|
Going to start a blog about my overlock process from my budget PC that I built. Some of you may remember this thread from Tech Support:
http://www.teamliquid.net/forum/viewmessage.php?topic_id=188036
Specs
+ Show Spoiler +
Athlon II X3 450 3.2ghz 'Rana'
Biostar A770E3 MB (770chipset)
MSI GTX 460 768mb 192bit Twin Frozr II
2GB Kingston HyperX 1333mhz 9CL
WD 160GB 7.2RPM 8mb Recertified
Antec Earthwatts 430D
NZXT Gamma
Later on I added:
+2GB more HyperX
4x Yate Loon Mediums
Hyper 212+ HSF
NZXT Sentry 2 Fan controller w/ temp sensors
Initial Build = $350
+ Extra RAM, HSF, x4 YL Fans = +$95
So I don't exactly know what I'm doing, but I've read a million guides. Unfortunately, every guide today about overclocking is either about overclocking Phenom's using only an unlocked multiplier (which is stupid, because a full overclock needs to use the FSB, and you're going to end up lowering the multiplier below stock to use a higher FSB anyways), refers to DDR2 RAM (ie outdated), or uses software instead of bios (read: is retarded). In fact, I've found the best guides to be Intel guides, despite being irrelevant in many ways, they were more comprehensive in general and leased good information here and there that I could use.
So what I basically understand is that a system is composed of 3 equations and 1 memory issue:
FSB x CPU Multiplier
(FSB x CPU_NB Multiplier) > (FSB x CPU_HT Link Speed)
FSB x RAM frequency (not totally correct but similar)
Memory Timings
In short, the FSB dictates things. Memory timings, simply keep at stock to maintain stability, or 'tighten' if you want. Or 'loosen' for stability.
What I have done in my overclock process, is simply raise the FSB as high as I can, and then drop the multipliers of everything else to very low. Now, who knows what my FSB really maxes at, but I've found that 260 is more than enough to tinker away. In fact, my CPU can't handle it's stock multiplier of 16 at a 260 FSB, much 13, unless volting is applied. So 260 is my sweet spot for FSB.
Now, because I don't have a black edition, my CPU voltage, CPU-NB, and RAM are actually locked as well. Fortunately, by raising my FSB so high, this isn't an issue. In fact, it should never be an issue, to anyone, AFAIK. I've found that by lowering my multipliers on everything with a 260 FSB, I've found 'stock' values for my HT Link, CPUNB, and RAM. So it's a process of slowly raising my CPU Multiplier up.
It's also crucial I have my 4th core unlocked, as unless I'm able to hit over 4ghz, which is unlikely given I want to avoid major overvolting and stay on air cooling, 4 cores is better than 3. I've also found my 4th core is pretty stable, and that my 3 core limit is the same as my 4 core limit, and vice versa. So.
I'm easily willing to do a +1 voltage increase offset (remember I have a locked CPU, so I must use my motherboard for this) and found my limit is between 260 x 13.5 and 260 x 14, or 3.51ghz and 3.63 ghz. When I lowered my CPU voltage by .5, and then raised my Mtoerhbard CPU offset to +2 (a way of overvolting when my CPU won't allow it due to being locked) I am now on 3.77 ghz (please I hope my system doesn't crash while writing this LOL).
I probably should've started this blog earlier, but oh well. here's my CPU validation for 3.63 ghz, i was stable here and ran for about 7 hours on prime95 blend torture test. I'm currently typing this out on a +.5 voltage with 3.77ghz, and see where I can go with this. i don't really want to hit 1.55voltage, but 1.5 is okay (of note, my CPU-Z says 1.488 but bios says 1.502).
