Overclocking The AMD Duron

JellyBeansOnToastWithStrawberryJamonTop writes: "I noticed in Sharky Extreme and Tom's Hardware coverage of the AMD Duron processor that they both have details on the overclocking capabilities of the processor. Looks like the new Celeron 300A, eh?" It's cool to see that AMD has not locked out overclockers with their new chips. But where are the dual Athlon motherboards, please? Updated 3:19GMT by t: Apropos of overclocking, check out Feedmag's not-uninteresting take on overclocking culture.

Comments

Nuclear Meltdown?

[Amoeba Protozoa]

I hate to sound coy but these processors should maybe not be called, "durons," as they consume a lot of power, which makes me really wary about the lifespan of these products due to the extra heat; less actually overclocking one of these suckers without a refrigeration system. Here is a quote from the sharkyextreme article:

The CPU is made of 25million transistors and takes up 100mm^2 of die space. In comparison, the Thunderbird uses 37 million transistors and takes up 120mm^2 of die space. Power comes in at 1.65V at a maximum of 25A, which means the Duron consumes up to about 41W of power. That is over twice the power consumption of an Intel Celeron 600MHz, which only consumes 18W. It is also a bit over half the power consumption of the AMD Thunderbird, which consumes up to about 79W... yikes! We will discuss power usage more a bit later on.

Sharkyextreme goes on to mention that the 16k L1 cache size may attribute to the extra power consumption-- but for a new chip with little air-time it makes me worry.

Re:Hi, I have a question

[Perdo]

Here is a price list from sharky's. It includes shipping costs. We overclock to get a 950mhz processor for 159 bucks.

Duron 600 $92 Spartan Technologies 888-393-0340

Duron 650 $125 PC Progress 888-727-7647

Duron 700 $159 United Micro 800-943-7255

Athlon 950 $581 Econo PC 888-326-6660

P3-933/133 $777 Atacom Inc. 877-228-2266

If you burn two processors shooting for 950mhz you still have not spent 581 bucks for a "real" 950mhz processor.

For those of you wondering about multi processor setups, imagine using two 950 Durons with mo-bo for less than the price of a single "real" 950 without a motherboard.

Re:OFF TOPIC

[Cliffton Watermore]

Hmmm:) Unless you're running it on a new Duron, I don't really know what this has to do with the topic and I'm surprized you haven't been moderated down already, AC. Still it seems like a genuine question.

I'm interested in the Duron, btw, and am going to try and run Solaris/Intel on it. Linux is very cool, but unfortunately neither myself or Johnny, our compiler expert - can compile our custom written software for anything other than Solaris. We still have hiccups with Solaris/Intel/cc! Anyway, it will be very interesting to try it with the Duron, I guess we'll need an externel compiler person to help us out though .

As I said though, your question seems sincere, so here goes:

I have Linux (mandrake 7.1) installed

Aha! The latest Mandrake.

I have a Linksys PC Card nic (ec2t) in it. I can't get it to see the other machines on my home network, or to even ping them, and I am really not even sure it is working (works ok when windows is installed).

Ok, I don't have much experience with PCMCIA cards under Linux or in general, but what you need to ensure is that PCMCIA support is compield into the kernel.

Go to /usr/src/linux

Type "make menuconfig". Configure everything appropriately, under networking, do add support for PCMCIA. Save the config by "Exiting", then type "make dep". After that, type "make bzImage". Copy the resulting bzImage from /usr/src/linux/arch/i386/boot into /boot, and edit "/etc/lilo.conf". add an entry for the new image. Once you have that going, your PCMCIA should be detected when you boot the new image.

I don't have any experience with Apples, but there's definitely a way to share drives through SMB (the protocol used by Microsoft for file/print/etc sharing, including Network Neighborhood). You'll need a package called "Samba". If you don't have it, you can download it from samba.org. Once this is installed, you will edit the /etc/smb.conf. Ensure that the domain/workgroup is the same as the one on your Windows machines. Also ensure that your Windows machine name has a Linux user account with the same name, then connect to the Linux machine using it. If you can't connect but do see the machine in your network neighborhood (the Linux machine from your Windows hosts), then you'll need to download the Windows 98 Registry Hack to Enable Cleartext Passwords. Perhaps with the latest Samba they've got built in support, I don't know.

Once all this is running, you'll need the latest Kernel. Get it from www.kernel.org. It has all the latest stuff in it. I'm running it, 2.3.99pre9. To do this, go into your /usr/src directory and type "lynx www.kernel.org". Scroll down and press enter on the "2.3.99pre9" link and type "D" for Download. Once you have it, click "save". Press "enter" for the default directory, /usr/src

Now, type "tar -zxpvf linux-2.3" . Now, proceed the same way as with your 2.2.15 kernel source code. Don't forget to backup your kernel images, btw. You should have a failsafe option in Mandrake 7.1 anyway. That's about it.

