Intel Xeon D-1541 or Xeon E5-2630L V3 What wins?

Which is faster, Xeon D-1541 or Xeon E5-2630L V3

  • Xeon D-1541

    Votes: 22 55.0%
  • Xeon E5-2630L V3

    Votes: 18 45.0%

  • Total voters
    40
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
I have a little machine running a single E5-2630L V3. It is a 22nm part, 55W TDP. Overall power consumption is higher than the D-1541 because of the C602 PCH and higher TDP. Base clock is better on the D-1541 as is the architecture (Broadwell v. Haswell.) The E5-2630L V3 does have slightly higher max turbo clocks, more cache and quad memory channels.

Many folks that ask for larger systems I recommend the E5 over the Xeon D line. Usually, this is due to the higher-end expansion capabilities of the E5 V3/V4 platform. I do really like the Xeon D, for smaller systems, but the E5 V3/ V4 UP line gets more attractive as you move up the Broadwell-DE stack to higher priced SKUs.

I had not done the comparison between these two chips before so I wanted to see which folks think would be faster.

Here is the ARK side-by-side.

I am going to tally based on 4 points of comparison:
1. Python Linux Compile Benchmark Speed
2. c-ray 1.1 speed
3. UnixBench Single thread Whetstone/Dhrystone
4. Monero Mining performance (L3 cache/ 2 threads to maximize Xeon D performance.)

Which do you think wins?
 

cheezehead

Active Member
Sep 23, 2012
723
175
43
Midwest, US
Would be interesting to see what the performance per watt and performance per dollar differences are on the setups....might get a different outcome beyond the raw performance numbers.
 

MiniKnight

Well-Known Member
Mar 30, 2012
3,072
973
113
NYC
According to the blogosphere the D-1541 will not just be faster, but it'll move mountains and make lattes too.

For the in-hindsight record, I'm happy you steered me to a E5-2620 V4 instead of Xeon D. I added a 25Gb SFP28 card yesterday and that wouldn't have fit with the HBA and NVMe drives in Xeon D.

@cheezehead it gets more complex. E5 also gets 2-3x the number of RAM slots and much higher RAM capacity. Also what's the board costs and are you using used NICs and what speeds? The V4 version in this segment Amazon.com: Intel Computer CPU 2.1 8 BX80660E52620V4: Computers & Accessories is only $420 so even with a mobo initial price is close but better on the E5. If you've got more than 6 drives, the E5 doesn't need a HBA to connect 10 SATA 3 drives.

I voted Xeon D but I want to see the E5 V3 win.
 

Netwerkz101

Active Member
Dec 27, 2015
308
90
28
D1541 wins, i think, based on ARK stats alone.
Will only 2 memory channels be used on the E5 system?

I learned a while back not to even bother with "L" chips on newer Xeons.
 

PigLover

Moderator
Jan 26, 2011
3,184
1,545
113
Its really workload dependent - but my money says the 2630L v3 wins almost every time. Even for single-threaded workloads the "big core" gets more ops/clock through the scheduler, cache is bigger and faster and memory bandwidth is higher. With TurboBoost 2.0 it should be able to kick in turbo on all cores when needed.

Frankly I don't think this will be much of a contest at all. Maybe a few isolated workload examples to the Xeon D, but the race overall goes to the 2630L
 

cheezehead

Active Member
Sep 23, 2012
723
175
43
Midwest, US
it gets more complex. E5 also gets 2-3x the number of RAM slots and much higher RAM capacity. Also what's the board costs and are you using used NICs and what speeds? The V4 version in this segment Amazon.com: Intel Computer CPU 2.1 8 BX80660E52620V4: Computers & Accessories is only $420 so even with a mobo initial price is close but better on the E5. If you've got more than 6 drives, the E5 doesn't need a HBA to connect 10 SATA 3 drives.
He's looking at the V3 version....given the passmark's I'm seeing, it will likely be a push (+- 5 points) with the D performing better at heavily threaded workloads while the E5 is better for single threaded. This assumes none of the workloads take advantage of TSX....if they do the D will clobber the E5.

D-1541 advantages
- lower power consumption
- 10GB and M-ITX configurations available

E5 advantages
- more expansion
- been around longer, available on secondary market for less.
- socket-able CPU
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Am using the same 4x 32GB RDIMMs but in quad channel mode on the E5.
 
Last edited:

Jeggs101

Well-Known Member
Dec 29, 2010
1,529
241
63
Broadwell is 7-10% faster right? @cheezehead TSX support is for certain DB apps. None listed in OP. I'm going with the D-1541 blogger's delight.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
I will give this a little longer. Just wrapped up the test runs.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
@PigLover will post results this afternoon. After compiling data I did another set of 5 runs on each.
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Ok 60% said the D-1541 would be faster. Here are the results:
upload_2017-1-11_15-16-55.png

Everything was run 10x so we did let the processors cool between runs.

That was certainly an interesting result and a bit of a mixed back. On our larger script's results I think the D-1541 does eeek our wins in other areas, but some of the E5-2630L V3 wins are impressive.

Certainly not the most exciting exercise, but it does show why I push folks (including myself) building larger than 2U single node systems to the E5 V3/ V4 lines.
 

Evan

Well-Known Member
Jan 6, 2016
3,346
598
113
Would have been interesting to see the power consumption as well, I am guessing the e5 used more also... but by how much is the question...
 

Patrick

Administrator
Staff member
Dec 21, 2010
12,511
5,792
113
Would have been interesting to see the power consumption as well, I am guessing the e5 used more also... but by how much is the question...
Yea the spare single socket machine I had for the E5 was not a fair comparison. It had a LSI SAS3008 controller onbaord. I know I could have disabled it but getting publishable power figures would have required changing out components to use the same PSUs, chassis cooling and etc.
 

Evan

Well-Known Member
Jan 6, 2016
3,346
598
113
From a home Lab / small business / dedicated device, point of view I still think the Xeon-D is perfect.

Having said that enterprise for us is 100% HP 2 socket e5 (lots in single socket configuration though) and I don't see that changing at all as it's always easier to just have one common model in use.
 

PigLover

Moderator
Jan 26, 2011
3,184
1,545
113
I feel good that this pretty much matched my prediction (see above post - though I did predict a clean sweep the Broadwell-D wins are close enough that I'll claim my prediction of 'most workloads' is covered).The Linux complie is basically a big memory pump problem so that result is unsurprising. I am a bit surprised about the floating point benchmarks (c-ray and whetstone) tipping to the Broadwell-D. I'd imagine this is due to Broadwell core improvements and would expect a v4 2630 would completely erase this. I am surprised at the impressive win on the Monero Mining run - might have expected that to be a bit closer given the c-ray/whetstone result - but to memory bandwidth difference and large cache on the E5 likely explain this.


Thank for running it. It is at least amusing ;)
 

Jeggs101

Well-Known Member
Dec 29, 2010
1,529
241
63
I'm reading the above as a "get the E5-2620(30) V4" post or am I reading into this too much?
 

PigLover

Moderator
Jan 26, 2011
3,184
1,545
113
I'm reading the above as a "get the E5-2620(30) V4" post or am I reading into this too much?
I think you read to much into it. The E5 will likely cost more to configure, will consume more power and won't be available in small for factor configs (though SM does have a mATX board for it now). Everything is a trade off...