Monday, March 27, 2017
9:50:10 AM
Users online: 0   You are here >> Home > CPUs and Cooling

Forums | CPUs and Cooling Forums search
Forum FAQ
Difference between Xeon, Passive, Active and 1U CPU's?
5/9/07 2:24:03 PM

So what the difference between the Active, Passive and 1U descriptors on the following processors please?

Intel XEON E5310 Quad Core, 1.60GHz, 8M Cache, LGA771, 1066FSB, Passive

Intel XEON E5310 Quad Core, 1.60GHz, 8M Cache, LGA771, 1066FSB, Active

Intel XEON 5050 Dual Core, 3.00GHz, 2x2M Cache, LGA771, 667FSB, Active & 1U

....disclaimer, i know nothing! my posts are evidence of this---

5/9/07 2:31:16 PM

The only thing I can think of is the cooling meathod. Active cooling is where you have a fan, or something that moves the warm air away manually. And Passive is the heatsink only, mainly ised in HTPCs for silence.

But i'm not sure that's what's being reffered to here.

If A is a success in life, then A equals x plus y plus z. Work is x; y is play; and z is keeping your mouth shut.

E6420 @ 3.2GHz | 2GB SOE | 965P-DS3P | HD 2900 XT

5/9/07 5:03:15 PM
Yes the passive and active are refering to whether or not the bundled cpu heat sink comes with a fan - the passive units should only be used in well ventilated cases.

The third cpu active & 1U is refering to the fact the bundled cooler is designed to be used in 1U rack mountable cases.

The two first cpu's are the same except for the cooler, the third however is an old cpu architecture not based on the core series therefore i'd avoid it if your building a high performance workstation/server.



5/9/07 8:50:34 PM

cheers guys,
most informative :-)

....disclaimer, i know nothing! my posts are evidence of this---

2/9/08 2:25:55 PM
1u is not a cpu, it's a form factor for case dimensions for servers and it's 1 3/4" thick.

Active cooling, low(er/est) profile heatsink, shortest fins and a fan on top to fit inside a 1 3/4 inch case. This works well for a cpu that the active heatsink was designed for. It's for lower load servers that don't run 100% cpu utilization for extended periods. What happens when you start dualing and quading cores ? Well a single core running at 100 % full speed is anywhere between 1.6 or 3.0 Ghz. Dual or Quad that and you run that same heatsink on effectively 3.2-6 Ghz (dual core) or 6.4-12 Ghz (quad core). If the heatsink and fan you have is designed for the 1.6 Ghz core cpu it'll cool effectively. Throw the 3 Ghz cpu in there and it'll probably cool decently at idle, especially if you are using a server load balance and performance power scheme. Run that 3 Ghz cpu at 100% cpu utilization for extended periods and expect to see the temps rise very closely to the thermal max that the cpu is designed for.

I ran into this problem with an AMD with a pair of Dynatron A75g. Great cooler for an Opteron 250 @ 2.4 Ghz. Upgraded to dual core 275's @ 2.2 Ghz. With Powernow the cores idle @ 1 Ghz each and you can't tell the difference. Put 100% load for 10 minutes and the 30* C idle temps shot up to 58* C. The good news, it handles it because Opt 275's thermally are designed to 70* C before they shut themselves down or fail. But what I rationalized happened. The A75g can cool a single core 2.4 Ghz cpu well even under 100% utilization. But when the 275's ramped up to 2.2Ghz under 100% utilization, the 2 cores were effectively running at 4.4 Ghz, or that same A75g was expected to cool (2) 2.2 Ghz cores/cpus. It can't or at least not like a passive heatsink with blowers. BTW, the A75g has a paltry 9.39 cfm output for the fans, Those who use more powerful desktop coolers use 80, 120 and larger fans that produce 40+ cfm of air flow.

On to passive heatsinks. These are beefier pieces of copper with longer fins to radiate and dissipate heat. These are actually better because the bigger the copper, the faster your cpu and core(s) can be. There is overhead for more cores and faster speeds. Active or passive refers to the heatsink itself, because even a passive heatsink uses airflow from bigger and more powerful blowers that move more air and are channeled with baffles inside the 1u case. The baffles are created and made of plastic/kydex, but they channel the air over the passive heatsinks, effectively cooling them and being just as active as screwing the fan onto the heatsink. Passive heatsinks are better for production servers that are constantly under high loads for extended durations. Think of a database server that has to run an automated program overnight because there are complex criteria and large amounts of data. For a desktop user, run a Microsoft Access query that takes 5-7 minutes to run, you'll see the cpu utilization in task manager approach 100%. Gamers, run your benchmarks, those take several minutes and you'll see your temps increase even with a desktop, same happens to a server.

Anyway, my 2nd server, was single core Opteron 248's 2.2 Ghz, upgraded to dual core 275's also. They have SuperMicro blocks of copper. With the upgrade my 2nd server noticed only a slight hit on higher temps. The Dynatron's work to keep the cpu's under 58* C at 100% cpu utilization. While the 4 blowers and passive heatsinks under that same load get no warmer than 45* C.

For a few hours, I couldn't figure out for the life of me why server with active heatsinks was runnning 10-13* C hotter than the other under identical loads. I first tried it with power scheme "always on" and the active heatsinks stayed right at 46* C for both dual core cpu's and as I ramped up the load, it got warmer to 50* C just surfing the internet and finally benchmarked to 58* C. The other server with "always on" power scheme idled at 36* C, internet surfing load was 38-40* C and the full load benchmark took it to 45* C.

After that I loaded AMD's cpu driver and then made sure the server cpu performnce was balanced in Windows 2003 was selected and that's where the bulk of my research in this post came from. at idle, the passive heatsinks are slightly cooler and when you ramp them up to 100% cpu utilization, I'm back to the heat differentials. Fortunately for me, the servers I have are not running full bore for brief or extended periods of time. If that were the case, the passive heatsink system would get that traffic.

I realize that the cpu's listed are dual or quad core already, so the cooler designed for those would have to be scrutinized as to whether they could cool faster cpu's and more cores ? A 1u heatsink works well with a 1u case because the case is designed to help channel air flow in most applications. It would have marginal success in a 2-5u case application. In those bigger boxes, get a desktop cpu cooler that has a huge, taller block of copper and even a higher output fan in terms of cfm. The blowers in a 1u server can turn at high rpms, they simply don't have the 40+ cfm fan output an 80 mm or larger fan produces. And they are noisy, but with blade servers and 1u's the name of the game is density of cpu processing power. Putting a bunch of towers in a server room is counter productive to that. Blades are larger than 5u, because each blade slides in is effectively a motherboard and memory, so density is maintained. With blade servers, the empty slots need to be inserted as they help channel air flow and cool a blade system.

Hope this post helps others to understand cooling for a 1u rackmount solution.


Forums | CPUs and Cooling