I just read that all of AMD's new CPUs SKUs will have a CRAZY 64 gen3 pcie lanes! Thats enough to put eight GPU's on one rig and maintain pcie 3.0 x8 performance, as long as the motherboard allows, or a more reasonable four GPU pcie 3.0 x16. This is the first time in a long time i have actually be excited for an AMD product. Ryzen was nice but nothing about it screamed at me like this monster.

bcavnaugh

Chris21010
I just read that all of AMD's new CPUs SKUs will have a CRAZY 64 gen3 pcie lanes! Thats enough to put eight GPU's on one rig and maintain pcie 3.0 x8 performance, as long as the motherboard allows, or a more reasonable four GPU pcie 3.0 x16. This is the first time in a long time i have actually be excited for an AMD product. Ryzen was nice but nothing about it screamed at me like this monster.

One Problem, no one has as MB that will support it.

Yet. I expect Asus, MSI et al to have mobo's available at launch 

I was bummed that they pushed Vega back until the end of July... Who knows what actual availability will be after that! Looks like i'll just go with 1080 ti again and fork over way too much money

bcavnaugh
Well taking a look at the chip first, we have dual channel memory, with support for up to 64GB of DDR4 (supporting 2400 MT/s officially, and up to 3600 MT/s on the board), 24 PCIe 3.0 lanes, 16 dedicated to graphics, 4 for PCIe SSDs, and 4 for chipset communication, and support for up to 4 USB 3.1 devices solely on the chip. 
 
On top of that the X370 chipset, supports an additional 2 USB 3.1 Type C, 6 USB 3.1 Type A, 6 USB 2.0, 4 SATA 6Gb/s ports, 2 SATA Express ports (or another 4 SATA 6GB/s), and an additional 8 PCIe 2.0 lanes as well.

Except that this isn't X370.   AMD's spec's as revealed at Computex last night are: up to 32 CPU threads (16 cores), quad channel memory, 64 PCIe 3.0 lanes.   From AMD: The processors will run on the new X399 platform which rely on a totally new chipset and already have motherboard configurations announced by the likes of ASRock, ASUS, Gigabyte and MSI. 
The CPU will launch at some point in the June, July, August period, with pricing details to be released closer to the launch date.
 
I'm not finding any actual specs on the X399 motherboards as yet, but ASUS, Asrock, and Gigabyte already have X399 boards being shown at Computex with four PCIe 3.0 X16 slots on board.

It's up to 64 pcie lanes. Remember that AMD still lags in single core performance per megahertz. The other factor is the more cores, the lower the clock speed. Still, look at the newer Intel x299 lineup, prices are definitely looking good (still way out of my budget).

This should be interesting but costs of the motherboards and processors still are not known at this time.

Skylake and Kaby Lake had pretty much zero improvement on IPC over Haswell or Broadwell.
They're comparing to Broadwell because there aren't any Skylake or Kaby Lake CPUs that have 8 cores (yet).
 
 Realistically though, the IPC on the CPU doesn't MATTER to folding - folding isn't CPU intensive, just need enough CPU to keep the data flowing.
 
 
 There are quite a few benchmarks that have been published showing the Ryzen matching or beating same-core-count Intels - though usually by a SMALL margin - while running the same core clock or the Intel running a little MORE clock speed.
 Intel does still have a significant advantage on clockspeed though, which is why they're winning most of the Gaming-related benchmarks (but not by a LOT any more).
 

My 3.4Ghz Haswell gets pegged to 100% on each core a folding unit is using. That's on code that is generally optimized for Intel cpus (all compilers target them by default). My system is storing all folding data on an ssd so that shouldn't be the bottleneck. The problem AMD has always had is that they don't or can't conform to the optimizations that Intel put into their chips. They need more fine tuning of code to work optimally with their chips and don't have good automatic compiler optimizations out yet. It's mostly hand tuning right now. Also for some reason their interconnect between core complexes runs at half memory speed so it really matters what speed you run your ddr4 at. The latency between complexes is murder on games and other multiprocess workloads that need to sync up regularly.
For games there is a compelling case that consistent framerates are better on Ryzen despite this, although the epeen numbers are lower (avg and max fps). Even multithreaded workloads only beat Intel due to the increased number of threads.
In practice servers are awesome at using individual threads that don't need to sync up all the time. Compiling and other synced threaded workloads will still favor Intel. That's just the way it is.
So for some BOINC workloads Ryzen will excell at (up to 4 threads in the same complex) and others will be behind Intel's well optimized hyperthreading and sync features. Intel has had so much more time to optimize their systems while so far ahead of AMD in the high end cpu market. It does keep getting better as AMD works on their bios and microcode, but the slight deficit to Broadwell and above will probably not be dealt with until Ryzen 2 (the awakening).