|
Hi. Some of you guys may remember one of the first blogs I made, my first overclock blog when I first built my PC. I originally joined TL to help me build a budget PC - I did not have much money at the time, and my way of 'thanks' back to the community was to make a Belial's Budget PC Guide (parts are outdated, mostly GPUs, but information is still up to useful and up to date for budget builds, although imo now pentiums and i3s are better than amd at lower price range).
Most people around here know me by my Starcraft 2 strategy input (see profile for the best ZvX guides out there), but don't realize my start is tech.
Anyways, I'm going to be writing a 2nd overclock blog! It's been almost 2 years since I built my original build:
+ Show Spoiler +
Athlon II X3 450 Rana OC to 3.41ghz/2.53cpu-nb@1.488v
2x2GB RAM Kingston HyperX 1333mhz 9-9-
GTX 460 768mb OC to 922core/1844shader/1987mem@1.087v
Biostar A770E3 6.3 Red&Black Motherboard 770/710
Antec Earthwatts 430w
NZXT Gamma case
Refurbished Caviar Blue 160GB
$300 at the time, closer to $250 now, $200 if you find good deals or used parts easily)
NZXT Sentry 2 Fan Controller and Temp Sensors, CM Hyper212+, 6 Yate Loon Mediums for cooling (~$50)
You may also recall some 'quirks' that I did, such as the following:
Jerryrigging heatsinks to my 3+1 mosfets
+ Show Spoiler +My motherboard only has a 3+1 VRM design, and the VRM design and mosfets themselves aren't particularly the best either. Now Biostar doesn't have a bad track record with blowouts (unlike MSi, never buy an MSi motherboard, dont take it from me, just search on vrm msi nuclear catastrophe), but a 3+1 phase is NOT going to support a x4 overclocked processor! So, I posted on overclock.net and stated I was going to mix elmer's glue and TIM together 50/50, and stick on a sawed off heatsink to my VRMs. Not the best idea in the world... but a fun one, and it's better than nothing. A budget build gives you freedom to do these kinds of things, to push the limit. A nice man at OCN mailed me, free, some thermal tape, and I secretly went into my roommate's room and took his am2 hsf off his computer and swapped it with my am3 (ooh im evil). An am3 stock heatsink doesnt exactly saw in a 1cm wide piece smoothly. Applying thermal tape to my 3 mosfet arrays. ![[image loading]](https://lh4.googleusercontent.com/_Vsbun8_dF5Y/TX-UvwThfMI/AAAAAAAAAjQ/Z8o_aJHYhpw/s640/IMG_20110315_052543.jpg) It took about 30 minutes to an hour to saw this 1cm wide chunk off the square, stock am2 hsf, with a finetooth saw. ![[image loading]](https://lh6.googleusercontent.com/_Vsbun8_dF5Y/TX-UXdp5uHI/AAAAAAAAAiw/62rA8SJMMWg/s640/IMG_20110315_052554.jpg) Stuck onto my VRM! ![[image loading]](https://lh5.googleusercontent.com/_Vsbun8_dF5Y/TX-UsZCQmgI/AAAAAAAAAi8/9uTvXh1tuYA/s640/IMG_20110315_064932.jpg) And my motherboard has yet to blow out, and my temperature diodes at my VRM heatsink never show temps above 40*C, so running very cool despite so much load! Bear in mind, that even though Biostar's website says this motherboard is rated for 125TDP, a 3+1 array is something you see in like.. micro-atx boards of shitty quality, low price, that aren't meant for overclocking or quadcores at all. So the fact I'm running over 140TDP on a board that shouldn't go over 95w, for almost 2 years now, safely, should tell you something. That I'm just so awesome!
