System Temperatures

[ddrueding]

I'm redoing the cooling for my system and am hoping for some feedback on my setup. Speedfan 4.35 Beta 24 is giving the following temps at idle:

CPU: 40C
Core 0: 49C
Core 1: 48C
Core 2: 42C
Core 3: 43C
HD1: 39C (this is actually from some part of the Seagate RAID-5 array, not sure how)
GPU: 50C
System: 18C
AUX: 38C

My first question is why the CPU and Core numbers are so different? Where are they being measured from? I'm assuming that AUX is the Northbridge and System is the motherboard somewhere (that is about ambient).

 

[ddrueding]

I'm running some stress testing at the moment (Encoding a movie using AutoGK while running one copy of RTHDRIBL on each monitor and a 100GB file copy from the Raptors to the large array. Power draw is around 600W sustained.

Now that things have stabilized, here are the load temperatures:

CPU: 61C
Core 0: 69C
Core 1: 66C
Core 2: 58C
Core 3: 63C
HD1: 39C
GPU: 59C
System: 43C
AUX: 39C

Well, there goes my theory about the System temp; it spiked under load more than anything. I'm fairly certain that GPU is the 6600 and not the 8800GTS; the GTS is too hot to touch at the moment.

Would anyone be worried about core temps in the low 70s?

 

[Mercutio]

My 3.2Ghz Q6600 spikes up to 62C or so in a P180 while using all four cores to play City of Heroes (which can use all four cores).

70C seems a bit much though.

 

[Mercutio]

My 3.2Ghz Q6600 spikes up to 62C or so in a P180 while using all four cores to play City of Heroes (which can use all four cores).

70C seems a bit much though.

 

[ddrueding]

Yeah, I'm not really happy about it. Right now the HSF is a Zalman 9700 with the fan removed, I think a Scythe Ninja would work better with the more general airflow I have in the case. Is there something better than the Scythe Ninja Plus for passive cooling?

 

[ddrueding]

Mercutio, which CPU temp sensor hits 62C? Because my "CPU" sensor gets about that hot; only the core sensors go over 70C.

 

[P5-133XL]

The first thing I observe is that there is a lot of variability between the different cores. Assuming a level load, that means your thermal greese isn't spread out evenly and you can do better by re-applying. Mine don't vary more than 2 degrees between each other, but it took about 10 times of reapplying the fan to get there. There is an art to applying thermal greese.

I use Real Temp v.2.6 for my temp measurements. The reason I like it is that it lists distance to T-Junction Max which is the important temp measurement. When that goes to 0 is when the CPU starts doing emergency clock drops and CPU HLT's to force a decrease in temp. I've gotten into a habit of simply trying to keep the distance to TJ Max around 20C, regardless of the temps being reported. For my Q6600 the TJ Max is 85C so I try to keep temps lower than 65C. For My Q9450 the TJ Max is 95C so my max temps are kept below 75C. For Intel's Mobile pricessors TJ Max will be either 100C (65nm) or 105C (45nm) so for those a core temp between 80-85 shouldn't be a problem.

 

[P5-133XL]

My choice of 20C distance to TJ Max is entirely arbitrary. I have nothing to back it up ...

 

[Mercutio]

Mercutio, which CPU temp sensor hits 62C? Because my "CPU" sensor gets about that hot; only the core sensors go over 70C.

From memory, CPU0 is the one that typically gets hottest. The others will be in the neighborhood of 58 - 60C. I really only pay attention to temps when I'm gaming, mostly because I can feel heat wash off the side of my case (my legs are probably around 18" away from the PC), but ironically most of the waste heat seems to come from the 8800GTX (I've had three of those die, by the way, nice cooling setup or no) or the supposedly very efficient 750W Seasonic PSU.

I can run all four cores at 100% utilization while ripping an HD-DVD and 60C CPU temps are fine and normal and not uncomfortable, but in late July and August I have to curtail my City of Heroing because I can't stand to sit near the PC. The extra involvement of the graphics hardware (and associated increase in power draw) seems to make all the difference.

My cooling setup in that rig is a Big Scythe cooler + a 1200rpm Noctua fan. I just replaced all the Antec-provided TriCool fans in my P180 with new ones. I run them on "medium" in the summertime. The whole thing is sitting on a Gigabyte P35 motherboard.

 

[ddrueding]

P5-133XL, it shows my TJ-Max as 95C in Real Temp 2.70 (new version BTW). I'll keep an eye on it next time I load up the system and see what happens.

Mercutio, When I'm gaming the system sends way more power to the 8800GTS, but fortunately, that has a separate air supply to the CPU (there are three other expansion cards between the GPU and CPU and they have their own fans). Another bonus to my rig is that the heat comes out the top of the stack; about 4 feet over my head.

