of the cores while running Prime95 Small FFT and the CPU fan sped up to all of 900RPM or so. The side of the case was off, but that's still pretty cool for a 115W TDP chip.
It happened a few times while I was messing with the system this morning testing the BIOS.The Standard fan setting in the BIOS doesn't behave any differently than Optimal. It should run the fans faster per the manual, but the PWM duty is exactly the same (32.71%) with either setting. The baseline PWM is supposed to be 50% with Standard and 30% with Optimal. I gave Supermicro my feedback.
I have a few Thermalright TY-147A fans between my various computers on their heatsinks so I will see if I have one that runs a little faster with the same PWM input. Otherwise I'm going to get a small DC/DC step up regulator and increase the voltage to the fan slightly so it runs just a little bit faster than it does now. I'd expect it might not even take an extra 0.5V to speed it up a few RPM.
I again noticed that the CPU doesn't seem to get very hot. With crappy white thermal compound CPU temps were barely cracking 50C on any of the cores while running Prime95 Small FFT and the CPU fan sped up to all of 900RPM or so. The side of the case was off, but that's still pretty cool for a 115W TDP chip.