Jerryrigging a fan onto my northbridge heatsink
+ Show Spoiler +The number one cause of motherboard blow-outs, is when people switch from stock, radial cooling (which blows down onto the board, and what motherboards are spec'd to use, and who's TDP is rated against), and tower cooling (awesome for CPU cooling, but totally neglects your precious VRM, mosfets, RAM, CPU-NB, and RAM VRM). When you switch into tower cooling, your cpu temps may go down 20*C, and that's great. But other components will rise in temperature, possibly critically so, even if you have good airflow. And so, I found myself with a 60*C motherboard northbridge suddenly when i switched to tower cooling in the cm hyper 212+ from stock am3 hsf. So I took my stock, 80mm hsf, and used copper wire to attach it to my cpu-nb (with electrical tape on certain areas to make sure no contact with the motherboard). Now, my cpu-nb is very cool, and it provides exhaust for the top of my gpu, and extra cpu cooling and airflow! ![[image loading]](https://lh4.googleusercontent.com/_Vsbun8_dF5Y/TXf84-mVdFI/AAAAAAAAAiU/1-vJHqY8pVk/s640/IMG_20110305_050630.jpg)
This has been a great system, and I've been able to stream my 1360x768 Ultra Graphics starcraft 2 on 720p@45fps on a regular basis (twitch.tv/belial88 for proof, although ive been testing more recently so if you really need a high quality video, twitch.tv/belialtester88).
So what is going on? Why have I assembled you all here?
I found a Phenom ii x4 955 BE for $31 on Ebay
You don't believe me?
http://www.ebay.com/itm/AMD-Phenom-II-X4-955-3-2-GHz-Quad-Core-HDZ955FBK4DGI-Processor-/170932444396?_trksid=p2047675.l2557&ssPageName=STRK:MEWNX:IT&nma=true&si=O3pygsuMdYheVgmNELV3td+JyS0=&orig_cvip=true&rt=nc
Still don't believe me?
![[image loading]](http://i.imgur.com/NPKmC.jpg)
It looks a lot better than it was, otherwise I probably wouldn't have ordered it lol. It's missing 3 pins, and 2 of the pins are bent so far out, that they were impossible to straighten without breaking. As in, they weren't just bent, which is an easy fix, they were crushed and folded. I didn't appear to see any of that in this picture, it just looked like some naturally bent pins, a challenge.
In the end though, it worked out. I unbent all but those broken 5, in which case I took an ethernet cable, cut it, then harvested the thin, copper wire, then took a small piece, folded it in half, and stuck it in, about the same size as a pin, then very carefully, stuck them into the corresponding motherboard socket holes (by 'sticking' the wire to my finger and very very very carefully sticking it into the right hole), closed down the lever to 'lock' the cpu stocket, then jam jam jam the wire tip poking out with a + screwdriver to avoid it contacting other pins and ensure good contact with the correct pin.
And it worked, first try! I mean it took hours to do all of this, but the phenom ii 955 works just fine, so far. I've been running p95 for a bit.
Now, the reason I'm posting this blog is because, overclocking has a lot of fine tuning involved. You can read as many guides as you want, but I'm pretty sure about 90% of the overclocks out there, are limited by something inane, goofy, and irrelevant like "Enable Floppy Controller" being on Disable instead of Enable, or 260 FSB being limited being a chipset/motherboard issue that could be solved with more volts rather than the limit of the cpu, or that so on and so forth. So I'm posting this blog here, to show exactly what settings I'm tweaking, for my overclock.
I've done quite a bit of overclocking and tweaking on my AMD sysytem (sorry intel, i can't help you guys out, although you might learn something here, tbh ive learned more from intel overclock guides than amd ones), so maybe those mystery settings in bios, you'll understand more, or understand how to push your overclock more. It may also be a way for people to talk about what they got on their phenom ii's, if they'd like to, or help me if i get into a rut.
You can check out my previous overclock blog, to see how this one will go.
So, onto the content. Now my Athlon ii x4 3.41ghz/2.5ghz cpu-nb had a 'passmark' bench of 4,141. That's equal exactly to the core i3-3240. That's about 100 points above a stock phenom ii 955. It is most closest to a phenom ii b60 (basically a phenom x4 deneb, which is any x2 or x4, overclocked to 3.3ghz). The reason I find this processor desirable, however, is i have 7 case fans, a hyper 212+ hsf with push/pull Yate loons, and 5 temp diodes. Temperatures is not an issue, and my athlon ii, unfortunately, would never go past 40*C but was very stability limited (.2v would not yield another 20mhz, literally).
