Slashdot Mirror
Athlon 64 In-depth Overclocking Guide
jmke writes "Everything you ever wanted to know about Athlon 64 overclocking, and then some. If you are confused about HTT, LDT, memory dividers and relationship between these settings, then read on. This in-depth overclocking guide will show you how to get the maximum from your brand new Athlon 64 system"
Idiot would do that to this sort of NEW, EXPENSIVE hardware?
Would you overclock a Z-Series IBM server? Would you overclock a 20 4-way xeons in a cluster?
Give it a while. Its not like the MOST OF US will need that speed...
Hell, I use a 1 GHz machine and develop on a 500 MHz machine. Yeah, 500 MHz because many users are still stuck on 300's and 450's.
yet another liquid cooling story is on the way!
Hardly anyone doing professional work on a machine will overclock it. Generally it just makes your system unstable and prone to crashing and making murphey's law become a reality on your precious data. Overclocking IMHO is more for gamers that want to take out as much juice as possible from their processors, and even so, it won't make that much of a difference. Generally its just for bragging rights.
I've already overclocked my 64 to a 96. W00T!
what about trying some 64 bits games and software, of course if you use *nix you do not need to worry about that, but windows seems to be trully lacking in that field.
Of course the real laugher is what the overclockers do with their "extra" cycles. Nothing useful, let me assure you. At least I've never seen a claim of utility. Moore's Law has given us cycles out the wazoo, and the overclockers are just silly fools, like the guy in the cheese shop with no cheese.
You want to improve the world? Write better software. God knows there is VAST room for real improvements there, and no help from Moore's Law.
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
discussions ranged on weather the A64 would even be overclockable at all.
/. editors :-)~
Or are they just edited by the
I have to say that those things are not overclockable. They won't even run at the default 204MHz bus speed. I've even tried three different brands of motherboards. They do run fine at 200MHz.
/proc/cpuinfo:
Important lines from
vendor_id : AuthenticAMD
cpu family : 15
model : 7
model name : AMD Athlon(tm) 64 FX-55 Processor
stepping : 10
cpu MHz : 2605.988
cache size : 1024 KB
anyone know how to overclock opteron 240s on a tyan s2875anrf? ive heard stories of people overclocking em, but no real evidence on the net. or maybe im not looking hard enough.
Overclocking is not something that is generally done professionally; it is a hobby, simply to show off and make one feel important. The same is true of 'tricking out' a motor vehicle, or modding a PC case, or the entire industry of do-it-yourself interior decoration.
When rendering, and presumably other activities that might theoretically benefit from increased performance from overclocking such as data analysis and science simulations, your are often leaving the computers working overnight or over the weekend, and the last things that you want are crashes or visual errors due to unstable hardware. Sometimes I even underclock my rendering system, for it is far better for a render to take a few extra hours or days than to have the whole render wasted because somethign went wrong with your elite hacked overclocking with ten percent enhanced performance.
Overclocking also reduces the life of the components, noticeably when they are rendering at full capacity nearly 24/7 for most of the year.
There are certainly professionals that overclock, but they have either carefully weighed the cost benefit ratios and decided on the most logical course of action, or have had a series major setbacks and mistakes and are desperate to finish before the deadline next monday; the boss will not be happy when he finds out that another project is late because of your bumbling incompetency, Jones, so you had better move right in to your cubical for the next week, or you will find yourself moving right out, permanently. You can have that little wife of yours bring you meals; I certainly would not mind having her around the office. Maybe she will finally see reason and bail out on that train wreck of a career that you are conducting, and set her sights up closer to where her standards should be. And if not, she will still be something nice to look at. Tell her to wear something that will cheer you up...Man, this is going to be a Hell of a week. Now get back to work, Jones.
If it ain't broke... overclock it?
can i overclock my amd k6-2/350 w/ 192mb of ram and a 4.3gb hdd?
seriously, this is my desktop running win2k pro sp4. its ran win2k3 server edition sp1. my linux box is a p2/400 w/ 128mb of ram. can someone overclock that too?
keep on because of a rumor that after the 4GHz barrier you can create a tear in the time-space continium... :)
I've taken down a few clocks off the walls here at home and put them under my system case. How may clocks does it take to see an improvement in system speed? I'm not seeing it. Do I need to put fresh batteries in the clocks to see the improvments???
