Intel Xeon E5-2600 v4 Broadwell-EP Review

Supermicro And NVMe RAID Testing: 3 Million IOPS And 21 GB/s

Supermicro SuperServer SYS-1028U-TN10RT+

We already had the slender 1U Supermicro SuperServer SYS-1028U-TN10RT+ in our enterprise storage lab, so after a quick BIOS update we slapped in the Xeon E5-2679 v4 CPUs and went to work.

Don't let its diminutive stature fool you; the SuperServer is a little powerhouse. It features 10 hot-swap NVMe bays that we populated with ten 400GB 2.5" Intel DC P3700 SSDs. Normally, 16 counter-rotating fans provide copious airflow over 24 Samsung 16GB DDR4-2133 ECC DIMMs, but we removed 16 of them to level the playing field. Even with 384GB of DDR4 at our disposal, we are not populating the server to capacity; it supports an amazing 1.5TB of DDR4 ECC LRDIMMs. 

The NVMe SSDs mate to the PCIe backplane, which Supermicro routes to a pair of switchless expansion cards in the rear. These cards provide direct PCIe 3.0 x4 connections to the CPU, enabling incredible performance. The back of the chassis features two 10GbE ports and one management port, along with dual Titanium-rated 1000W redundant power supplies.

Need a slice of parallel processing in the storage sandwich? You also get three auxiliary eight-pin power connectors, allowing the server to accommodate either two full-height full-length graphics cards, or one double-width board and a low-profile offering.

We configured the 10 Intel NVMe SSDs into a RAID 0 volume (mdadm) in CentOS 7, demonstrating how the E5 2697 v4's 36 threads come into play under heavy I/O operations. The colored spaghetti strings in the picture above indicate the load of each core during a 32-thread 4KB random read workload.

Incredibly, we're pushing just over 3 million IOPS during this capture with an aggregate CPU utilization of 90.9 percent. To put that into perspective, you're seeing nearly 12 GB/s of random throughput. Now you see why very fast multi-core CPUs are so important to extracting maximum performance from NVMe-based SSDs in servers.

These results are hard to beat; we recently took 64 of the fastest PCIe SSDs out for a spin in 64 PCIe SSDs, 120TB Of Flash And One Stop Systems: The FSA 200 Review. The FAS 200 topped out with 3,055,735 IOPS in a system that monopolizes 5U of rack space. Meanwhile, the 1U Supermicro server musters even more random read performance thanks to its switchless design. We reach 814,351 IOPS during the 4KB random write workload, which is excellent scaling considering that Intel specs each drive at 75,000 random write IOPS.

The SuperServer continues to impress with a beastly 21 GB/s of sequential read throughput (yes, gigabytes), easily surpassing the 13 GB/s we achieved with the FAS 200. 

It is hard not to like this much power packed into such a small footprint; it's the personification of both performance and storage density. Now that we have seen what Intel's current-gen SSDs can do, let's take a look at the next generation.

This thread is closed for comments
22 comments
    Your comment
  • utroz
    Hmm well we know that Broadwell-E chips must be coming very very soon if Intel let this info out.
  • bit_user
    Wasn't there supposed to be a 4-core 5.0 GHz SKU? Single-thread performance still matters, in many cases.
  • turkey3_scratch
    328798 said:
    Wasn't there supposed to be a 4-core 5.0 GHz SKU? Single-thread performance still matters, in many cases.


    In most server applications it doesn't matter as much as multithreaded performance. If you need single-core strength, getting a consumer chip is actually better, but you probably aren't running a server if single-threaded is your focus.
  • PaulyAlcorn
    Quote:
    Wasn't there supposed to be a 4-core 5.0 GHz SKU? Single-thread performance still matters, in many cases.

    I read the rumors on that as well, but nothing official has surfaced as of yet to my knowledge.
  • bit_user
    1712875 said:
    328798 said:
    Wasn't there supposed to be a 4-core 5.0 GHz SKU? Single-thread performance still matters, in many cases.
    In most server applications it doesn't matter as much as multithreaded performance. If you need single-core strength, getting a consumer chip is actually better, but you probably aren't running a server if single-threaded is your focus.
    Try telling that to high-frequency traders. I'm sure they want the reliability features of Xeons (ECC, for example), but the highest clock speed available.

    And the fact that Intel even released low-core high-clock SKUs is an acknowledgement of this continuing need. Clock just not as high as I'd read. With the other specs basically matching the Haswell version, the only difference is ~5% IPC improvement. Seems pretty poor improvement, for a die-shrink.
  • firefoxx04
    Would nice to have a quad core xeon that turbos at 4.4ghz just like the 4790k. I had to go with a 4690k when building an autocad system because it only uses one core and needs that core to be fast... this means i have to sacrifice ecc support.
  • bit_user
    2074532 said:
    Quote:
    Wasn't there supposed to be a 4-core 5.0 GHz SKU? Single-thread performance still matters, in many cases.
    I read the rumors on that as well, but nothing official has surfaced as of yet to my knowledge.

    On wccftech (not the most reliable source, I know), they claimed:

    Model: Intel Xeon E5-2602 V4
    Cores/threads: 4/8
    Base clock: 5.1 GHz
    Turbo clock: TBD
    L3 Cache: 5 MB
    TDP: 165W

    Given what we know about 2.5 MB/core of L3 Cache, the 5 MB figure sounds suspicious. It's conceivable they could disable some to hit the target TDP, I guess.
  • firefoxx04
    We cant get skylake to consistently hit 5ghz... why would a xeon chip suddenly hit 5ghz?
  • JamesSneed
    211300 said:
    We cant get skylake to consistently hit 5ghz... why would a xeon chip suddenly hit 5ghz?


