Ryzen - Return of the Jedi

[chew*]

I getting ready for my new water-cooled build.

Does the higher binned chips clock any better? Memory OC etc

chew* can you give me an advice of which chip to buy.

1700 vs 1700X vs 1800X

Thx

1700 for a daily rig.

1800x if benching ln2.

[Opteron96]

Thank you.

[Opteron96]

Thx.

[lowdog]

@ Chew,

You may want to let Asrock know that latest bios based on Agesa 1.0.0.0.6b has vcore issues when using P State overclocking. It will either immediately push vcore .05v + above what the same settings in Agesa 1.0.0.0. 6a bios is delivering or it will insidiously creep up in the background to .05 + over time, also it tends to get "stuck" at this max volt with very little undervolting when the cpu down clocks in idle.

This is all tested on X370 Fat Pro Gaming which as we know is pretty much the same board as the Taichi. Others on the Asrock forums have described similar overvolting/spiking with the Agesa 1.0.0.0.6b coded bios as well.

[chew*]

Due to changes required to fix the multi bug and from what I was told bios is in qualification and since the bug was related to vcore and or multi it makes it pointless to debug or report anything until those changes are implemented and we see how it effects things.

The inability to oc is a priority fix is basically what I am saying. Until fixed and have fix bios i would just be wasting my time and theirs.

[lowdog]

Due to changes required to fix the multi bug and from what I was told bios is in qualification and since the bug was related to vcore and or multi it makes it pointless to debug or report anything until those changes are implemented and we see how it effects things.

The inability to oc is a priority fix is basically what I am saying. Until fixed and have fix bios i would just be wasting my time and theirs.

Sorry, just for clarification; are you saying they Asrock are aware of the problems with voltage over volting and overclocking with last 1.0.0.6b bios or they are unaware of the issue with voltage over volting and overclocking with P-states with last bios?

[chew*]

I'm saying until the changes are implemented for the newer cpu's and the multi bug which appears to be related to some voltage discrepancies screwing with anything or having them make changes until that priority bug is fixed is a waste of time.

cpu's stuck at 1500mhz or 2200mhz = priority 1

Fixing that will change things.

It's like throwing a waterpump on a car with a bad headgasket.....Problem 1 is headgasket fix that first, change waterpump after.....

Even if I wanted to report it I need to show them anyway.

My board with a bugging cpu is in Taiwan. My hands are tied until it returns.

[drmrlordx]

Soooo Pinnacle Ridge (Zen+) in maybe March 2018? Who wants in?

[adamsleath]

depends on the improvements. some tweaks here and there ..slightly higher clocks. can only be good. but i hope they take the opportunity to fix up their design as much as possible. such as:

Improved IMC
Improved scheduling
Improved turbo (Ryzen turbo is quite fussy)
Improved inter-CCX communication
Improved branch prediction
AVX and other instruction set support/optimization
Critical path optimizations
Tweaking for lower power
Tweaking for higher clocks
Decently revised process.

although zen2 is more likely to have more attention....but thats another year away at least....best to make themost of the PRidge revision...

but the amd roadmap says that Pinnacle Ridge is "Summit Ridge architecture" ...so i'm guessing fat chance its anything but a process shrink....but im hopeful that its more than that, with some minor design refinements of the cpu/stepping

[chew*]

I dont think ryzen turbo is fussy. I think it works exactly as intended. A quick bolt on of phase change verifies this.

Since amd has forever made a server chip and downgraded from there for desktop one must look at the environment server boards are specced to run at.

A climate controlled cold room.

A quick trip outside this winter shoving pc in window dialing system in stock should verify what i saw on phase change. Turbo speeds that "stick" while running workloads mainly because the turbo feature is based on temp sensors.

[charged3800z24]

I think Intel still has a more aggressive Turbo. I am also curious to see some core tracking while running benchmarks and compare AMD vs Intel and see how the clocks and thread scheduling compares.