Gives you another few months, you start thinking about that shiney new GPU CPU and stuff and salivating, but you know it's going to drop to .3 of the price in 2 months.
So you overclock. If you bought the low end last generation you can keep going WAY LONGER!.
I had a 9000 pro and was able to overclock to survive DOOM 3 and CS source... didn't need a 9600 pro or XT and wasn't tempted until the 600gt showed up... now I'm good for a few more generations unless it's another really awsome one (like the 9700 pro).
My A8V, AMD64 3200, oced to 2.4 Ghz, and 2x1G Ram, running Linux 86_64 and Mathematica 64, is TWICE as fast as my 2.8x2 Xeon oced to 2x3.15, 2G running Mathematica on either Windows or Linux. Not bad for $650 toy - MB+CPU+2G.
Benchmarks at: http://smc.vnet.net/timings50.html
It is also 10% faster than an FX53, 512M Ram, running Linux 86_64.
And yes, I care about stability especially when some calculations take a few days.
Overclocking also reduces the life of the components, noticeably when they are rendering at full capacity nearly 24/7 for most of the year.
Component life only seriously degrades when part voltage is raised more than a little, and even then, some parts are more forgiving than others.
If you're not doing any of that, it's free performance. If you are competent, why not, even if just a little (for a home user; obviously not on most enterprise equipment)?
As someone who does real-time generative audio/video processing, I have to say that oftentimes what a 2GHz machine cannot run comfortably, a 2.5GHz machine can do satisfactorily. 12FPS may not be good enough, but 15FPS may be passible. People often dismiss overclocking as something just for gamers, but in reality, it can be useful to anyone doing processor-intensive, real-time processing. I feel that some of the anti-overclocking opinions here are a bit unjustified, and more of a knee-jerk response to a loosely correlated l33t culture. The only games I play are Clan Lord and Civilization III... hardly a reason to overclock. However, for video processing, I need all the power I can get despite my modest budget.
"Is $150 for a lower end Athlon 64 really THAT EXPENSIVE?"
For an Athlon64 (939) yes. Plus about a $100 more* for the matching board. I believe it also takes a more expensive memory as well.
*Compared to say the Athlon64 (754) boards.
LOL.
The kind of person who, 10 years from now, when he gets his amazing new 200,000 GHz 512 bit processor with a terabyte of RAM, will say, "How do I overclock it?"
"However, for video processing, I need all the power I can get despite my modest budget."
The GPU as general processor story must have had you salivating then. Or maybe since you have a modest budget. You can pick up some Microway transputer boards off of eBay and go to town with them.
And here I got one of those new Athlon64s because I wanted to be able to underclock it.
It reduces stability, it decreases component life, and it increases power usage and heat. If you want to do it, I'm not going to stop you, but I'm not going to complain if Intel and AMD come up with a way to effectively prevent it, and I'm still going to think it's stupid.
I rarely criticize things I don't care about.
"yet another liquid cooling story is on the way!"
With all the beer geeks drink. I bet the next fluid to be tried will be urine.
Kids these days. Back in the old times, you didn't go under or over clock, you went through it!
Overclocking on the Pentium I was fun. CPU speeds were still far far away from the GHz levels and in school, it had amazing brag value. Never matter that the recursion programs we wrote in TurboC (it was way back ... I was a kid) never seemed to compile any faster.
c hmarks/super_pi/did calculate PI to 512K decimals in 49 seconds (It was 52 seconds earlier). Didnt make much of a difference to anything else that I use. (Am an MBA now ... what i use is powerpoint and outlook ... I sold out!!)
... the results are just incremental, but they do give the kicks. Very zen!
Overclocked my HP Athlon 2.2GHz upto 2.5 Ghz. Noteable difference? Well, super pi http://www.computerbase.de/downloads/software/ben
The fact remains that overclocking is not a performance enhancement
Remember the "Turbo" button on the machines those days?
I volunteer to come and kick Jones' boss' ass. Damn apostrophes...
I was just happy that I could plonk a NEC V20 in my IBM XT - going from a 4.77 MHz 8088 to a whopping 8 MHz!
