E5-2640v1 to?

EngineerNate

Member
Jun 3, 2017
64
14
8
32
Hi all,

My lab/NAS/Firewall setup currently consists of a Cisco C220 m3 and an Intel S2600gz system sold under the Penguin Relion badge (running stock Intel bios). Both have been rock solid stable with E5-2640v1 processors and IIRC 48gb of ram per.

Both are running ESXi 6.7 and I have my Sophos UTM firewall and FreeNAS (haven't upgraded to the new images under the TrueNAS name yet) running on them as my primary permanent services for the household. I tinker with other stuff off and on, and I have some other basic services running in docker via PhotonOS and Portainer.

Now that E5v2 processors are coming well down in price I'm curious if it's worth the upgrade or of I'd be better off leaving them as is until I can move up to a newer set of servers that can take v3 and newer processors? I have a set of E5-2650v1 processors I never installed because I haven't needed the cores and the 2640s have a higher clock speed, but I'm wondering if there are any performance advantages unlocked with the v2s or if my power usage would improve on the newer CPUs.

Thanks!
Nathan
 

EngineerNate

Member
Jun 3, 2017
64
14
8
32
I'm capable of comparing the specs, it's more the practical side I'm curious about.

v3 would take a new server correct? I believe that jumped to a new socket.
 

nthu9280

Well-Known Member
Feb 3, 2016
1,626
465
83
San Antonio, TX
Power saving is definitely a + if you can keep the delta cost of swapping CPUs to a minimum.

I think the other benefit going to v2 on Intel S2600CP boards was to gain wider PCIe Gen3 compatibility. v1 on those boards only supposed select cards at Gen3 speeds and rest at Gen2 speeds. Not sure of the GZ line, but assume the same limitation applies.

If the current setup meets your needs, I don’t see a big reason to move to E5-V3. DDR4 is still expensive, though it has come down quite a bit from its peak.
 

EngineerNate

Member
Jun 3, 2017
64
14
8
32
Power saving is definitely a + if you can keep the delta cost of swapping CPUs to a minimum.

I think the other benefit going to v2 on Intel S2600CP boards was to gain wider PCIe Gen3 compatibility. v1 on those boards only supposed select cards at Gen3 speeds and rest at Gen2 speeds. Not sure of the GZ line, but assume the same limitation applies.

If the current setup meets your needs, I don’t see a big reason to move to E5-V3. DDR4 is still expensive, though it has come down quite a bit from its peak.
Could this (The PCIe limitation) explain some intermittent 10gbps networking speed issues I have on that box? I have ConnectX-2 cards in there and it randomly gets stuck at ~2.5gbps speeds in iperf in one direction only. (I can't remember if it's testing to/from that server that's the issue)
E5-2640v2s are cheap. Not sure if the .5Mhz per core less clock speed matters much. The extra cores and instruction sets would be nice.

I'd probably break about even if I sold the existing CPUs and the extra set of 2650s I have. Assuming they sold.
 

gregsachs

Active Member
Aug 14, 2018
432
127
43
Could this (The PCIe limitation) explain some intermittent 10gbps networking speed issues I have on that box? I have ConnectX-2 cards in there and it randomly gets stuck at ~2.5gbps speeds in iperf in one direction only. (I can't remember if it's testing to/from that server that's the issue)
E5-2640v2s are cheap. Not sure if the .5Mhz per core less clock speed matters much. The extra cores and instruction sets would be nice.

I'd probably break about even if I sold the existing CPUs and the extra set of 2650s I have. Assuming they sold.
The GZ and GL are the same, except for 16 or 24 dimm slots.
The CX-2 will be enforced at pci 2;
see here:
for the pci 3 list.
I was able to get a pair of 2650v2 for $75 or so shipped last year, that was the sweet spot of power/dollar.
 
  • Like
Reactions: EngineerNate

EngineerNate

Member
Jun 3, 2017
64
14
8
32
The ConnectX-2 is a PCIe 2 card so I guess that's not it. I'm curious if maybe I put it in an electrically PCIe 1x slot or something. Though that wouldn't explain the asymmetrical nature of the problem.