• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD FX Overclocking

I find for 100% stability when running stress / torture tests with Cool & Quiet and C1e turned on I need fractionally more V-core than with them turned off.

Doesn't seem to matter on Realbench but Prime requires extra voltage with them enabled.

Rich
 
From my own experience, disabling these has given me a slight gain in benches, not by a lot though so could be margin of error. For me, I'm not bothered about saving a few pennies on power and aim for max performance, power saving comes later if I can be bothered (Like I said, I'm lazy). disabling them also eliminate them as possible causes of instability (no matter how small a chance they actually will). I will however be trying these settings again now you have said they shouldnt make a difference. I want to see just how much difference they make again, if any. Hopefully I can report back my findings.

As for the offset, I've never tried as setting the volt manually is so easy and predictable once LLC is set correctly, and I should imagine its much easier to explain to newcomers and for them to understand too. Most have never progressed further, myself included (Probably because the system is up, running and stable using manual)
Your guide Davedree will help as you seem to have the skill and knowledge to help people like me to use these underused settings. ;)

I appreciate everyone has their own experience and advice to offer in these threads. I have 15 years experience with Amd, starting off with a K6 II 500 and a Via Kt400a. I've tried and messed around with every Amd except the original phenom and Kaveri. So anyone who can provide benches or goes out their way to testing settings and providing before and afters gains both experience and expertise, has the right to state their opinion.
Please provide a variety of benchmarks with those settings on/off it would be great to see, also an idle power consumption in comparison to the performance measurement would be great too. Unfortunately I sold off my fx8320 and I only have my old phenom II to play with now, but I did carry out a big test on Vishera vs Deneb clock for clock in benching and gaming.

a lot of my experience enabled me to push an fx8320 very hard using a 4+1 phase motherboard, when most people said it couldn't be done, or still believe it'd only be a matter of time before the board failed, other guys on overclock net also use this board on overclocked fx6's fx8's.

I'd happily write a proper guide up but Is it now a little too late to teach people to suck eggs lol? I'll explain as brief as i can and I'll write a guide up to follow.

Whilst overclocking with cpb turned off, explaining offset voltage is a little technical for the first timer but actually isn't hugely different to manual overclocking once you know two things, 1. Your cpu Vid, 2. The v-core bump up from your LLC setting.
In the gigabyte bios for example the V-core set would look like this.



The V-core is set to Normal, and the Dynamic Dvid will add or subtract from the v-core value you see above it. If you had LLC disabled then to get 1.5v you would enter 0.0125v in the Dynamic setting. But then fully loaded you would experience Vdroop. So depending on your LLC bump Level you will have to recalculate your Dynamic V-core, as extreme could make it jump way above 1.5v. That's why it's easier to use lower clock speeds just to see how much LLC bumps up the voltage when using hwmonitor. Then once you know how much your LLC setting always bumps it up by, you then lower your dynamic by that amount and recheck for vdroop this time with the cpu at max frequency. If LLC extreme always added .040 V then you would actually set dynamic dvid as -0.0275v.
But even then some fine tuning has to be carried out when finally setting up the Dynamic Voltage.


On the older Bios it's much easier to work out dynamic it looks like this.
Vid is 1.2625 + and offset of .100v = 1.3625v




Here is an example of a Gigabyte, the bumpup of +40v from LLc EXTREME which was used for my phenom II. The Bios set was the Vid+ the offset.
Note I can't remember I could be wrong but I think cool and quiet voltage for an Fx doesn't scale upwards it remains at a fixed lower voltage.


 
I appreciate everyone has their own experience and advice to offer....................*snip*

Great post!

Note I can't remember I could be wrong but I think cool and quiet voltage for an Fx doesn't scale upwards it remains at a fixed lower voltage.

When Cool and Quiet is enabled on my FX8350 the board only drops the voltage to whatever I've set as V-Core, under load it climbs dependant on LLC settings.

Rich
 
Last edited:
Cool and quiet doesnt affect stability on k10 or fx, it just enables the cpu to idle at a minimum p-state. You are aware the fx processors have 6 p-states with predetetmined voltages and multipliers.
The rumour of instability dates back to socket.939, 754 when amd cpu's only had 2 p-states and instability could occur,

Cool n' Quiet should only be used once you've found a stable overclock.