I've always like to overclock my PCs - and I've never sacrificed stability. It's true that the hardware will fail earlier in theory as a result of overclocking, but unless you are running crazy timings with way too much voltage or whatever, the hardware is still likely to last until it's practically worthless. If you can overclock with just a modecum of skill, you literally get more performance than you paid for. I ask you: how can it be wrong, when it feels so right?
I think that overclockers tend to overlook the danger of data corruption when they experiment with overclocking. If you boot into Windows (or any other OS) with flaky CPU/memory, you risk corruption of any filesystems/data that the system works with while it is operating in an "unstable" condition. If the system would just freeze or reboot when the hardware fails, there'd be no problems with corruption (or evaluating stability), but... Just make a Ghost image of the filesystem before you experiment with timings, and restore that image after you are done with your overclocking adventures. Way less headaches this way.
It spins around it's own axis.
"its".
I am amazed by the number of people against overclocking. If you get a new CPU and overclock, well I would guess you are in the world where you are happy to play with hardware just to see what it can do. If it breaks it's your own fault and presumably if you are happy to try you don't mind loosing the money - if the cpu life is shortened do you care, probably not as you are going to be a early adopter of the next quickest speeds!
The Turbo button acctually made the CPU operate at slower speed. It was really a "Turbo Off" button:l
http://www.pcguide.com/ref/case/switchTurbo-c.htm
- Peder
Turn the power supply on.
Honestly, every one i've seen is so insanly overpowered, it isn't even funny...
On top of that, people will try to overclock a cpu when the problem lies elsewhere... RAM, drives, etc.
An 4Ghz 64-bit cpu is nearly worthless if you mate it with 64 megs of ram and a 3600 RPM laptop drive...
..for not supporting Linux on K8T800Pro chipset!
I've got Abit AV8 with amd64 3200+. Now, I'd sure like to see how far the CPU can go, but the damn motherboard does not show any temperatures. Gkrellm does not support that chipset, nor does lm_sensors.
Only a madman would try overclocking without seeing how hot the CPU is. Overclocking and hoping that my cooler can remove the excess heat is not very reliable way to do it, I'd say.
If you know a way to see the temperatures on AV8, let me know. I haven't found any information on that (except the pages that say it cannot be done).
"This in-depth overclocking guide will show you how to get the maximum from your brand new Athlon 64 system"
With a minimum of warrantee... to be exact NONE.
I'd like my processor to last for a while, so I don't take stupid risks for a little extra speed.
WOOHOO 20% higher clockspeed ! ;)
It gives 10% more realworld performance.
And 10000% higher failure/error rate!
Well having 100x higher error rate might go undetected most of time since errors in cpu are not that common without overclocking.
Anyway. I'm thinking that what the overclocking gives is minimal increase not worth the potential problems. Not everyone get problems, and not everyone who got problems realize them.
How you tell if the X server crashed by change of a bit at some place or by software failure if the error happens rarely enough? Or some other bit crashing.
It simply goes that they burn em and test them to fit at certain point. Few people get lucky because they need to mark some chips at lower speed than they are truly capable and they keep certain margins on the chip timings to ensure it works. But getting rid of those margins just gives a potential like.
If certain datapath gets used commonly enough, it gets hot, and it slows it down. The different bit patterns may show what kind of changes makes it hottest. Now those problems can be at ANY circuit location. It can be parts of alu, FPU, instruction decoder, cache. ANYTHING that just cannot keep with certain changing bit patterns when it becomes hot. And don't say that your cooling helps this a lot. Well the problem is INDIVIDUAL TRANSISTORS temperature not average chip temperature. And those transistors are becoming so small for dissipation.Plus it might be that the cycle time is not enough even at normal temperatures to keep up with worstcase changing bit patterns. Billions of different things CAN go wrong, with overclocked chip, and we don't know what it takes to burn certain paths. Then there is electromigration which worsens with voltage increases and so on. Basicly you are trading the lifetime and reliability of results for getting 10% more results.
I've though thats not worth it for me, I'm more happy with my A64 chip, staying at 1Ghz when I don't need its performance and powering up to 2Ghz when load goes up. And having PC thats more quiet than other sound sources around here, like flow of water trough the pipes for heater, is more important me than the bragging rights of 10% higher performance
Emacs is good operating system, but it has one flaw: Its text editor could be better.
