$90 Xeon E5-2670 2.6Ghz (8cores / 16threads)


Recommended Posts

1 hour ago, DoeBoye said:

 

Nope. Your G2120 scores a 1640 Passmark for a single core, while the 2670 gets a 1597 Passmark score. Not enough of a difference to be noticeable. And of course, if you look at the combined score for the dual 2670 vs the G2120, the difference is enormous (at the cost of more power, of course) :).

 

With 2 VMs, a bunch of dockers, and 15 or so plugins installed, I rarely see cpu usage above 15%. What I really love with having so many cores available, is that I can dedicate cores to specific tasks (Vms and Dockers) without concern for usage overlap or affecting the core functionality of unRaid.

 

[Edit] Also, for the record, if we're talking the 2670, it runs at 2.6Ghz, not 2.0 :)

[Edit 2] And, your G2120 runs at 3.1Ghz, without turbo. The 2670 has turbo boost to 3.3Ghz when required...

 

@DoeBoye  I was a little more curious about the gap in E5-2650 vs G2120.  Using the metrics you mentioned, it looks like we're comparing 1640 to 1315 for a single core.  Maybe I need to do a bit more research into understanding how CPU speed a cores effect performance so I can get a better grasp on what differences I may notice under different uses...

Link to comment
1 hour ago, Living Legend said:

I was a little more curious about the gap in E5-2650 vs G2120.

Ahhh! That makes more sense. That is definitely a more significant gap. Are you looking at an application that is single core dependent? It may be a non-issue if you are using apps that can take advantage of multiple cores. Some research into what you plan on using it for and the threaded-ness of those apps is definitely a good idea

 

If you're concerned, the price difference is only $84, spread between 2 processors (looking at the 64GB packages), so it might be worth the extra bucks to invest in the 2670.

 

Passmark numbers aside, I can't possibly see how a Pentium-class consumer chip could be anything but blown out of the water by a dual processor, 16-core, hyperthreaded monster like the ones you are considering. Especially if you consider multi-tasking. With 16 cores (32 with hyperthreading), you can dedicate a few cores to each docker, a few to unRaid, and some to a potential VM, and never have any cpu-performance issues because the apps will never be fighting for cycles.

 

Whatever you end up doing, good luck and enjoy the process!!! :)

Link to comment

@DoeBoye Ultimately, you're right.  The new dual CPUs with 16 cores and 32 threads is going to blow my single Pentium CPU out of the water.  It won't even be close.

 

I don't have any specific applications/dockers/VMs in mind which concern me.  I'm more or less trying to get a better grasp on the big picture, and how two 8 core, 16 thread CPUs at 2.0 GHz would function compared to a single core, single thread chip at 3.2GHz.  Are there instances where an easy workload would actually be handled better by the 3.2GHz processor over the 2.0GHz, multi core processor?  Again, this is probably where I need to do a little research to better understand how multi-core/multi-thread CPUs work.

 

Surprisingly, the Pentium processor has held its own for 3+ years now for my purposes.  I run numerous dockers and do the occasional Plex stream.  But I think now I understand why Plex can be flaky and intermittent.  It's likely that I've hit my threshold when I attempt to transcode while running all the dockers I have, along with a new VM I have on from time to time.

Link to comment
20 minutes ago, Living Legend said:

Are there instances where an easy workload would actually be handled better by the 3.2GHz processor over the 2.0GHz, multi core processor?

I'd be interested in that too! :)

Also, don't forget, the Xeon 2650 chips may have a slower default speed, but they have turbo boost as well. Perhaps in a single-threaded processor intensive task, the turbo will make up the bulk of the difference? I'm not sure how the Passmark value takes that into account (especially when comparing a turbo vs a non-turbo chip)

Link to comment
4 minutes ago, DoeBoye said:

I'd be interested in that too! :)

Also, don't forget, the Xeon 2650 chips may have a slower default speed, but they have turbo boost as well. Perhaps in a single-threaded processor intensive task, the turbo will make up the bulk of the difference? I'm not sure how the Passmark value takes that into account (especially when comparing a turbo vs a non-turbo chip)

 

What determines whether a task is intended to be run on a single thread, or over multiple threads if available?

Link to comment
35 minutes ago, Living Legend said:

What determines whether a task is intended to be run on a single thread, or over multiple threads if available?

