Intel Xeon 6 Series

[bson]

Came across this well-leveled video (meaning it's not simplified to gamer or consumer desktop level, nor does one have to be an engineer active in the field to get it) - https://youtube.com/watch?v=-Pq_nFLL9n0

They use a Linux kernel compile time benchmark, which I personally find excellent.  It would help if they provided info on exactly what config (drivers, etc) they build to, but maybe that's available somewhere.  It makes me curious how it relates to my 14900K desktop.

From the spec (clocks, etc) it looks like the 6 Series possibly has simplified P cores intended to run mostly in some sort of E mode.  This would make a lot of sense as the power and footprint budgets are favorable (more perf/W, more perf/mm²).

500W sounds like a lot, but the multi-die surface is huge so can probably comfortably be air cooled (kind of a necessity in a 4U rack unit).

[Someone]

Intel branding/prodcut codes is a complete mess these days, in your example from intel:

The Intel Xeon 6 processor family encompasses four series of processors
Intel Xeon 6900-series processors Maximum performance ideal for the most demanding cloud, AI, and HPC environments
Intel Xeon 6700-series processors Enhanced performance ideal for a wide array of data center and telco environments
Intel Xeon 6500-series processors Essential performance ideal for mainstream server and edge environments
Intel Xeon 6300-series processors Entry-level performance ideal for small/medium business environments

So 4? 5? (more?) different architectures?

Or are you wanting to talk about the parts with increased core parallelisation with simpler cores?
https://www.eevblog.com/forum/general-computing/hyperthreading-or-not/msg5649297/#msg5649297
Which are the 6700-series (for now)

[bson]

Who cares about their marketing.

[Someone]

Who cares about their marketing.

You said Intel Xeon 6 series. I point out that covers many different architectures and chips. Do you mean one of those architectures or all of them?

[bson]

I mean the ones mentioned in the video:  6960P, 6952P, 6972P, 6980P.

[brucehoult]

 $17800 For 6980P (128 cores).

> They use a Linux kernel compile time benchmark, which I personally find excellent.

~130 Kernel 4.4.2 compiles/hour on a dual 6980P machine (so $36k just for the CPUs).

What a strange benchmark. And that's a kernel from February 2016!

My i7-13900HX laptop, which I paid about US$1750 for, takes 1m15s to do a defconfig x86_64 build of the current HEAD as of today.

So that would be 48 compiles/hour if I did them strictly sequentially, but there are periods of 7-8 seconds at the start and end of the build when there is only 1 or 2 cores active, so if you overlapped the builds it might be closer to 60 compiles/hour.

I checked out v4.4 but using current gcc etc the build errors out almost instantly. So I don't know what parameters they are using for their Linux kernel build benchmark.

Just over twice the performance from $36,000 worth of CPU (plus RAM, disks, PS, case etc) vs a laptop that cost 1/20th as much doesn't seem like good value, so I'm not sure what I'm missing here?