Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
AMD

Athlon Overclocking - The AfterBurner 193

NoWhere Man writes "Over at RB Computing (an AMD-only shop in Ottawa, Canada), they are distributing the AfterBurner, an Athlon Overclocking card, developped by Golden Fingers. It offers on-the-fly frequency and core voltage modifications, that is a reasonable alternative to building your own, as shown at Tom's Hardware Guide. "
This discussion has been archived. No new comments can be posted.

Athlon Overclocking - The AfterBurner

Comments Filter:
  • Is this really cost effective? I mean, you could probably get a better performance/cost ratio by SMPing two processors, each 3/4 the speed of the one processor you were going to overclock.

    Or, am I missing something?
  • Well, we all are pretty much aware that Intel and AMD are having a speed war. It seems just about every few weeks, they push their chips faster to claim to have the fastest.

    I know that the AMD chips that are currently coming out overclock pretty well, but I'm wondering if the new "Merced" or whatever Intel is going to call it now is going to be so responsive to overclocking. My friends claim that P2s and P3s are pretty crappy for overclocking, but Celerons seem to be good. Is the new chip from intel going to overclock well?

    "You ever have that feeling where you're not sure if you're dreaming or awake?"

  • I spoke with the owner, and he said the afterburner works quite well. He actually sells systems pre-configured with the device in place.

    Rock-on, but won't this (can this?) hurt AMD's sales for the high-priced, high-speed chips?
  • There's nothing to not be compatible with as there is no software involved. If you're asking if you can run Linux on an Athlon, then the answer is yes. -ctepher
  • Since it requires no software support (even the hardware dosen't know it's there), it should work with linux, any of the BSDs, windows, V2OS, Mach, DOS, any of the Windows (including 1.0), CP/M, App-specific OSes such as Kings Quest 1 and 2, and Zork 1-3, and any other x86 operating system you'd care to mention.
  • by Fjord ( 99230 ) on Friday January 14, 2000 @02:07PM (#1370760) Homepage Journal
    Here's the Hard OCP [hardocp.com] review of the Afterburner (from the link).
  • Well, I don't think any SMP Athlon motherboards are out there yet. This is something you can do right now.


    ---

  • i doubt it but he could be talking about the following...


    from http://www.rbcomputing.com/ab/afterburner.html
    UPDATE: Word is, software will be coming available to adjust the L2 divider through software shortly. RB Computing will keep you updated!
  • There are about a half-dozen of these adapters available for sale. They range in price from $20 to $125 depending on options (voltage tweaks, build quality, etc.).

    The hard-wired multipiler lock of the Athlon (and awful motherboard support, you can't argue that) were the only reasons keeping me from upgrading to an Athlon. Instead I'm using a Coppermine 500E and I have it overclocked to 700MHz (5.0*140) on a Soyo 6BA+ III motherboard with an IWill Slotket II.

  • by DiningPhilosopher ( 17036 ) on Friday January 14, 2000 @02:10PM (#1370764)
    The cache divider is not controllable through the Athlon's edge connector. In some cases, higher overclocked speeds may be possible by changing the cache divider from its default setting of 1/2 to 1/3.

    I'm no overclocking master, but are they suggesting you cut the L2 cache speed from 1/2 core to 1/3 core? Why on earth would you do that? Let's say your core frequency is 800 MHz, and your L2 runs at 400. If you overclock it to 900 but your L2 cache is only running at 300, surely you're getting worse performance overall than you were before...

    Is this just another example of the blind worship of the almighty MHz? I think this is the first time I've seen anyone sacrifice performance for higher core processor frequencies...

    Or have I just forgotten everything from my architecture class? :-)
  • AFAIK, the biggest bottleneck in SMPin' is the OS. According to my blurry recollection, BeOS is supposed to be the most efficient at utilizing the multiple processors, NT does a half-decent job, and Win9X doesn't do SMP at all. I don't know anything about how well Linux does, but I suggest you do some more research if you want hard numbers.
  • The people who really benefit from high clock speeds are the 3Dgamers. AFAIK there aren't any games that take advantage of SMP.

    ID was looking into it, but I don't believe they shipped the hacks with Q3A. Anyone know what happened with this?
  • I thought about getting one of these things for my Athlon 700, but the idea of taking the plastic casing off the cartridge always scared me off. I'm not generally all that squeamish, but we're talking about a pretty expensive CPU here. It'd really suck to screw it up. I wish AMD would provide some sort of safe access to the pins for overclocking.

    Cheers,
    Perrin.
  • I've had my Athlon 550 for a few months now, fully-loaded with the ASUS K7M.

    Just about every computer store in Ottawa (Canada) has the K7M in stock. Even the onboard VIA audio just recently got inserted into the Linux Kernel (although I prefer the good ol' Sound Blaster).

