Intel Xeon Bronze 3104 v E5-2603 V3 and V4 Three Generations Compared

[Patrick Kennedy]

We compare the Intel Xeon Bronze 3104 to its two immediate predecessors: the Intel Xeon E5-2603 V4 and V3 generations to see how much performance improves

The post Intel Xeon Bronze 3104 v E5-2603 V3 and V4 Three Generations Compared appeared first on ServeTheHome.

Continue reading...

 

[Evan]

Interesting
V3/v4 grouped in some tests, v4/scalable in others and some you just see an incremental advantage at each generation.

The entire Intel Xeon Bronze 3104 system used a maximum of 85W at the wall running a heavy AVX-512 workload. Given the SSDs, onboard SAS3 controller, PCH, 10Gbase-T PHY, fans, power supply loss, and six DIMMs all in the platform, we can conclude that the CPU is using significantly less than that. We would also state that this is at least 10% less than what we would expect to see on a similarly configured Intel Xeon E5-2603 V4 system.

When you say similar system 10% less is this assuming addon cards or assuming you could find a board with both SAS and 10G on board for v3/v4 CPU's ?
What did you test platforms idle at even given different config ? Just curious.

 

[Patrick]

When you say similar system 10% less is this assuming addon cards or assuming you could find a board with both SAS and 10G on board for v3/v4 CPU's ?
What did you test platforms idle at even given different config ? Just curious.

Comparing onboard SAS3 + AIC X550 10Gbase-T.

On the power, assume you can run sub 60W idle.

 

[Evan]

@Patrick so do you see scalable as more efficient than v4 in general even given the new big TDP PCH chipsets ?
I could have only imagined them going backwards with the chipset and still 14nm processor not really much newer, I Guess maybe the new small cache etc we see reduced power consumption?

 

[Patrick]

@Patrick so do you see scalable as more efficient than v4 in general even given the new big TDP PCH chipsets ?
I could have only imagined them going backwards with the chipset and still 14nm processor not really much newer, I Guess maybe the new small cache etc we see reduced power consumption?

The test platform C622 PCH has more SATA 3, a much faster backhaul (PCIe 3.0 not 2.0), 10Gbase-T and etc compared to older generations.

Cache is not that much smaller really. You went from 256K of L2 + 2.5MB L3 to 1MB + 1.375MB. So you are not losing that much cache. Plus, you have inclusive v. exclusive cache so it is not apples to apples.

Made a diagram here: Intel Xeon Scalable Processor Family Microarchitecture Overview on why Intel did it:

That huge area between 0.25MB and 1MB is why Skylake is intriguing.

I am not sure why people are under the impression that Skylake-SP uses so much more power than E5 generations?

 

[Evan]

@Patrick The impression comes from the PCH TDP. At a glance you could assume idle up 10w before you add on all the 10g etc.
The cpu it's self would assume similar to v4 mostly.

Now we all know TDP and other metrics don't absolutely translate to power consumption.

I am most curious because the figures I have seen for scalable under load look better than expected but idle is interesting. Even in enterprise DC all the Intel servers a lot spend a lot of time idling.... looking at cpu stats on servers is borderline depressing to see such Low utilisation rates.

Of course will run my own tests on enterprise kit soon but that certainly does not compare the the supermicro type systems as our networking is all AIC for 10G+ speeds. (Ie not using Intel onboard NIC)