4770 vs 4770s for 24/7 server

[mikmort]

Hi folks,

I'll be running a server machine that will be on 24/7. It will monitor IP security cameras, run Windows Media Center, provide storage, and do other home automation tasks.

Since it will be all on the all time in a tech closet I'd like to build a PC with as little power consumption as possible. I do need a powerful CPU, as the IP security cameras will useBlue Iris software that does consume a decent amount of CPU per camera for motion detection.

Would a i7-4770s use less power than the i7-4770 in this scenario (running the same load)? Or would be the power be the same and the i7-4770 has more headroom? I'm guessing at idle they would be the same, but I'm curious if the 4770s would do any better at load.

 

[Arkaign]

The S is clocked lower, 3.1Ghz vs. 3.4Ghz. They're binned to be stable at a lower voltage and so use less power. Most of the time you can manually underclock a processor to match that however.

Given that, I would recommend the 4770 non-K, non-S standard variant, or perhaps the Xeon equivalent which are sometimes a better value (same socket). B85 is probably the chipset you want. Without overclocking, Z87 becomes meaningless, so save your bucks there as well.

 

[erunion]

The 4770 and 4770s will likely use the same amount of power.

Someone posted about swapping out the 65W Celeron in his home server for a 35W Celeron, thinking it would save power. It ended up using the same amount of power, because his 65W Celeron was already down clocking itself. At idle the 65W and 35W Celerons drew the same power.

So it depends on what demand your server is going to have. A 4770 seems like over kill. Maybe an i3, instead?

 

[zir_blazer]

Xeon E3-1245 V3 > Core i7 4770. Its 20 U$D cheaper, you only lose 100 MHz on Turbo, but get ECC Memory support if you were to use it in a Server Motherboard with proper ECC Memory Modules, ideal for 24/7 operation. Also, the Xeon brand name means serious business, which add a bonus point to it.

 

[milee]

If the idle to load ratio is high I would choose a comparable Xeon E3 v3 like 1230, 1240 and even 1270 v3.

If it's low enough I would use an E3 1230Lv3.

On a C222 equipped MB, with some low volt DDR3-1600 and SSDs.

 

[ShintaiDK]

The S one will not use less power. Its only designed for acses with a max TDP limit. Same with the T. Both S and T versions however will be slower and take longer time to complete a task vs the non S/T.

 

[Arkaign]

The S one will not use less power. Its only designed for acses with a max TDP limit. Same with the T. Both S and T versions however will be slower and take longer time to complete a task vs the non S/T.

Are you sure? That must have changed :

http://hardware-review24.com/load/c...ridge_intel_core_i5_2400s_cpu_review/1-1-0-37

That's just one case where the 2400S was compared to a 2400 vanilla, and although idle power draw was identical (that makes sense), loaded power draw was a bit different, 55.8W vs. 68.6W

I've seen similar tests in forums confirming that for 3770 vs. 3770S using kill-a-watt and swapping the cpus in the same systems.

Now it's not a fair comparison because of the clock speed difference, so I suppose one could say accurately that the 4770S won't do the same work at a lower power usage, but it will clock lower while using less volts, meaning that in usage it *should* use less power from the wall compared to 4770.

 

[ShintaiDK]

Are you sure? That must have changed :

http://hardware-review24.com/load/c...ridge_intel_core_i5_2400s_cpu_review/1-1-0-37

That's just one case where the 2400S was compared to a 2400 vanilla, and although idle power draw was identical (that makes sense), loaded power draw was a bit different, 55.8W vs. 68.6W

I've seen similar tests in forums confirming that for 3770 vs. 3770S using kill-a-watt and swapping the cpus in the same systems.

Now it's not a fair comparison because of the clock speed difference, so I suppose one could say accurately that the 4770S won't do the same work at a lower power usage, but it will clock lower while using less volts, meaning that in usage it *should* use less power from the wall compared to 4770.

You forget the most important part. Energy used on the task. At the CPU burn they use. The 2400 gets what, 3.3Ghz. And the 2400S might not be above 2.8Ghz.

Essentially the non S/T finishes faster and goes back to idle. While the S/T version is still crunching.

 

[Arkaign]

You forget the most important part. Energy used on the task. At the CPU burn they use. The 2400 gets what, 3.3Ghz. And the 2400S might not be above 2.8Ghz.

Essentially the non S/T finishes faster and goes back to idle. While the S/T version is still crunching.

That's a definite truth, but I think that would be more applicable if we were talking about someone who was encoding, then stopped, idled, did a bunch of near-idle stuff, then went back to encoding/etc.

With those IP camera feeds, they keep things pretty steadily loaded. Not maxed out kind of load in my experience, but fairly steady. The question then would be would the 4770 non-S peg at 3.4Ghz while doing that work? And would the 4770S also peg at 3.1Ghz? If so, then the S would indeed use less power over time. Turbo shouldn't come much into play with highly threaded apps like BlueIris.

Regardless I do think the vanilla 4770 (or Xeon) is a better value, but it would surprise me to see the 4770S not actually use slightly less power to do the task described in the OP. I haven't personally used BlueIris, but I checked some forums, and apparently it does load 8 threads with enough cameras feeding. Of course the hdd performance is also something to consider with things like that.

The actual power usage gap would be pretty damn small, small enough that I think it would be pointless. However, if this is going to be in a confined space without great cooling/ventilation, it might be worth another look. Of course manual underclocking of a Xeon / 4770 would bring the same results (or better!) with the right settings.