I'm beginning to wonder if the pendulum has swung to making underclocking the smarter move. Certainly I've never had as stable, cold and quiet a machine as when "Cool 'n' Quiet" (on my MSI NForce 4) is kicking in (dynamically lowering the multiplier). You might say I should have bought a slower cheaper machine in the first place but just sometimes (DAW stuff, those VSTis can be hungry beasts) I need the grunt, but not all the time.
Plays violent online games as: Nerfherder76
Overclocking a processor does far less unnecessary damage to our air quality than driving an SUV or ferrari to work or soccer practice. Unless you destroy and replace a lot of processors this way.
CPUs are batch produced. before testing there is no mark on the CPU, their frequency are not determined. under strict testing, some will fail. then the remaining CPU will be marked to some certain frequency which they can handle, with little or quite some headroom. in case the market needs more lower end CPUs, some CPUs might get marked lower than what they could handle.
if U don't know these procedures, don't EVER talk about overclocking, or undervolting, or anything.
with PC3200+ RAM on a SIS(AFAIK lowest price chipset there is) based mommaboard.
Running CPU at stock 2.2 GHz and mem(HTT) at stock 400MHz.
It's unbelievably cool with 42 degrees C reported after sustained 92% user 8% system operation.
If I understand the article, the optimum OC path is to keep the HTT at 400 by upping the CPU clock and upping the divisor, unless I want to worry about all the other system clocks and peripherals running beyond speced speed. If this is correct, let me know and I'll have a look at BIOS to see if this is possible with this extremely inexpensive motherboard.
The current redhat / centos kernel out right now has a bug that made my functioning temps unavailable on my office's quad opteron server. It's a known bug, just wait for the next release.
SearchIRC - Now with live chat directory!
Okay, maybe not the most friendly subject line I could come up with, but you should hear yourself, all of you...
Last week, I bought an A64 3000+ with Venice core and a Neo 4 Platinum motherboard. A 3000+ runs stock at 1.8GHz. I bought this core with the very purpose of overclocking it. And well, it did damn well. This CPU is now running @ 2.65GHz, without upping the core voltage, without dangerously high temperatures. Actually, I will be buying a better cooler one of these days to keep it cooler that it runs now, and possible to run @ 2.8GHz.
Why, you ask? Well, why do people buy an A64 4000+ ? I can use it for games, for compiles, etc. And don't be telling me that there is no measurable difference between my processor @ 1.8GHz, and now @ 2.65GHz. It is blazing fast, as fast as when I would have spent 500 or something for a 4000+.
Oh, and you can throw whatever stability test you want at it. It is rock stable.
And why would it reduce the lifetime of my CPU drastically? My now retired 1700+ has ran its entire life @ 2GHz (up from 1.466GHz), and it will live on in my girlfriends PC.
Also, what do you think is the difference between a 3800+ @ 2.4GHz and an overclocked 3000+ @ 2.4GHz? That's right, ABSOLUTELY NOTHING. All the current A64's do 2.4GHz anyway, so speedbinning is virtually non-existant, the processors are the same.
The higher HTT? Oh come on. And everything (PCI-E, PCI, whatever) is running at standard speeds, even with an overclocked system (well, except for the CPU ofcourse, and the HTT).
Talk all you want, but it is VERY clear that loads of people here don't have any clue as to what they're talking about.
Bit of brag...
Athlon 64 3200+ (754) retail box which included cooler.
Motherboard, SIS based but hey, they all build to the reference.
1gig RAM
160 gig 7200RPM SATA drive, WD but for $50...
New power supply
total $370 including tax at Fry's after rebates.
put it in a surplus 4 U rackmount for FREE bandwidth via stealth install on internet at local biz(with managers permission)
Do not know if motherboard will allow overclock but it runs so cool at 2.2GHz that next excuse I can find for a reboot I'll have a look at BIOS options.
Underclocking is as stupid in that situation as overclocking. The biggest stability issue with clocking apart from lack of voltage and then heat, is improper support for your CPU because you told the software that it's something it's not.
Although I have always been an Intel man, the good words I keep hearing about AMD has me considering building my next system with a CPU from them. But frankly, I don't know anything about AMD.