Here's where prime95 was stable:
http://valid.canardpc.com/show_oc.php?id=1694085
here''s where I am now, and will continue upping my multiplier until I crash:
http://valid.canardpc.com/show_oc.php?id=1694810
Since I can't do multipliers, my Bios actually allows me to tinker with speeds, such as listing my HT LinK Speed as 1.6, 1.8, or 2 - with an increase in FSB, these are essentially ways to tinker with my multiplier. I have this in bios:
HT Link Speed: 1.2ghz // 16bit
DRAM 'Limit' 1066mhz
Timings: 9-9-9-24-33-4 with everything else on Auto, on "both"
CPU-NB 1600mhz
ACC Per Core 0.0.0.+6% (did this for the unlocked core, probably unnecessary, will go back to 0 once I find limit, unless I find it's needed)
CPU +2 voltage
CPU_VID 1.35 from 1.4 (so with offset this is 1.5 vid essentially)
CPU Multiplier is 14 right now
Will keep this updated.
|
So I've been messing around, it seems my limit between 3.64ghz and 3.77 (the above post probably shows 3.77 is stable enough for browsing, but not for cinebench benchmark). I've been running Cinebench CPU test as a quick 1 minute benchmark, as I've noticed a lot of my overclocks will crash when I run it, usually a second after I but the button, but sometimes about 6 seconds into the test.
Anyways, given that 3.77 is my limit, what I did was divide 3.77 ghz by 14 (as I figured 3.77 as a limit by doing 260x14.5). So, my limit with a 14 multiplier is between 260 and 269.28 FSB. I started testing 269, 268, 267... and at 267, at boot the screen showed "Too Many Cold Reset due Bios Post. System might be incorrectly overkclocked [sic]. Press DEL to resume.
It went into BIOS, and I simply lowered the FSB to 266 (probably would've crashed anyways, and at 266x14 I'm sure I'll crash after I post this and run cinebench). I don't think my motherboard is limiting me - true I raised everything, the HT link, the memory, the NB, by about 10mhz by going to 270 FSB, but I doubt that it's enough to make an impact, considering my CPU limits are right abouts there, and 260FSB is roughly stock settings, give or take (HT link slightly lower, NB slightly higher, RAM slightly higher).
just wondering what that was all about. Hope I'm not doing undue stress by repeatedly crashing the system o_O
|
Be thankful you can overclock at all... Whenever I try my computer just shuts down. Even if it's just increasing my multiplier by one.
|
On March 06 2011 19:54 StallingHard wrote:
Be thankful you can overclock at all... Whenever I try my computer just shuts down. Even if it's just increasing my multiplier by one.
You sure you've upped the voltages?
I couldn't get my 4th core working or any unclocks at all without raising the voltages a bit on my phenom II x2 555 BE.
Had to settle at slight overclock around 3.5 Ghz cause my motherboard and cooling is trash ATM and I don't think it can take it.
|
Be thankful you can overclock at all... Whenever I try my computer just shuts down. Even if it's just increasing my multiplier by one.
System specs, what did you try to do? If your FSB is 200, that means an increase in 1 on your multiplier is actually a pretty big jump. I'm not surprised it shuts down - assuming this is how the story goes. You may be better off by raising your FSB and then lowering the multiplier, for a net gain of maybe 50mhz, at first. I went straight ahead and upped vcore by +.1, but I really doubt my system would be stable if I simply gave it another 200mhz from stock voltages either.
I couldn't get my 4th core working or any unclocks at all without raising the voltages a bit on my phenom II x2 555 BE.
Had to settle at slight overclock around 3.5 Ghz cause my motherboard and cooling is trash ATM and I don't think it can take it.
Actually, that's a problem I've heard a lot about and surprisngly, few peole know the answer to. On the entire internetz, the answer to this problem is in a single Rebelshaven X3 unlocking guide, very hard to find (not exactly as popular as overclock.net or extremeoverclock, but more useful and 'hardcore' and specific). The solution is lower your HT Link Speed.
I couldn't get an unlock at 200 FSB/stock settings until I lowered my HT Link speed from Auto (which means 2.0ghz) to 1.6ghz. HT Link speed has no impact on the system performance (i'll hold back from saying 'at all' but its pretty insignificant, my understanding is you either have enough or you don't, as long as its above 1.2ghz or so your good).
Also, ACC values let you raise the voltage on each core independently, to help. I haven't had any issues with my 4th core in p95 tests, leading me to believe I probably just have a disabled x4 chip, but there no reason you 4th core is bad. 3.5ghz is a great OC though.
Anyways, back to me me me...