    I fail to see how the motherboard issue remains.
  • $125 for three switches, an Rpak, and a connector?
    Wow, that seems steep considering the cost of the
    parts.

    I've modified a couple of Athlon 500s (which both
    turned out to actually be 650's based on the legend on the chip itself) to overclock at 750.
    It just involves moving a few SMT resistors. This
    board just gives you easier access to selections that are already possible.

  • The Athlon has 128k of L1 cache, so reducing the the L2 cache speed may not have as much effect as on a PII/III chip which has 32k of L1 cache (?).
  • http:/ /www.siliconinvestor.com/stocktalk/msg_multireplie s.gsp?msgid=12564826 [siliconinvestor.com]

    Seems to be rising. Anyone with other numbers?

  • You are not nessecarily going to be getting worse performance. How big of an impact cache speed has on your application depends on the application and the circumstances. Quake2 for instance is not much impacted by the cache as it is the raw power of the cpu and as such would see increases in framerates with this method. It all depends on what you are doing.
  • It is much better to start with a slower chip, say a Athlon 500, and then clock it up as necessary while lower the cache divider.

    I bet a fair number of 500Mhz Athlon cores can do 750Mhz but are held back by the cheaper L2 cache they run. Getting a 750Mhz chip at the cost of a 500 aint too bad, even if it isn't as fast as the "real" Athlon 750s.

  • I tend to agree with this, but public perception is based on the higher MHz. It doesn't matter that you're using a 3600 RPM 15ms drive, or 16 MB of RAM (ugh), as long as it's equipped with an 800MHz Athlon, people think it's a rocker of a machine.

    The amount (and speed) of RAM as well as the speed of the HD and accompanying bus are equally important. Otherwise, Linux (or Windows 2000) will load just as slow as a Pentium 200 with 128 MB. But most people ignore this.

    The almighty MHz. Some people's friend. Some people's ennemy.
  • You still can't run Windoze over DR-DOS on it.

  • I read Ars Technica and Toms Hardware Guide often enough to know that it's all fun and games, and there are clans and cults of overclocking out there, etc. And if I wanted to, I could get into it and build myself a monster machine.

    Well, I don't wanna. I'm too lazy. I've got this aging Pentium Pro/200 system that's slow by todays standards, but which has served me quite well, and actually I just don't wanna mess with hardware anymore.

    This doesn't mean I don't want to *own* the fruits of such activities, though. I'd love to have the absolute screamingest machine that a couple G's could buy, and I'm sure there's someone out there that would be happy to provide such an elite box o' power for a small price.

    Point is, does anyone know of any companies that build these sorta monster boxes, or is it just better to go with a good quality hardware vendore like VA Systems or something like that for my 'leet hardware needs?

  • I was JUST reading this today. Looks very interesting.

    Not sure I'd buy it though, $125 is pretty expensive.

    I just got my new Athlon system, 500mhz on a Asus K7 motherboard. It's great. A lot of the modifications towards overclocking can be done through the BIOS setup - very cool

    This is really a great chip and I'm happy to give my business to AMD. I looked at the Intel chips and everything pointed to the Athlon.
  • Outside Loop [outsideloop.com] as it says on the card... and they sell it for $80 US... (in the US)...

    The $125 Canadian from the Ottawa store is cool for us Canadians though. ;)
  • Overclocking, today, is pretty silly. It's rather expensive and doesn't really provide that great a benefit. It's mostly an exercise in macho tinkering, done to brag about the top speed more than to actually run the thing.

    However, it was entirely sensible when Intel released a whole pile of Celerons which were perfectly capable of running at half again their stock speed, with no special cooling hardware.

    It didn't make sense not to to overclock, in that case. Intel's marketing department decided to lie to everyone about what these chips could do so they wouldn't cut into their high-margin market.

    However, chip manufacturers have now learned their lesson: a few people will always test to see if their chip is really as slow as the spec say, and if they learn otherwise they'll tell everyone else over the internet. So they will build their chips to run slower if they want a slow chip to sell at a cheap price, and make damn sure that there is no cost-effective way to run it faster. The golden days of overclocking are over.
  • Okay, so you want a fast system. You go out and buy an Athlon. Then you buy an overclocking kit to squeeze every last CPU cycle you can. The only problem: your machine runs just as slow as it did before because the rest of the system can't keep up with the processor. I don't understand why everyone worships processor speed when in reality it's bus, memory, and hard drive speed that have the greatest influence over system performance. Processor speed will never be the limiting factor in overall system speed. I'm not going to drool when the 1 GHz Athlons come out; however, I might get very excited when that $100 motherboard has a 500MHz backside bus.
  • I have this motherboard as well using the 500mhz chip. I LOVE this motherboard, very sweet. Just got the system last night. Have you tried overclocking it?
  • For the same amount of money, Trinity Micro will install switches on your Athlon, including a switch to change the cache clock multiplier.
  • by HomerJ ( 11142 ) on Friday January 14, 2000 @02:37PM (#1370785)
    I jumped on the BP6 bandwagon when they were released with a couple of celeron366's at 458. BTW I got off-week celery's at a good price and hoped for the best for 504 and didn't work.