Good luck

.

Change Mhz while running

[dam_ned]

I had an idea, when you want to overclock, you got to reboot everytime you change the speed. Perhaps we should change this and make it possible to change speed using buttons. Remember those good old days with those 'Turbo' buttons.

Re:Change Mhz while running

[Idaho]

This is exactly what the Crusoe does, even automagically (depending on how much work it has to do).

This results in massive energy saving abilities in case of the Crusoe, but am I the only one wondering if it can be used for overclocking too? After all, the control of the speed is software based, and therefore it should be possible to hack it, methinks...:-)

There's something i don't understand

[fluxrad]

It seems to me that overclocking has often been frowned upon by a lot of people. It seems that no one comes out and say "you damn well better not overclock a-hole!" but it's a silent thing....you just get the feeling that some people are of the opinion that 'overclocking is stealing processor cycles'

Idunno - i'm glad that AMD is still making their chips 'overclockable' - or at least isn't denying anyone the ability to do so. That, in addition to the fact that AMD has been bitchslapping Intel in the processor market lately, makes me even more brand loyal than i was before. I'm on my third AMD (K6-2)...and i'm not about to switch any time soon!

VIVA LA PROCESSOR RESISTANCE!

off-topic: has anyone notice their karma decreasing for no reason since the new /. code was put in place??


FluX
After 16 years, MTV has finally completed its deevolution into the shiny things network

Re:There's something i don't understand

[fluxrad]

not so sure about the self-destructive behavior.

overclocking is the technological equivalent to putting a turbo kit on your car. The same equivalents can be said for new exhaust systems, new cylinder heads, tranny, etc. No one is accusing these guys of being self-destructive even though they lower the life span of their vehicle's by up to 100k miles or so. Of course, rather than beef up your car, you could just buy a newer, faster one for more money, but you don't get the same satisfaction and you certainly don't get the same cost to performance ratio as you would by fixing up a slightly older or slower vehicle.

When a guy fucks with his car to run a quarter mile in 13 seconds...that's initiative. When a guy fucks with his processor to make it 50-100Mhz faster, is that self-destructive??? I think it's frugal

Secondly, who the hell are you to tell me that just because i sniff glue i'm less of a person than you are. Of course sniffing glue is fun...but...er....wait...i better go...


FluX
After 16 years, MTV has finally completed its deevolution into the shiny things network

Overclocking and distributed.net.

[Lita Juarez]

While we're on the subject of overclocking, I wonder how many people are running a distributed.net client on an overclocked machine, against distributed.net's wishes? This piece from the distributed.net states clearly that distributed.net do not like people running their client on overclocked machines, since there is a possibility that the overclocked processor could produce an incorrect result, and thus jeopardise the entire project. The people who run distributed.net clients on overclocked boxes are very selfish, as they do not care that their actions may mean that millions of hours of other people's processor time may be wasted. Considering all the processor cycles I have donated to distributed.net, I would be furious if I found that the project has been spoiled by some kiddie running the RC5 client on an overclocked box.

I'm not going to directly say that overclocking is wrong, but I think that overclocking is a childish pursuit and for many people it is nothing more than a dick size contest. Why not just go out and buy a fast processor? Considering the cost of some of the cooling devices that are used (Peltier coolers etc), it would probably be cheaper to buy a fast processor in the first place.

--Lita (member of Team Slashdot)

Re:Overclocking and distributed.net.

[jidar]

And as a longtime overclocker you know all of this personally right?
This is total BS. An overclocked processor isn't significantly more likely to create an incorrect result in RC5 than one a normally clocked one. If something fails it will fail the same way a normal CPU will, it will screw up enough stuff such that the client wont function at least, and probably bring the OS to it's knees too.
As for the Dick size contest, that is sometimes true, but not usually. I Overclock because I make $8/hour (in other words, im poor) but I still like to play Quake competitively online. A Celeron300A cost me ~$100 when I purchased it and I achieve framerates in Quake2 which are identical to the framerates a p2-450 gets you. This was -without- having to put any money into extreme cooling measures, nothing but a stock heatsink and fan. At the time I purchased that Celeron it cost me about 1/5th of what a p2-450 would have cost, and by overclocking I got the same performance.

Stability? Rock solid, and over a year later, its still going strong.

I plan on getting one of the new flipchip Celeron 533's for about $100 here in the next week or two, drop that into a FC-PGA to Slot-1 converter ($20) and according to everything I have read it should do 800mhz with stock cooling.

And again.. why am I doing it? Because I'm poor. And yes, I have a small penis, but that is unrelated.