 

[ddrueding]

It is cooler this morning, so all the temps are lower, but the discrepancy is significant. Speedfan 4.35b24 is reporting Core0 to be the hottest at 67C, while Real Temp 2.7 is reporting the same core to be 62C. The coolest (Core2) is at 57C and 52C respectively.

Any ideas?

 

[P5-133XL]

With Real Temp, you can go to the effort of calibrating the sensors to give accurate numbers. However, according to the creator, it really only matters at low temps...

What I'm using to cool both my main machines is a Noctua U12P in a push-Pull config inside a P180 Case. The vertical 120mm case fan was removed and used as the 2nd CPU fan. They Push-Pull in the direction of the top side case fan which is only a few inches away from the cooler. The video cards are all two slot designed to exhaust air out the second slot. The case fans are set to medium. The bottom case fan (next to the PS) has been removed and placed in the middle for cooling HD's while the PS fan is now used to pull air over the bottom HD's. By no means is this configuration silent (that was not my goal), but it seems to do a very good job of cooling even when OC'ed on very hot days (100+ Farenheight).

If I need silent, I should replace the standard P180 case fans: That's where most of the noise comes from and if I drop them to low it is much quieter but it does not cool nearly as well.

I specificly went to a lot of effort to maximize cooling because I wanted to OC the heck out of the machines. However, I wasn't willing to go to the effort and cost of liquid cooling. I'm happy with how it turned out.

 

[ddrueding]

Thanks Mark. In my case...er...situation, the fans are farther away. There is plenty of air velocity, but it is more important for the fins to be farther apart so that the lower pressure air can still move through it.

 

[P5-133XL]

Well, its my understanding that you like to do passive CPU cooling and that's a whole different beast than what I'm doing.

 

[ddrueding]

Well, its my understanding that you like to do passive CPU cooling and that's a whole different beast than what I'm doing.

Yup, I agree. And I also agree that if noise isn't an issue, your configuration makes the most sense. I'm still aiming for that "silent" thing.

 

[Santilli]

David:
http://www.swiftnets.com/
ntroduction:

The MCX-VPro™ heatsink is a high performance-low noise thermal solution for Intel® Xeon™ processors. It replaces the MCX603-V and MCX604-V heatsinks, and includes retentions mechanisms for all know Xeon™ compatible motherboards.
I've found that with these coolers, the fans I use kick on at start up, and, then pretty much are either off, or low speed the entire rest of the time. This includes running both cpus at full speed, or very close, in certain tasks...

They are heavy, and, I think they may eventually 'bend' a motherboard...

 

[ddrueding]

Thanks for the suggestion Greg, but that heatsink design really won't work well for my configuration. Imagine a motherboard mounted to the wall with the USB/PS2/etc connectors on the top and the expansion cards aligned vertically. The airflow is all vertical; from well below the motherboard (hard drives are 2-4ft lower, fans are below that) or well above (the "pull" bank of fans are 1ft above where the CPU is).

In this configuration, I want to work with the vertical movement of the air. Using a tower-style heatsink will allow the heat to be scrubbed off without changing the direction of the airflow or using an additional fan. Even if I do end up adding another fan, I would prefer it to work with the other fans in the same direction (up).

The other thing to keep in mind is some basic fluid dynamics. The fans are far below and above the heatsink, and as such have the choice to flow around the heatsink instead of through it. By choosing a heatsink with the fins spaced farther apart, less air pressure is required for the air to go through, therefore exposing a larger surface area, and keeping the temperature difference lower.

I'm sure many of the folks here know more about this stuff than I do, so if I oversimplified or got it flat out wrong, please correct me.

 

[Santilli]

Hi David

What I was trying to point out is the Swiftechs are so good at pulling heat away from the processor, by their very thick base plate, and 'fins'
that they work as well, or better, then the Scythe Ninja Plus. I've found the Scythe requires more airflow, in other words, cranking the fans up, then the Swiftechs, which run on my Xeons without the fans moving. Just think of them is little radiators, that would pull heat from your processors, and, move it away from the motherboard. They would HELP your flow design, by creating hot air that would flow from their 'fins', and, suck cooler air in from below. I'd only install a fan as a last resort. I've installed a number of pretty much silent fans, and really turned the noise level down on my box. The cross flow from the front and back panel exhaust enough air, hot air that comes from the Swiftech heat sinks, so that the fans on the heat sinks rarely, if ever come on. I had one on the first box that IIRC, you built, and, that one I could watch the fan on the processor turn on and start up, then shut down, and stay shut down, in the same case I have to keep the big fans cranked up on, with the Scythe. IIRC, the first one you built was a 3000, and the second a 3200, both in the same case, with the same fan setup.
I got to compare both coolers, and, when we/you build another, it will be with the swiftech. It worked MUCH better.

My only concern is that they may have reduced the thickness of the plate/heatsink that contacted the processor, and, that might reduce the sinks ability to pull heat away from the cpu, and transfer it to the 'fins'.