 

The software needs to be coded to support multiple threads. It used to be fairly rare to see support for it, but in this day and age, as multi-cored cpus have become the norm, it's become much more common...

Link to comment
4 hours ago, DoeBoye said:

I'd be interested in that too! :)

Also, don't forget, the Xeon 2650 chips may have a slower default speed, but they have turbo boost as well. Perhaps in a single-threaded processor intensive task, the turbo will make up the bulk of the difference? I'm not sure how the Passmark value takes that into account (especially when comparing a turbo vs a non-turbo chip)

 

 

My passmark score for my dual 2670s is 19500.

 

i had dual 2620 with a mark of 13500 on the same motherboard.  I believe the 2650 will come in between 13500 and 19500.

 

The turbo mode should kick in when a threshold load is reached.  The passmark suite can't turn off or on turbo mode per se.

 

Gary

Edited by garyhgaryh
Link to comment
  • 2 months later...

This question is about ram compatibility.

 

This is my first build (ever) and I am trying to follow the techspot article closely with the exception of RAM due to its price (components listed at the end) and because it seems that ECC ram is better for scientific computing. The computer will primarily be used for machine learning for academic research, some deep learning, and various other tasks (browsing, writing, etc.).

 

As a newbie, I'm finding RAM compatibility to be incredibly confusing. The Xeon spec list says that the total possible is 384 gb, accepting types ddr3 800/1066/1333/1600. The motherboard spec list says there are 8x240 pin DDR3 DIMM slots accepting up to 256GB DDR3 R/LR DIMM or 64 GB unbuffered UDIMM. There is a memory QVL which gives a small list of "approved" types of DDR3 (mostly Kingston brand).

 

My question is: Will any ECC registered 240 pin DDR3 work in this set up? I see lots of good deals on ebay, but I hesitate pulling the trigger because I am unsure about compatibility. I am looking for 64 gb or 128 gb kits.

 

Thank you!

 

Parts list:

CPU: 2x Xeon E5 2670

Coolers: 2x Noctua i4

Motherboard: Asrock Rack EP2C602

PSU: Corsair RM 850x

GPU: GTX 1070

RAM: ???

 

Link to comment
38 minutes ago, JC59 said:

Will any ECC registered 240 pin DDR3 work in this set up?

Maybe. The only sure bet are the part numbers on the approved list. Others are a crapshoot. Should work, yes, but no guarantee.

 

When dealing with server grade equipment, stability is king, price is secondary, so the motherboard manufacturers generally fully certify and work with a very short list of modules. It's not like desktop stuff where timings can be relaxed and a crash here and there isn't the end of the world.

Link to comment
1 hour ago, JC59 said:

My question is: Will any ECC registered 240 pin DDR3 work in this set up? 

 

 

 

I'd suggest looking at the M/B recommended or supported Ram and buy that regardless of if you can find a "deal" or not on the recommended units. Always go with the recommended parts.  (Less of a headache later)

 

Cheers,

 

Link to comment

For the record, for my ASRock EP2C602-4L/D16 board, I bought this exact ram from this exact ebay seller and it works without issue:

 

http://www.ebay.ca/itm/64GB-8x8GB-PC3L-10600R-DDR3-1333-ECC-Server-Memory-RAM-Dell-PowerEdge-R410-/201380486027?hash=item2ee3365b8b

 

That said, going with what the manufacturer recommends is always the best move, no question. Definitely saves you from potential issues down the road.

 

For me, the price difference between the recommended ram, and what I bought, was enough to justify the purchase. Prices have changed substantially, so the difference may not be as much anymore (in which case, manufacturer recommended is the smart move).

 

 

Link to comment
13 minutes ago, DoeBoye said:

For the record, for my ASRock EP2C602-4L/D16 board, I bought this exact ram from this exact ebay seller and it works without issue:

 

 

 

 

 

Very good point!! Nowadays things really do just work out for the most part. :) Cheers!

 

But still, yeah, better safe than pulling your hair out. 

 

Cheers!