    q3a supports SMP, only in WinNT. I bought q3a expecting SMP code to be in there for linux, but Carmack doesn't think it's nessecary to have SMP in linux. That rant is for another day though =) I grabbed Win2k early to get a natvie DirectX and SMP support. The only thing I saw was complicated aps taking 50% cpu power. I never got the voodoo 3 driver to work in WIn2k quite right, so I didn't benchmark q3a with SMP(the drivers from NT game palace is you must know)

    Kernel compiles in linux only take a couple minues, but that's about the only real use I get out of SMP. Linux distributes the processes wekk enough, but I rarely do something that really taxes the machine, besides q3a and compiling kernels. And GCC is the only thing that uses both. About the only REAL benafit, is that I can run GCC with only one job, and it doesn't tie down the system.

    With just about every OS now supporting SMP, including WIn2k, OS X, linux, etc., when will companies start writing apps that take advantage of it? Is Win9x holding SMP back because it doesn't support it?
  • I've said it once, and I'll say it again! The biggest bottleneck for SMPs is the concurrency supported by the cachememory link. Not bandwidth, not latency, not capacity, concurrency.

    If you don't match the concurrency of your memory link with the concurrency of your clients (i.e. processors), you're hosed for any demanding application.

    What do I mean by memory link concurrency? It could come from crossbar versus bus, or multi-ported memories, or from multibanked (interleaved) memories.

    Cray has zillion-banked memories. Processors now have multi-banked caches, because there are lots of things going on at once inside out-of-order issue processors!

    It's all about concurrency matching!!

    nick
  • I'd rather buy this:

    http://www.kryotech.com/Products/superg/Tech_Spe cs/tech___specs.html

    It's a barebones athlon based PC overclocked to 1Ghz, and it comes with a 1 year warranty.

    I'd rather have someone else to blame if I happened to fry my machine.
    Mike


  • This is cool, except for the fact that it's 120$ canadian. Right now, actually, I am building my own athlon OC'er with the circuit on tomshardware.

    I finally got 5 of those connectors after about 2 months of trying to find them. No one really sells them in non-bulk supply. Now I'm just waiting on my resistor packs and diodes. So far I've spent maybe 20$ on the whole thing and that's all I'm going to have to spend. If you know how to make one you can save a lot of money!

    Btw, if anyone wants to buy 1 or 2 of the actual 40-pin connectors email me at jpal@linuxfan.com. They're a bitch to find, and I'm willing to sell 1 or 2 of them if someone else wants to build their own.

    -John
  • 180 $ for Athlon 500. + 70 $ for NinaMicros overclocker. = Athlon 700 with slightly slower cache. For under 200 bucks, I have a slightly slimmed down Athlon 700. I don't see a PROBLEM with this. Do you? (And yes. Most reports indicate that you can overclock an Athlon 500 up to 700 and have it run just as reliably as a real 700.)
  • A lot people still build celeron overclocker systems. It's pretty common practice in the gaming scene. Overclocking isn't a fad at all. It may not be as cost effective with some of the newer processors, but for a lot of people it's still worth it. And don't forget, there is a lot more you can overclock than just your CPU. A lot of people overclock their video cards too. Why do chip makers care about overclocking so much? Why do they care what I do to the chip? It's MY CHIP. I bought it. If i wanna void my warranty and run the thing as fast as it can then why should they try to make that harder for me to do? I believe chip makers should make it as easy as possible to overclock just out of courtesy to their customers.
  • dangit! I always post too early. I meant to add that most commodity SMPs don't concurrency-match well at all. In fact, the concurrency of most commodity SMP subsystems is about 1!

    In contrast, the concurrency of "good" SMPs, like the convex exemplar or origin 2k is huge. For example, the exemplar has (can't remember exactly) like 8 banks or 32-way interleaved memory, attached to processors over a crossbar!

