Need help with server build: D-1528 or D-1541

[MikeS]

I'm about ready to build a server, but need a little help in choosing the CPU. I'm trying to decide between the D-1528 and D-1541. I'll get the corresponding Supermicro X10SDV Mini-ITX board.

I'll be running ESXi and anywhere from 7 - 12 linux VMs and 2-4 windows VMs. Not all of these will be online all the time, but there may be times when many of them are and I don't want to end up with a CPU performance problem. The workloads won't be anything too intense...vulnerability scanners, splunk light, elk stack, MS Office - no graphics or heavy number crunching.

I'm leaning toward the D-1541, but if the D-1528 can do the job without any issues I'd be happy to put the $200+ difference in price towards the rest of the build.

 

[T_Minus]

Go 1541 for sure.

My 02

 

[j_h_o]

Do you have an existing server with some of this workload?

You can compare CPU benchmark numbers for these CPUs vs. your existing one to get some idea of performance difference.

 

[MikeS]

Do you have an existing server with some of this workload?

You can compare CPU benchmark numbers for these CPUs vs. your existing one to get some idea of performance difference.

I've been running the VMs on my desktop i5-4670k with VMware workstation, but it's time to move to a dedicated server. I'm running the i5 at 4.4GHz which is a lot faster than the D-1541, but the i5 only has 4 cores vs. the 8/16 for the D-1541.

 

[Sergio]

Interesting post as I'm in the same situation

I started with the possibility of the 1518 (4c), then @Patrick remind me of the 1528, just between the 1518 and the 1541. My build is going to be a home lab, just experimenting and learning, however will not be operating (most probably ) 24x7. It will be turn on/off according to my needs. Being said that, the price is the most limiting factor for me, as ideally I'd like to build a cluster with at least 2 nodes of this.

Take a look to the prices in Europe:

- 1518 (mobo 600 $ )
- 1528 ( mobo 800 $)
- 1541 ( mobo 1200$)

The price difference from the 1518/1528 it's not that big but the is huge IMO with the 1541. I could get by with the 1518 but the 1528 it's right in the middle regarding performance/money.

What are you thoughts if the equipment doesn't perform 24x7?

 

[j_h_o]

I ran Passmark PerformanceTest on my D-1537 and D-1540. I'm using CPU Mark as some gauge of performance.

Xeon D-1540 Benchmarks · Justin Ho

 

[Evan]

D-1537 take a fair performance hit, turbo must be somewhat limited by the 35watt tdp.
On the other hand it idles less and full power a lot less.

I will compare my new d-1540 power consumption compared to a 35watt shortly to see if the newer d-1540 is any different than the original ones tested by people like @Patrick

 

[MiniKnight]

If you can run these VMs on your i5, you don't need a lot of RAM. 64gb in 2x 32gb on a Xeon D will be more than enough.

With how taxing having way more VMs than CPU cores is, if you are running everything on your desktop, those are very light use VMs. The D-1528 will save money up front and money/ power down the road. With how much more then D-1541 is and how little memory you need you aren't far off from looking at 2x D-1518's. That'd let you really start playing with clustering VM hosts.

 

[MikeS]

If you can run these VMs on your i5, you don't need a lot of RAM. 64gb in 2x 32gb on a Xeon D will be more than enough.

With how taxing having way more VMs than CPU cores is, if you are running everything on your desktop, those are very light use VMs. The D-1528 will save money up front and money/ power down the road. With how much more then D-1541 is and how little memory you need you aren't far off from looking at 2x D-1518's. That'd let you really start playing with clustering VM hosts.

I've thought about doing multiple lower power systems (I'm all for clusters, VMotion, etc.), but multiple lower power systems require multiple cases, power supplies, storage, etc. and would end up using more power than one more powerful box. I'm starting to lean more towards the 1541. I may not need/use all of its horsepower today, but I should get 2-4 years out of it and it should hum right along even if I throw more workload at it.

Also going to throw a RAID card in it and then use my existing Synology NAS as a backup target.

Along with the regular systems it'll run I'll probably also run a media server and maybe a home security/surveillance server, too. The list keeps growing...at this rate maybe I'll end up with 2x D-1528s.

 

[j_h_o]

If you're concerned about power usage at all, you should also check power supply efficiency. At 35w-70w range, a lot of supplies might not hit their efficiency (80PLUS, etc.) ratings.

 

[IamSpartacus]

I ran Passmark PerformanceTest on my D-1537 and D-1540. I'm using CPU Mark as some gauge of performance.

Xeon D-1540 Benchmarks · Justin Ho

Wow that's a bit of a surprising result. According to cpubenchmark.net the D-1528 (6-core at 1.9Ghz) beats your CPU Mark score for the D-1537. CPU Passmark must crave higher clocks.

 

[j_h_o]

I'm suspicious of my results in general, actually:

• Submitted results may be run in machines that aren't completely idle. Windows Update could be scanning for updates in the background, etc.
• The sample sizes of submitted results are too small.
• Rerunning the benchmark resulted in quite a bit of variance, on the exact same hardware.

 

[IamSpartacus]

I'm suspicious of my results in general, actually:

• Submitted results may be run in machines that aren't completely idle. Windows Update could be scanning for updates in the background, etc.
• The sample sizes of submitted results are too small.
• Rerunning the benchmark resulted in quite a bit of variance, on the exact same hardware.

I see. When I picked up my two D-1537's I imagined the CPU passmark would be somewhere in between the D-1528 and the D-1540 so 9500ish.

 

[j_h_o]

I'm sure Intel would clock the chips to ensure that that is the case.

I guess I'm saying: Avoid comparing the numbers too much. Squint a bit as you look at the results

 

[IamSpartacus]

I'm sure Intel would clock the chips to ensure that that is the case.

I guess I'm saying: Avoid comparing the numbers too much. Squint a bit as you look at the results

Can cpu passmark be run on Linux? If so I'll gladly run some benchmarks when I get home next week.

 

[MikeS]

I've decided on the D-1528. I'll do the build and post some pics when done. All the parts I need are on order. Thanks for all your input and suggestions!