Temps during mprime stress testing

[pincorrect]

I've been running mprime in stress-testing mode and watching my CPU temperatures with IPMI view. I hoped to get some feedback as to whether the temps I'm seeing are OK.

I have the Supermicro X11-SSH-F motherboard with a Xeon E3-1240 v5 cpu. It has the stock cooler, but I removed the stock thermal compound and used Arctic Silver. I ran it long enough during disk testing to get past Arctic Silver's recommended 100 hour break-in period. (And of course I used their recommended application method. I also used their cleaner and surface purifier on the CPU and heat sink first.)

Anyway, at the peak, IPMI View says my CPU temp is 82 C. The datasheet says that TCASEmax, the max temperature at the cap of the processor is 74.1, but the TJunction-max is 100. I'm not sure if the temp IPMI view gives is the same as the case temp. I've just run mprime an hour, and it's running without any errors.

My feeling is this is a little high, but I'm not sure if an aftermarket cooler is warranted. I've heard the stock cooler is good. What do you think.

 

[nightshade00013]

Don't know what stock cooler it is but if the fan blows down towards the cpu it is likely recycling hot air. For those coolers to work right you need a ton of air movement through the case when they are under load.

I prefer the heat pipe coolers that can help to exhaust the heat through the back of the case or just some passive supermicro's with a fan zip tied on over the stock coolers.

I wouldn't be running them that hot though, if something gets over 65c to 70c I evaluate the cooling and reduce the loads.

 

[DrKK]

82C is hard to imagine. Are you sure there is no error in your application of thermal compound or heatsink application? If your fan is operating at all, on the intel stock cooler, on a Xeon, and you got to 82C, then something is wrong, sir.

Edit: Do Xeons even come with a stock cooler? Whatever, something is wrong.

 

[joeschmuck]

Do Xeons even come with a stock cooler?

Mine did. It works great too.

It has the stock cooler, but I removed the stock thermal compound and used Arctic Silver.

This was your first mistake. You should have just used the stock cooler as it was provided, the thermal compound was fine, at least on my CPU it was.

Now for the real question, how often are you planning to run the CPU at 100% for any length of time? If you are running FreeNAS then I don't think you have any concerns even though it's not perfect.

Also as mentioned earlier, you could have case cooling issues which should be fixed as this could also affect hard drive temps which will become a big and expensive issue. If the side of the case was open while doing your tests and the room was cool then you likely do have a thermal paste issue, however you never mentioned how long it took to get to the 82C temperature. If it took 3 hours okay, if it took 3 minutes then bad.

EDIT: I'm not saying to ignore this issue if it's the thermal compound, if it were me I'd fix it.

 

[pincorrect]

Thanks guys. As far as the thermal compound goes, maybe I should have stuck with the stock compound, which was a paraffin based patch supplied on the base of the stock cooler. However, I had used Arctic Silver numerous times before and liked it. I followed the instructions to the letter, including cleaning both surfaces with their two-part cleaner, "tinting" both surfaces with a little paste, then applying the compound in the method they recommend for the specific processor.

The stock fan pulls air up, away from the processor (I felt it to check). My case has good ventilation. It's a Cooler Master "high air flow" case which has fans pushing air in from the front (past the drives) and side, and fans pulling it out on the top and back. The air coming out of it always feels cool.

I ordered a Cooler Master Hyper 212X from newegg. I'll put that in and see what that does. Looks to be a PITA because it has a backer flange under the motherboard. However my case has an access hole on the MB tray just for that, so if I'm lucky I won't have to unmount the MB.

BTW, I only saw it get that hot during mprime testing. That is an unusual load, but that's the whole point of the test. During memtest86 tests, it would sometimes climb into the 50's (C), but usually stayed in the 40's.

 

[joeschmuck]

The stock fan pulls air up, away from the processor (I felt it to check).

You might want to check that again. I have never seen a heatsink pull air out as you are indicating. Look at the blades of the fan and the rotation, this will tell you the air flow. If you are feeling air flow and not just turbulence, provide a photo of your heatsink/fan (system powered off). I'd like to see what is going on.

 

[Ericloewe]

82 degrees Celsisus is not that weird when running mprime. My Core i3-4330 got to the mid 70s - keep in mind that it's two cores vs four and a more anemic cooler than the stock cooler that ships with 80ish Watt Xeons.

 

[DrKK]

My G3220 has never once been above 32C on the stock cooler ;)

 

[Ericloewe]

My G3220 has never once been above 32C on the stock cooler ;)

I'll bet it's never run mprime then. The temperature shoots up to the maximum value within seconds. It's almost disturbing.

 

[pincorrect]

You might want to check that again. I have never seen a heatsink pull air out as you are indicating. Look at the blades of the fan and the rotation, this will tell you the air flow. If you are feeling air flow and not just turbulence, provide a photo of your heatsink/fan (system powered off). I'd like to see what is going on.

I think you're right that I was feeling the turbulence. This is the stock fan from the retail cpu package. The fan comes pre-attached to the heatsink in a plastic cage. There was no need (and it might not even be possible) to remove it during installation. Photo below. The blades rotate clockwise, so it looks like it pushes air down toward the CPU.

I'll see how the Cooler Master cooler does. Mprime seems to be a particularly stressful test. Highest I saw before was during memtest86, where the cpu temp got into the low 50's sometimes.

