Is this the wrong CPU

Woomack said:

Looks like it's only catzilla behaviour. HT is generally not helping in anything ( it's also mentioned in about every good FX OC guide ).

Click to expand...

That is what I am thinking...

Johan45 said:

That was a one off trial, I had a goal to beat my 4770k. Didn't try it with anything else to be honest. Maybe I'll revisit it this winter when I have the time and see if there really is something to it or just the benchmark. My gut feeling is it does, how much remains to be seen. I have my HTPC set up for gaming now and once I have a few more titles downloaded I can do a FPS comparison I suppose. It has normal cooling , the 770's and an FX 8320 in it. That Chip is fairly new to me so I don't know it's limits yet but I'm sure I could do something and games would likely have the most real world difference.



Yes but they're not talking about SLI set up either.

Click to expand...

I also wonder if that translates to single card increases as well?

Anytime I've heard of it helping anything was only in respects to dual card setups. I'll find out over the winter and post something up if there really is any gains.

So... I started doing the OC thing on this desktop again from scratch, have memory stable at 2133 and stock clock, everything looks good nothing in the bios is set to auto anymore, it all has some sort of number in it now. I go to bump some thing up FSB from stock 200 to 210 leave the multiplier set at stock 21.0increase voltage to 1.368 increase HT to 2316 and nb to 2316 with my thinking if they are in tune they should operate more efficiently, not sure how much higher I can go with them. Also is there a rule of thumb that you want your nb and ht to be roughly 1/22 of core physical clock or better, say core is 4.8 I want my nb and ht to be at least 2400?

But the problem is that when I start p95 in the blend test I get two failures from cores' 1 & 3 or 2&4, just says error no warning, I did notice that when I increased the fsb it increased my mem speeds, could this lead to a problem causing the errors and two cpus to stop?


Edit: this was only a 10min load so far.

Sure DDR3-2133 running at DDR3-2246 could be a problem. Such clocking of all the other busses when clocking with the FSB/HT Ref Clock is why we generally suggest when starting out to clock with cpu multiplier alone and not mess with the other busses by using the FSB. When you use other than the multiplier alone...it takes some variables out of the mix and makes understanding a problem easier.
RGone...ster.

t1nm4n said:

So... I started doing the OC thing on this desktop again from scratch, have memory stable at 2133 and stock clock, everything looks good nothing in the bios is set to auto anymore, it all has some sort of number in it now. I go to bump some thing up FSB from stock 200 to 210 leave the multiplier set at stock 21.0increase voltage to 1.368 increase HT to 2316 and nb to 2316 with my thinking if they are in tune they should operate more efficiently, not sure how much higher I can go with them. Also is there a rule of thumb that you want your nb and ht to be roughly 1/22 of core physical clock or better, say core is 4.8 I want my nb and ht to be at least 2400?

But the problem is that when I start p95 in the blend test I get two failures from cores' 1 & 3 or 2&4, just says error no warning, I did notice that when I increased the fsb it increased my mem speeds, could this lead to a problem causing the errors and two cpus to stop?


Edit: this was only a 10min load so far.

Click to expand...

Read the bulldozer (and descendants thereof) OC guide on this site. It'll tell you, like any good OC guide, that you never screw with multiple things simultaneously. If you do that, and things don't pan out, how do you know where your problem is?

For example,

If I change vcore, multiplier, ram voltage and ram clocks all in one go, and then I fail P95 blend. What do I change? Exactly. I don't know what to change.

If I had just screwed with one thing, lets say, vcore, I would know that P95 had failed because I didn't have enough vcore. If i had just upped the multiplier and it failed, again, I'd know I needed more vcore, and so on.

Start simple, and take your time. On AMD you don't go "KAPLOW! OVERCLOCKED!" the way you can with Intel. There are a lot more things to tweak that actually make a difference and it's worth tweaking them, you can only tweak one thing at a time though or you'll get lost in a maze of your own creation.

Hope that made sense and Merry Christmas, or whatever holiday you celebrate. Should have just said Happy Holidays but I'm sort of against that... so if you don't celebrate Christmas please understand my reasons for saying it anyways.

@RGone, I came to that understanding when my mobo would post but I kept having bsod and realized it was a mem problem, set mem to auto config no issues booting now. I'll play with mem later. I was reading in another thread ( I believe you started) about what progs to use for different OC's or if this forum could come up with a standard, and no such conclusion was found, so I'm using a variant of different progs, OCCT and P95 not sure how long I should be running these though. The only thing I gathered was that its really personal opinion on what you find stable and what you are looking for in a OC, it can fail at any given time, so I'll run P95 (blend) and OCCT for an hour each, that should saturate the lil h100i cooler I have to give me an idea of how much heat can be dissipated at the most extreme loads of the CPU. Then I will just use the pc like i normally would to see how it holds up, if I get failures then I will look at lowering clocks or seeing if its software related. With your experience, does this sound like a wise idea or not enough testing to be sure?