Hopefully, I can take advantage of my system cooling, with the phenom, and overclock the crap out of it. Secondly, I can't afford a motherboard upgrade, and $30 for a small CPU upgrade from athlon ii x4 3.4 ghz to a, i dont know, a phenom ii x4 3.6ghz at the very least (a mild estimate of an overclock on a c2 stepping), is definitely worth... the $30. Third, it was a spur of the moment purchase, $30 for a phenom ii, yes sir. I'm hoping this will turn out to be a really big upgrade as I overclock this thing to it's limits, though.
So onto what's up.
The first problem I had, was then when I enabled "Custom P-states" in Bios->Performance (basically the overclock section of the bios) -> CPU FID/VID Control, which allows you to tinker or not with cpu vid, nbvid, and fid aka multipliers, my system wouldn't post, or get to the boot-up screen. just black on powerup. I must have CMOS reset my computer at least a dozen times, as I was also trying to make sure it wasn't one of the various motherboard settings in BIOS I always do (things like disable legacy usb support, ide detect time to 0, numlock on start up, etc)...
It took a long time to figure out custom p-states was the issue, and it was quite depressing at first. I got a phenom ii that I can't overclock at all, for whatever reason. I wasn't even increasing the clock or anything, i was simply just enabling the menu for overclocking. If I can't access this menu, it means no raising or lowering the multiplier, and no raising or lowering of core vid.
I'd be limited to overclocking via the fsb, which is fine, I actually prefer overclocking through the fsb than through multipliers, because amazing power-saving and chip preserving features like C1E (lowers volts and clocks based on usage state) and PowerNow! (aka coolnquiet, lowers volts and clocks based on load) are not able to be used if you overclock via multiplier (so all you black edition people, yea, black edition is stupid). Of course, when tweaking your system, turn these features off, and overclocking via increasing the multipier is an easy way to quickly narrow down your overclock, but to really fine tune your overclock, a combination of FSB and lowering your multipliers is the best way to do it. Multiplier only overclocking means you can only finetune by 100-200mhz, which is a huge margin. I like to have my overclocks tight within 5, especially on the cpu.
Then, I decided screw it, up the voltages on all the things! Chipset voltage (motherboard northbridge) to 1.18, cpu-nb voltage to +.1v, memory to 1.65, and +.2v cpu over voltage (different from vid, its where the motherboard supplies extra voltage).
And it booted! One by one i reduced voltages back to stock, and found out that for some odd reason, I need to add cpu voltage through the motherboard. As in, enabling p states and increasing core vid would not work, but enabling p-state and then adding motherboard overvoltage to cpu does.
So right now I have the system with +.1v cpu over voltage, and just a very conservative x17 (from x16) multiplier so my cpu is overclocked to 3.4ghz. I'm basically making sure that this chip is stable at stock value (id be very shocked if it failed an 8 hour prime95 blend).
Will make sure the missing pins isn't an issue, basically, and update this as i go.
edit: on a side note, Advanced Clock Calibration (ACC) causes and caused my my cpu temps to be displayed as total whack and wayy under ambient. This was an issue with my athlon ii was well, i didnt realize acc was to blame for it. Odd thing, in bios my cpu temps were listed perfectly accurately, just any monitoring program would show core voltage as 0-10*C. I just realized it was ACC's fault because when i first booted up without changing any bios settings, my cpu temps seemed normal and ambient+, but when i tweaked it all, it was 0*C idle and i realized it must have been one of the bios settings and was able to narrow it down (odd, because at default, bios was acc auto with my athlon ii... must be because of some bios update i did since i overclocked my athlon ii). Will have to email biostar about that one, i had a looong back and forth with their tech department trying to figure that out and we never did lol.
On a quick p95 test, my VRM temps didnt go above 40*C by the way. The temp diode is underneath the heatsink on the vrms. CPU temperatures maxed at 44*C and idle around 30*C. I understand that I dont want to go above 55*C on load (my system is set to shutdown at 60*C).