    And all this concurrency for a measily 16 processors :)
  • It's only 180$ or so for an athon 500, right? And about 70$ for ninjamicro's keen overclocker. That's under 200$. And most people have been able to clock up to 700 Mhz with no problems. Small minus for the cache slow down.
    So you've got a CPU that's almost as good as a 600$ CPU for less than 200$. Hell, get two of them, if you want an SMP setup so badly. :)
  • Try here instead: http://www.ninjamicro.co.uk/ It's 65$. And they'll ship it anywhere in the world with no shipping charges. (Dip switch driven.)
  • I have an SGI Indy, and supposedly the bus is capable of 2.3 Gb/s (on a 32-bit system, this would be like a 616 mhz bus i think) ... the system is only 1/6 the speed of my intel box (in terms of processor speed in MIPS), yet the two machines often run neck-in-neck. The only thing particularly slow I notice on the Indy is heavy disk access, but that's probably just my hard drive ;-)
  • The effect of the cache depends mostly on the application you are running. Some applications will benefit more from higher megahertz, others will benefit more from a faster clock rate. With an overclocking card, you can actually go from one state to the other fairly easily, depending on the application you are running. Some applications are more memory or cache dependent, others are more megahertz dependent. Also remember that the Athlon has 128K of L1 cache, and a damn good branch predictor, so on some apps the L2 cache may make very little difference.

    Some other points:

    1. This is not the first overclocking device of this type. There are a couple of others, the most prominent being one from Trinity Micro.

    2. There is also a 2/5 cache divider option, which allows you to get to higher megahertz without going all the way to 1/3. Also, there is a way to set the divider in software, so you don't have to solder to change the divider. I think that is why AMD didn't give access to the cache divider on the "golden fingers" connector on the top of the processor. The program to do this is still not posted in a public spot on the net, as far as I know. "Soon"

    3. The L2 cache is sometimes more than a little bit of a limitation. Let's suppose your processor only does 600 mhz at 1/2, but it does 900 mhz at 1/3. That's a 50% increase in megahertz, both ending with a 300 mhz cache, so in this case you would DEFINITELY be faster.

    4. Athlons should soon have onboard cache, at full speed, so you won't need to do this. There is supposed to be a version of this with up to 2 megabytes of cache, called the Mustang, or "Athlon Ultra". This chip should kick some serious ass.
  • Can you SMP Athlons..?
    I am considering getting a SMP box running and right now I have 2(370 Overclocked Celerons) in mind maybe I should change this to Athlons.
    Come lets here you suggestions for the best non Intel SMP machine....
  • If you troll with your real ID, maybe we'll all chip in and get you a fishing pole. I bet Rob is numb to you by now, so I'm not sure if you just do this to irritate everybody, or whether you're just bored. Anyhow...
  • Of course, switching to java would make it impossible for me to view slashdot in lynx.
  • that's $125 canadian dollars. It should be much cheaper in USD.
    "You want to kiss the sky? Better learn how to kneel." - U2
  • Overclocking is great. Heck, any hardware hack is great. And it's not about 'sticking it to the man' either.. it's simply about knowing how to *use* technology instead of being enslaved by it.
    When I buy a PIII, I'm not paying intel for the right to use it at a certain frequency, I'm paying intel for a chip that *they* have guaranteed will run up to a certain speed. Over that speed, and you are on your own.

    Now.. when the Celeron 300A was out, and you could easily clock it to 150%, heck, that's fantastic. A real money saver... spend $30 on extra gear to cool it, and you were set.
    Now.... do I spend $75 on extra fans/heat sinks, when I could buy a chip that's rated at a a similar higher speed for about the same added cost? Sure.. it might cost me a few dollars more.. but then I *know* it will work too.

    Like.. Kryotech.. now, those Cool athlon 1Ghz jobs have major geek cool factor, I'll admit, and I'd love to have one.. but realistically, I could be 2 other full machines for the price of just their base model, each machine being around 600Mhz anyway.... so why would I bother? What good would it do?
  • I want one that mounts the switches on an unused drive bay, kind of like Creative is doing with their sound card connecter bay 'Live Drive' or something like that. Then I'd need some big assed knobs that go up to 11 like Spinal Tap had. That way everyone knows I can crank that Athlon up anytime I want to.
  • by Jeffrey Baker ( 6191 ) on Friday January 14, 2000 @03:03PM (#1370803)
    Actually the PIII is WAY overclockable. The PIII-550E can be clocked up to 825 MHz according to some reports. Consider that a PIII-550E costs $350 and an Athlon 800 costs $904. An 825 MHz x86 processor cannot even be purchased on the open market yet. Anand's [anandtech.com] has a pretty good report on overclocking the PIII.

    -jwb

  • According to Anandtech, who did a review of overclocked Athlons (including ones with the L2 cache switched to 1/3), you still get a performance boost if you up the MHz and change the L2 cache multiplier to a lower setting. Granted, it's not linear, but you still *do* see a significant boost. Here's the article -A.P.
    --


    "One World, one Web, one Program" - Microsoft promotional ad

  • The point was that overclocking is a result of intentional under-rating for marketing purposes. There may still be some stuff out on the market for which it is practical, but now the chip makers are clued in to the fact that they can't get away with just underrating their own chips in the documentation. Someone will catch the lie.