As I mentioned earlier benny. Regardless of it not effecting instability I have found it to be best to use this method even if dave thinks its just a placebo. ;)
 
Thont, Thont, Thont. I think this is the second time I've suggested you get a pair of glasses, or maybe get your cataracts removed.
..
\snip - just other guff.
Oh wait a minute. Maybe you consider Cinebench a game? That would explain it. Have plan playing that dude.

Please just get the hell out of this thread already, you've dragged it off-topic too much already. You expressed your opinion and that's fine, however this is the "AMD FX Overclocking" thread. Coming in and basically telling everyone that's bought an FX processor that they're a fool is not a good idea.

Dave is still using his 20/20 to only read the parts that are relevant to his opinion. He doesn't contribute much else, but hey ho! :)
 
I appreciate everyone has their own experience and advice to offer in these threads. I have 15 years experience with Amd, starting off with a K6 II 500 and a Via Kt400a. I've tried and messed around with every Amd except the original phenom and Kaveri. So anyone who can provide benches or goes out their way to testing settings and providing before and afters gains both experience and expertise, has the right to state their opinion.
Please provide a variety of benchmarks with those settings on/off it would be great to see, also an idle power consumption in comparison to the performance measurement would be great too. Unfortunately I sold off my fx8320 and I only have my old phenom II to play with now, but I did carry out a big test on Vishera vs Deneb clock for clock in benching and gaming.

a lot of my experience enabled me to push an fx8320 very hard using a 4+1 phase motherboard, when most people said it couldn't be done, or still believe it'd only be a matter of time before the board failed, other guys on overclock net also use this board on overclocked fx6's fx8's.

I'd happily write a proper guide up but Is it now a little too late to teach people to suck eggs lol? I'll explain as brief as i can and I'll write a guide up to follow.

Please write a guide Dave. At least your a positive Dave! :D

Off topic nit picking posts are just not needed.
 
I've currently got my FX-8320 at 4.51 (20.5x220), though I currently have all the power saving features disabled. Will I have to back off on the in a bit to enable cool 'n quiet etc? Would be great to be able to still utilise the power saving features whilst still getting the speed bump when needed.
 
Are you manually setting the voltage?

I am yes - on a Sabretooth 990FX rev2.0.

Would I be better doing it offset?
Its currently at 4.7 at 1.39, jumps to 1.41 under full load.
Power saving stuff is enabled.

I'm pretty out of practice with overclocking, just trying to get back into it.

Rich
 
Ok well an update on testing CnQ.
Enabling CnQ and C1E (individually or both)...No noticeable difference in any benchmark I tested or during general use. No stability issues seen and no bump in voltage required for stable testing. However, it does not seem to have affected anything else in a positive way either. Everything at idle still appears exactly as before (temp, volt etc) I'm not sure its even functioning TBH (Maybe I have a fault?). Anyway, I'll be leaving enable from now on as there little reason not to.

Have now moved from manual volt to offset (set to same level) as well and so far have had some good promising results and even slight less fluctuation with regards to vdroop and llc kicking in. Temp improvement on VRMs and idle temp as well. Looks good so far.

Davedree, you mentioned Gigabyte having offsets as a early feature. This made me realise on my older 970a-ud3 (no LLC but still able to holding a 1055T @ 4ghz) only has offsets as an option. Manual simply isn't an option at all. I was only using that system a few month ago too. Such a bad memory I have. Anyway, this just confirms to me that offsets are the way to go. Thanks for the guide (although I shouldn't have needed considering)
 
I am yes - on a Sabretooth 990FX rev2.0.

Would I be better doing it offset?
Its currently at 4.7 at 1.39, jumps to 1.41 under full load.
Power saving stuff is enabled.

I'm pretty out of practice with overclocking, just trying to get back into it.

Rich

If you manually set your voltage then even with the c1e and cool and quiet on when the cpu idles to the lowest multiplier it won't drop the voltage down to below the voltage you have set manually, Offset however will drop the voltage lower.
Offset in my opinion is the way forward, but more so in the sense that power saving at idle should be quite a lot better, with less heat on the vrms and phases and cpu at idle.


Ok well an update on testing CnQ.
Enabling CnQ and C1E (individually or both)...No noticeable difference in any benchmark I tested or during general use. No stability issues seen and no bump in voltage required for stable testing. However, it does not seem to have affected anything else in a positive way either. Everything at idle still appears exactly as before (temp, volt etc) I'm not sure its even functioning TBH (Maybe I have a fault?). Anyway, I'll be leaving enable from now on as there little reason not to.

