• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Ryzen: Strictly technical

Page 50 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

SpecChum

Member
Aug 16, 2007
31
8
81
I can't get my 1700 to even boot with 3200 RAM speed set.

I just get the 5 "F9" power cycles then it reboots with 2133 default settings.

I've got the G.Skill 3200C14 everyone else seems to have and I too think it's a CPU limitation; you're overclocking the fabric so that's probably as much "silicon lottery" as the external speed.
 

looncraz

Senior member
Sep 12, 2011
716
1,637
136
I can't get my 1700 to even boot with 3200 RAM speed set.

I just get the 5 "F9" power cycles then it reboots with 2133 default settings.

I've got the G.Skill 3200C14 everyone else seems to have and I too think it's a CPU limitation; you're overclocking the fabric so that's probably as much "silicon lottery" as the external speed.
Fabric speeds may well play a role in this. I think AMD expected this fabric to run much faster (@The Stilt said something about a 2x multiplier, IIRC). If it's based on HT 3.1 (probably is in some manner, Wikipedia seems to think so, anyway) then it was likely supposed to run at > 3GHz.

I can't help but to imagine a 3.2GHz data fabric with DDR4-3200 CL15 1/2T with Zen ver2 being a formidable system... and possibly doable. Even just a 5% IPC bump and 5% OC bump along with that would make an enticing platform, and I doubt we'll see anything that low given all the low hanging fruit we've already found in Ryzen.

With the improved software ecosystem, dedicated Ryzen memory ecosystem, and a mature platform... AMD has a bright future.
 

SpecChum

Member
Aug 16, 2007
31
8
81
Saying that, there's a user on Overclock.net you says his chip as a "hole" from 3200 to about 3350 I think it was; he can't post at 3200 but can at 3400.

Just have to wait and see what the new AGESA enabled BIOSs bring I guess.

(off topic, but how the heck do I remove/change that signature, it's....kinda old haha)
 

Zucker2k

Golden Member
Feb 15, 2006
1,160
521
136
With the ram overclocks, if your bioses allow multiplier/bclk adjustments, you could dial those raise the base frequencies while dialing down the multipliers to see if that makes a difference. It's much easier if you're not pushing two components extremely at the same time. After achieving a higher clock on the ram, you can then start raising the multies till you strike a nice balance between cpu and ram for a higher ipc per clock. This was how it used to be on Intel, back in the day,
 

Blake_86

Junior Member
Mar 13, 2017
21
3
36
In theory, that should be the case. This is probably what they have planned for the May update.

Ryzen has a RAM-learning implementation that seems to cause more trouble than anything (five power cycles, two soft resets, and a post-BIOS soft reset is a pretty insane process to have to go through every time you change a memory timing or frequency that the system finds not to be "stable").

Ryzen has crazy-sensitive error detection - I turned some of that off and hit 4.15GHz before having stability issues at 1.395V (highest CPU-z I did during that run was 4.05Ghz). I ran the DDR4-2400 divider because it was unstable at anything about 4GHz with 2667.

I have a feeling we are being limited by the data fabric more than the cores or process. That'd be a really good sign that clocks will improve with microcode updates and with Zen ver2. We're basically adding more voltage to prevent relay errors.
Interesting, maibe we can see improvements just in aprile when r5 cames out. Maibe too early..
 

Blake_86

Junior Member
Mar 13, 2017
21
3
36
With the ram overclocks, if your bioses allow multiplier/bclk adjustments, you could dial those raise the base frequencies while dialing down the multipliers to see if that makes a difference. It's much easier if you're not pushing two components extremely at the same time. After achieving a higher clock on the ram, you can then start raising the multies till you strike a nice balance between cpu and ram for a higher ipc per clock. This was how it used to be on Intel, back in the day,
Many people say that rysing bclk is a bad idea
 

looncraz

Senior member
Sep 12, 2011
716
1,637
136
Many people say that rysing bclk is a bad idea
Everyone except ASUS, who recommends it. I've only tried a little BCLK overclocking, but I hit 4.15GHz when I did (not stable, mind you, but it took a minute to crash :p). I also disabled data fabric error checking and a few other bits I wouldn't run on any form of production machine (which is what this is - so I'm careful with my 7 connected drives... (excluding optical... that would take it to 9).
 
  • Like
Reactions: Drazick

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Everyone except ASUS, who recommends it. I've only tried a little BCLK overclocking, but I hit 4.15GHz when I did (not stable, mind you, but it took a minute to crash :p). I also disabled data fabric error checking and a few other bits I wouldn't run on any form of production machine (which is what this is - so I'm careful with my 7 connected drives... (excluding optical... that would take it to 9).
You never mention it so just gonna ask, have you raised the SoC voltages?
The way you are talking about the fabric and DRAM clocking issues, makes me wonder if you aren't pushing it without feeding it with some extra V. Note that on the C6H it was critical to lower SoC V to default or lower before flashing.Not sure if the rule still stands but better check.
 
Last edited:
  • Like
Reactions: Dresdenboy

looncraz

Senior member
Sep 12, 2011
716
1,637
136
You never mention it so just gonna ask, have you raised the SoC voltages?
The way you are talking about the fabric and DRAM clocking issues, makes me wonder if you aren't pushing it without feeding it with some extra V. Note that on the C6H it was critical to lower SoC V to default or lower before flashing.Not sure if the rule still stands but better check.
No, I've only played with VCore and VMem. I like to take it slow and ensure I have a stable platform before getting too carried away.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
No, I've only played with VCore and VMem. I like to take it slow and ensure I have a stable platform before getting too carried away.
You can't quite push memory and the DF otherwise.
Just take is slow and ,in theory, up to 1.15-1.2V should be "safe" but don't jump straight to that.
 
  • Like
Reactions: looncraz

looncraz

Senior member
Sep 12, 2011
716
1,637
136
My P2 settings. Pretty sure it was running at 3.9, though. Suppose I should verify that by setting the P2 to match P1.
 
  • Like
Reactions: Drazick

looncraz

Senior member
Sep 12, 2011
716
1,637
136
Good to know. I'm just sensibilized by the Geekbench 3+4 reported clocks for XV and Zen. ;)
I've noticed the CPU creating invisible P-states for if I set the P2 state very low (such as 1Ghz). Not seeing much of any power reduction (maybe 1W) since Ryzen already has absurdly low idle power (the CPU parts, anyway... the SoC power draw is ~15W). One thing is for sure: if I use stock settings, my CPU only ever runs at 3.5GHz max. Never sees dual core turbo or XFR. Also verified that using just two cores (1+1) does not allow me to get an higher overclock. Too bad we can't disable the caches for some playing :p

Data fabric and the caches have their own C-states as well, so they enter some power saving mode when enough cores are inactive.
 
  • Like
Reactions: Drazick

SpecChum

Member
Aug 16, 2007
31
8
81
For us unfortunate ones that cannot hit 3200 even with the fabled G.Skill 3200c14 B-Die, do you think this will be permanent?

I've tried both the straight multiplier to run the RAM at 3200 with 100 blck, and with raising the blck, but every time I end up in an "F9" POST loop which eventually resets the RAM to 2133 and it proceeds to boot fine.

Here's hoping it's a recoverable issue and not the the Data fabric on our chips just can't run at 1600Mhz.
 
Status
Not open for further replies.

ASK THE COMMUNITY