  
|
Wow, that stuff about the CPU pins was awesome. I would have binned it for sure : D
|
^ Thanks. It sure is quite a score to get a working 955 for $31. I now have a Phenom ii x4 that will probably get around 4+ghz, 4gb ram, gtx 460, system, that costs about $200. Probably a bit less considering the mb, ram, psu, case, are down a few bucks in price nowadays too.
So my method is going to be prime95 blend ~8 hours at night, and then change the settings in the morning for the next day to be the next overclock I want to test. Since I want to see where the limits lie with this cpu right now, I'm testing 3.8 ghz, which I think is where most people get their c2 955's on stock voltage or higher voltages, not sure exactly, but i think its a good high-end spot to start.
I'll use 3.8 ghz for the day, and p95 blend 8 hours tonight. I know I bag on multiplier overclocks, but my system is stock now and im just doing multiplier to get a rough idea where I am at, since I disable powernow! when fine tuning anyways. Core FID x 19 = 3.8ghz @ stock 200 fsb. (Bios->Performance->Core VID/FID -> Custom Pstates enabled, Core FID = x19/3.8ghz ;; Performance -> CPU Over Voltage +.2v)
Interest to note - when I raised core VID to 1.45 (with the +.1v motherboard overvoltage), it wouldn't post (as in black screen upon start, didnt get to boot up screen or anything). 1.45vid+.1v over voltage should put me around 1.53vcore. Thinking it might be a similar issue that i had earlier, I changed my vcore the other way around - 1.35vid(stock)+.2v over voltage. And it worked just fine. This is 1.503vcore (i believe VID is a little higher than the actual supplied voltage).
I could probably test 4ghz, 1.55vcore, but I want to make sure my temps dont get too high, so that's why I'm testing 3.8ghz@1.503vcore. I'm going to run a 5 minute p95 blend priority 10 test just to make sure my temps aren't too high. I'm currently around 32*C cpu idle, 32*C VRM idle. My VRM's got around 50*C last night.
|
Nice, it'd be cool if you could make a table of some sort with stock speed, the update it as you improve the speed of each componant.
How come you don't get watercooling if you're trying to get really high overclocks?
|
Because I'm not trying to spend a lot of money, and water cooling neglects your entire system. It's not my CPU, RAM, or GPU temps I'm worried about, it's my low-quality motherboard. I'd replace my motherboard with higher quality VRM/mosfets with an asus 8+1 before I got better cooling.
The benefits of air cooling is that I'm trying to get cooling across all of the motherboard components. I've pushed my GPU to 1.2volts by flashing the bios with a custom bios, and I couldn't really raise my GPU core clock more than an extra 20-30mhz (which is a huge voltage increase from 1.087, that's double increase from stock, and 1.087 is still covered under warranty and is just the top limit through msi's afterburner, which is perfectly safe, but above 1.1v is where it gets dangerous). My temps never got acros 70*C on my GPU, I have a beasty twin frozr ii gpu heatsink (although the horrible design with the card is the GPU VRMs do not have a heatsink on them, wtf), so temps were never a problem, just stability, and I'd be cautious pushing high volts on my GPU since there is no GPU VRM heatsink. As long as I'm within the limits of afterburner, that should be perfectly safe anyways.
Spot cooling on my chipset-NB keep it cool, and RAM nevers goes above 35*C, they are quite cool.
So that's why. A combination of temperatures are never an issue to me, and the fact my motherboard is quite low quality that going to water cooling would probably result in a blow-out. Like I mentioned above, you have to be careful with tower cooling, you might blow out your NB or VRMs because they are no longer getting cooled anymore with tower cooling. Maybe i could get VRM water cooling... but that wouldn't really make much sense to get vrm water cooling as opposed to a nice, asus, 8+1, rds on board with a 8-pin cpu power port and solid state capacitors. .
Although this phenom seems to get kind of hot, so it might become on. I'll worry about that when it becomes an issue.
