RIVF 3820 OC template help!

Summary
--------
vCore 1.365v
VSSCA 1.15v
VTT CPU Voltage 1.15
CPU PLL 1.8v
Strap of 125
bclk of 128
multi (by all cores) of 37.
cpu llc to medium
cpu capability 120%
vccsa lcc to regular
vccsa capability 120%

RAM divider 2048 (rated for upto 2133)

Temps under load hit 68-71 across the cores.

Where would I naturally go next to see if I can eek more out of this processor under my Antec 920?

Should I drop the bclk down to 125 and raise the multiplier to 38? Or try and up the bclk further, 129, 130 etc.
Guides suggest 130 is prob the max on strap 125...
 
Last edited:
I would go up BLK with PLL at 1.65v.

Hmm...my CPU PLL is 1.8v at the moment (forgot to list it so added it in), Does decreasing that help with increasing BCLK?

The guide over at rog suggests a +- 7 on the BCLK so I'm at +3 at the moment (128 raised from 125, using strap 125).

I'm not sure what the fundamental difference is in approaches...other than BCLK affects RAM frequency too.
 
BLK tends to lead to less error correction needed due to modulation in the CPU Phase Lock Loop. Often due to this less CPU voltage is needed for a given clock if you find the sweet spot for your CPU. Around 130 for most!!
 
BLK tends to lead to less error correction needed due to modulation in the CPU Phase Lock Loop. Often due to this less CPU voltage is needed for a given clock if you find the sweet spot for your CPU. Around 130 for most!!

So I tried 129x37 and got a bosd STOP 0x000000101 in prime pretty quick. That's usually an indication of too lower vcore correct?

Currently at vCore is at 1.365 which I thought would be enough. Are you suggesting that if I move my PLL down from 1.8v perhaps as opposed to trying to give it more vCore?
 
Last edited:
I never got anything stable going beyond 128 with upping vcore from 1.365v or lowering CPU PLL from 1.8v

So I gave the multiplier a shot.

Same volts with 125x38 and stable, then upped to 39 with 1.38v at the moment. 1.38v is in the bios, but my LLC is set to high 140% so I think that is upping it to 1.39 when under load (as seen in picture) which is still below the 1.4v so I think that is good.

What is crazy is that since going this route temps seem a little cooler. Although my basement might be cooler tonight...

So running stable @4.875GHz!!!

Highest core 76c lowest core 68c.

That looks pretty good yes?

I'm going to knock my voltage down to 1.375 and see how that fairs.

And also play some games....

Now I've achieved this I wonder whether I really need it to run at this speed! I don't leave the machine on 24/7, only when I game tbh, so maybe I will.

475125x39.png
 
Last edited:
I've settled on 1.39v (chip has broken in now I think), with LLC set to high and 130% its going to 1.408v idle, and 1.4v when under load.

Is that too much do you think? Temps in prime average 73 (after 20mins), when I use my machine for what I actually need it for temps are 61c.

Just wondering if I should knock it down to 4.75 and a tad off the vcore.

I don't use the machine 24/7 just a few hours here and there.
 
I've been playing around and I dropped my CPU PLL down to 1.65 from 1.8v and I can get 4.75 (38x125 on1.25 strap) with 1.36v with LLC still set to high. This pushes the vcore up to 1.368-1.376 under load.

8Pack...can you explain why dropping PLL helps with reducing vcore? You and on another thread someone mentioned this. I've not done much more testing, but I couldn't go this low on vcore before at 4.75
 
Asking a few questions now...

I have set the core LLC to medium, which gives me some vdroop (High always increases vcore under load, so figure that isn't the point of it) I have vtt and vcssa set to 1.15v. CPU PLL set to 1.65.

Running on a RIVF I can get 4.75GHz with 1.37v on the core.

However, to get to 4.875GHz I have to set the vCore to 1.41v. So that is +0.04v extra to get IBT to run 30times on High. Underload it vdroops down to about 1.392-1.400v

Does that sound like I've hit the wall of the processor when it needs so much more vcore?

Temps with an Antec 920 are 78-71-70-74
 
Back
Top Bottom