Overclocking: Multiplier VS Bus Speed

[apcs13]

Hi everyone, I am kind of a beginner overclocker, and until recently I had my FX-6350 sitting at 4.4 GHz by just changing the multiplier on the CPU in the BIOS. However, I saw a guide on YouTube that required almost no changing of the multiplier, and mostly changed just the bus speed, and altering the RAM, NB, and HT frequencies as close to default as possible. The author of the guide, I believe it was JayzTwoCents, said that OCing this way VS just changing the multiplier provides better performance in each individual core, and that just adjusting the multiplier only increased overall performance in multithreaded applications and wouldn't help too much in each individual core.

My question is for you guys: Does it really matter wether you overclock by changing the multipler on the CPU or by altering the bus speed? If so, which one provides a better, more powerful, and more stable overclock?

Also, I am currently running the 6350 at a so-far stable 4.6 GHz just by upping the multiplier on my Asus M5A97 R2.0 if you didn't notice my specs in the signature, without upping the voltage at all. However, even though I set it to the default 1.380 volts manually in the BIOS, it seems to automatically bump it up to 1.404 volts when the CPU is under load. Is this normal/okay?

Thanks for answering my questions!

 

[schmuckley]

FX doesn't like over 220..eh 7..-ish
lot of info here: http://www.overclock.net/t/1139726/amd-fx-bulldozer-piledriver-owners-club/0_100
that voltage is fine and you could do more

 

[Beezy]

I personally have found that no overclocks are stable on my M5A97 R.2.0 with the BUS set to anything but auto. 200 works fine but auto goes to 200.63 for me anyways since its the default. I would say with a better board for overclocking like most 990 chipsets you would find success. but for me with my experience with the m5a97 r.2.0 that would be too many factors to tweak to stability, im sure it can be done but it will take a long time to find stability once you mess with the n/b stuff. kinda picky board in my opinion, perfect for light OC's at a good price

 

[apcs13]

I personally have found that no overclocks are stable on my M5A97 R.2.0 with the BUS set to anything but auto. 200 works fine but auto goes to 200.63 for me anyways since its the default. I would say with a better board for overclocking like most 990 chipsets you would find success. but for me with my experience with the m5a97 r.2.0 that would be too many factors to tweak to stability, im sure it can be done but it will take a long time to find stability once you mess with the n/b stuff. kinda picky board in my opinion, perfect for light OC's at a good price

Eh, I just got it and don't plan on buying another board and reinstalling everything just for that. Besides, I think I have found a stable overclock at 4.6 GHz by just touching the multiplier, not even upping the voltage. I'll probably run P95 for a few more hours or so, 4-6, then do a blend overnight.

 

[Beezy]

at 4.6 Ghz you should be matching a stock 8320. nothing to complain about there

 

[apcs13]

at 4.6 Ghz you should be matching a stock 8320. nothing to complain about there

Damn it, I finished the test and actually had to use my PC and while playing a game i got a BSoD... oh well I backed it off to a safe 4.4 GHz and will run a longer P95 test overnight with 4.6 and up the voltage a little, I was still running 1.380V, which is the same as what is used for 3.9 GHz. I might just leave it at 4.4, it works fine for me.

Getting a BSoD from overclocking doesn't mean you killed a part or anything right?

 

[Beezy]

temps permitting you can go up to 1.55 V to get what you want. LLC on that motherboard sucks so I leave it on auto or disabled. enabled it starts to skyrocket the temps under load. I think LLC is broken or something . you've been through the "bulldozer piledriver overclock guide" on overclock forums?

 

[apcs13]

temps permitting you can go up to 1.55 V to get what you want. LLC on that motherboard sucks so I leave it on auto or disabled. enabled it starts to skyrocket the temps under load. I think LLC is broken or something . you've been through the "bulldozer piledriver overclock guide" on overclock forums?

Yeah I took a look at them and used them as a general guideline. Currently I am sitting at 4.4 GHz with 1.416 volts on the vcore, I will probably (time permitting) up to 4.5 and bump the voltage up the smallest increment possible or even try that with 4.6 GHz.

 

[apcs13]

Just revisiting this old thready again, I decided to overclock again and I have found that I get most stable results with LLC Enabled on CPU and set to Auto on CPU/NB value. The auto on CPU/NB usually sets it to lower or disabled values, which gets me better temps and lower voltages by a small amount, and the enabled on CPU allows me to retain a pretty constant VCore which, with it disabled or even on auto, I suffer core failures in P95 but with it enabled all goes well. Don't see much over volting, however at 4.6 GHz and manual voltage set to 1.380125~ (what it actually sets it to when I try 1.380), during full load scenarios such as P95 tests, the voltage goes up to a perfectly acceptable 1.404V and is constant on that value.

So to anyone trying this and reading this, I recommend from my experiences on this specific configuration CPU LLC Enabled and CPU/NB LLC set to auto. Try that, and play around with that if it idoesnt work.