I'm going to go straight ahead and overvolt to 1.55v, just going to test tonight p95 8 hours to make sure temps aren't out of control at 3.8ghz @1.5v. Or I suppose I could just do a burn test to make sure temps are okay, and then start pushing 1.55v tonight... either way, I'm going to work backwards. I'm going to find my max overclock with 1.55v, finetune it by doing FSB overclocking (reduce multi on cpu-nb, htlink, and ram, then increase fsb or adjust as necessary so i can adjust it +/- 5mhz), and then i'll work on downvolting it to see if I can get it stable at that clock at a slightly lower voltage.
With a c2 stepping though, temperature will definitely be a concern with this chip, especially with the L3 cache. I mean I already got up to 44*C on a minor overclock. My athlon ii never got hot at all...
|
OCN member mdocod has found that as of 3 March 2011, at least 71% of the VRM cooling failure incidents in the compiled list of horror stories have happened on a cooling that deviates from stock cooling. This value may be higher due to the amount of situations where cooling was not described. "Stock" CPU cooling is designed to blow down onto the motherboard components, including VRMs. Aftermarket cooling, which includes: tower cooling, any sort of water cooling, is usually not. Remember, TDP rating on all boards is done with processors at stock and with stock cooling. That means your 4+1 phase or even 3+1 phase (on AMD platform) may actually be fine for a more power consuming (i.e. 125W TDP) processor with stock cooling & at stock speed, but deviate any one of these and you're on your own. ....At least 70% of all VRM failure incidents happen with aftermarket CPU cooling installed.http://www.overclock.net/t/943109/about-vrms-mosfets-motherboard-safety-with-125w-tdp-processors/0_100
|
|
|
^ Thanks!
Does anyone know what the max safe VRM temps are? I know the temp 125*C is sometimes tossed around, but those are for high quality boards. This is a 3+1 board. Now, I took photos and wrote down (somewhere on the internet, i think in original oc block or maybe in my pc building guide) what my VRMs and mosfets consisted of, so I suppose I could check on the parts but...
They hit 80*C when I tested 3.9ghz@1.552v.
I'm really surprised I'm nearing 60*C with just 1.55v, 3.9ghz. I mean i guess not, that's near the limit of what people get with the phenom ii and all, and it is c2... but i never thought my temps would get so high with 7 case fans and a hyper 212+ :X.
Will be doing a blend95 tonight of 3.9ghz@1.552v. Temps are becoming quite restrictive.
|
http://www.overclock.net/t/947924/incorrect-temps-displayed/0_100
You know, I've been checking out all the threads I posted when I first built my athlon ii system, and was overclocking it. I made a lot of posts. there were lots of problems lol.
And even with ACC disabled, my temps were whack under my Athlon II.
I'm starting to think that maybe my ~55*C temperature readings for cpu on this phenom ii may not be correct. That seems a little high, even if im at 1.5v+. I'll just touch the heatsink, see how hot it feels. It maybe just looks the most correct because it idles around 20-40*C.
I mean right now, 36*C idle temp. i dont know how correct that sounds? maybe it is. i just recall temps being whacked out on the athlon ii, probably can't trust these ensors either.
the problem is that despite having 2, discrete temperature diodes, I can't really attach them to my CPU though. The 212+ heatsink sits squarely on top of the am3 processor, so its not like I can slip it 'under' the 212+, and there isnt enough surface area to really tape it to the side. The hsf mounting bracket makes it near impossible to mount a sensor onto the base of the HSF as well. But, with the sensors stuck into the fins or snugly tucked into the base under the mounting bracket, they register like 30*C max. I can't quite tell what the right temperature is...
|
![[image loading]](http://i.imgur.com/IKh4x.png)
Note that CPU-Z says 125w TDP, HWMonitor shows 113.4.
VRM temps around 75*C.
Last night I tried to run 3.8ghz stable, but it required 1.3875+.2v CPU Over Voltage = 1.568vcore, which put CPU temps around 60*C (where it rose/fell around 59-61) and VRM at 90*C, I was a bit worried so I cut it down to 1.375+.v2=1.552vcore, and it failed with a BOSD x124 (thank god for blue screen view) around 5 hours in.