 

[joeschmuck]

Unfortunately the one disadvantage of using the new heatsink will be loss of air flow over the motherboard voltage regulators. This is also a good feature of the Intel stock cooler.

 

[pincorrect]

Unfortunately the one disadvantage of using the new heatsink will be loss of air flow over the motherboard voltage regulators. This is also a good feature of the Intel stock cooler.

Yes, but maybe my case will make up a little for that because the side panel which faces the top of the motherboard has a 140mm fan blowing air straight down at the motherboard.

 

[nightshade00013]

Another fan moving air over the components would be a big help in this situation and put as many fans in the case as you can fit. If there is space you can even do double fans, stacking two together will increase the flow. The Fans in some of the Supermicro rack server cases are actually two counter rotating fans packaged together.

But the muffin coolers like that one really either have to have a TON of air moving through the case or they recycle air that has just been ran through the cooler especially bouncing off the ram and anything else nearby. Dealt with the same problem on some desktop builds using AMD coolers, I solved the issue by adding a second fan that forces the air through the cooler towards the back of the case on a Phenom II X4 965. On the FX8350 it was so bad I had no choice but to change coolers.

 

[pincorrect]

Another fan moving air over the components would be a big help in this situation and put as many fans in the case as you can fit. If there is space you can even do double fans, stacking two together will increase the flow. The Fans in some of the Supermicro rack server cases are actually two counter rotating fans packaged together.

But the muffin coolers like that one really either have to have a TON of air moving through the case or they recycle air that has just been ran through the cooler especially bouncing off the ram and anything else nearby. Dealt with the same problem on some desktop builds using AMD coolers, I solved the issue by adding a second fan that forces the air through the cooler towards the back of the case on a Phenom II X4 965. On the FX8350 it was so bad I had no choice but to change coolers.

That's why I like the particular case I have (CM HAF 932). I actually got the fan size wrong. The one in the back is 140mm. The one one the side blowing air down onto the MB is 230mm, but the panel has mounting holes so you could replace it with 4 120mm fans if you want. The top also has a 230mm fan (blowing out) which also could be replaced with 4 120mm fans. There is also a 230mm fan in front blowing air in the 3.5 inch internal drive bays. I had to put 3 of my drives in the 5.25 inch bays, so I added another fan to blow on them. It's reasonably quiet with the 5 case fans, plus cpu and psu fans. I could also add another fan on the bottom of the case, next to the psu, blowing upward.

 

[joeschmuck]

That's why I like the particular case I have (CM HAF 932).

I have the mid-size case and it is fantastic at flowing air, but I only have three running fans on my system, two 120MM fans in the front of the case and the CPU fan. All is very cool. I do have a 140mm fan on the side and another 120mm fan in the rear but they are not connected to power because they are not required. My front 120mm fans are connected to 7VDC so they run at a nice slow speed. And my PSU is mounted upside down so it pulls air from within the case and exhausts it. But honestly, most of the air comes out the top vents. It's a good case although I wish it were filled with SSDs vice 3.5" spinning rust, it should would be a lot lighter to carry.

 

[Stux]

FWIW, i use a Noctua cooler on my e5-1650v4, that's a 140W TDP CPU and it runs at 65C during an mprime test.

Intel doesn't supply stock coolers for e5 CPUs.

I'm currently burning in an overclocked work station with an i7-6900k. I've got that boosting to 250W and then dialing back to 200W after 60s of full mprime load with a corsair h110i. It hits 84C under mprime before cooling down to 75C with the 200W tdp cap.

My point is that the Noctua will dissipate 140W easily and the h110i 180-200W.

What's the TDP of your CPU?

You need a better cooler if you want to run mprime small ffts cooler. But 84C Is actually okay if you don't get motherboard heat soak. I found my test system would crash after 2-3 days at 85+C.

 

[mcflytfc]

I have a similar CPU to you, e3-1245v5 and my temps don't go above 59c when testing and high 30's when transcoding one plex stream. That's with the intel supplied cooler. It idles at 25c with a 22c room temp.

I was originally going to toss in one of these coolers however the stock cooler did well for me so it wasn't required.
http://overclocking.guide/gelid-solutions-siberian-pro/

 

[Stux]

There's a big difference between testing with anything else and mprime avx in my experience. Did you use mprime?

Here's an actual log from my mprime testing:

https://forums.freenas.org/index.ph...o-x10-sri-f-xeon-e5-1650v4.46262/#post-315465

 

[pincorrect]

Thanks for your replies. Addressing several at once here. My E3-1240 v5 is an 80W TDP cpu. I tested with mprime, but just in blended mode. I'll wait for my new cooler before trying the more stressful small fft mode. The Coolermaster Hyper 212X looks a lot like the Noctua pictured in the thread above. I ordered the dual fan model. I seem to remember using a Noctua in the past for another system, but this time I ran across the CM cooler first.

I think I'll use Arctic Silver again. They recommend different application patterns based on the cpu model. I'm pretty certain I'm doing that right.

 

[Stux]

I ordered the dual fan model. I seem to remember using a Noctua in the past for another system, but this time I ran across the CM cooler first

Interesting comparison: intel stock vs amd stock vs Cooler Master Evo 212

(the Hyper 212X is supposed to be better than the Evo 212)

http://www.anandtech.com/show/10500/stock-cooler-roundup-intel-amd-vs-evo-212

TLDR; the Evo 212 is 23C cooler than the Intel E97378 (circa yours) at 100W load.