Is there a website that spells everything out for newbies? I'm talking about information for all the different AMD chips and how they compare, socket types; the sort of thing. (The sites I have seen so far seem to assume that you are already knowledgable about AMD.)
I am building a A64 system - should say built a system and are now troubleshooting. The BIOS will only recognize about 3.4G in a 4x1G configuration.
I'm running a SLI-DR board (rev AA0) with 4x1G of RAM and a revision E (Venice core) AMD64 CPU. The BIOS only recognizes 3,407,334 KB RAM. I've tested each stick, ran them in pairs, and tried them in another machine - so I don't think this is a RAM issue. I will be dual booting, but 64-bit Gento and Win2K (not moved to Win64 yet, but will when the MSDN kit shows up later this week) but neither recognize all the RAM. I've tried a couple BIOS versions - the latest, the latest beta, a couple hand rolled builds I found here - but no joy. Memtest-86 shows 3328M cached, 257M reserved. Add that up, and it puts me almost exactly 512M short of what I expected. I do have a 256M PCIe video card, but that is all the extra peripherals added in so far. BIOS reports all the RAM if I go 4x512 or 2x1G.
Another annoying thing is the board clocked down my DDR 400 RAM to DDR 333 and 2T timing when I added it in pairs. Granted, not the most uber leet stuff out there (CAS 3, with 3-4-7-4 stock timing, running 1T) Bumping it back up to 200mhz (DDR 400) from 166mhz (DDR 333) count as an overclock?
+++ UGUCAUCGUAUUUCU
You can overclock it in software using -funroll-loops and -O9.
So why are so many of you Fear, Uncertainty, and doubt filled computer newbies on here trying to bash overclockers?
Most of you dont know what youre talking about in the first place.
In the overclocker community "24 hour Prime95 stability" is the benchmark against which all overclocks are measured.
"It's not stable unless it's prime stable."
My compaq pos at work couldnt even pass prime95 for 24 hours..but my amd64 3200+ @ 2.8 ghz home machine sure did.
Being ignorant about somthing and talking bad about it, is just about the same as racism.
Now sit down, shut up, and let the grown ups talk.
"Comedy's a dead art form. Now tragedy, that's funny."
The thing is, enterprise hardware is typically overengineered to handle extra heat+whatever that would come from overclocking, while, the average home user might try to use a stock heatsink+fan that defintely won't cut it...
Have you ever known Athlon 64 that was sold and the advertised clockspeed was 1Ghz? My chip is 2Ghz chip which runs at 1ghz when I surf the net. But when I look at AVI it goes to its nominal speed of 2Ghz.
Emacs is good operating system, but it has one flaw: Its text editor could be better.
Allright I couldn't let this one go past.
Overclocking is BRILLIANT, I mean you take a stock part from the manufacturer, they never quite clock all of their chips as high as they can, and they cut corners. So what ya gonna do? Crank it up a bit and get more performance for free, right??
Wrong. Theres a damn good reason why people who have something SERIOUS to do with computer hardware dont overclock their systems. Why?? Because you dont get extra speed for nothing. It's running faster - but - it's GETTING THINGS WRONG!
Thats right. The error rate in the calculations your system's pushing goes up. By the time you notice BSODs and games crashing, your PC is working so far over what its capable of, that its fricking retarded.
Manufacturers VERY rarely give away free chips. So what if your processor is a 2500+ and its got the same core as a 3000+... Know why its clocked slower? It's not cos the manufacturers lazy. They designed a chip to do 3000+, they they produce them by the truck load. Each chip gets tested, if it doesnt pass at 3000+ clock rates, they'll lower it, until they find the speed it can work at.
Manufacturers arent stupid. They want money. Why throw defective (not as high quality) cutting edge chips in the bin if they can be sold at a slightly reduced speed? Thats the old step 3, proffit.
SO yes you might get that 2500+ and ramp up the juice and it claims to be running at 3000. There is a slim chance it failed the testing and that was a one off. Maybe. More likely your getting reduced overall ability (speed vs correctness) and you just arent pushing it hard enough. Try compiling source code on it for a couple of days and look for random bail outs. Try complex rendering that takes weeks. If it works, great, you got lucky.
