Intel pours pros a cup of Coffee Lake with Xeon E CPUs

Intel's Coffee Lake CPUs are fine performers for applications where single-threaded performance is paramount. Though workstation integrators like Boxx have built Coffee Lake systems, those chips don't tick every box a workstation user might want. Most notably, they lack support for ECC RAM, and that makes them a no-go for many professional users. Intel has brewed a new round of Coffee Lake CPUs for professionals who need that single-threaded mojo, and it's pouring those folks a cup today with its Xeon E processor family.





Turbo Boost






UHD P630






TDP Price
E-2186G 3.8 4.7 6/12 Yes 12 MB Two channels

DDR4-2666 with ECC

95 W $450
E-2176G 3.7 80 W $362
E-2174G 3.8 4/8 8 MB 71 W $328
E-2146G 3.5 4.5 6/12 12 MB 80 W $311
E-2144G 3.6 4/8 8 MB 71 W $272
E-2136 3.3 6/12 No 12 MB 80 W $284
E-2134 3.5 4/8 8 MB 71 W $250
E-2126G 3.3 6/6 Yes 12 MB 80 W $255
E-2124G 3.4 4/4 8 MB 71 W $213
E-2124 3.3 4.3 No $193

These ten CPUs comprise four-core and six-core parts that can be had with or without Hyper-Threading and with or without Integrated graphics to fit into any bill of materials or thermal budget. Conveniently, Intel uses the “G” suffix to indicate whether the chips' UHD Graphics P630 IGP is enabled. Don't confuse that label's meaning with the same G suffix used for the Core i7-8809G and friends, though—these Xeons do not have a Radeon Vega M GH or Vega M GL graphics chip on-package. Like consumer Coffee Lake parts, these chips will all provide 16 lanes of PCIe from the CPU.

The Xeon E family drops into motherboards with the C246 chipset. This silicon appears to be a tweaked version of the same silicon that underpins the H370, B360, and H310 consumer chipsets. Most notably, it includes the vPro support critical to any business-ready Intel PC. The C246 chip can be tapped for up to 24 chipset PCIe 3.0 lanes, as many as six USB 3.1 Gen 2 ports, as many as 10 USB 3.1 Gen 1 ports, and as many as eight SATA 6 Gbps ports. The C246 chip also has integrated Wi-Fi connectivity that can be paired with an Intel CNVi Wireless-AC module for businesses that want to cut the Ethernet cord. Devices from the chipset will communicate with the CPU over the DMI 3.0 bus and the four lanes of PCIe 3.0 connectivity that underpin it.

Intel says OEM systems from Dell, HP, and Lenovo will spearhead the introduction of Xeon E processors to the business market, and other OEMs will doubtless take advantage of these parts with time.

Comments closed
    • Krogoth
    • 1 year ago

    Funny that Intel is cannibalizing their own lower-end Skylake-X LCC SKUs with these chips.

      • chuckula
      • 1 year ago

      AMD is smart enough to avoid that problem.


        • ronch
        • 1 year ago

        Well, AMD is the Smarter Choice, after all.

          • Kretschmer
          • 1 year ago

          Yeah, except from 2011-2017.

      • jihadjoe
      • 1 year ago

      Skylake-X LCC was a mistake from the start!

      • blastdoor
      • 1 year ago

      Are these cheaper than those?

      • NoOne ButMe
      • 1 year ago

      Those can support up to twice the memory, and the cost per core isn’t that far off.

      And extra PCIe lanes from die.

      I don’t see it.

      edit: i lied. Forgot that i9 parts don’t support ECC. So it is a massive price drop on the Xeon W line, but that Xeon W line also supporst half a GB of RAM, and all offer 48 PCIe lanes direct from the CPU.

        • Krogoth
        • 1 year ago

        It is because of stock clockspeed.

        The Coffee Lake Xeons have higher base and turbo speeds then lower-end LCC chips (6-8 cores models). The only reason to go lower-core count is because you need more clockspeed for your workload.

        If you need more PCIe lanes and memory capacities chances are that you are going to want more cores.

