Is going for an L5630 server worth it?

[Darkytoo]

I'm trying to upgrade my servers to give more ability to add ram. I've already added an LGA2011 server and a xeon e5 -2620v2 chip, ended up being about $700 total, and Now I see I can just get an lga1366 board with two L5630 chips, and it will draw the same power as a single 2620v2, and cost less than half of that. So what am I giving up? Besides going for used 3+yr old parts?

1. Looks like I lose total memory. my lga2011 board will hold up to 512mb, while the 1366 board maxes at 192. Not as big an issue as this is for a lab anyway. it will also only use 48gb of unbuffered / non ECC, while the 2011 board will hold 64GB (again, minor inconvenience)
2. Loss of sata 3 and usb4, not that big of a deal as most of my data is on an infiniband SAN, and I guess I could just add a SATA 3 controller if needed.
3. Running hyper-v 2010r2, will it cause issues with the "different processor version" check? Can i live migrate without restricting processor features?
4. Is it true it only uses ~40 watt per chip?

 

[PigLover]

The L5630 might draw less power than the E5-2620v2 when running at full load. And even then you have to say "might" because you are comparing TDP (Thermal Design Power aka heat dissipation load) and not raw power draw (they are similar - but no the same). Also note that the L5630 will be coupled with a 5520 chipset - one of the hottest and power hungry chipsets Intel ever designed, while the E5 will have a much simpler IO controller that draws very little power (most of the chipset functions moved on-die with the E5 chips).

Lots of factors to consider, but at full load it is very likely that the total system power draw would be comparable.

At idle or light loads it is a totally different story. The E5-2620v2 system will draw less - perhaps much less - than a comparable L5630 based system. Remember - most systems, even in the Enterprise Data Center environment - spend most of their lives at idle/light load.

Note that the E5-2620v2 will do more work per unit time than the L5630 (clocks and core-counts are similar, but with pipeline improvements and advanced instruction sets the E5 is still quite a bit faster). For any given workload, the E5 will spend less time "working" and more time "idle" than the L5630.