2630L v2 vs 2620 v2 - What am i missing here?

[james23]

Im trying to get the lowest power DP CPUs (only for @ idle , not at load) for my new Freenas system.

e5-26xx v2 cpus are what i have to choose from.

in my test setup over the past few months, I had been running 2x e5-2620 v2, thinking those were the lowest power consumptoion options when at idle- but one of them has now failed, so im looking at buying a new cpu (or 2x). however, when i look at the 2630L v2 cpu, its base clock + turbos are higher than the 2620v2 (both are 6c/12t w 15mb cache), yet the 2630L v2 has a TDP of 60w vs 80w for the 2620v2 (and a lower Tcase temp too). How is this possible?

(see data in images and links below pls)

If im going for the lowest idle power usage should i just forget the TDP #, and go for the lowest possible clock'd at idle cpu ie 2.1ghz on the 2620v2 ? (or @ near idle, as this box will be sitting at idle 98% of the time). the assumption being the lowest possible clock the 2630L v2 can ever run at, is its higher base clock of 2.4ghz.

(or is there some magic at work on the "L" cpus? Everything below matches up as it should, except the lower TDP #)

thanks

Intel ARK comparsion of the 2x cpus (direct link):
Intel product specifications

turbo states:


passmark:

 

[nthu9280]

I think you are better off getting the most economical higher binned CPU within the same CPU series and not bother with the L.

I've not compared myself but, I've read here and elsewhere as well that the idle power draw is the same across the same series i.e., E5-26xx V2. Moreover e5-26xxL V2 does not mean idle is any less just the max is capped for use in TDP limited chassis applications.

Edit: initially I misread the question as V3.

 

[james23]

> I've read here and elsewhere as well that the idle power draw is the same across the same series i.e., E5-26xx V2.

thats pretty interesting, as after my post i did some tests as best as i could, and i also seemed to see *about* this to be the case (even regardless of core count!). I think im going to stick with 2620v2

i can post these 2x data points from my own notes. sorry if they are confusing, but they are as apples to apples a comparison as i could do (ask ? if want more details):

(note this first sys has 3x hba, and 22x HDDs, bottom set is almost only MB):

2x 2640v2 (8c) !!! ~250w IDLE (ePDU)
2x 2620v1 (6c) !!! ~260w IDLE (ePDU)
----

(on backup/test x9dr3 board w 1x stick of ram + 1x 120mm fan + 1x 4tb RED )
both tested using my YUMI boot usb

1- sitting at the YUMI screen
2- started up Active@ 2018 ISO via YUMI and sitting at the screen asking for LIC key)

1x 2630L (v1)
1- = !!!_77-79w (wallMTR)
2- = !!!_71w (wallMTR)

1x 2620 (v1)
1- = !!!_76.8-77.5w (wallMTR)
2- = !!! 70.5w (wallMTR)

 

[Markess]

As @nthu9280 mentions, other than being thermally capped, the "L" series processors don't have any unique power saving features. Intel is tight lipped about their binning process, but some folks on the internet speculate that "L"s are just regular processors that weren't stable/efficient enough to clock up to the speeds of the regular product line. This could be true, since the silicon for a given line all starts the same. If so, then the "L" series, while having a lower TDP, could in fact be less efficient at a given clock speed. All you are guaranteed with an "L" processor is that it won't dissipate more than the rated amount of heat.

> I've read here and elsewhere as well that the idle power draw is the same across the same series i.e., E5-26xx V2.

thats pretty interesting, as after my post i did some tests as best as i could, and i also seemed to see *about* this to be the case (even regardless of core count!). I think im going to stick with 2620v2

i can post these 2x data points from my own notes. sorry if they are confusing, but they are as apples to apples a comparison as i could do (ask ? if want more details):

(note this first sys has 3x hba, and 22x HDDs, bottom set is almost only MB):

2x 2640v2 (8c) !!! ~250w IDLE (ePDU)
2x 2620v1 (6c) !!! ~260w IDLE (ePDU)

It may be relevant to note that the E5-26xx V2 series used three different dies. So the 2620 v2 and 2640 v2 were, in effect, completely different processors with their own thermal characteristics, rather than being essentially the same with differing clocks.

CPUs up to 6 Core used one die (E5-2620 v2 for example)
CPUs up to 10 Cores used another different die (E5-2640 v2 for example)
CPUs up to 12 Cores used a third die

It's even possible that Intel made E526xx CPUs with lower core counts out of higher core count dies that had failed cores in testing. So, in the end, other than a general envelope of performance, your exact measurements can often vary from unit to unit.