Re:Overclocking and distributed.net.

[darkstar-x]

Overclocking is not a matter of a dick size contest as such.

its getting your _true_ value for money. sure, i can go and get a AU$2000 cpu, but tell me why i should? when a) they are extremely hard to find, and b) with overclocking, i can get close to, or the same speed as that $2000 chip, for only $400 + 75 for a cooler.

see if you can justify those costs.

also, tell me what is rc5 trying to prove anyway? yes, i am a very mild contender in rc5, but it doesnt exactly hold any scientific value. seti@home for example, is where the consern should be.

since you say overclocking is a "how big my dick is" _contest_, what about rc5? i get nurmorous amounts of people saying quotes like "ive got such and such blocks done, im so leet, ph33r my hardcore skills"

overclocking is a childish pursuite? it maybe to a constricted mind, like yours, but if it wasnt for the remarkability of overclocking, the celeron range would be scarce, dead, or simply not preforming as well. the intel celeron got popular, and made intel money because of overclockers.

As for you knocking peliter coolers, overclocking for these people is a hobby. they are proving a concept to themselves and going beyond whats nessasarily required, needed and so forth. take the 1ghz $2000 cpu for example again. remember ho w my cooler and 600e cpu, costs me a total of 475. Ad a peliter to that, and hold on a second, whats this? only an extra $50. a total of $525 to get the preformance of a chip that costs 4 times as more.

A person wants to by an old car, lets say a classic or fairly moden, etc... and do it up. why dont they just go buy a new car? simply because THEY DONT WANT TO. there is no enjoyment going and buying something that already preforms well, when building something, creating something to be proud of is more valuable, both in price and emotional attachment.

Take hacking (NOT cracking) for example, now why dont people just learn the minimum required? Because they feel as if learning the minimum is stupid. i bet you many people can get by with out learning a programming language, but they learn it because they are intrested in the benefits it holds, and they are expanding thier knowledge.

next time when you oppose something, try to research and examine carefully as to what you want to get across, and what your knocking.

oh, just for the road, another analogy for you. any car can go to the required road limits. why do people buy a car that is more appealing to them, buy a peice of machinary that is somewhat exquisite? take your time to answer, and see if you can realise that people have free will, if they have an ambition to gain a certain status/value and so forth, then they would do so with a strong passion.

if it wasnt for overclocking, i doubt that processor companies would be where they are now, hell even amd overclock their processors for the high end athlon classic range.

--
----
"What you SEE is sound, what you HEAR is the Desfness of Society" - darkstar.

Re:Hi, I have a question

[oingoboingo]

This is equipment that is expensive enough for most of us to be somewhat annoying at the very least should we wreck it.

while i'm not one of the 'hard-core' overclocking crowd (you know...building systems that are cooled by a bath of refrigerated mineral oil), i've overclocked every system i've ever owned...and never had an overclocking related failure.

admitedly, my efforts aren't all that adventurous, but i've had a P200-MMX to 250MHz, a K6-266 to 300MHz, a PII-400 to 500MHz, dual Celeron 366s to 525MHz (my home system), and a PIII-450 to 600MHz (my work system). they all work great and are just as stable as the 'spec' versions.

why wouldn't a user want to get extra performance out of their systems? i think the risks of damaging CPUs by overclocking are are a bit overhyped.

AMD vs Intel

[andri]

Actually, this overclocking race shows us how far the companies stand from eachother :P

When Pentium II 300MHz and Celeron 300A ruled the known universe, they had a lot of slack. One could put it up to at least 450MHz without any hiccups, quite often even higher.
At that time, AMD was having a hard time. Their K6-2 sucked, it is a well known fact that they don't go up as easily as Intel's CPUs...

So, Intel was WAY ahead of AMD in the MHz race, because it could introduce CPUs with higher speed more often than AMD.

Now, the tables may have turned. Duron 600MHz goes up 350MHz (to 950MHz), has a 200MHz FSB et cetera, while Intel Celeron can barely reach the same frequencies - 733MHz -> 1GHz, that's only 266MHz...

So, AMD is holding the MHz race flag. For how long? This competition thing, I rather like it :)

Changing the multiplier

[JabXVI]

I don't think Sharky's review mentioned this, but according to Tom's article, Asus will soon be releasing a Socket-A board with jumpers and dipswitches for changing the multiplier! This is the board Tom used for overclocking four Durons to 950MHz, all of which were stable he said. More motherboard manufacturers will most likely do the same as Asus.

This is very exciting, because with AMD's Slot-A CPUs you have to open the Athlon's case, and do some soldering (or you can buy special cards that attach to the Athlon for the purpose of changing the multiplier).

