New Flagship Xeon E5 and E7 Processors

Intel has recently released two, new “flagship” Xeon processors, one for the E7 v4 product family, and one for the E5 v4 product family. The new Intel Xeon E7-8894 v4 processor has 24 physical cores, and runs at a slightly higher base clock speed of 2.4GHz, compared to the 2.2GHz base clock speed of the previous flagship Intel Xeon E7-8890 v4 processor.

All of the other specifications of the E7-8894 v4 are identical to the earlier E7-8890 v4. One big difference between these two processors is the price. The new Xeon E7-8894 v4 is $8898.00 while the older Xeon E7-8890 v4 is $7174.00, which is a 24% price increase. While this seems like a pretty significant price increase by Intel, I think that most organizations that have a need for this type of hardware are not going to be very sensitive to that difference in hardware cost.

From a SQL Server 2016 Enterprise Edition license cost perspective, each physical core license is $7128.00. A four-socket Dell PowerEdge R930 server would require 96 core licenses, which would cost $684,288.00. The added $6,896.00 hardware cost of four E7-8894 v4 processors vs. four E7-8890 v4 processors is pretty trivial. The base clock speed increase is 9%, which means better single-threaded performance, which actually makes that large investment in SQL Server 2016 licenses more acceptable. Getting 9% more CPU capacity and 9% better single-threaded performance for less than 1% of the total hardware and license cost is actually a pretty good ROI. Table 1 shows some comparative metrics for a four-socket system using either of these two processors.

 

Picture1

Table 1: Comparative Metrics for Xeon E7-8894 v4 vs. Xeon E7-8890 v4 Processors

 

Back in Q4 of 2016, Intel made a similar new flagship model introduction in the Xeon E5 v4 product family with the rollout of the Intel Xeon E5-2699A v4 processor. This new flagship SKU has 22 physical cores and a base clock speed of 2.4GHz, compared to the 2.2GHz base clock speed of the previous flagship Intel Xeon E5-2699 v4 processor. Again, all of the other specifications for the E5-2699A v4 are identical to the earlier E5-2699 v4. There was also a significant price increase for this new flagship processor, with the new SKU costing $4938.00 vs. $4115.00 for the older flagship SKU, which represents a 20% price increase. This also seems like a case of price gouging from Intel, but is is actually acceptable from a SQL Server 2016 license cost perspective.

A two-socket Dell PowerEdge R730 server would require 44 core licenses, which would cost $313,632.00. The added $1,646.00 hardware cost of two E5-2699A v4 processors vs. two E5-2699 v4 processors is even more trivial. Table 2 shows some comparative metrics for a two-socket system using either of these two processors.

 

Picture2

Table 2: Comparative Metrics for Xeon E5-2699A v4 vs. Xeon E5-2699 v4 Processors

 

In both cases, my standard guidance about selecting the fastest available processor for a given physical core count for SQL Server usage still stands. The added hardware cost for getting the fastest processor core is really insignificant compared to the total system cost, including licensing costs.

The fact that Intel feels justified in charging 20-24% more for just 9% more performance is just a sad fact that stems from them not currently having any viable competition in the server CPU space from AMD. I really do hope that the next round of AMD Opteron processors based on the Zen microarchitecture are successful, and start to give Intel some decent competition.

Still, as a SQL Server DBA, getting 9% more capacity and 9% better single-threaded CPU performance for less than 1% higher system cost is actually a pretty good deal.

 

 

Intel Xeon E7 Processor Generational Performance Comparison

Intel has a fairly recent document titled Accelerated Operations for Telecom and Financial Services which is also listed under Accelerate OLTP Database Performance with Intel TSX. It describes the “performance” increases seen with the AsiaInfo ADB from moving from 2.8GHz Intel Xeon E7-4890 v2 (Ivy Bridge-EX), to 2.5GHz Intel Xeon E7-8890 v3 (Haswell-EX), and finally to 2.2GHz Intel Xeon E7-8890 v4 (Broadwell-EX) processors, as shown in Figure 1.

 

Figure 1: Speedup from Successive Processor Generations

 

This workload is described as “AsiaInfo ADB Database OCS k-tpmC”, while the AsiaInfo ADB is described as “a scalable OLTP database that targets high performance and mission critical businesses such as online charge service (OCS) in the telecom industry”, that runs on Linux.

The reason I have performance in quotes above is because what they are really measuring is closer to what I would call capacity or scalability. Their topline result is “Thousands of Transactions per Minute” as measured with these different hardware and storage configurations.