So I tried 269, 268, all the way to 263, and it got steadily and steadily more 'stable' - cinebench would run a few seconds longer until crashing, BSOD crashes became resets... But at 263FSB, was able to run through a complete benc hmark of Cinebench! So I guess my system's limit at 1.502 voltage is 3.682 ghz, although I haven't ran a p95 test yet so I imagine my real system stability is probably just slightly less than that (ie its workable stable, maybe not rock solid 24 hours stable). Anyways, as you can see from this whiny thread:
http://www.teamliquid.net/forum/viewmessage.php?topic_id=195906
I am unable to correctly view my temps (all programs say below ambient, but same low ambient temp at same time) but all my temps are still nonetheless below what my system reported with stock cooling and stock settings on load. My point is that temps don't seem to be an issue.
So what I don't really understand, is how voltage comes in to the equation. I sort of understand more voltage, more overclock/stability - so do crazy phase cooling/liquid nitrogen/dry ice setups just run like crazy high voltages and get good overclocks and keep temps low? Because I thought there's a certain 'maximum' limit to how much voltage a chip can take - I mean I can't see it being even theoretically possible you could put a nuclear reactor to an athlon ii so long as you kept it cold, right? Anyways, heat isn't an issue, but I hear over 1.55 is 'unsafe' and I'm already at 1.5, and I absolutely do not want to damage this processor. I understand CPU's are 'tanks', so halving the life just means it only lives to a huge 5 years instead of 10, so I understand I have room to play, but at the same time I want to take care of my system, ya feel? Anyways, I may try a bit more voltage, see where that gets me, but I'm pretty sure 1.5v is about as much as I go as far as big gains per volting.
|
assuming 263x14 and a vcore of 1.502 is where I want to stop, my next step would be to p95 for 10 or so hours (ideally 24, but ill wait until I have cpunb, ram, and ht overclocks settled before that final test), then I would see about overclock/volt on CPU-nb, and then I would probably just leave ram and ht at stock (they dont affect performance, why chance stability on no noticeable gain).
But I'm going to play around with some voltage jumps, maybe even 1.55. Maybe a dangerous thing to do since my temps are inaccurate, but so far my temps are way below stock with stock cooling (yes, the inaccurate temps, but I think they are just falsely low, Im sure they still pick up increases and decreases of heat).
|
so just a quick try of bios=1.6volt // cpu-z and hwmonitor = 1.58volt, temps are still very low, lower than stock cooling on stock settings. 3.81 ghz, something impossible without the extra .08 voltage. gonna try cinebench, see where temps go.
|
crashed on cinebench lol. so clearly 1.58 voltage doesnt net much gain, couldnt do 113mhz increase for an increase of .076v
i'll see where increasing my voltage just .125 increments from 1.504 up to 1.55 does, but i imagine my 3.682 with a 1.504 v is what i'll settle for. it's nice to know temps aren't an issue (i think).
|
T.O.P.
Hong Kong4685 Posts
On March 06 2011 21:00 Belial88 wrote:
So what I don't really understand, is how voltage comes in to the equation. I sort of understand more voltage, more overclock/stability - so do crazy phase cooling/liquid nitrogen/dry ice setups just run like crazy high voltages and get good overclocks and keep temps low? Because I thought there's a certain 'maximum' limit to how much voltage a chip can take - I mean I can't see it being even theoretically possible you could put a nuclear reactor to an athlon ii so long as you kept it cold, right? Anyways, heat isn't an issue, but I hear over 1.55 is 'unsafe' and I'm already at 1.5, and I absolutely do not want to damage this processor. I understand CPU's are 'tanks', so halving the life just means it only lives to a huge 5 years instead of 10, so I understand I have room to play, but at the same time I want to take care of my system, ya feel? Anyways, I may try a bit more voltage, see where that gets me, but I'm pretty sure 1.5v is about as much as I go as far as big gains per volting.
More voltage = more power to the processor. With more power your processor can go faster.
I read on Intel's site before that the power consumption formula is around
Power = (Voltage^2) * (Clock Speed)
|
i5 750
4GB 1600 DDR3
Gigabyte mobo don't know the exact model number
1100W PSU
2x GTX460s in SLI
don't know much else to tell. Thanks for any help that comes of this
|
More voltage = more power to the processor. With more power your processor can go faster.
Actually I know this isn't exactly correct. I'm nub but I have searched, believe me, I wouldn't be posting questions if I couldn't find it elsewhere, part of the reason for this blog.
