ES Xeon Discussion

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

mouse

Member
Mar 3, 2016
44
1
8
54
I'm planning to use it for 4vm
1 win 10 for tvserver
1 pfsense
1 xpenology
1 Linux for squeezeserver and oscam

Inviato dal mio PRO 5 utilizzando Tapatalk
 

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
I'd still recommend the v4 due to more modern architecture and increased overall efficiency.
O/c knowing your applications single thread frequency requirements would be a tremendous help in deciding;)
 
  • Like
Reactions: mouse

Mr.Harmony

New Member
Apr 1, 2017
2
0
1
38
Hello everyone, new to the forum. First a quick thank you for the outstanding information gathered in this thread.

As many other here, I'm intending on building a new workstation pc (dual cpu build). As I'd like to try to get as much bang for my buck as possible, I began seriously considering getting hold of some ES xeons.

As far as I can tell many have already dealt with the japanese trader "icomputer_parts_international" with mostly positive experiences. I'm interested in a couple of chips they're offering and your collective expertise on those would be very welcome:

E5 2666 v3
This one kinda hits my sweet spot on core count and per core speed. However, it's not listed by intel. Further investigation showed, it might be a custom made chip for amazon and is therefor not listed by intel. I'm not sure how trustworthy this chip is. According to the cpu-z pic it's a close to retail revision. However, it's named "Core i3/i5/i7", so the information on this one is pretty much all over the place.
Any help on what to make of this offer will be very much appreciated!

E5 2690 v3
E5 2680 v4
With only the available information, I suppose those two can be considered rather early engineering samples, or am I mistaken?

And from another trader located here in Germany (at least according to ebay):

E5 2658 v4
E5 2658 v4
Two auctions for the same model. I think someone already posted somewhere in this thread that the B0 revision for those is misleading. Seeing the reduced per core speed compared to the retail version I suppose it has to be an early engineering sample right?
Btw, what sorta puzzles me is, that the product description is pretty much the same as with the japanese trader. Might they be affiliated and just work with various accounts in various 'official' locations?

My next step would be to contact the sellers and try to get further information on those chips, i.e. pictures of the two actual chips I'd end up buying and screenshots of what hwinfo64 has to tell about those chips.

Lastly, a little background information on what my machine is intended for. I'm mainly going to use it as an audio workstation. My current PC (i7 3930K) has served me well for the last 4 1/2 years. However, with my latest projects I've constantly been on the edge (sometimes over) of what my machine can provide. I've been hesitant of just building the same machine with up to date parts (i7 6850k for instance) since I'm not sure wether this would leave me enough headroom in computing power to feel comfortable for the next 3-4 years. Thus, the consideration of building a dual CPU machine. The difficulty I'm facing is, that the official information you'll get on audio recording and editing software most of the time says a higher per core speed is to be prefered to a higher core count in terms of overall performance. And it's really difficult to find xeons that are similar to their 68xx/69xx counterparts but simply dual CPU capable.
Apart from that, some casual gaming will also be the case, and from what I can tell of this thread so far, a higher base clock is less likely to leave me with an unwanted bottleneck.

Any constructive feedback on my usage scenario will also be welcome!
 

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
2666 looks good for your application with All Core Clock of 3.2. Cpu-Z is only as good as the data that is being fed to it, so if there is little known then they can't enter it. It of course has no Hyperthreading, but that might not be an issue in your case. And Base Clock is irrelevant, "All Core Turbo" is the value to look for, because thats the "base" clock the CPU will reach if its not thermally constrained.

I have bought from "icomputer_parts_international" before and was quite happy with it, they are reachable via some app as well to actually chat with (and or complain) and that worked reasonably well.

Sellers in Germany rarely are the original 'recipient' of the CPU but usually just are reselling what they got elsewhere. Some are just looking for a quick buck (copying other sellers auction descriptions, claiming later stepping [especially the 2883v3's are prone to that atm] or not mentioning ES stepping at all), some are of course genuine. Whether thats good enough for you is yours to decide.

Regarding your build plan - you should evaluate whether your audio applications are truly multithreaded or not. If yes - then you can get lower clock more cores (slow in short workloads, good in large ones) raising up to the clock/core ratio you can afford. Also true if you run multiple jobs at the same time o/c.