The key point to keep in mind with these types of benchmarks is whether they are actually comparing relatively comparable systems or not. In this case, the systems are quite similar, except for the core counts of the successive processor models (and the DD3 vs. DDR4 memory support). Here are the system components, as listed in the footnotes of the document:

Baseline: Four-sockets, 15-core Intel Xeon E7-4890 v2, 256GB DDR3/1333 DIMM, Intel DC S3700 SATA for OS, (2) 2TB Intel DC P3700 PCIe NVMe for storage, 10GbE Intel X540-AT2 NIC

Next Generation: Four-sockets, 18-core Intel Xeon E7-8890 v3, 256GB DDR4/1600 LVDIMM, Intel DC S3700 SATA for OS, (2) 2TB Intel DC P3700 PCIe NVMe for storage, 10GbE Intel X540-AT2 NIC

New: Four-sockets, 24-core Intel Xeon E7-8890 v4, 256GB DDR4/1600 LVDIMM, Intel DC S3700 SATA for OS, (2) 2TB Intel DC P3700 PCIe NVMe for storage, 10GbE Intel X540-AT2 NIC

The baseline system has a total of 60 physical cores, running at 2.8GHz, using the older Ivy Bridge-EX microarchitecture. The next generation system has a total of 72 physical cores, running at 2.5GHz, using the slightly newer Haswell-EX microarchitecture. Finally, the new system has a total of 96 physical cores, running at 2.2GHz, using the current Broadwell-EX microarchitecture. These differences in core counts, base clock speeds, and microarchitecture make it a little harder to fully understand their benchmark results in a realistic manner.

Table 1 shows some relevant metrics for these three system configurations. The older generation processors have fewer cores, but run at a higher base clock speed. The newer generation processors would be faster than the older generation processors at the same clock speed, but the base clock speed is lower as the core counts have increased with each successive generation flagship processor. The improvements in IPC and single-threaded performance are obscured by lower base clock speeds as the core counts increase, which makes the final score increase less impressive.

 

ProcessorBase ClockTotal System CoresRaw ScoreScore/Core
Xeon E7-4890 v22.8GHz6072512.08
Xeon E7-8890 v32.5GHz72102114.18
Xeon E7-8890 v42.2GHz96129413.48

Table 1: Analysis of ADB Benchmark Results

 

Table 2 shows some metrics from an analysis of some actual and estimated TPC-E benchmark results for those same three system configurations, plus an additional processor choice that I added. The results are pretty similar, which supports the idea that both of these benchmarks are CPU-limited. From a SQL Server 2016 perspective, you are going to be better off from a performance/license cost perspective if you purposely choose a lower core count “frequency-optimized” processor (at the cost of less total system capacity per host).

This is somewhat harder to do with the Intel Xeon E7 v4 family, because of your limited SKU choices. A good processor choice for many workloads would be the 10-core Intel Xeon E7-8891 v4 processor, which has a base clock speed of 2.8GHz and a 60MB L3 cache that is shared by only 10 cores.

If you could spread your workload across two database servers, you would be much better off with two, four-socket servers with the 10-core Xeon E7-8891 v4 rather than one four-socket server with the 24-core Xeon E7-8890 v4. You would have more total system processor capacity, roughly 27% better single-threaded CPU performance, twice the total system memory capacity, and twice the total number of PCIe 3.0 expansion slots. You would also only need 80 SQL Server 2016 Enterprise Edition core licenses rather than 96 core licenses, which would save you about $114K in license costs. That license savings would probably pay for both database servers, depending on their exact configuration.

 

ProcessorBase ClockTotal System CoresEst TPC-E ScoreScore/Core
Xeon E7-4890 v22.8GHz605576.2792.94
Xeon E7-8890 v32.5GHz726964.7596.73
Xeon E7-8890 v42.2GHz969068.0094.46
Xeon E7-8891 v42.8GHz404808.79120.22

Table 2: Analysis of Estimated TPC-E Benchmark Results

 

The Intel document also discusses the “performance” increases seen from moving from Intel DC S3700 SATA drives to Intel DC P3700 PCIe NVMe drives. This is going to be primarily influenced by the advantages of being connected directly to the PCIe bus and the lower latency and overhead of the NVMe protocol compared to the older AHCI protocol.

Finally, they talk about the “performance” increases they measured from enabling the Intel Transactional Synchronization Extensions (TSX) instruction set and the Intel AVX 2.0 instruction set on current generation Intel E7-8800 v4 series processors.

SQL Server 2016 already has hardware support for older SSE/AVX instructions as discussed here and here. I really hope that Microsoft decides to add even more support for newer instruction sets (such as TSX) in SQL Server vNext.

 

 

More Intel Broadwell Server Processor Families Available

Back on March 31, 2016 (the last day of Q1 2016), Intel released the 14nm Xeon E5-2600 v4 Product Family (Broadwell-EP), which is currently the highest single-threaded performance product family for two-socket servers. You can read about some of the specific enhancements in this family in this document: Intel® Xeon® Processor E5-2600 V4 Product Family Technical Overview. These processors are socket compatible with the previous Haswell-EP family, so they work with (and are currently available for) existing server models from all of the major server vendors.

If you are going to purchase a new Xeon E5-2600 v4 system (which I think is the best choice for most people), then the model numbers shown in Table 1 are your best choices for SQL Server usage, for each physical core count. The eight-core Xeon E5-2667 v4, the six-core Xeon E5-2643 v4 and the four-core Xeon E5-2637 v4 are the “frequency-optimized” models that would be my preferred choices in most situations for SQL Server usage.

