E5-2696 V2 CPUs for $325!

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

MBastian

Active Member
Jul 17, 2016
205
59
28
Düsseldorf, Germany
cinebench r15 for e-2686 v3 stock 1 cpu: 1988
So with dual cpu I should be able to easily beat 1950x. but that comes at the price of having to go with higher price of 2011 v3 and DDR4
According to Wikipedia this is one of the rare v3 CPUs that works with DDR3 RAM. So I could re-use my existing 8x 16GB DDR3 Sticks in an 2011-3 board until prices for DDR4 and the more beefy E5 v3/4 parts go down?
 

wildpig1234

Well-Known Member
Aug 22, 2016
2,198
443
83
49
According to Wikipedia this is one of the rare v3 CPUs that works with DDR3 RAM. So I could re-use my existing 8x 16GB DDR3 Sticks in an 2011-3 board until prices for DDR4 and the more beefy E5 v3/4 parts go down?
Where are you going to find a MB that will have ddr3 slots and 2011 v3 cpu slots? You do realize that ddr4 slots wont take ddr3 mem right?
 

Davewolfs

Active Member
Aug 6, 2015
339
32
28
The 2696 is nothing special with single thread which I don't like - at least TR is decent with single thread.

The closest matching Intel rig to TR seems to be a dual 2687W v3 setup.

What kind of workloads are you running on your workstation - I'm still torn between going for higher clocks or more cores.
 

skelleton

Member
Nov 4, 2015
47
10
8
124
So I just got my 2x 2696v2 from cuba4sale. Both of mine are working in a Supermicro X9DRi-F with the latest BIOS.

Edit: based on a quick test (48 threads in stress and lscpu). The all core turbo is 3.1Ghz But i am not 100% that one. As lscpu also shows me 3.1Ghz in idle.
 
Last edited:

skelleton

Member
Nov 4, 2015
47
10
8
124
So I checked on my 2x 2680v1 system and it shows 2,4Ghz when mostly idle and 3Ghz when all cores are at full load.
So I would suspect that the values are also accurate with the 2696v2.

This means the mainboard CPU combo either does not like to do any power saving or such, or my Proxmox did not like that I changed from Sandy Bridge-E (2x2620v1) to Ivy Bridge-E.

But its running for now, and I have to figure out the Fan noise before I am going into power consumption.
So if any had issues with fan control on Supermicro X9 Series mainboards let me know how you solved them. (Based on the noise compared to my other servers the fan speed might be reported wrong and I can't seem to manually control the PWM from Linux)
 

wildpig1234

Well-Known Member
Aug 22, 2016
2,198
443
83
49
So I just got my 2x 2696v2 from cuba4sale. Both of mine are working in a Supermicro X9DRi-F with the latest BIOS.

Edit: based on a quick test (48 threads in stress and lscpu). The all core turbo is 3.1Ghz But i am not 100% that one. As lscpu also shows me 3.1Ghz in idle.

you mind running cinebench r15 and userbenchmark and post the score? I want to know how dual 2696 v2 is against 1950x. thnx
 

skelleton

Member
Nov 4, 2015
47
10
8
124
you mind running cinebench r15 and userbenchmark and post the score? I want to know how dual 2696 v2 is against 1950x. thnx
I have Proxmox running on that server and it's also my main file server, best I can do if this has to be fast, is running a Windows VM with all cores assigned.

Otherwise i might be able to take it down for maintenance on the weekend and throw windows on a spare disk.(Though I can't promise that right now)

Is the VM OK for you?
 

wildpig1234

Well-Known Member
Aug 22, 2016
2,198
443
83
49
I have Proxmox running on that server and it's also my main file server, best I can do if this has to be fast, is running a Windows VM with all cores assigned.

Otherwise i might be able to take it down for maintenance on the weekend and throw windows on a spare disk.(Though I can't promise that right now)

Is the VM OK for you?
will be a little slower but sure.. thnx
 

skelleton

Member
Nov 4, 2015
47
10
8
124
Ok I did the tests.
KVM VM set up as 2 sockets 24 cores each
drive set up as IDE
network as e1000
about 64000MB RAM

The CPU clock is read as 2.5Ghz in Windows and 3.1Ghz in Linux
When the VM is running and not doing the benchmarks linux is reporting cpu speeds between 1800-3300Mhz, oddly it seems stable at 3.1Ghz when no VMs are active.

