Linked by Thom Holwerda on Mon 5th Nov 2018 23:14 UTC
Intel

Despite having officially launched back in July, Intel's Xeon E desktop platform has yet to see the light of day in systems casually available to users or small businesses. This should change today, with the official embargo lift for reviews on the parts, as well as the announcement today that SGX-enabled versions are coming for Server use. The Xeon E platform is the replacement for what used to be called the E3-1200 family, using Intel's new nomenclature, and these parts are based on Intel's Coffee Lake (not Coffee Lake Refresh) microarchitecture. We managed to get a few processors in to test, and today we'll start by examining most of the six-core family.

Another great and detailed benchmark by AnandTech.

E-mail Print r 1   5 Comment(s)
Order by: Score:
The Ryzen effect
by crystall on Tue 6th Nov 2018 10:55 UTC
crystall
Member since:
2007-02-06

Compared to the previous generation these new Xeons offer more cores, higher clocks _and_ integrated graphics for lower prices (especially at the high end of the range). Ryzen's ECC support made the old Xeon E3 line woefully uncompetitive and forced Intel's hand. For my purposes (software development) a Ryzen 2700X workstation build is still a better choice. I still get ECC memory but with more cores and for a lower price.

Reply Score: 4

RE: The Ryzen effect
by Kochise on Tue 6th Nov 2018 12:37 UTC in reply to "The Ryzen effect"
Kochise Member since:
2006-03-03

And support AMD's innovative R&D

Reply Score: 2

RE: The Ryzen effect
by Troels on Tue 6th Nov 2018 18:12 UTC in reply to "The Ryzen effect"
Troels Member since:
2005-07-11

I am genuinely curious, why do you need ECC memory for software development? I didn't even know my 2700X can do ECC...

Reply Score: 2

RE[2]: The Ryzen effect
by crystall on Wed 7th Nov 2018 08:21 UTC in reply to "RE: The Ryzen effect"
crystall Member since:
2007-02-06

I work primarily on Firefox which is a very large codebase. This machine has 32 GiB of RAM which is enough to run a concurrent build with 16 jobs and keep most of the sources in memory. It also stays on or in standby most of the time being rarely turned off. Having that much memory without ECC in a single box is asking for trouble.

Even discounting transient errors memory tends to go bad after some time. I've seen it happen on a number of machines the last of which was my laptop where a single bit went bad in one of the two DIMMs. When it happens you usually run into random crashes and data corruption. In my laptop's case I was traveling and wasted two days of work trying to figure out why sometimes a build wouldn't work, then wasted hours to run a memtest, check all the installed packages and the Firefox sources to ensure no data corruption had occurred and finally had to work with only half the memory until I could get a replacement DIMM.

With ECC memory that kind of problem only shows up as a message in your machine logs without hampering your workflow and without corrupting your data. I find the small difference in price well worth the peace of mind it offers.

Reply Score: 2

RE[3]: The Ryzen effect
by Alfman on Wed 7th Nov 2018 11:01 UTC in reply to "RE[2]: The Ryzen effect"
Alfman Member since:
2011-01-28

crystall,

I work primarily on Firefox which is a very large codebase. This machine has 32 GiB of RAM which is enough to run a concurrent build with 16 jobs and keep most of the sources in memory. It also stays on or in standby most of the time being rarely turned off. Having that much memory without ECC in a single box is asking for trouble.

Even discounting transient errors memory tends to go bad after some time. I've seen it happen on a number of machines the last of which was my laptop where a single bit went bad in one of the two DIMMs. When it happens you usually run into random crashes and data corruption. In my laptop's case I was traveling and wasted two days of work trying to figure out why sometimes a build wouldn't work, then wasted hours to run a memtest, check all the installed packages and the Firefox sources to ensure no data corruption had occurred and finally had to work with only half the memory until I could get a replacement DIMM.

With ECC memory that kind of problem only shows up as a message in your machine logs without hampering your workflow and without corrupting your data. I find the small difference in price well worth the peace of mind it offers.


Out of curiosity, could you post your output for "edac-util -v"?

edac-util -v
mc0: 0 Uncorrected Errors with no DIMM info
mc0: 0 Corrected Errors with no DIMM info
mc0: csrow0: 0 Uncorrected Errors
mc0: csrow0: mc#0branch#0channel#0slot#0: 0 Corrected Errors
mc0: csrow0: mc#0branch#0channel#1slot#0: 0 Corrected Errors
mc0: csrow0: mc#0branch#1channel#0slot#0: 0 Corrected Errors
mc0: csrow0: mc#0branch#1channel#1slot#0: 0 Corrected Errors
mc0: csrow1: 0 Uncorrected Errors
mc0: csrow1: mc#0branch#0channel#0slot#1: 0 Corrected Errors
mc0: csrow1: mc#0branch#0channel#1slot#1: 0 Corrected Errors
mc0: csrow1: mc#0branch#1channel#0slot#1: 0 Corrected Errors
mc0: csrow1: mc#0branch#1channel#1slot#1: 0 Corrected Errors
edac-util: No errors to report.




I've read that cosmic rays are theoretically a problem:

https://en.wikipedia.org/wiki/Cosmic_ray#Effect_on_electronics
Cosmic rays have sufficient energy to alter the states of circuit components in electronic integrated circuits, causing transient errors to occur (such as corrupted data in electronic memory devices or incorrect performance of CPUs) often referred to as "soft errors." This has been a problem in electronics at extremely high-altitude, such as in satellites, but with transistors becoming smaller and smaller, this is becoming an increasing concern in ground-level electronics as well.[83] Studies by IBM in the 1990s suggest that computers typically experience about one cosmic-ray-induced error per 256 megabytes of RAM per month.[84] To alleviate this problem, the Intel Corporation has proposed a cosmic ray detector that could be integrated into future high-density microprocessors, allowing the processor to repeat the last command following a cosmic-ray event.[85]


Still, I've never witnessed it on my servers running 24/7. And although I've had bad ram before, I've never actually experienced good ram becoming bad over time. Maybe the memory is getting too hot or the power quality is marginal? Statistically errors must happen some times, and ECC is great to have just in case, but I don't think you should consider it normal if it's happening regularly. You may have had a marginally defective part from the manufacturer. It's not a bad idea to run tests on new ram without waiting for symptoms to occur so you know it's good to begin with.

Reply Score: 4