Why aren't you willing to just adjust one variable at a time, and thus, know your recourse and remedy right off the bat if something doesn't work?

I think that would be the wisest course of action here. That's what everyone else does and it works.

An hour of P95 is not enough to say you have a solid OC. I would say "blend" for 3 hrs as a minimum.

@theocnood I should of been a lil more clear there, I didn't do that all in one sitting, I went for safe windows load til I was comfortable with the numbers, I just didn't realize how much fsb affected mem until I went to do the p95 blend and it was failing on two different cores at the same time, or giving errors, which is a fail I guess. I am the worst kind of noob, I WANT IT NOW and not later, lol. I'm taking things in small steps, maybe larger than what others would take, I hold noon accountable for my actions other than myself, but you got to make mistakes to learn something new, right.

@theocnood I should of been a lil more clear there, I didn't do that all in one sitting, I went for safe windows load til I was comfortable with the numbers, I just didn't realize how much fsb affected mem until I went to do the p95 blend and it was failing on two different cores at the same time, or giving errors, which is a fail I guess.

Definitely a fail yes.

I am the worst kind of noob, I WANT IT NOW and not later, lol. I'm taking things in small steps, maybe larger than what others would take,

Yes, and you are creating problems for yourself in the process. It's like people who buy motorcycles and want to do wheelies and stoppies that first night and end up dropping the bike and scratching the crap out of it. Not to mention the concussion. Why? Why I ask you?

I hold no one accountable for my actions other than myself, but you got to make mistakes to learn something new, right.

That's a given, but you are making the wrong mistakes, and learning the wrong things. Learn to do it right, or don't do it, would be the best advice I can give you.

@Theocnoob I get what your saying, but people learn at different levels. I don't know if I'm making the wrong mistake, and your not sitting over my shoulder watching what I'm doing, so you can't really say I'm learning the wrong things.. well I guess you can since you did This being my first OC ever really, I don't know where to start, the multiplier idea is good and all but (this is for me) I want to know how everything works in relationship to one another, and for me to learn that I have to tinker.

After reading the BD OC guide for a 5th time I understood more, now that I have made some minor mistakes and more of that makes sense, so the problems I created for myself lent to me learning some good.

I didn't know there was a 'right way' to OC, seeing how the entire purpose is to push your components above what they were designed for, OC'ing in its nature is dangerous, there might be a safer way, but I would say I'm not hurting anything by testing the boundaries here or there.

t1nm4n said:

I didn't know there was a 'right way' to OC, seeing how the entire purpose is to push your components above what they were designed for, OC'ing in its nature is dangerous, there might be a safer way, but I would say I'm not hurting anything by testing the boundaries here or there.

Click to expand...

Overclocking is not, by definition, a dangerous thing to do. If you respect the reasonable "safe" voltage limits of your parts, keep temperatures reasonable, and use the right parts (in this case, the right motherboard), the overclock will not be dangerous at all. Sure, it'll degrade your CPU faster than running it stock will. One day, 4 years from now, your 4.5Ghz OC may suddenly not want to hold stable in Windows, and you may need to bump up Vcore or bump down multiplier. That's how these things work. Take it outside it's envelope, it degrades. I've lost 1X multiplier on my 2600K from 48X to 47X since May 2011. It degraded. Doesn't mean it's going to die any time soon, and I probably have another 4 years before it noticeably degrades further, by which time it will have been sold to somebody else who has use for such old things.

There definitely is a "right way to OC", and that is to isolate each "overclocking factor" be that the vcore, or the multiplier, or the ram clock, or the ram voltage, or load line calibration, or whatever, and tweak it on its own. This means leaving everything else at known safe stable settings while you tweak each new part. Now, some things, like the aforementioned load line calibration, come in series, after other steps. For example, with LLC on an Intel platform, what it would be doing is preventing voltage drop from what I punched into the bios. Without LLC, my 1.35V in bios could become 1.33V in Windows. LLC can even that out. You would apply LLC after setting your voltage and multiplier, and finding that your voltage was dropping. You wouldn't just go into bios with a stock CPU and turn LLC up. THat would be silly. That's why it is important to understand each factor and setting going into this, and know when to implement what and how it will play with everything else. Believe it or not, the best way to do this truly is by messing with 1 thing at a time.

