I know I called this thread "hype train", but that was meant to be tongue-in-cheek, and I was honestly hoping that we'd get more solid information along the way, and less constantly-shifting rumors and leaks. I can't plan upgrades based on things like rumors of theoretical peak boost clocks.

The lead-up to the 2X00 series CPUs felt a lot less wishy-washy. Oh well.

Mobo manufacturers have announced that they'll be debuting Ryzen 3X00 compatible boards at Computex. I'd be really happy if this means that AMD will come out and do a real 3X00 announcement, with individual CPU SKUs and specs.

"jim : https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=5010

AMD has made significant changes to their CPU architecture which help deliver twice the throughput of their first generation Zen architecture. The major points include an entirely redesigned execution pipeline, major floating point advances which doubled the floating point registers to 256-bit and double bandwidth for load/store units.

https://wccftech.com/amd-ryzen-3000-cpus-7nm-...epyc-rome-q3-2019-launch/

Is this of much use to the LHC projects? I have not seen much comparison between Intel and AMD here."

I'm waiting on the 16 core, 32-thread variant's specs to come out. Hopefully announced in June.

https://www.techspot.com/news/80453-amd-intro...re-ryzen-9-3950x-cpu.html

It took me until Thursday to get my hands on one, but here (better late than never) is my overview of the 3900X, as it compares to the previous generations of Zen CPUs, and with an emphasis on underclocking and undervolting it -- because it's too hot to run in a stock configuration in my apartment! (TL;DR: it underclocks/volts amazingly)

https://firepear.net/grid/ryzen3900/

Since I'm still doing WCG subproject testing, the doc doesn't have any numbers for that in it yet. Here's a peek at what I'm seeing so far:

FAH2: avg runtime 04h 25min 01s, vs 04h 37min 00s on the 2700 (1.05x faster)
MCM1: avg runtime 02h 59min 51s, vs 03h 25min 36s on the 2700 (1.12x faster)
MIP1: avg runtime 01h 03min 34s, vs 02h 07min 37s on the 2700 (2.01x faster)

And here's an overall count of WUs crunched in the past 24 hours for my 1600, 2700, and 3900X:

1600:  102
2700:  116
3900X: 242

Edit: the reason for the enormous differential in WUs/day is that the 3900X's cache (and AVX improvements?) let it plow through MIP1 WUs at double the rate of the other CPUs, so overall throughput is increased.

If I weren't crunching MIP, the improvement wouldn't be so impressive.