Have now moved from manual volt to offset (set to same level) as well and so far have had some good promising results and even slight less fluctuation with regards to vdroop and llc kicking in. Temp improvement on VRMs and idle temp as well. Looks good so far.

Davedree, you mentioned Gigabyte having offsets as a early feature. This made me realise on my older 970a-ud3 (no LLC but still able to holding a 1055T @ 4ghz) only has offsets as an option. Manual simply isn't an option at all. I was only using that system a few month ago too. Such a bad memory I have. Anyway, this just confirms to me that offsets are the way to go. Thanks for the guide (although I shouldn't have needed considering)

Thanks for taking the time to test and write back on your findings, as mentioned above, with cool and quiet on and c1e on manual voltage won't allow the v-core to drop below the voltage you've manually selected in the bios. You didn't have a power meter to hand to compare the settings did you?
I'm interested in you mentioning about how you're seeing more stable vdroop, I'm not sure how that would really change but it's interesting you've noticed that. As i've always overclocked using the offset way for quite some years now.


I've currently got my FX-8320 at 4.51 (20.5x220), though I currently have all the power saving features disabled. Will I have to back off on the in a bit to enable cool 'n quiet etc? Would be great to be able to still utilise the power saving features whilst still getting the speed bump when needed.

Cool and quiet and c1e won't affect you maximum potential overclock, You just need to tune the dynamic-voltage along with the level of llc so as to get the correct voltage when stress testing, whilst enjoying the benefits of sipping power when idle.
 
I'm very tempted to build myself a cheap AMD rig to learn and have fun with going into detail with Overclocking, use it as a testing rig so to speak.

It's brilliant fun especially when you use the ht-ref to overclock your cpu and memory, then muck around with ram timings and Nb-cpu and Ht link in relation to your ram speeds.

Whilst I still have a 4.8ghz 2600k, it's so boring for mucking around with things in bios.
 
If you manually set your voltage then even with the c1e and cool and quiet on when the cpu idles to the lowest multiplier it won't drop the voltage down to below the voltage you have set manually, Offset however will drop the voltage lower.
Offset in my opinion is the way forward, but more so in the sense that power saving at idle should be quite a lot better, with less heat on the vrms and phases and cpu at idle.

Any experience of that on my board? Not too sure of the settings.

If I read your posts right I'd be setting CPU Voltage to 'offset' and applying a positive offset?
Would I apply enough offset to take it to the 1.39V or or the 1.41? Taking into account for the LLC, set at Ultra High at the moment IIRC which is 75% - not sure what its 75% of though haha

Rich
 
Last edited:
It's brilliant fun especially when you use the ht-ref to overclock your cpu and memory, then muck around with ram timings and Nb-cpu and Ht link in relation to your ram speeds.

Whilst I still have a 4.8ghz 2600k, it's so boring for mucking around with things in bios.

Yeah i mean I've overclocked my 4690k and it's great and all but i'm scared to push it as I don't want to ruin it.

But i want something i can put through it's paces and learn what all the settings do for future overclocks and getting the most out of them.

I think I may have to do it in the future.
 
Any experience of that on my board? Not too sure of the settings.

If I read your posts right I'd be setting CPU Voltage to 'offset' and applying a positive offset?
Would I apply enough offset to take it to the 1.39V or or the 1.41?

Rich

Last asus I used was a m5a97 evo r2, I think you change cpu and nb voltage to offset mode, then offset sign to +, to calculate the correct offset you need to know the vid of the cpu if you scroll from offset mode back to auto it should show you it. Once you know the vid add the offset to equal the v-core you used before for your overclock, (totalling 1.39v.
Then you would be best to boot into windows and using hwmonitor stress test checking your llc isn't overvolting from your previous 1.41v.
Fine tuning can be carried out by adding or reducing the offset.
 
Last asus I used was a m5a97 evo r2, I think you change cpu and nb voltage to offset mode, then offset sign to +, to calculate the correct offset you need to know the vid of the cpu if you scroll from offset mode back to auto it should show you it. Once you know the vid add the offset to equal the v-core you used before for your overclock, (totalling 1.39v.
Then you would be best to boot into windows and using hwmonitor stress test checking your llc isn't overvolting from your previous 1.41v.
Fine tuning can be carried out by adding or reducing the offset.

That makes sense cheers, pretty sure the Vid was 1.375 but I'll check.

When you say check that the LLC isn't overvolting do you mean pushing it past the 1.41 mark?

Rich
 
Back
Top Bottom