I currently have hwinfo logging and coretemp logging (hwinfo does not allow automatic logging, thats why).
I might try at it again with 3.8ghz@1.568v tommorow, but let's just see if this thing is stable at all with an overclock first, so I'm running 3.7ghz@1.502v. So far it's run about 6 hours in prime95, but it's been off and off priority 1 and 10, as I've been browsing the internet a bit while testing, but also watched some tv on another computer.
|
I'm a bit hesistant to accept that your copper-wire fix worked, especially given that it was more likely that the broken pins are not used for data.
However, it's nice to see people still playing with old hardware so optimistically. I have a Q6600 running at 3.4GHz (comparable in performance to an overclocked 955 BE) from 2008 (suicide clocks ~4GHz for benchmarking). I would have called it quits by now.
|
I've done a TON of searching on this subject in the last few days, so I'm very hazy on where I read this, and it doesn't help that since am2, AMD considers pin layouts secret trade material...
But basically, something like of the 938 pins that a deneb has, less than 1% are unused... like it's just a couple pins that are redundant or ground pins. The chances that one of the pins you broke, is one of these maybe half dozen of 900+ pins, is just really small. Let alone 5 of them.
I'm pretty sure my copper wire fix worked. At worst, maybe the wire fixed only 4/5 of the pins and I got super super lucky on the last pin, but given that I'm able to stably overclock to 3.7ghz and 3.4ghz, I'm pretty sure my fix worked.
I've been running an Athlon ii x3 450 Rana at 3.41ghz@1.488v, and I was able to stream sc2 on ultra on 720p@45fps with absolutely no problems, so I'm quite optimistic in this phenom ii. I'm sure I'll still be nearing 90%+ loadout on all 4 cores even with a phenom ii 955, but it should be a noticeable improvement - particularly single threaded performance for sc2, and multi-core/encode performance.
I had no plans to upgrade my athlon ii based system anytime soon, I probably would have stuck with it for at least another 3 years, so it's cool that this worked out. It's really the only 'budget' upgrade I could do - an x6 would surely fry out my 3+1 power phase if overclocked, and offers zero gain on single threaded/dual threaded applications like games, particularly sc2, so the only significant upgrade option is a phenom x4, unless i invested in a new motherboard and a 100+ cpu.
maybe i can buy an intel lga 1155 with missing pins one day for a sb/ib upgrade
|
So last night, while prime95 was on priority 1 and i was just surfing, and had had p95 for about 7 hours, it froze (3.7ghz@1.504v).
I set it to 1.536v, but that failed around 3 hours later.
I think these failures have something to do with the temperatures, the so called 'heat wall', where heat is a limiting factor in stability. I guess my core temps are accurate.
I really need to back off the aggressive overclock, I really just need to see what exactly the range this CPU is capable of, so today I'm running 3.6ghz@1.504v, and I'll run prime95 later on. If it works, tommorow I'll work on getting 3.7ghz to work somehow.
But I might not have a choice, I'm really pushing 55*C+ with these 1.5+ volts.
Here's a message I got from Biostar when I emailed them today though, when I asked them what is the maximum safe temp for VRMs. I have 2 temperature diodes under the hottest 2 phases, I only have a 3+1 power phase, and one of them routinely gets 70-93*C depending on 1.502-1.568vcore respectively (i didnt stay on 1.568vcore long, the cpu temp was 60-62*C and vcore at 93*C quite quickly).
Technically the phase designs can handle 100C, but definitely not recommend for long periods of time. At 90C there will be PCB discoloration. But at those high temps, pretty good odds will shorten the lifespan of the components.