        Poor binned Skylake-X LCC chips are in a very tough spot.

    • Chrispy_
    • 1 year ago

    Our workstations and renderfarm doesn’t need ECC, but I’m curious what ECC-requiring niche this serves.

    If you need multi-threaded performance, Ryzen’s been winning in the consumer ECC space for a long time.

    If you need a mix of single-threaded speed and multi-threaded performance, Ryzen’s been winning in the performance/$ and very competitive on overall performance and performance/Watt for a long time.

    If you need single threaded performance, it looks like Intel don’t offer anything with low core/thread count and high clock, you need to pay top dollar for the high-end models.

    Back in the old days (Skylake) Intel simply satisfied this market with i3 processors. If people have the budget for lots of cores, they’ll likely get themselves a Threadripper – unless of course they’re vendor-locked to an anti-AMD OEM.

    • crystall
    • 1 year ago

    Quick comparison with the previous generation (or the “Ryzen effect”): the top SKU has 50% more cores, almost 20% higher turbo clock, includes a iGPU and costs 26% less than the Xeon E3 v5 top SKU. Thanks AMD!

    • JosiahBradley
    • 1 year ago

    You’d be better off with any Ryzen CPU at these prices. AsRock boards support ECC and so does consumer Ryzen. Why pay extra for these Xeons? This odd artificial limitation is annoying. I guess if you really need AVX then this is your only option but 450$ for a six core? We’ve got a 2600 on sale today for <200$.

      • srg86
      • 1 year ago

      I’d use it for the iGPU (plus I generally prefer Intel over AMD).

      Also I’d trust that Xeon motherboards would have more stringent ECC validation than a consumer board, even if it says supports ECC supported on the box.

      • thecoldanddarkone
      • 1 year ago

      Last time I checked Ryzen is not actually certified on ecc. I’m not sure how it’s odd either, it’s only been this way since at least Sandy Bridge.

      • synthtel2
      • 1 year ago

      When you require ECC and maximum possible performance for a workload that doesn’t scale well beyond 6T? It can’t be *that* rare a scenario, and if it’s what you need for some particular business reason, $450 for the best possible CPU may be much cheaper than any other way of handling it.

      AMD and Intel support the same SIMD instructions, Intel just runs the 256-wide ones faster. That’s very different from AMD not having the support in the first place.

      • MOSFET
      • 1 year ago

      ASRock does not replace SuperMicro, not this year. I would never put an ASRock board into production.

      • jihadjoe
      • 1 year ago

      Ahem. Coffee Lake is better for gaming, and I want to make sure my games run error-free.

      • jihadjoe
      • 1 year ago

      The E-2176G on the list is has exactly the same specs as an 8700k (3.7GHz base, 4.7GHz turbo) with lower TDP (80W) and costs about the same ($362). It’s only the top bin that’s more expensive.

    • Leader952
    • 1 year ago

    Only a lousy 12MB L3 cache and that is only on the 6-core processors.

    My X5687 from 2011 has the same 12MB L3 cache and is only a 4-core processor.


      • srg86
      • 1 year ago

      You’re confusing a Xeon E and Xeon EP product line, they are different.

        • Leader952
        • 1 year ago

        Still stuck with only 2MB L3 per core over 7 years is stagnation.

          • jihadjoe
          • 1 year ago

          Or it could be that 2MB L3 per core is the optimal amount for the current 64-bit instruction set. Even AMD’s Ryzen architecture has 2MB L3 per core.

          You have to realize that needlessly increasing the size of a set-associative cache also increases its latency because the entire data set in the cache has to be searched.

          Ever since caches moved to occupying the same space as the CPU die there really hasn’t been a difference between how L1, L2 and L3 is constructed. The increasing latency as you go from L1 to L2 to L3 is entirely down to the size of the cache.

      • derFunkenstein
      • 1 year ago

      what’d you expect? It’s the same Coffee Lake die as the i7-8700K, only with a few buttons and knobs tweaked.

Pin It on Pinterest

Share This