    The reason they would make it hard for people to overclock is that they would rather sell you the more expensive chip. Courtesy to customers is in damn short supply, which is why low profit margin cars are built to disintegrate in time for the new model to come out ($6000 construction cost for a $10000 car that lasts 5 years, or $10000 construction cost for a $50000 car that lasts 50 years: looked at as individual jobs, the latter is much more profitable, but to a long-term industry, they are close to equal; fairness doesn't even come into it). In theory, competition is supposed to wipe out these tactics, but industries always have these little understandings that member companies will follow even to their own demise, like a daimyo refusing to arm his troops with modern guns and change the face of Japanese feudal society even when he'll be defeated otherwise.
  • According to Anandtech, who did a review of overclocked Athlons (including ones with the L2 cache switched to 1/3), you still get a performance boost if you up the MHz and change the L2 cache multiplier to a lower setting. Granted, it's not linear, but you still *do* see a significant boost.

    Here's the article

    -A.P.
    --


    "One World, one Web, one Program" - Microsoft promotional ad

  • Rob, slashdot's broken. Fix it. - A.P.
    --


    "One World, one Web, one Program" - Microsoft promotional ad

  • does overclocking produce fumes that induce highness?


    how can you compare an overclocked chips price to a non-over clocked, but overclockable chip?

    that makes no sense

    especially since clock speed is not really a great measure of anything other than clock speed....

    it used to say how many instructions per time period, but it doesn't do that anymore w/ all the different execution units....

    not that overclocking is a bad thing but clock speed fixation is

  • well perhaps it would be more cost effective, but as of now, there is no SMP atlon board, so its a moot point

  • Already a couple of years ago I thought about how nice it would be to have a gas pedal under the desk; when playing or compiling, you could speed the system up like a car... of course it's pretty idiotic idea. And now, it seems, anyone can do just that.

  • Actually...according to the first article written on Tom's Hardware guide about overclocking the Athlon, the case is not easy to open.
    Of course...I think he was trying to overclock a 500Mhz version or something...might have changed since then.
  • Unfortunately buying one and sticking it on the card yourself and turning a few knobs if about as easy as it comes...just be thankful they came out with a better way then moving resistors around.
  • Completely offtopic, but the content of the HTML is independent of the type of server generating it. Thus a Java servlet engine can generate content suitable for any HTML browser.

  • As a techie the first question I asked myself was why not?. Why wouldn't you want to overclock an Athlon? True the other limitations of the computer bring down the effectiveness....but when you realize the difference between what we have...and what we could have...you'll know why...

    Its a driving force in us all...
  • the problem is, they don't want the people selling you your system to overclock it without telling you....

    if you buy a system that is specificially overclocked its cool

    but if you buy a pIII 600 system, and its running a pIII 500 chip, w/ no extra cooling, in a piece of crap case it has a higherpossibility of overheating and then who do you blame?

    intel, cause their chip overheated

    then you go get an amd chip, and intel has lost market share

  • At least you can encode your cds to mp3 [freshmeat.net] on both CPU's :)
  • This is why AMD is gonna beat out intel on the chip market. Intel just locks their chips at one speed. AMD sets the Athlon to a speed, but makes it possible to OC by way of an add-on card which also generates revenue for those willing to manufacture/distribute it. They're opening a whole new avenue of sales in the PC arena.
    =======
    There was never a genius without a tincture of madness.
  • I was looking at that a year back and what I am disappointed about it that they used to offer a system case you could buy yourself and put your own computer into. Not as effective because you don't get the same expertise, but still pretty good.

    And if you combined the AfterBurner...with the Kryotech system, well, let's just say I wish I had a few grand lying around.
  • 125.00 Canadian dollars = 86.41 US dollars

    Exchange rate: 0.691300
    Rate valid as of: 1/10/2000
  • and why not ? i was wunning windoze 3.1 fine with drdos 6 & ndos 7
  • Perhaps it's also getting harder to overclock these CPU's as they're getting faster, faster than the FSB is getting faster.

    Take for example: with a clock multiplier of 6 on a 100 Mhz FSB (600Mhz chip), upping the FSB to 133 MHz boosts the CPU to 798Mhz (198Mhz gain). Compare with a 400Mhz chip (multiplier of 4 on a 100Mhz FSB), upping the FSB to 133 will boost the CPU to 532MHz (132MHz gain). In both cases the CPU speed has gone up be a factor of 1/3. But perhaps CPU's are not effected so much by the factor, but the shear amount.