If not, multiple 3GHz cores will not help that much and there is little use in having 10 or 20 of those. Then it might make more sense to go high frequency single cpu with few cores, CPUs like E3/i7/Ryzen would be good for this (if you don't need the memory).
 

Nashten

New Member
Mar 18, 2017
11
1
3
28
Hello everyone, new to the forum. First a quick thank you for the outstanding information gathered in this thread.

-snip-

Any constructive feedback on my usage scenario will also be welcome!
LinusTechTips built a massive 20-or-so-core workstation computer meant for photography and video work a year ago maybe and found out that an i7 beat the $4200 Xeon CPU he had simply because the core clock on the i7 was much, much higher.

Games prefer high core clocks as well. Very very few games use more than 4+ threads/cores. Ashes of the Singularity and Cities: Skylines like a lot of CPU cores, for instance. DirectX 12 titles will use more cores/threads, too.

I'm building a workstation myself and I found an ES i7 CPU that was labelled as a Xeon on Ebay. I confirmed it was an i7 at a local computer shop the other day (and with the help + screenshots of another member here) and have my motherboard and RAM on order. My workload is going to involve a lot of rendering in Autodesk Revit because I'm designing the architecture/buildings for a video game. Revit loves using a lot of cores for high-quality renders with a sprinkling of light sources. My i7 4770k takes 6+ hours to render the sample building included with the program in high quality with all light sources present... and I have it at 4.2GHz.

What programs do you use? Determine whether they scale better with cores or clockspeeds and tailor your choices from there. Is ECC RAM a must? I would have liked to get ECC on my Revit workstation... but oh well.
 
Last edited:

Jointer

Member
Oct 27, 2016
30
0
6
41
My general purpose Dual Xeon have 2667 V4, that is 3,1GHz max (32 threads total) and about half year ago cost 350 USD/piece, however hard to find these days.
I was not able to find faster V4 ES which would be affordable or even exist, some 2687W/269x might be bit faster, but those are not available usually. Search for my posts for details about the build.

When comes to gaming, i have Titan X Maxwell on water (two in fact - recent price drop, but no SLI yet and no water on the 2nd) and run at 2560 x 1600, Fallout 4 seems to be bit CPU constrained (.INI magic did not helped), Division is on the other side, very fast loads and 16+ threads load for sure, Skylines are strange, they seems to run at 30 FPS anyway with 200% oversampling mod, but not CPU hungry even on my old 4770K with TX Pascal at 1920x1200.

I noticed some apps seems to have very slightly longer startup time, for example a Java app i use every day, compared to 4770K (no OC).
I have Samsung 960 EVO NVME in both.
So you should really find out if your apps benefit from more cores and if not, find the highest clocked Xeon or go for standard desktop CPU, where the Ryzens might work better for less money.
And i hope games multithreading will improve with the 8 core Ryzen now being "normal" for higher end desktops.

You can probably use Task manager and affinity as simple way how to detect your CPU usage now.
 

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
My general purpose Dual Xeon have 2667 V4, that is 3,1GHz max (32 threads total) and about half year ago cost 350 USD/piece, however hard to find these days.
Which stepping is that?
 
Last edited:

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
Ah so its 3.2 single and 3.0 All Core turbo. Ok, I wondered b/c 3.1/3.2 base clock would have been pretty close to retail :) Still nice @3.0:)
 

nthu9280

Well-Known Member
Feb 3, 2016
1,628
498
83
San Antonio, TX

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
We need a disclaimer for this thread: If you want CPU recommendations please state your use case :p
As always - depends on what you want to do with it - if you just need 'a cpu' then buy the cheapest v3 you can get.
If you have a use case - which one is it? :)
 

nthu9280

Well-Known Member
Feb 3, 2016
1,628
498
83
San Antonio, TX
As always - depends on what you want to do with it - if you just need 'a cpu' then buy the cheapest v3 you can get.
If you have a use case - which one is it? :)
Great question. Wasn't thinking while trying to digest as much from this thread and having an adult beverage in hand when I posted the request for advice. [emoji482] [emoji3]

While I want the cheapest, it may not be the optimal for the long run.

My main use case would be to run homelab practice VMs for database and development servers for SAP and Business Intelligence, Analytics, Some hadoop and machine learning etc.
I would think more cores/threads and RAM vs clock speed is preferred.
Lower heat & power consumption - guess goes without saying.





