Load line calibration (LLC) on AMD CPUs?

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
So the only real data I've read on LLC is for Intel CPUs.

Does the same data apply to AMD CPUs?

I'm wondering whether to use it on my FX6100. The settings on the ASUS M5A97 EVO I'm using is so confusing...it's like playing Vcore roulette lol...I never know what actual vcore I'm going to get until I boot Windows and check. I have LLC on currently but am wondering whether to just turn it off?

I have C1E and Cool n Quiet enabled so my idle voltages drop down. I have it OCed to 4.4GHz at 1.45v (loaded). So should I turn LLC off or leave it on?

EDIT:
Well, I tried with LLC off using manual voltages, but then C1E is disabled.

I get about 0.1v of vdroop. I guess I'll just leave LLC on. I'm only running at about 1.45v anyway.
 
Last edited:

Soulkeeper

Diamond Member
Nov 23, 2001
6,712
142
106
It should be the same principle on either platform

My understanding is it's a tradeoff between voltage stability and extra stress/temp on the VRMs
maybe someone else could provide a more detailed explanation
 

hans030390

Diamond Member
Feb 3, 2005
7,326
2
76
I'm also wondering how to optimize my FX6100 OC with my M5A97 (non EVO). I currently have LLC on and all of the power saving features such as C1D and Cool n Quiet. It's not too hard to compensate for the voltage drop, but I'm wondering if it is better to leave it off. And realistically, do the power saving features really impede on overclocking ability? I like the power savings they offer when my computer isn't at load.
 

Soulkeeper

Diamond Member
Nov 23, 2001
6,712
142
106
having it on will likely increase your cpu/mb temperatures a small amount and decrease lifespan.
but it can be used to stabilize an overclock if you don't want to raise the voltage anymore

having it disabled is perfectly fine and should be within intel/amd specs

I suppose you can see it as "something to try" if your system is almost stable and you don't mind extra stress/temp on the mb components
some motherboards from asus have elaborate ways of implementing this stuff ie "digi vrm"
I guess they use a little chip to dynamically adjust things depending on load

To be honest i've never seen any noticeable benefit from enabled llc on a motherboard.
 
Last edited:

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
The basics of it is that it adds more voltage at load so you have less vdroop.

It doesn't actually prevent vdroop, it simply increases voltage applied to raise voltage under load.
 
Dec 30, 2004
12,554
2
76
I disagree that it will decrease the lifespan. There is a long thread way back where we all discussed it. The damage comes from high voltage transients (at the moment a workload is stopped/begun they occur) wearing stuff out, but they are so short that I don't believe they cause problems. Besides, there's a ton of capacitors on the motherboard, and right under the CPU itself to keep the voltage supply stable.
 

Soulkeeper

Diamond Member
Nov 23, 2001
6,712
142
106
I disagree that it will decrease the lifespan. There is a long thread way back where we all discussed it. The damage comes from high voltage transients (at the moment a workload is stopped/begun they occur) wearing stuff out, but they are so short that I don't believe they cause problems. Besides, there's a ton of capacitors on the motherboard, and right under the CPU itself to keep the voltage supply stable.

I was going to add "Negligible", but my spelling wasn't up to par :)
but regardless the mosfets are transistors so the same physics should apply as a cpu's transistors.
not to mention temperature increase to surrounding components

I doubt anyone on these forums cares if their computer lasts 10yrs instead of 11yrs, so point taken
 
Last edited:

frostedflakes

Diamond Member
Mar 1, 2005
7,925
1
0
Here's an explanation of it for Intel CPUs, but same basic idea for AMD.

http://www.masterslair.com/vdroop-and-load-line-calibration-is-vdroop-really-bad/

Like soccerballtux explained Vdroop is supposed to prevent voltage from overshooting what you set it to during changes in load. May make sense when running at stock, but when you're overclocking and pushing your CPU to the limit, it seems to do more harm than good, so motherboard manufacturers added load-line calibration to allow OCers to adjust Vdroop.

Just be glad your motherboard even supports LLC, my rev 1.0 GA-990FXA-UD5 doesn't. Have voltage set to 1.5625V in the BIOS, droops to 1.36V while running Linpack. Really makes OCing more difficult having to deal with such a large Vdroop, and I'm afraid to go much higher because at idle/low load when the voltage doesn't droop as much processor voltage gets up to like 1.49-1.51V.
 

BD231

Lifer
Feb 26, 2001
10,568
138
106
I've never had anything close to that kind of v-droop frosted, what are you using to monitor your voltage?
 

frostedflakes

Diamond Member
Mar 1, 2005
7,925
1
0
HWMonitor, CPU-Z, other programs.

If you read around online the UD3 doesn't have near as bad of Vdroop as the UD5 and UD7.
 

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
^Yeah that's some horrendous vdroop. I've never had anything that bad with any of the last 2 Gigabyte mobos I've used.
 

frostedflakes

Diamond Member
Mar 1, 2005
7,925
1
0
Well if your previous Gigabyte boards were AMD, keep in mind that AM3+ is the first AMD socket that has Vdroop built into the design. I never had any Vdroop on my AM2 Gigabyte 785G board, for example, because AMD didn't design voltage regulators for AM2 CPUs to do that. On these AM3+ boards the Vdroop is normal and is exactly what AMD intended when they designed the socket and voltage specs, I'm not blaming Gigabyte for that, they just followed AMD's design guidelines. But with a ~$200 enthusiast motherboard, you would kind of expect them to implement features like load-line calibration that allow overclockers and power users to tweak and adjust the Vdroop. ASUS AM3+ boards, for example, have supported LLC since they were released and you don't hear Sabertooth or CHV users complaining about the Vdroop problems 990FXA-UD5 and UD7 users have.
 

thilanliyan

Lifer
Jun 21, 2005
11,848
2,051
126
Is it supposed to droop THAT much though?

I can see a minimal amount being built in but that is a LOT.
 

frostedflakes

Diamond Member
Mar 1, 2005
7,925
1
0
My understanding is that the droop is proportional to how much current the CPU is drawing. Higher power means more voltage droop. You can notice this just with how much you load the CPU, for example Linpack puts a pretty big load on the CPU and causes voltage to droop as low as 1.36V according to HWMonitor, but with something not quite as stressful like Fritz Chess Benchmark the lowest droop I see is 1.39V. The system consumes about 40W less while running Fritz than it does with Linpack.

My X4 955 @ 4GHz, for example, had some droop on this board (probably about 0.04-0.05V), but not nearly as much as the FX-8120 at 4.4GHz, which is a pretty power hungry chip. The 8120 also doesn't have a ton of Vdroop at stock either, it's only once you start overclocking and overvolting especially that it gets crazy.