Still doesnt exactly say what the max safe limit is ;/
So I'm guessing around 80's, so I'l continue working with vcore ~1.552, although maybe i should step it down since it seems to be heating things up too much.
|
You do the exact same things I like to do with AMD systems. People call amd cpus shit but they are far from it. and they are SO cheap. Usually whenever I go to build a pc, I take a look at my budget and then go straight to ebay and literally buy the BEST parts I can for my budget. My last system was a Phenom II x6 1045T (oc'd it to 3.4ghz), 4gb ddr3, a really nice asus mobo (forgot the name but it also supported am3+), a gts250 (1gb version), a thermaltake psu (550w), and a 320gb hdd for $225. If you dont mind doing a little hunting, you can build just about any pc you want on ebay for almost half the cost.
|
^ You know, I don't really feel AMD is ever worth it. The only reason I felt AMD was better at the ~$120 and lower price range was because of the cores, but I didn't realize how far behind their single threaded performance was or how well intel's hyperthreading is.
You *might* be able to make an argument that an overclocked am3 phenom x4 is better than a dual core pentium G860 or something, maybe i3, in streaming, but a sandy bridge dual core like pentium or i3 would be wayy better in gameplay for sc2. I also wasn't really aware that intel made competitive chips below the i3, i thought it was just i3, i5, i7, i didnt realize sandy bridge also included a bunch of pentium sandy bridge chips that are actually really good.
And to boot, pentium sandy bridges run on the lga 1155 socket, so one day you can get a significant upgrade to an i3, an i5, an i7, not to mention you could go with an older or newer generation i3/i5/i7, ie sandy bridge, ivy bridge, XXX or XXXX series.
Whereas with Am3, maybe you go with an athlon ii, which is a great chip, sure, you can get 4 cores for a great price, but your only step up from there is a phenom ii, which, at the same clock, is basically like 300mhz stronger only because of a L3 cache. Hell, deneb chips are really just propus chips with no level 3 cache on them, it's not like the phenom is a totally improved chip over propus.
And thuban, it's not even that much over a deneb, and the problem is that 5th and 6th cores don't mean shit. Single, dual, even triple or quad threaded performance on a thuban x5 or x6 is the same as a phenom x2.
The only real improvement i see with am3, is if you go athlon ii propus, to phenom ii deneb with c3 stepping. To get a major upgrade on the same platform, you really gotta get a c3 revision deneb and overclock the shit out of it, which, to be fair, c3 revisions overclock amazingly. But if you got a c2 revision deneb, well, your screwed (unless you got it for super super cheap... like i did, $31 for 955 c2).
I loved my athlon ii, and im loving my c2 deneb 955, don't get me wrong. They are great chips. And intel motherboards are a little more expensive. And the 4 logical cores I have is great for streaming, which I happen to do. I wouldn't say I regret buying AMD, overclocking AMD is a lot of fun, intel are kind of bastards in that regards. But I don't think I'd ever recommend AMD, even at the sub-$100 level. The pentium 860 is just so much stronger at the same price level, and it's such an outdated chip too.
It doesnt really look like AMD is going anywhere with piledriver either, which sucks. A phenom ii is better single threaded performance, dual threaded performance, and frankly, better per core architecture. Not that it matters, it's a totally different platform, and at that price level you'd be stupid not to go i3/i5/i7 unless you have a very specific reason for needing AMD's quasi-8 core.
The idea that AMD is better at the lower price level is BS. AMD chips are great, i love them, but intel is just better. Quite frankly, any processor these days will play starcraft 2, or any modern game for that matter, just fine. Streaming takes a little more, and takes use of multiple cores, so that's really the only instance where you need more power, but a phenom ii or even althon ii can stream quite nicely if you tweak the settings right.
i dont mean to start an argument here - i loved my athlon ii, and i just got a phenom ii (bought it off ebay for $31, it had 5 missing pins, i stuck copper wire into my motherboard socket and got it to work and currently overclocking it around 3.7ghz). But i think intel's pentiums might be the go-to at the low end price level. the architecture is just too good to look over just because they are dualcore. both amd and intel are great when you really get down to it. but it's most just athlon ii/phenom ii, and everything by intel.
|
I've been struggling. I think temps might be a real limiting factor to me.
I failed prime95 after like 4-5 hours on 1.35+.2v=1.520vcore on 3.6ghz. Ughhh the humility (at least I know 3.4ghz@1.488 is rock solid, if i need to fall back).