What stops you from OCing, then reading the BIOS temp monitor after its been sitting a while. When its in the BIOS the CPU temp usually rises enough to see what its like under load.
ABit motherboards come with their own auto-overclocking utility that works like a dream? My former roomate has one, and I've seen the results. Quite impressive. Also, you can override the auto-overclock and manually set it yourself, AND monitor your computer at the same time. The Guru software also can auto or maunally overclock your nVidia graphics card.
ABit is on the ball on this one, I have to hand it to them.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
people will try to overclock a cpu when the problem lies elsewhere... RAM, drives, etc.
My thoughts exactly. Show me a computer with "overclocked" RAM or HD, and THEN I'll be impressed.
I have two computers, at 500 and 600 Mhz each. The CPU is rarely the bottleneck. More often it's the slow HDs, or the fact that I'm on dialup.
BTW: I do all my gaming on my PS2, so I haven't really had the urge to upgrade in a long time. It takes about 10 seconds to compile programs I write, and that's fast enough for me.
"That's so plausible, I can't believe it!" - Leela
For the first time, I have a system that I don't feel overclocking will help. Applications start instantaneously, games run smoothly. At the moment, CD and hard drives seem the bottleneck. In short, my Athlon64 is ***fast enough already***. Granted, I don't think this will last too long, but it's been quite nice for the 9 months I've had this box.
Granted, if I was running an antivirus suite, I'd probably welcome overclocking...
-- I prefer the term "karma escort."
MemTest86+ helped me solve a memory voltage issue (my mainboard defaults to 2.5V on RAM, but it gets errors until I hit 2.7V). Also, scaling back my RAM (a generic DDR400 Samsung 512MB) to DDR333 lets me have lower memory timings (2.5-3-3-7) & overclock my HT to 220x4. My Athlon 2800+ is running @ 2.16GHZ: a 20% overclock.
http://www.wolfsheep.com/technical/rocko.html
Life is irony, and nothing ever goes as planned.
All the chips come off the same assembly line.
They first test all the chips at a "safe" configuration to check for clearly defective chips (usually about 50% when you talk about 90nm process).
Then, they test as many chips as necessary at a speed margin above the rated speed to fill in the "speed bin". As soon as this quota is met, they run the rest through a test at lower rated speed.
Since they go from high to low, it's quite conceivable a chip that was not reviewed to fill a higher speed quota will fill the quota of a lower tested speed.
Testing takes time, so they don't like to test chips multiple times... they will test them the minimum times necessary to ensure all speed rating bins are filled for a batch of chips.
Hence you get chips rated a lower speed which were never tested at a higher speed (wasn't needed to fill predicted demand)... and a good chance of overclockability.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
I have money to burn... literally.
This signature is part of a balanced post.
I've got my Model 1 running at a butt-kicking 2.1 Mhz! Yeah baby! That's, like, over 18% faster to run the Folding client.
I'm thinking of upgrading to a Z-80B, so I can really rock and roll.
Now sit down, shut up, and let the grown ups talk. Right... Being ignorant about somthing and talking bad about it, is just about the same as racism. Erm, no, having a dim view of overclocking is not "the same as racism". I think its disgusting you are trying to equate a computer hobbyist issue with someting that has blighted millions of lives. It doesn't say much for your maturity either bucko.
Up untill my cheep ass PSU killed my previous system (yes, I know, my bad) I had overclocked my 1.6Ghz AthlonXP 2000 to 2Ghz stably, I had a watercooler from a previous setup which was forever overheating, so I installed that, the result? A cpu that idling ran at 35C and under load at 39C (after overclock, though the temps were about 1C lower in each case prior). That speed boost for most things meant that windows booted up a tad quicker, and was slightly more responsive, but when gaming, the effect of theoverclock really made itself known - playing things such as Call of Duty, whilst perfectly playable before, had a certain amount of smoothness that simply wasn't present before.
Had I not been such a tight wad with regards the psu, I'd still be using that system. As it is, I'm about to build myself an Athlon64 system, sadly the watercooler won't be used this time around - it was socket A/370 only, shame. I also suspect that that chip had the potential to go much faster also, but since the overclock was achieved purely through fsb (the board had no provision to change the multiplier, which was locked in any case) I was stopped by running out of rungs on the motherboard.
regards, the_leander