Cinebench is are the best 3 of a couple of runs. No idea how good/or bad these scores are.

 
  • Like
Reactions: wildpig1234

Davewolfs

Active Member
Aug 6, 2015
339
32
28
Ok I did the tests.
KVM VM set up as 2 sockets 24 cores each
drive set up as IDE
network as e1000
about 64000MB RAM

The CPU clock is read as 2.5Ghz in Windows and 3.1Ghz in Linux
When the VM is running and not doing the benchmarks linux is reporting cpu speeds between 1800-3300Mhz, oddly it seems stable at 3.1Ghz when no VMs are active.

Cinebench is are the best 3 of a couple of runs. No idea how good/or bad these scores are.

About what I expected. If you don't mind could you run passmark single + multi or geekbench 4.
 

skelleton

Member
Nov 4, 2015
47
10
8
124
About what I expected. If you don't mind could you run passmark single + multi or geekbench 4.
Passmark is a no go, i bluescreens after a few seconds with a directio error that i don't want to debug.

Here is geekbench:


QEMU Standard PC (i440FX + PIIX, 1996) - Geekbench Browser

Edit: there are currently two video streams running from samba on the host and the streams are located on a zpool with 17 encrypted disks, so this may have some impact on performance.
 

wildpig1234

Well-Known Member
Aug 22, 2016
2,198
443
83
49
thanks for the numbers. just what I expected. it's almost about equal to stock 1950x for multithread.
 

e97

Active Member
Jun 3, 2015
323
193
43
Like others, looking to upgrade my E5-2670 to V2, mainly for PCI-E 3.0.

2017-09-04

E5-2670 v2 ~ $190
E5-2696 V2 ~ $310

I think theres still a ways to go as V2 reach EOL at the end of this year.

As other mentioned, ThreadRipper seems like a better option at this price point, plus 64 PCI-E 3.0 lanes and NVMe which is the biggest boost in noticeable performance.
 

skelleton

Member
Nov 4, 2015
47
10
8
124
Well 128GB DDR4 is a bit more expensive than with DDR3 and as far as I can see there are no TR4 Mainboards with ECC + IPMI. And I prefer my servers to have both. If I was buying new hardware, id probably go with 2x Epyc 72xx. Though I have not looked into the new Intel processors.
 

wildpig1234

Well-Known Member
Aug 22, 2016
2,198
443
83
49
Like others, looking to upgrade my E5-2670 to V2, mainly for PCI-E 3.0.

2017-09-04

E5-2670 v2 ~ $190
E5-2696 V2 ~ $310

I think theres still a ways to go as V2 reach EOL at the end of this year.

As other mentioned, ThreadRipper seems like a better option at this price point, plus 64 PCI-E 3.0 lanes and NVMe which is the biggest boost in noticeable performance.

It depends. If you do have some more money to spend then 1950x might be a better option but I would wait some more for better MB. there is no pro WS MB for TR right now, i.e. lack of good ECC memory support, only support up to 16gb ram modules or 128gb total ram.

The limitation of no more than 128gb with support only up to 16gb ram modules with dodgy ecc support at this point is a real problem for anyone with real professional use with large dataset.

For a cheap budget, if you need the performance right now, I would really consider dual 2696v2 which does give you about the same multithread performance as a stock 1950x since you can put together a system for much cheaper with also much cheaper ram. the only caveat is that 2696 v2 is likely yr last cpu upgrade for that platform.

you can step up to 2011 v3 dual 2686 v3 which give you significantly faster multithread performance than 1950x with better and higher capacity ram and ecc support but your system cost won't be as much of a bargain like 2696 v2 . well, the ram might actually be slightly cheaper since 16gb ddr4 ecc ram looks a little cheaper than 16gb non ecc ram modules.
 

Davewolfs

Active Member
Aug 6, 2015
339
32
28
Where are you looking for DDR4?

I recall paying about $100 for 64GB of DDR3. DDR4 seems to be about $110 for 16GB of the SM certified stuff.