More voltage means the current increases, or is 'faster'. It means logical 1's are registered easier, whereas in an overclocked yet unstable CPU what's going on is that the electric on/off signals aren't being registered because the signals aren't being sent as fast as the processor is reading. By increasing VCORE, you are making the electric signals faster. An easy analogy is that when you turn on a hose, it doesn't instantly flow, it sort of builds up. In a CPU, data may be read or misread because that buildup should be registered as a 1, but is only registered as 0. In gets complicated when you start to factor in that a voltage signal of 1.5 may be seen as ON, but a clearly ON signal of 1.4 isn't registered as 1 because that signal wasn't fast enough to reach 1.5 before the system logged in a 0 and is waiting for the next piece of data. But that's just semantics, I'm sure that's what you meant 
i5 750
4GB 1600 DDR3
Gigabyte mobo don't know the exact model number
1100W PSU
2x GTX460s in SLI
don't know much else to tell. Thanks for any help that comes of this
What exactly are you trying to tell me?
Anyways, considering my temperatures are okay, why can't I simply go past 1.55 vcore (what is usually recommended on forums as 'safe')? From what I understand, raising voltage does 2 things - increase temperatures, which is horrible, and increases electromigration, which isn't really good but is very minor and won't really be noticeable (voltage/overclocking doesn't kill, heat does - as the saying goes). Then what I don't get, is how are like 6+ghz overclocks achieved? These crazy lab setups, are they like running extreme voltages or something?
I also read, though I can't be sure, that 'voltage isn't everything' and some crazy phase overclock was achieved with 'relatively' low voltages, although who knows what that means. I think the point was that some CPUs go faster when below freezing, but that isn't really practical for me/anyone. So like does everyone with crazy watercooling and stuff just have huge voltages, and only do so because they can keep temps low?
I mean you realize I have a Hyper 212 + in push/pull, 4 more case fans, a 40mm cpu fan mounted onto my northbridge (ghetto style with copper wire, had an issue because hyper 212 isn't blowing air like the stock hsf was onto the nb), as well as a pretty well ventilated case. Now, like I already said, i'm unable to monitor temps correctly since they are reporting below ambient, but I do see my temps are below what was reported when stock cooling w/stock settings (ie they read a low/high of 16-50) and now its 10-35)
|
That was supposed to be a PM rofl. Sorry for going off the road hahahaha
|
pm about what? pm me the answer lol.
yea still dont get the overvolt thing. I clearly see a overvolt of 1.504 w/3.68ghz on unlocked x3 to x4 is best return, as .1v increases won't even net 100mhz now, but why not overvolt more if temperature is the only thing to be worried about?
|
Just a quick thing - most of my crashes are BSOD. However tinkering with higher multi/lower fsb or lower multi/higher fsb generally yielded crash/stability at the same points. Just wondering if it's a NB issue? I hear BSOD is generally NB issue, but my CPU-NB and DRAM speed is about stock, my NB about 104mhz higher at 2104 and ram about 35 mhz faster at 701mhz - also, system tests stable at 260 FSB until i raise multi on CPU..
I mean bsod can be cpu too right? has to be, unless im doing something clueless here. Or maybe higher cpu speeds affect nb, thereby causing crashes, so I need to up voltage / lower speeds on other things while increasing cpu speed? im trying to stay at about stock speed with overclock.
|
It seems there are stability issues 
So i know this is perfectly stable:
![[image loading]](http://valid.canardpc.com/cache/screenshot/1694085.png)
I thought I had 263x14=3.68ghz, 1.504v stable, but it crashed on p95 pretty quickly, and even raising up more voltage didn't really work. Right now still tinkering with more voltage, even lowered HT and NB. To the guy who said they couldn't get more than 200mhz for one more multiplier, yea, I'm not surprised, I'm maxing out voltage by .2 and can't even get a 400 mhz gain. It seems it's a bit hard to overclock.
You really have to max out volts I think to get around the 3.8+ mark, but from what I read 1.55 is the most you'll go - although I'm not sure why people say that, if it's for temps or other reasons.
