Return to Level1Techs.com

The RYZEN 1000 Thread! Summit Ridge - General Discussion

amd
ryzen

#1168

Simply Googling “VDDP ryzen” already shows your answer as first answer

So, 1,5 dram, 1,2 SoC, 1.0 VDDP

:sparkler::fireworks::sparkler::fireworks:
explosions
:fireworks::sparkler::fireworks::sparkler:

I know that one tick up from 1.5 ram doesnt go well, and havent had reason to rise over 1.15 SoC


#1169

I would take SoC voltage as low as possible.
It’s power draw goes mental with higher voltages and I suspect it really can’t take it long term.


#1170

To be seen where I organize that VDDP, but as I noted you sometime ago, that 1.143 SoC did correct odd tCWL behavior, and lower stable ram voltage like tick or two

Kinda have to do it

  1. VDDP
  2. ram
  3. SoC

#1171

https://community.amd.com/community/gaming/blog/2017/05/25/community-update-4-lets-talk-dram

TLDR: It’s very finicky.
VDDP is basically the Ram Mood adjustment knob.


#1172

#grumpyram is like

Screenshot_20180222_120823


#1173

Shit senses tell me that the mood switch will stay identical so I could aswell find lowest working one for these regular 3200 sticks


#1174

Very interesting, once again.
AMD might split up cores and controllers into separate dies with ZEN2.


#1175
Paramters Input
Die Width [w] (mm) 9.68
Die Height [h] (mm) 22.06
Horizontal Scribe Lane [sh] (mm) 0.2
Vertical Scribe Lane [sv] (mm) 0.2
Wafer Diameter (mm) 300
Edge Loss (mm) 5.00
Defect Density (#/sq.cm) [Worst Case] 0.5
Manual wafer placement [horizontal shift] 0
Manual wafer placement [vertical shift] 5
Die Centering / Wafer Centering check

That vertical shift can net on average 3 more working dies for Ryzen’s rectangular shape :wink:


#1176

Cant find cldo_vddp from x370 taichi, just regular vddp, which I bravely degraded from 0.9 down to 0.8 and it seems to work, should probably run that prime95

Then there are some weird apu and soc vddp’s over the messy hex side panel, to which options are auto-down-up :smiley:


#1177

Stop raining on my parade. :cry:

Yeah, I figured it wasn’t possible, but it’s fun to speculate about. Thanks for the diagrams and explanation. Really cleared it up.


#1178

Why not one of these:
https://www.titancomputers.com/AMD-EPYC-Based-Workstation-Computers-s/1021.htm
http://www.velocitymicro.com/promagix-hd360a-epyc-workstation-pc.php

They only start at what, US 4K? :sweat_smile:


#1179

Them regular 3200 sticks

soc 0.95
dram 1.345
vddp 0.81

Vddp 0.8 froze warframe, and that 0.81 succesfully does the things


#1180

Ok I was wrong. I just got to play with a Ryzen 2400G.
And I previously stated that the GPU would share the CPU Vcore Bus power rail. Except…

Oh god WTF AMD.

It seems the Ryzen APU’s are actually powering the Vega GPU core from the rather seriously limited SoC voltage rail because the GPU core needs lower voltages than the CPU core. ~1V GPU vs 1.4V CPU.

So they chose to use the lower SoC voltage supply.
This is of course a compromise due to technical limitations/necessity.

But this means that it can lead to huge variation in performance based on mainboard SoC voltage supply and on cheap dirty mainboards will put a lot of strain on the SoC VRM.

It also means that the Ryzen APU’s may run at higher SoC voltages and thus hotter than necessary.
But as it appears, this isn’t something you can easily work around without redesigning the mainboards specifically for Ryzen APU’s.

Now it’s still well within the limits of any modern VRM’s. However any overclock or long term wear can have a much more significant impact on system stability/ performance and operational lifetime.

Average 2 phase SoC VRM (around 25A per phase) assuming 1.2V = 60Watt
Vega APU Stock = 30-35Watt.
Vega APU MAXED to 1600Mhz = ~55-60Watt + CPU SoC power draw and you notice that it’s running over the budget.

Which means stick proper heatsinks on all the low end boards or go nuts on the SoC phases.
And that’s what full platform compatibility gets you :smiley:


#1181

Well this is interesting…

I wonder if @MisteryAngel wants to go over best SoC power configs now? :crazy_face:

/s?


#1182

After searching things about vddp and that cldo_vddp,
it seems like some take that SoC voltage down with these “tbh not sure which one”, as you suggested, but its still unclear what have others done

So, what I am trying to say is that if that ram is doing that 3200 CL14 as easily as mine seems to be doing,
you could still get everything to around 1.0v’s, right?

Think mine defaulted to 1.1v SoC 0.9v vddp 1.35v dram
Which I pulled down to 0.95v SoC 0.810v vddp 1.345 dram


#1183

Actually after thinking about it for a bit, is this another piece of the puzzle for why overclocking the CPU and GPU on RR see limited performance increases after a point?

I’ve been watching TechEpiphany’s videos on the APU’s and he seems to have given up on a CPU OC of any note (all running at 3.6GHz lately) but the GPU is running up to 1675MHz with very nice results, suggesting that the CPU is not limited in terms of frame delivery but opening up the GPU with more bandwidth (high RAM speeds) helps a lot. Maybe it’s just getting all the watts that were otherwise stabilizing the SoC?

I don’t know enough to know whether this makes sense or not.


#1184

Yeah wenn you first mentioned that in the other topic,
I asked you about some sources on this.
Because it didnt really sounded kinda right to me.
Unless AMD used some kind of a Fivr ish concept for those apu’s that i had not heard about yet.
And i didnt had the time to lookup any additional information on it yet.


#1185

I could do that eventually wenn i have some time.
i kinda doubt that these apu’s are going to become very popular doe.
But i will see if i can make a list of “budget” boards with reasonable soc vrm’s.

But yeah the Asus Crosshair 6 Extreme has the best SOC vrm lol. :crazy_face:


#1186

Mostly I’m curious about the mITX boards. I ordered the Gigabyte AB350N and it seems they run fine from reviews like MindBlank:


#1187

Basiclly they should run fine on pretty much any am4 board.
I dont really see a huge problem yet.
But of course cheap boards with lower quality mosfets will run a bit hotter.