Xeon E5-2683 v3 2.3GHz ES seems stable. How to stress test

[pclausen]

So I rolled the dice and picked up the ES variant of this chip for a decent price.

This is for my Emby media server running on FreeNAS. I had a E5-1620 v3 before, and it was running out of steam anytime 3 or more clients were watching high bitrate content that had to be transcoded.

When I received it, I was surprised by how "production" it looked, no "Intel Confidential" stamp on the die, and the "SR1XH" stamp also seemed to indicate a retail unit.

And the back side even had a Q.C. PASSED sticker on it.

My motherboard is a Supermicro X10SRL-F running the latest BIOS (1.0c) dated 9/9/2015. I'm running 4 x 16GB Samsung DDR4 2133 MHz ECC ram. I dropped the ES CPU in and put on a decent sized Noctua cooler (replacing my much smaller Supermicro one I was using with the E5-1620 v3).

I was glad to see that it posted right away:

I booted into Windows temporarily so that I could run a few CPU ID utils and some benchmarks to see what I got:

So it does appear to be a stepping one ES unit...

Intel's processor id util. This one is indicating processor speed is 2.3 GHz where CPUZ was showing 2.399 GHz:

So a stepping 1 revision 14. Not sure if that is close to production or not. Anyone know?

Time for some benchies:

Memory:

The CPU scores look excellent, but the memory seems to lag somewhat. Perhaps there's a penalty for running ECC?

Cinebench:

Just for grins, I checked out Cinebench's online database, and the score of 1573 is the 3rd highest in their database for single socket motherboards. Here are the top 5:

During all of the above tests, CPU temps never went above the mid 60's and I didn't have a single hiccup with the machine locking up or freezing or doing any other abnormal behavior.

Satisfied that the CPU was at least somewhat stable, I went ahead and booted into FreeNAS and let all my plugins come up. I then started 6 transcoding sessions of 1080p @ 25 Mbps material (un-compressed Blu-Ray rips), and the server handled it with no issues and was still plenty responsive. I could have probably throw another half dozen transcoding sessions at it.

Again, temps remained in the mid 60's and not a single stability issue.

So it seems that the chip I got is decent. I booted back into Windows and started running Prime95, which has been running all night at this point with no issues and temps holding in the mid 60's.

So my question is this, what tests, that I have not already run, should I throw at this CPU to see if I can break it?

And also, is the memory performance I'm seeing 'normal' or are there some settings I should change? It appears to be operating at 2133 MHz according to the BIOS and my IPMI. Although CPU-Z showings the following:

The specific RAM I'm running is this: SAMSUNG 16GB 288-Pin DDR4 SDRAM ECC Registered DDR4 2133 (PC4 17000) Server Memory Model M393A2G40DB0-CPB

It has a Cas latency of 15.

 

[Araxie]

Cancel Prime95 ASAP if you want to keep your chip working. since haswell and haswell-E Prime95 is no longer recommended to run in those chips as is dangerous and can damage the chip or even the motherboard...

in the other hand there you got your RAM bottleneck. you probably need to configure manually the timings of the RAM. they are all wrong over the place.. if they are CAS 15 the timings should be something similar to: CAS 15-17-17-35 or 15-15-15-34 as example.. row refresh cycle time at 278 clocks and command rate at 1T or 2T at worse.. you have a big bottleneck there bud..

as for recommended apps.. I would highly recommend ASUS Real Bench.

 

[Nathan_P]

That is a retail cpu in the photo, not an ES, all ES chips have a stepping code that starts with a Qxxx.

If cpu-z is reading it as a ES chip its either an older version of cpu-z, or its a final stepping ES chip that will be the same as the retail ones.

Have you tried HWinfo64 to see what that identifies the chip as?