With a Socket-A motherboard like Asus', overclocking the Duron (and Socket-A Athlons) will be a piece of cake. Now we just need some SMP Socket-A boards... :)

Overclocking is frowned upon

[linuxonceleron]

One day I was with a few friends, and we were in Philadelphia, and we walked by one of those neighborhood asian computer stores. At the time I was looking for a new CPU so we stopped in. I asked the woman behind the counter, "Do you have any Celeron 300As, you know the ones that can overclock?", to this she responded, "no overclock, no overclock" while waving her finger back and forth, after that we got thrown out of the store. I thought these places would be pro overclocking, but I guess not, I didn't want to start an international incident, I just wanted a new CPU!

Re:Hi, I have a question

[jidar]

This is an easy one to answer.

"Is a couple hundred dollars so hard to come by that we are unwilling to spend it for the peace
of mind and stability of a processor we know is not about to overheat?"

The answer, yes. Newsflash; a couple hundred bucks is rich man territory if you look at the state of humanity on a global scale. Hell we don't even hav eto leave the good ol US of A. You have even sat down NEXT to people in restuarants who consider a couple hundred bucks quite a bit of cash. Amazing huh?

That aside, the truth is, we get more performance at -significantly- reduced cost, not just a few hundred dollars. Those Durans that are doing 950mhz are $100 CPUs and outperform 1000mhz Intel chips. :P

Dual Durons "later this year"

[Meerkat299]

In the conclusion of Ars Technica's review of the Duron, Loki cites a couple of articles reporting that SMP support will come out for the Durons later this year:

On the SMP tip, the Duron according to this news post at 2cpu.com will support SMP. This shouldn't come as a big surprise since it's built on the Athlon core. The SMP-ability would have to be specifically disabled, which is likely more trouble than it's worth. Later this year, AMD will be releasing the AMD 760MP chipset which will have ATA100, DDR SDRAM support, and SMP support. I'd love to see a board that I could start on dual Durons and kick up to dual Thunderbirds later on.

HTH

Off-topic, but, well, it's Saturday...

[Wah]

I have a theory on the correct clock speed to recommend to friends and family when they inevitably ask me what they should buy.

It goes like this. Take the fastest speed you've seen, divide by two, and get plenty of RAM.

This allows for both their interest and investment in technology (if the fastest speed they have seen to date is 600mhz, they don't need to spend $1,000 on a box of metal.). The RAM should be plentiful, depending on the day. NIC for sure. The peripheral should be based on need...and that's about it.

OS....do what you can, recommend Be if it works.
--

Differences between Dr. Tom and Sharky Extreme

[aabrown1971]

First off... This thread is supposed to be about Dr. Tom's and Sharky Extreme's articles, not about the overclocking debate. If you do it, great! If you don't, *who cares*! I've been running OC since I bumped my 486 DX2-66 up to 80Mhz, and have _never_ blown up a chip. Nowadays, with temp monitors, and screeching alarms, you'd have to be a complete moron to blow up your chip!! Ok, enough of that. Did anyone else notice how the two articles seemed to have completely different takes? Dr. Tom seems to consider this chip the new holy grail of price/performance, and Sharky seems to compare it more to the 66Mhz bus celerons. Also, did you notice how Dr. Tom had all chips running at 950Mhz with no problems, and Sharky couldn't get past 770? I wonder what the deal was there? Maybe sharky did not want to bump up the voltage. Hopefully, we'll get a Duron to test at Redhat, but until then, I'm crossing my fingers that Dr. Tom is correct (as he almost always is!). .95Ghz for $112! Awesome!

Re:Differences between Dr. Tom and Sharky Extreme

[Pont]

Actually, I heard that the multiplier is locked on Durons*, so the only option is to overclock the FSB. Therefore, it is valid for Sharky Extreme to do their overclocking tests in the way normal people will have access to.

Review sites often get engineering samples, which are not clock-locked, which is probably how Dr Tom Pabst got his Duron to 950. Also, Dr Tom pulls more weight than Sharky in OEM and IT circles, so if I were AMD and I had two engineering samples, I'd probably send the better one to Dr Tom too.

*OCI claims to be releasing a Socket A motherboard that lets you change the CPU multiplier on any Duron/Thunderbird, but most won't be able to. With Athlon Classics and Slot A Thunderbirds, you have to use a GoldFingers device to change the multiplier (or a saudering(sp?) iron). It's kinda hard to stick a GoldFingers onto a socketed CPU.

Everyone always says that speculation isn't worth shit, but I'd really rather someone gave me their speculation than their shit.

Cooking food on an AMD Mustang (funny images)!

[antdude]

Speaking of overclocking AMD chips... Check this thread out: http://discussions. hardwarecentral.com/Forum11/HTML/008273.html :)