Link to comment
1 hour ago, DoeBoye said:

For the record, for my ASRock EP2C602-4L/D16 board, I bought this exact ram from this exact ebay seller and it works without issue:

 

http://www.ebay.ca/itm/64GB-8x8GB-PC3L-10600R-DDR3-1333-ECC-Server-Memory-RAM-Dell-PowerEdge-R410-/201380486027?hash=item2ee3365b8b

 

That said, going with what the manufacturer recommends is always the best move, no question. Definitely saves you from potential issues down the road.

 

For me, the price difference between the recommended ram, and what I bought, was enough to justify the purchase. Prices have changed substantially, so the difference may not be as much anymore (in which case, manufacturer recommended is the smart move).

 

 

Awesome, great information! I'll consider this as well as I keep looking. Thanks!

Link to comment
  • 3 weeks later...

Hey guys, I built a dual xeon 2670 with the Intel S2600C motherboard from Natex.  I put it in a Norco 4224, and I am using two Noctua NH-U9DXi4 for the CPU cooling.  I've got esxi running a couple ubuntu VM's and when I check my sensors, coretemp is reading 100.0*C for every single core, never fluctuating. My VM has 10 cores assigned out of the 16 available, and the sensor output is only showing 10 core temps.  The heatsink is cold to the touch, and there's no burning smell, so I'm not actually sure if it's idling at 100.0*C. 

 

Is it possible that the heatsink isn't installed properly and the chip is hitting a thermal limiter?  I haven't upgraded the BMC, Bios, FRUSDR and ME, would this have an effect on the temp sensor?  Can I upgrade these things without wiping my system?

 

Thanks!

 

 

coretemp 100C.png

Edited by seanap
Link to comment
26 minutes ago, seanap said:

Hey guys, I built a dual xeon 2670 with the Intel S2600C motherboard from Natex.  I put it in a Norco 4224, and I am using two Noctua NH-U9DXi4 for the CPU cooling.  I've got esxi running a couple ubuntu VM's and when I check my sensors, coretemp is reading 100.0*C for every single core, never fluctuating. My VM has 10 cores assigned out of the 16 available, and the sensor output is only showing 10 core temps.  The heatsink is cold to the touch, and there's no burning smell, so I'm not actually sure if it's idling at 100.0*C. 

 

Is it possible that the heatsink isn't installed properly and the chip is hitting a thermal limiter?  I haven't upgraded the BMC, Bios, FRUSDR and ME, would this have an effect on the temp sensor?  Can I upgrade these things without wiping my system?

 

Thanks!

 

 

coretemp 100C.png

 

 

first thing is shut that puppy down, re-check the fans, did you remove the plastic from the heat sink , use good thermal paste?

 

I would also update the board would not hurt.

Link to comment

It has been this way idling for about a month now, and I'm just noticing this now.  Fans are all spinning, everything feels nice and cool, I did remove the plastic from the heat sink, and I used Arctic Silver 5 thermal compound.  I will shut this down and try to reseat the heatsink.

 

Will upgrading the board mess with any of my VM's or esxi?

  • Upvote 1
Link to comment
57 minutes ago, seanap said:

Is it possible that the heatsink isn't installed properly and the chip is hitting a thermal limiter?  I haven't upgraded the BMC, Bios, FRUSDR and ME, would this have an effect on the temp sensor?  

your picture is look like from Ubuntu. if you run it under ESXi, then you are ok - i have some Ubuntu VMs under ESXi, and all the sensors are the same as yours - that mean, you can't read it from VM this way. Does this mobo have some IPMI? i can read my Supermicro mobo sensors with impitool for example inside VM.

  • Upvote 1
Link to comment
1 minute ago, uldise said:

i have some Ubuntu VMs under ESXi, and all the sensors are the same as yours

That is very reassuring, this is my first time playing around with esxi and VM's, I'm sure its something simple that I over looked.  I havent configured anything for IPMI.  Is there a way to check the coretemp in esxi, or do I have to look somewhere in bios? 

Link to comment
7 minutes ago, seanap said:

That is very reassuring, this is my first time playing around with esxi and VM's, I'm sure its something simple that I over looked.  I havent configured anything for IPMI.  Is there a way to check the coretemp in esxi, or do I have to look somewhere in bios? 

i don't think coretemp will work from VM inside, so:

First, look in bios for BMC LAN Configuration, configure it, and then read sensors inside VM with:

ipmitool -H <ip> -U <usr> -P <psw> sensor list all

  • Upvote 1
Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.