I'll give you an example:

I recently overclocked a Pentium G 3258 for my HTPC. It was a bum chip (I lost the silicon lottery big time) and was able to get a miserable 4.4Ghz at 1.439V. Given that 1.45V is the max any sane person would push through this chip, I was unable to go for 4.5Ghz because I was already at the "safe voltage wall/limit".

What I did was first try a few mild overclocks- 4Ghz, 4.2Ghz, and first try them with auto voltage to ballpark what the CPU was going to need, then fine tune the voltage while leaving everything else alone to get the OC stable. Once I was stable, I upped the multi to 44 and screwed around with the voltage again, figured out what I was going to need to get it stable at 44, added some LLC to equalize what I was putting into the bios with what I was seeing in Windows, and I was done. Just ran P95 blend for 4 hours and it passed so I consider it stable.

That's another thing-- stable for WHAT? If you're going to be doing pinned-to-the-wall work on the CPU 24/7, an overclock that only tested p95 stable for 3 hours may not prove stable for that work. Even if you're going to be doing CPU intensive gaming for long periods of time, that OC may not prove stable. In my case all this computer is doing is streaming video from Netflix, Youtube, etc, downloading torrents, using VLC media player, surfing the web, etc. Simple stupid stuff. Stuff that doesn't push the hardware.

If you want to know what something means like "HT" on your platform, look it up. Watch a video. You don't need to screw with it to know what it is. I understand how planes stay in the air but I've never flown one, and I'd like to think that if I did fly one, I'd crash it... despite thousands of hours in FSX.

Seriously. Do it the way everybody else does it, or you'll never get truly stable, you'll give yourself headaches... you might even brick a bios chip (thankfully your board has 2 and they are replaceable)...

ONE thing at a time. Get it rock solid, move on.


This video may be helpful.

EDIT: Nope he's just screwing around without knowing anything. Try this one.


You should look up (it'll be in forums here) the TJMax and max safe 24/7 vcore for your CPU as well.

Was any of that helpful or relevant? Possibly not, as I had a colonoscopy and was sedated with propofol 8 hours ago and am still flying a little bit... I don't know if you even HAVE LLC on AM3+ as I'm an Intel guy. I only build AMD for "other people" like my mom and I tend not to tinker with them too much as I build low end AMD boxes. I don't mess with the high end AMD stuff. I disagree with the way they've implemented their cpu design, the power draw, the PCIE implementation, etc etc. I just don't mess with it. So take my posts with a grain of salt. The one thing you can take from my long, rambly posts that is 100% true across any platform and any CPU maker is that you tweak things 1 at a time.

The book you've just read is available for your Kobo reader or as an audiobook

@Theocnoob I don't know where you got the idea that I was taking these huge leaps, yes I went from 200 to 210, is that huge, not in my book, now going from 230 to 240, that would be HUGE, if/when I get anywhere near that number it will be 1 step, but going from 4.2 to 4.4 is not a major jump. the fact that I had the mem setting locked in to the XMP profile was had me screwed up and that's a noob mistake.

t1nm4n is all okay at your end man it seems like. We are typing and not over your shoulder as you mention above. It takes time to get where can read in words what someone is thinking. Yeah, I know nearly impossible, but it does sort of happen.

No really 200FSB to 210 is not a huge leap. It can be for the ram though depending how it overclocks. My GSkill will take that jump in ram speed pretty easily and most of us that are using 1.5V ram are running 1.55V on the ram just for that insurance as the ram gets pushed a little extra. AND we tend to bump CPU_NB voltage to 1.25V up from 1.18V as standard for most of these boards and their bioses even from different manufacturers. That 1.25V helps with those 4-4Gig sticks you have in the ram slots.

IN my mind you have the premier mobo for clocking FX processor. Certainly none better and maybe but not sure a few may be the equal in adjustability for an overclock. So you not at issue there.

Because we want to know the max cpu speed you can attain within temp and voltage limits, I still suggest using multiplier only and the other busses then should remain at default. You set your ram to its' rating speed and timings and voltage and if 1.5V ram...then set to 1.55V which hurts nothing but is insurance and most of us do it now out of HABIT.

Oh by all means you will have to finally learn something of how the other busses are related to rasing the FSB/HT Ref Freq, but that can wait until we know within 50 or 60Mhz of just how high your cpu will run with stability. WHEN you reach the limit with multiplier alone, then you can begin to drop the multiplier and raise the FSB and see if you can as some have, get a slight bit higher than using muliplier alone OR do the same speed with a shade less voltage to the cpu. You can also use the FSB to overclock the ram by moving FSB up and letting ram speed rise and adjusting the cpu multiplier to get the max stable cpu speed and adjust the multiplier of the HT Link Speed and the CPU_NB speed so they remain within a semblance of specs or still stable.