So I am lowering voltage to 1.3375 + .2v = 1.504vcore on 3.6ghz. I know that sounds counter-intuitive, how is less volts going to stabilize my overclock, but I think temps might be why i'm crashing. I mean I'm constantly hovering around 55*C, give or take 2-3 degrees, and it's not like im crashing as soon as I hit 55*C.
But with 1.504vcore, my VRM temps is only maxing out around 75*C (in a warm room, not like a cold room last night), and my core temps are 53*C, socket temps of 45*C, so I think maybe staying below 55*C will help stability.
We'll see. Would be pretty sweet if this holds true and i could go to like 3.8+ with lower vcore than ive been testing so far.
|
I'm still struggling with 3.6ghz, and there seems to be no correlation in crashes between 1.488v and 1.52v, as in the range of voltages, 1.488, 1.504, 1.520, all kind of crash after like 3+ hours. Hell i had 3.7ghz@1.504v fail at 7 hours, youd think that would have crashed earlier than 3.6 on more voltage.
So my conclusions working forward now are:
1. When the CPU runs at 55*C and above the entire time, I think it will fail after a few hours. I think this is why I'm seeing no correlation in my overclock fails. Like youd think the higher the speed is over where 'true stability' is, the sooner a crash would occur, or the lower the voltage, the less stable it'd be, but it doesn't appear to be the case.
2. I think I should run ACC to help stability.
I've done some reading and it looks like some people had gained some stability when using ACC to aid their phenom x4 overclocks. So I have ACC on Auto, I'm going to look into seeing how to manually set each core, but right now I'll run it on auto and see if I can run this 3.7ghz@1.488v - which is the lower limit of volts I've tested, and the higher limit of overclock (i was tempted to run it at 3.8 or 3.9, but i just need to confirm if ACC helps overclock or not, so if it can pass this, then i know acc is good, if i fail, well, maybe it helps, maybe it doesnt, ill try 1.488v at 3.6ghz since I had the system fail there after a few hours too).
3. I'm manually setting everything in my system. I know CPU-Z reports a 2ghz CPU-NB, 2ghz HT Link, and my ram timings are at default, stock settings, but I'm leaving nothing to chance and manually setting them all to stock values.
Which, I found out, if I set HT link speed to default 2ghz, or 1.8ghz, or any value, I won't post until I spam ctrl+alt+delete and reset button. Which was scary, at first, and kind of odd, so I left ht link speed at auto.
I'm basically sorta starting my OC results at scratch, and seeing if ACC helps my overclocks out. I know it's supposed to be built into phenom ii's these days, but it seems like a few people say it really helps. I also dont accept that my phenom is only good at 3.4-3.5ghz. I also think less volts might be more, so I'm running 1.488vcore since that keeps temps below 55*C, just barely.
The only problem with ACC is my core temp is totally whacked out. But with running on a voltage I know is below 60*C, it'll be okay.
|
|
|
I keep failing at the same FFT lengths. Generally, Prime95 fails after it's last pass is 2400k (looking up online what comes up next on blend, is 160k). I even did a custom test with 160k fft as min max and it failed after about 10 minutes.
Ughghhgh what does 160k fft length test that the other lengths aren't testing? What does it mean?
And i dont care afor 24/7 prime95 stable. i should try another stress testing program... a quicker one! This is for streaming, i dont care if all my files get deleted ~_~
|
I'm having an extremely difficult time getting 24 hour prime95 stable on more than 3.6+ ghz.
Looking on the internet, it seems to me that no one has gotten a Phenom ii x4 955 C2 revision 24 hour stable, and all the people saying 3.6+ ghz are just full of shit and didnt test prime95 24 hour stable, just something stupid like 10 runs IBT.
So has anyone gotten their C2 revision 24 hour stable, and at what settings?
I'll probably just go with 10 runs IBt stable, but before I do that, I want to know what my max prime95 24 hour stable overclock is first.
|
|
|
|
|
|
EPT