On an important note, my motherboard has TDP of 95w, and unlocking + overclocking is bringing me over 140w - is this a problem and how so?
|
I'm having a lot of trouble getting above 3.6ghz actually. I think my limit is somewhere between 3.5 and 3.68 ghz irregardless of voltage (unless I start pushing over 1.6v, which I may do but don't fully understand why people say 1.55 is the limit, because if its temps, well, my temps are very good atm!).
|
People generally say 1.55 is the limit for 24/7 use after seeing chips running higher than that degrade or die. It's not so much the temps, which can still be a factor, depending on your cooling solution, but rather the voltage.
How are you measuring that 140 watt? Generally, a motherboard TDP refers to how much power it's designed to cleanly provide to the CPU.
|
Is there a good overclocking guide for beginners using AMD processors? I have an Athlon I want to overclock. I used my motherboard's built in utility and I can get a stable +10% but I never know how much to raise the voltage. I raised the voltage 10% as well (that overclock isn't applied currently) but I don't know how much I need to raise it. Any guides out there that can help a beginner?
|
On March 06 2011 21:00 Belial88 wrote:Show nested quote +I couldn't get my 4th core working or any unclocks at all without raising the voltages a bit on my phenom II x2 555 BE.
Had to settle at slight overclock around 3.5 Ghz cause my motherboard and cooling is trash ATM and I don't think it can take it. Actually, that's a problem I've heard a lot about and surprisngly, few peole know the answer to. On the entire internetz, the answer to this problem is in a single Rebelshaven X3 unlocking guide, very hard to find (not exactly as popular as overclock.net or extremeoverclock, but more useful and 'hardcore' and specific). The solution is lower your HT Link Speed.
If his CPU wasn't getting enough power, whether that is because it had some slightly defective cores or for some other reason, simply lowering the HT link speed wouldn't have helped.
Also, ACC values let you raise the voltage on each core independently, to help.
Outside of its use in unlocking cores, ACC is a pretty mysterious feature, but I've never heard that one. Do you have a source for that?
|
Is there a good overclocking guide for beginners using AMD processors? I have an Athlon I want to overclock. I used my motherboard's built in utility and I can get a stable +10% but I never know how much to raise the voltage. I raised the voltage 10% as well (that overclock isn't applied currently) but I don't know how much I need to raise it. Any guides out there that can help a beginner?
No, there's not. All current AMD guides are either outdated and refer to AM2 or DDR2 RAM, only talk about rich kid Black Edition Multiplier unlock CPUs, or use software (lawl). If you want to learn how to overclock, my suggestion is to look up "belial overclock guide" and read what I've posted on various forums about the subject, and read Intel overclocking guides (you will generally know when and where it differs from AMD, and then search to pick up what to do elsewhere). I may not know everything but you can probably piece together everything from what I've posted. You will have to spend at least 20 hours searching the subject, minimum. It's a long process made longer by lack of any good guides.
If his CPU wasn't getting enough power, whether that is because it had some slightly defective cores or for some other reason, simply lowering the HT link speed wouldn't have helped.
yea probably. my suggestion was simply just that. It's mainly an Athlon II X2/X3 suggestions but may help other CPUs. There's no logic behind it, as far as i can tell, as my HT has been much higher than 1.6 with unlocked via higher FSB and been fine.
Outside of its use in unlocking cores, ACC is a pretty mysterious feature, but I've never heard that one. Do you have a source for that?
AMD actually has a guide of it.
http://sites.amd.com/us/Documents/AMD_Dragon_AM3_AM2_Performance_Tuning_Guide.pdf
Here's an overclock.net thread talking about it.
http://www.overclock.net/amd-bios/596792-acc-value-question.html
But you are right, it's pretty mysterious. It's not really easy to use google to figure out that one. Granted what I said may not be entirely accurate, but the idea is right - if a core is having issues, raise the ACC value on it. May or may not be additional or less voltage, but supposedly it helps by compensating in some way.
Anyways having a LOT of trouble with 3.68 in prime95. I think I was able to get 3.77 running but it would crash upon just opening p95. In fact, I'm having a VERY hard time getting anything in the 3.6 range running prime stable. I tested a high volt 1.55v 3.625ghz build and it crashed a couple hours in prime95, so clearly +400mhz is the upper limits of the Athlon II.
I'm starting to see other people say similar reports
|
|
|
|
|
|
EPT