    I'm not saying the 5Ghz rumor is true but Intel has always known which chips can hit higher clocks during certification if the chip is a top end or low end chip cores disabled etc. I'm sure they could cherry pick a few to sell for $$$ if they wanted. Now are they I have no real idea.
  • bit_user
    211300 said:
    We cant get skylake to consistently hit 5ghz... why would a xeon chip suddenly hit 5ghz?
    Well, I was surprised, too.

    There are obviously things you can do in chip design that allow one to reach different timing targets. And I was hoping they might've refined their 14 nm process, since the time the first Broadwells launched. So, I thought, with more TDP headroom afforded by this socket (roughly double what Skylake has to work with), maybe they could do it.

    I thought maybe Intel was addressing some pent-up demand for high clockspeed applications. That said, it seemed particularly odd in Broadwell, given that it generally seems oriented towards lower clockspeed / lower power applications.

    But maybe it was a typo, or even a blatant lie, in order to track down leakers.
  • alidan
    Quote:
    We cant get skylake to consistently hit 5ghz... why would a xeon chip suddenly hit 5ghz?


    proper binning and sold specifically as that because of what it hits, this could double/triple the value of the chip at least compared to other lower binned versions.
  • thor220
    Quote:
    Wasn't there supposed to be a 4-core 5.0 GHz SKU? Single-thread performance still matters, in many cases.


    A really high clock on a server platform seems like an overclocker's dream to me. Stability and performance. Not to mention that server processors use solder instead of that cheap paste Intel uses in their consumer processors.
  • RedJaron
    Doesn't sound right to me. A server chip binned that high would be ridiculously expensive, more than even the 5960X. I can't see then selling more than a couple hundred to the richest and most eccentric computer enthusiasts.
  • LudeMasta99
    How many FPS will I get in Crysis with this?
  • Adriano Bordignon
    How does Photoshop behave under this cpu?
  • bit_user
    570460 said:
    Doesn't sound right to me. A server chip binned that high would be ridiculously expensive, more than even the 5960X. I can't see then selling more than a couple hundred to the richest and most eccentric computer enthusiasts.
    FWIW, IBM introduced Power6 processors in 2007 & 2009 that were clocked up to 5 GHz. No doubt, they cost an arm and a couple legs.
  • Waldek
    Slightly off the topic, but... I was curious about the data centers' power consumption statistics. The article says 416.2 TWh per year. This is true. What the article says incorrectly, however, is that it would be more than 182 countries (of 192). The correct example would be that this gives the datacenters of the world 11th place in the power consumption ranking in the world. For example, the UK alone consumes 320 TWh (and is currently number 11 worldwide). The datacenters consume currently ca. 5% of the world's power usage...
  • sincreator
    Getting a chip to hit 5.0ghz or more stable is pretty rare to say the least. Silicon Lottery https://siliconlottery.com/collections/2011-3 specializes in picking out binned chips to sell, and they don't even have one model that is clocked that high.
  • PaulyAlcorn
    Quote:
    Slightly off the topic, but... I was curious about the data centers' power consumption statistics. The article says 416.2 TWh per year. This is true. What the article says incorrectly, however, is that it would be more than 182 countries (of 192). The correct example would be that this gives the datacenters of the world 11th place in the power consumption ranking in the world. For example, the UK alone consumes 320 TWh (and is currently number 11 worldwide). The datacenters consume currently ca. 5% of the world's power usage...


    The article does not state that it is more than the *combined* total of 182 countries, merely that it consumes more power than each of them compared individually. You are right,mentioning that it would place 11th in the world is probably a better way of stating the statistic.
  • bit_user
    248772 said:
    Getting a chip to hit 5.0ghz or more stable is pretty rare to say the least. Silicon Lottery https://siliconlottery.com/collections/2011-3 specializes in picking out binned chips to sell, and they don't even have one model that is clocked that high.
    Sure, but there's a difference between binning chips designed to run at a lower clock vs. actually designing a chip to hit higher clock speeds. There's no reason Intel can't make chips that clock higher, but they don't choose to because they think there's not sufficient market demand for something which burns so much power. AMD tried this with 225 W TDP Bulldozers, a few years back.

    I remember reading that the Pentium 4 was originally designed to scale up to 10 GHz, by the end of its production. Of course, back then, the only way they could hit those speeds was to use really long pipelines composed of very simple stages. Then, when they discovered that leakage of newer process nodes was higher than anticipated, they were left with a very inefficient architecture that was stuck below the clock speeds that would've made it competitive.

    These days, I think Intel could do it without such a drastic architectural tradeoff. But it still comes down to a power vs. clock, no matter what.
  • utroz
    328798 said:
    570460 said:
    Doesn't sound right to me. A server chip binned that high would be ridiculously expensive, more than even the 5960X. I can't see then selling more than a couple hundred to the richest and most eccentric computer enthusiasts.
    FWIW, IBM introduced Power6 processors in 2007 & 2009 that were clocked up to 5 GHz. No doubt, they cost an arm and a couple legs.


    Those IBM chips had a really long pipeline to allow clock speeds that high as well as an SOI process node basically built from the ground up for them. I wonder what version of 14nm Intel is using for Broadwell-E/EP/EX as I know they had one version they used for the Broadwell-U,Y,H,DT(C) and when they moved to Skylake they used an updated version of 14nm. Is it possible that Broadwell_E/EP/EX are using the updated 14nm process?
  • pastorpastor
    nice review, but I'm deceived, there is no important 3d rendering benchmarks like cinebench 3dsmax / VRAY