    I'm no hardware guy: I don't know the effects of increases MHz and heat on these increasingly smaller dies. Maybe somebody would like to dicuss this (and probably point out the error of my ways.)
  • Come on, you want anyone to believe that there are a quality selection of Slot A motherboards out there? Even the K7M as its problems: placement of the ATX power connector, an AMR Slot, onboard sound (with select boards), lack of overclocking via bus speed selection, and let's not forget the chipset itself - until the KX133 comes out Via motherboards are still in the doghouse for disk performance, memory bandwidth, a horrible AGP implementation, compatibility, and CPU execution. The BX trounces the Via even with the Via proclaiming a 200MHz EV6 front side bus.

    If you want to get critical, think of the FIC SD11, possibly the worst mainboard ever created. Did you know FIC's engineers dropped the ball on this one? The board is a preproduction sample. They had to rush their product to market so better boards won't hog all of the sales.

  • I tend to agree with you. There are better ways (of course, depending on how you use your machine though) of boosting performace. You can get a dual Mobo with 2 Celeron 500 for about $20 more than a single MoBo and a P3 500. In my experience Celerons are very good performers, and this is a really good way of getting something vastly faster than a P3 500 for roughly the same price.

    Incidentally, I've just ordered the parts for a dual P3 system (I got carried away when I came across a good deal whilst looking around at Slockets). Hopefully I won't be sitting around waiting for Visual C++ under NT for so long in the future. As I work from home, I can't afford take risks with my computer crashing due a minor instabilty caused by overclocking. I almost bought that dual Celeron hack the Abit-BP6, but I don't think that it is stable enough for me to risk not being payed for time spent recovering from a crash.
  • $125 CDN is pretty high, considering it's about $1 in parts.

    As per usual, I'll do it myself.
  • From what understand, Linux doesn't scale past 2 CPUs at all well. It has it's arse kicked by NT which scales better.
  • "With just about every OS now supporting SMP, including WIn2k, OS X, linux, etc., when will companies start writing apps that take advantage of it?"

    It depends what the app does really. A word processor doesn't benefit greatly from SMP (the CPU spends most of the time idle as the app is waiting for user input). MS Word is multithreaded in some areas though, for example, background printing. In this case, the multithreaded makes it Word feel more responsive, but as a side effect utilises SMP if available. As I understand it, an SMP OS like NT will automatically take advantage of extra CPUs when threads are in use. Just looking at my task manager, I see that the majority of processes have more than one thread (and thus could benefit in some way from an extra CPU), for example: Internet Explorer - 11 threads; McAffee Virus Sheild - 11 threads; the system - 31; SQL Server - 17; Yahoo! Messenger - 14; etc.

    I would think that SMP gives a real boost when two processes (or threads) pipe data from one to the other and work in parallel. Reduced context switches, and true parallelism rather than imitated parallelism (which takes twice as long).

    My point is: people already multithread their apps, so they will use SMP. Whether you see a benefit is another story - that depends on what you're doing.

    The other thing of course: it's much easier to quickly develop a single-threaded application. Companies are constantly rushing. On top of this, in my experience, a large proportion of the software developers can't handle concurrency. The amount of obvious multithreading bugs that I've had to fix is staggering. When a company realises that they could develop a not so nice and inferior piece of software in less time, they generally do so (unfortunately).
  • Go to 11? But couldn't you just make them go to 10, and make 10 be louder?

    --
  • Q3A does support SMP. You have to set a variable from the console to enable it (search Deja [deja.com] for the variable).

    Not a 2x performance increase but it does make a difference.

  • This is true; however, the course-grained semaphoring in Linux is at this very moment rapidly being replaced with fine-grained semaphoring. When this is complete Linux likely may start beating NT in the large-cpu configurations.
  • by Anonymous Coward
    Unfortunately, though you make a good case and have hit Intel right on the nose, you are wrong about AMD, at least about their Athlons. The chips they mail out as 500s roll off the same production line and have the exact same contents inside as a 700MHz machine except for a few resistors. Now, according to your theory that all chip makers have caught on, and I think they have, what would be the logic here?

    Easy, look how fast that one Celeron chip that was superoverclockable (and it was only one model boys and girls, your cousin never overclocked his 450 Celery to 900) sold out! Feed the overclockers a MASS of your 500MHz chips that can go to 700, make a huge profit in bulk there. Sell the overpriced 700MHz chips to power users in business and home that want power but don't really feel like cracking the case open on their chip... let alone simply opening the case which scares most people.

    E.
    I think this is probably the smartest thing AMD could ever do, even better than beating Intels chips in every single conceivable benchmark by a significant margin especially in the holy floating point arena...
  • You run X right? Well, on a single CPU box, your app and the X server constantly vie for the CPU, the app (well, actually xlib) queues up some drawing requests, then the X server takes over the CPU and draws them. Rinse, repeat.