 

ModelCores/L3 CacheBase ClockTurbo ClockPrice
Xeon E5-2699 v422/55MB2.20 GHz3.60 GHz$4,115.00
Xeon E5-2698 v420/50MB2.20 GHz3.60 GHz$3,226.00
Xeon E5-2697 v418/45MB2.30 GHz3.60 GHz$2,702.00
Xeon E5-2697A v416/40MB2.60 GHz3.60 GHz$2,891.00
Xeon E5-2690 v414/35MB2.60 GHz3.50 GHz$2,090.00
Xeon E5-2687W v412/30MB3.00 GHz3.50 GHz$2,141.00
Xeon E5-2640 v410/25MB2.40 GHz3.40 GHz$939.00
Xeon E5-2667 v48/25MB3.20 GHz3.60 GHz$2,057.00
Xeon E5-2643 v46/20MB3.40 GHz3.70 GHz$1,552.00
Xeon E5-2637 v44/15MB3.50 GHz3.70 GHz$996.00

Table 1: Preferred Xeon E5-2600 v4 Family Processors (Broadwell-EP)

 

On June 6, 2016, Intel released the 14nm Xeon E7 v4 Product Family (Broadwell-EX), for four-socket and larger servers. These processors are also socket compatible with the previous Haswell-EX family, so they work with (and are currently available for) existing server models from all of the major server vendors. I have to admit that this quick release of the Xeon E7 v4 took me by surprise, since I was not expecting it until late 2016 or early 2017, based on Intel’s past release history. Typically, there is anywhere from a six to eighteen month delay between the release of a new generation Xeon E5-2600 processor family, and an equivalent generation Xeon E7 processor family.

If you are going to purchase a new Xeon E7 v4 system, then the model numbers shown in Table 2 are your best choices for SQL Server usage, for each physical core count. The eighteen-core Xeon E7-8867 v4, the ten-core Xeon E7-8891 v4 and the four-core Xeon E7-8893 v4 are the “frequency-optimized” models that would be my preferred choices in most situations for SQL Server usage.

 

ModelCores/L3 CacheBase ClockTurbo ClockPrice
Xeon E7-8890 v424/60MB2.20 GHz3.40 GHz$7,174.00
Xeon E7-8880 v422/55MB2.20 GHz3.30 GHz$5,895.00
Xeon E7-8870 v420/50MB2.10 GHz3.00 GHz$4,672.00
Xeon E7-8867 v418/45MB2.40 GHz3.30 GHz$4,672.00
Xeon E7-4850 v416/40MB2.10 GHz2.80 GHz$3,003.00
Xeon E7-4830 v414/35MB2.00 GHz2.80 GHz$2,170.00
Xeon E7-8891 v410/60MB2.80 GHz3.50 GHz$6,841.00
Xeon E7-4809 v48/20MB2.10 GHz2.10 GHz$1,223.00
Xeon E7-8893 v44/60MB3.20 GHz3.50 GHz$6,841.00

Table 2: Preferred Xeon E7 v4 Family Processors (Broadwell-EX)

 

On June 20, 2016, Intel released the 14nm Xeon E5-4600 v4 Product Family (Broadwell-EP), for four-socket servers, which was also somewhat of a surprise. To be honest, I am not a big fan of the E5-4600 v4 Family. The Xeon E5 is really meant for two-socket systems, and it does not seem to scale very well in a four-socket system. If you really need the higher total core counts and higher total memory limit that you would get from a Xeon E5-4600 v4 vs. a Xeon E5-2600 v4 system, I think you are better off to make the jump to a Xeon E7 v4 system (which will also get you more RAS features).

All of the Xeon E5-4600 v4 Family processors have relatively low base clock speeds, and the higher core count SKUs are actually quite expensive, basically in the same range as similar core count Xeon E7 v4 processors. If you really want a Xeon E5-4600 v4 system, then the model numbers shown in Table 3 are your best choices for SQL Server usage, for each physical core count. The ten-core Xeon E5-4627 v4 and eight-core Xeon E5-4655 v4 would be my preferred choices in most situations, since they are somewhat “frequency-optimized”, and have lower overall core counts, which will dramatically reduce your SQL Server 2016 licensing costs.

 

ModelCores/L3 CacheBase ClockTurbo ClockPrice
Xeon E5-4669 v422/55MB2.20 GHz3.00 GHz$7,007.00
Xeon E5-4667 v418/45MB2.20 GHz3.00 GHz$5,729.00
Xeon E5-4660 v416/40MB2.20 GHz3.00 GHz$4,727.00
Xeon E5-4650 v414/35MB2.20 GHz2.80 GHz$3,838.00
Xeon E5-4640 v412/30MB2.10 GHz2.60 GHz$2,837.00
Xeon E5-4627 v410/25MB2.60 GHz3.20 GHz$2,225.00
Xeon E5-4655 v48/30MB2.50 GHz3.20 GHz$4,616.00

Table 3: Preferred Xeon E5-4600 v4 Family Processors (Broadwell-EP)

In many situations, you would be much better off to get two, very fast Xeon E5-2600 v4 based two-socket database servers rather than one Xeon E7 v4 based four-socket database server, assuming you can split your workload across two database servers.