Back in the spring of this year, we transferred the infrastructure for fresh Xeon Gold 6254. It's too late to do a detailed review of the processor - now more than a year has passed since the release of the "stone" on sale, and everyone knows the details about the processor. However, one feature is noteworthy, the processor has a base frequency of 3.1 GHz and 18 cores, which, with a turbo boost, can all simultaneously operate at a frequency of 3,9 GHz, which allows us, as a cloud provider, to “ship” always a consistently high frequency to virtual machines processor.
Nevertheless, we are still interested in evaluating its ability under load. Let's get started!
Brief description of the processor
As we wrote above, the processor is already familiar to everyone, but we will briefly give its specifications:
Codename
Cascade Lake
Number of Cores
18
CPU base clock speed
3,1 GHz
Maximum clock speed with Turbo Boost technology on all cores
3,9 GHz
Memory types
DDR4-2933
Max. number of memory channels
6
We conduct testing
For the test, we prepared a virtual server with 8 cores and 64 GB of memory dedicated to the virtual machine, the data is located on a fast pool based on an SSD array. We perform testing on a Microsoft SQL Server 2014 database, while the operating system is Windows Server 2016 and, of course, we can’t do without the most important thing - 1C: Enterprise 8.3 (8.3.13.1644).
We also paid attention to . If you haven't read it, then in short: four processors were tested there - 2690, 6244 and 6254. The fastest was 6244, and the result on 6254 scored 27,62 points. This experience interested us, because in early tests in our cloud in the spring of 2020, we received a spread in Gilev tests from 33 to 45, but it didn’t work out less than 30, perhaps this is precisely the feature of working with another DBMS, but this prompted us to take measurements on our own infrastructure. We spent again and will share them.
So let's start testing! What's in the results?
Test result
As we can see, on a MSSQL server with a Xeon Gold 6254 processor with turbo boost enabled, the result is 39 points. We interpret the obtained value into a Gilev score and get a result that is greater than the “Good” score, but not “Great” yet. We consider the result to be good from the point of view of evaluating this particular type of “parrots”. It is important to consider that we did not optimize at the level of the OS and SQL server and got the result as it is, if you wish, you can increase it a little more, but these are the subtleties of tuning, a topic for a separate blog entry.
Here it is also worth making a reservation that we do not call for evaluating the workload of productive bases according to the Gilev test and immediately drawing conclusions about the appropriateness of using a particular processor, but according to our statistics, processors with a frequency of 3 GHz or more will be more efficient when working with 1C, and Gilev's test can show different numbers, even in the conditions of one provider or in the local infrastructure. You can get high results on simpler processors, even non-server ones, but this does not mean that when you “feed” a load in the form of 1C ERP for 50-100 people or Trade, you will get consistently high results. Always pilot and test if possible.
Source: habr.com