    With SMP, the X server gets a CPU, as does the app. Much less context switching, and the cache doesn't get thrashed because the linux scheduler tries to keep processes on the CPU they last ran on.

    The old X FAQ actually recommends swaping workstation and $DISPLAY with a coworker on a fast network, to reduce the context switching overhead. You work on her machine, showing up on your local X server, and he does the same.

    SMP is the wave of the future. Between the dual celeron board, and $110 P-II 450s, SMP is the way to go for the fastest system available cheap.
  • Okay, for my peace of mind, my interpretation of what you said be (please correct me if I'm wrong)...

    Coarse grained being (sorry, C++ not C):

    someObj::someFunc()
    {
    myMutex.lock();
    ... // Lots of processing
    mySharedObj->doSomeSomthing();
    ... // Do more processing
    myMutex.unlock();
    }

    Whereas fine-grained would be:

    someObj::someFunc()
    {
    ... // Lots of processing
    myMutex.lock();
    mySharedObj->doSomeSomthing();
    myMutex.unlock();
    ... // Do more processing
    }

  • . . . I don't get it.

    Sorry for being dense but. why start with a slower chip?

  • What about it? It's a stable board, not made for overclocking (but no Intel-made board is)...my Coppermine runs exceptionally well on my BX Soyo 6BA+ III.
  • way cheaper.

    Athlon 700 ~$500
    Athlon 550 ~$200

    if they both can get to 650MHz (since they may be from the same fab lot (taking all sorts of things into account), you feel smarter and not so poor...
  • You'll have to pardon me here, but I don't see the purpose of this "workstation $DISPLAY swap" thing. Either way, you're running both an X server and a client on the same CPU, so you're still gonna be context switching quite a bit. Except after the swap, you've got network overhead too.

    Jeremy
  • That variable is r_smp.

    Type r_smp 1 at the console, and restart the game, and you've enabled q3's smp support..

    Also important to note is that 3dfx cards do not support q3 in SMP mode.

    Jeremy
  • by Animats ( 122034 ) on Friday January 14, 2000 @06:32PM (#1370862) Homepage
    Classically, overclocking was a terrible idea, because speed was a part selection. In other words, parts were manufactured to run at the highest speed, then tested. If a part was available in 100MHz, 150MHz, and 200MHz, a part marked 100MHz had failed the acceptance test at 150 and 200. But typically, only a few gates on the chip failed at the higher speed, so the chip almost worked at the higher speeds. And if its temperature was kept well below the upper limit of the rated range, it might work consistently. But in general, overclocking meant a system with a substantially higher error rate.

    As the fab for a given process became more mature, the defect level usually decreased. So at the beginning of a product cycle, you got more of the slow parts and fewer of the fast ones, and over time, more parts were produced with the higher speed ratings. Over time, then, the price of the high-speed parts declined.

    Then Intel reinvented itself as a consumer products company, and started pricing ICs the way GM prices cars. In the auto world, a luxury car costs maybe 30% more to build than an economy car, but sells for perhaps 3x as much. Intel started doing this for processors, with advertising-promoted brands at different points in the speed spectrum. The interaction between this policy and the way fabs actually work resulted in some deliberately undermarked chips, and the rebirth of "overclocking" as a semi-respectable enterprise.

    Then some distributors started shipping systems with overclocked CPUs. Some even printed fake part numbers on the chip package. This led to trouble. Intel may have lost some revenue, but worse, they were getting a reputation as an unreliable IC supplier. So they added holograms on chips, part ID info readable from software, and speed-checking (which is hard; CPU chips ordinarly lack an on-chip timebase.)

    Today, IC fab yields are so good that the part-selection approach is rare. If parts are failing, the fab has a problem. CPU speed and model has become mostly a market positioning thing.

    In the industrial computer world, underclocking is common; the temperature margins improve, and so does reliability.

    At this point, Intel and AMD are competing so hard on speed and price that neither can afford to undermark. So overclocking is a marginal idea at best. Gamers are probably better off getting a new graphics board.

  • Yeah, that's about the idea. An example is the difference in the linux kernels. 2.0 just put the entire kernel in a lock, so only one processor could be in the kernel at a time. 2.2 Added locks to the major subsystems, so one processor could be sheduling while another processor did some memory management. 2.4 promises even finer grained locks. For instance, 2.2 locks all wait queues as a whole, while 2.4 will lock each wait queue individually.