Sent from my Nexus 6 using Tapatalk
 

kijun93

Member
Mar 19, 2017
40
8
8
31
Lot of great info. Didn't think I'd come knocking on this thread. Only read 7+ pages so far. Now that I am getting the X10DRI from liquid8 but no more budget or WAF, need a low cost CPU and some DDR4 ram to test the MB during the standard return window.
Looking at the following options. Really appreciate any insight

Intel Xeon E5-2648L v3 QS 12C LGA2011-3 Compatible X99 i7-6950X E5-2650L v3 | eBay
or
Intel Xeon E5-2630L v3 ES LGA2011-3 8C Compatible with X99 i7-6850K 6900K 6950X | eBay
not 100% sure though there's a rumor that 2630l v3 might be able to lock up to 25x for all cores.
 

Mr.Harmony

New Member
Apr 1, 2017
2
0
1
38
Thanks a lot for the replies so far!

I'm working with Steinberg Cubase. Steinberg themself state that cpu speed is to be prefered to core count. However, it's pretty much impossible to find actual figures on how much or if at all this is the case. Also, Steinberg doesn't state wether there is a maximum number of cores that Cubase will be able to manage. So, theoretically it should at least try to work with all the cores I throw at it.

A friend of mine recently build a dual cpu machine with two 8x2,6 GHZ v2 Xeons. Theoretically his machine has about twice the computing power of my current PC. So we took one afternoon to set up my working environment on his new machine and see how the VST load (a combined value of all the factors relevant for real time audio processing) would scale compared to mine. Turned out it was 25-33% lower. So, program efficiency didn't scale linear to computing power in that case. Interestingly, when doing a cpu-z benchmark, his machine actually performed a little worse than mine, which made the result all the more confusing.

Another thing you folks might help me with, is how v3 compares to v4 in terms of computing power. I feel like I never got my head quite around that topic. I know efficiency increases from generation to generation so you can pack more cores and/or higher base clock on a cpu and end up with the same TDP. However, how would a v3 and v4 xeon with identical core count and per core speed compare to each other? Would they perform almost equally in various benchmarks, or would there be a definite increase from one generation to the next?

EDIT:
Just asked my friend again for his CPU-Z score. Turns out it's more than twice as mine. Either he optimised his settings by now or my memory failed me real hard here :p
 
Last edited:

Rand__

Well-Known Member
Mar 6, 2014
6,626
1,767
113
You can run Cinebench 15 - thats a good comparison value for single thread and multi core cpus
 

pippaman

New Member
Feb 25, 2017
1
0
1
32
Hello, i'm new here and i'm studying to buy a xeon rig. I've been lurking here for a while to gather informations and trying to make up my mind on what i should by.

I have a question though, i've spent quite a while comparing prices on ebay with the bench performances on cpu-monkey trying to find a good overall chip between single-multicore performance/$$$. And i'm finding some discrepancies that make it difficult to decide what to buy.

For example look at this comparison:

xeon.jpg
We have (from what i understand) the same architecture and core count and only the clock speeds differs.
Infact looking at the clock speeds the 2667 chip should win across the board, but the 2640 with a slower turbo core speed in single thread wins (and by a big stretch).
How can it be?

Thank you very much
 

diobgh1

New Member
Jan 24, 2017
28
5
3
42
We have (from what i understand) the same architecture and core count and only the clock speeds differs.
Infact looking at the clock speeds the 2667 chip should win across the board, but the 2640 with a slower turbo core speed in single thread wins (and by a big stretch).
How can it be?

Thank you very much
I've also found irregularities with online databases of CPU benchmarks.

If you look at the Passmark website for these cpus, you get:

E5-2667v3 = 16125 (overall score) 2048 (single thread)
E5-2640v3 = 14117 (overall score) 1959 (single thread)

Which seems a little more like we should expect.

I've taken to taking online databases of benchmarks with a grain of salt, is there a more reliable source? I'd love to know.
 

Klee

Well-Known Member
Jun 2, 2016
1,289
396
83
got a new workstation, doing DC full time... :)

With that many video cards plugged into that EP2C612 WS you should plug a molex connector in the black connector on the board, I had some stability issues when I was testing my dual RX 480's until I did.