I urge muliplier first simply since it should be simpler. It can give us a quicker idea of overall speed to expect. And it usually gives a quicker sense of satisfaction in reaching a pretty fair overclock. That said and how I normally coach...how you want to get to Max speed and how fast...again that is up to you. No I have not seen any errors that endanger jack-shett, so in that regard, you can progress as you see fit. But you have to establish a baseline not far into your overclock as a place and speed that IS stable; otherwise you may be flying blind and spend considerable time reaching for a speed that was never in the cards anyway.

I have taken this self-inflicted overclocking stuff more then a few times and it has only been about 20 mos since I lost my way in an overclock and was sure I had blown the cpu and had to tear everything down and swap back to my favorite mobo to get back on track. It is always best to know that there is a not so heavy duty clock that works time and again as a baseline to help get going forward again with confidence when that failure just kicks us in the teeth. Luck man and Happy Holidays.
RGone...ster.

@RGone, that's what I was missing from the bulldozer was what you just pointed out, and I started doing that this morning, raising the multiplier, after rereading these last few posts and the bulldover guide I kinda thought I might have to start with multiplier and see where it can get, keep temps at or below 43c for a 24/7 OC which this will be, I might bump it higher for a short period, but nothing to the tune of 5.3Ghz which to me is just insane.

The adjustability of this board might be a problem for me, cause wow is there a way to adjust everthing, and I mean everything, its kinda overwhelming. So starting from square 1 rev.a subsection2.3

Yes start with just using your multiplier if you get high enough in your OC then we will help guide you a bit more on how to use the FSB to help your OC along.

@bassnut that's where I'm at, still not sure which programs and for how long to run them, Theocnoob said 3 hours on p95 blend as a min and I guess that makes sense, that would definitely give complete saturation of heat throughout the computer and its components. But it doesn't seem to use all the instruction sets a cpu has to offer. I was playing CoD MW2 last night for several hours and on stock clock speeds attained higher core temps than with jacked-up (I say jacked-up cause I didn't go through the tried and tested steps at attaining it,) OC. I read elsewhere that Aida64 was a good burn-in utility. Linpak doesn't seem to run on my pc nr does linX I think it was called. OCCT is the only other utility I has other than 3dmark and loop physics tests, which don't fully utilize the cpu, but ( I am probably wrong here) do use other instruction sets the cpu has to offer.

Would it be detrimental to use more than one of these programs at the same time?

In the attachment under the multiplier is there always the bracketed multiplier settings? I have everything disabled so there should be no dropping of the multiplier, but do the brackets go away or is there something else I forgot to disable? Like I said, and RGone said this board got lots of adjustments.

Ignore those brackets ...... you are trying to make this more complicated then it is. Put you OC back to stock / optimal settings and lets start from scratch. Set your VCore at around 1.4v and start upping you multiplier only by0.5 and then save. Boot into windows and run Prime blend for about 20 / 30 min for now we will run it longer later to test stability. Watch you your temps in HWMonitor 70*C max socket ( CPU Temp in HWMonitor) and 62*C max Core (Package Temp in HWMonitor). If you pass your 20 / 30 min blend test with no errors then reboot and bump your multy again and re test. Keep doing this till either you get an error or you get over the max temps I have mentioned then get a screen shot of CPUZ and HWMonitor while Prime is still running.

Don't worry about setting anything else we can do that all later. Just worry about the Multi. I have not read through all the posts tonight but I would guess this has been mention all already so ....... Lets keep it SIMPLE. Just worry about your multi and temps the rest we can do later.

@bassnut Been doing that for the most part, cept I didn't bump volt up right away, I increased multy and ran a 20min bench, if windows locked up I increased voltage until I got a 20min bench. this attachment is after 30 min p95 blend no issues and the temps seem a bit low, but it is cold here tonight and the ambient is about 68f. I wasn't tryin to over complicate things, I was just concerned that it might be affecting temps cause if there is a setting I missed in the bios that allows the cpu to throttle and run cooler than I need to disable that.

Edit: how high do I want to go with the multy only? I understand that a personal answer, but ideas are welcome.

t1nm4n, please have HWMonitor open before you begin the stress test so we can tell max from min temps. If those are max temps that is pretty freaken cool at 4.8 ghz.