    Jeremy
  • ...
  • Ill say it again jus to reiterate. The PIII Flip-Chips (PIII 500E etc.) are HIGHLY overclockable. Almost intentionally so. They have lower multipliers than original PIII's and everything. There made with the smaller Micron Manufacturing process so they generate less heat. There core is 'flipped' upwards so the cooler cools directly to where the heat is being generated basically. All of that leads one to wonder.. And yes they can be made to SMOKE. Right now those sound like the most fun to OC. oh well

    Jeremy Allen
    jallen@idminc.com
  • It seems overclockability will only increase a chips popularity, in fact.
    Check out Kyrotech: http://www.kryotech.com/
  • Even with an MTH which transfers RDRAM requests to SDRAM the i820 beats out the Via. You want to talk about memory bandwidth and performance, the Via is the worst of the bunch.
  • Not only is it silly/expensive/low benefits, but one thing many people don't realize is that these reviewers who check out the overclocked Athlons aren't using them for an extended amount of time. Try using it for 6 months+ and your chances for more problems gets worse and worse. Testing it for a week (if that) might sound good to test speed, but your chip probably isn't going to make the long haul.
    - Mike Roberto
    -- roberto@apk.net
    --- AOL IM: MicroBerto
  • by costas ( 38724 ) on Friday January 14, 2000 @07:50PM (#1370880) Homepage
    I will repeat this until I am soar, 'coz, well, it's a soar point with us in the Beowulf community: Linux SMP sucks. If you want good SMP performance, you're better of with NT or Solaris.

    The big problem with Linux SMP, IMNSHO: NO CPU affinity. Which means roughly this: processes are rotated thru all available CPUs, instead of being assigned to one CPU and then being dynamically balanced (new jobs sent to the lightest-used CPU, when CPUs are imbalanced by some threshold %, move 1-2 smaller jobs that will balance them out).

    What does this mean? well, CPU cache is practically useless. Makes all that dough spent on Xeons instead of Celerons seem wasted --and it is.

    Don't get me wrong; I am all for Linux, and I am sure SMP will catch up pretty soon. But don't go spending $$$ on SMP machines expecting (n-1)*100% increase in performance.


    engineers never lie; we just approximate the truth.
  • Gee, then I guess the overclocked K6/233 (to 264) I've using for the last two years right now shouldn't be working...
  • Actually, AMD has been putting 600 and 650mhz cores into the chips they are selling as 500mhz processors. If you crack open the case you can see what the processor is actually rated for... the late week 500 athlons also shipped with some rather quick SRAM...
  • Wow! If you're right, the cpu allocation in linux is pretty primitive. I have a scary alternative idea, though...
    How about some of us set about crafting a neural or fuzzy scheduler? Neural nets have have an incredible way of optimising to a problem. Speed might be difficult at first, but a simple checksum or crc pattern stream might provide enough clue to a neural net to allow it to balance loads in a meaningful way. Cached past patterns could be used for future prognostications...
    I can guarantee you one thing: someone will try this strategy, at some point. Might as well be us!
  • Which version of Linux SMP are you looking at to make these comments? Last I read on the linux-kernel list, Linus detailed that linux threads do not switch CPUs.
    Also, From my last look at the scheduler, there is a big preference given to the previous processor a process was run on, so as to not incur a TLB flush.
    But on an overloaded box, TLB flushes are a huge overhead. But we are talking about where the # of running process == # of avail CPUs. So I don't see the "processes rotated thru all available cpus" that you claim.
  • Just a few things that may/may not have been said already.

    First of all, why is this particular product being highlighted? I've seen at least 8 other companies making these. Not to mention that making your own is a simple matter for a first-year EE student.

    Second, why do I see so much pooh-poohing of overclocking in this discussion? I think one of my happiest hardware moments was when I got my K6-2 380 to run stably at 500Mhz. Believe me there *was* a performance difference, and I don't really play 3D games. Now that I just got my Athlon 650, I'm assembling the parts to make my own "goldenfingers." I see many accounts of 650's doing 750 and 800 without a problem, and without modifying the cache divisor.

    Speaking of the Athlon, damn. I'm very impressed with this chip. I have never been this close to the "bleeding edge" before, and it's a neat feeling. It takes less then 6 minutes to do a complete kernel build (dep, clean, bzImage, modules, etc.), 3 SETI@Home packets a day, and I can turn on almost all the bells and whistles in E with no problems.

    If you dont want to overclock, don't. I wouldn't do it on a mission-critical machine, but it's a thrilling feeling the first time you boot up faster then before.

    If you want to see some crazy-assed OCing, check this out:
    http://totl.net/Eunuch/Eunuch2.html [totl.net]
    Now *thats* serious :)

  • Well, there are in the way... If you running NT then at least there will be much more processor-time for the game as some of the system threads will run on the other one. Not tremendous speedup but still something to consider.
  • There are about a half-dozen of these adapters available for sale. They range in price from $20 to $125 depending on options (voltage tweaks, build quality, etc.). I'm no overclocking master, but are they suggesting you cut the L2 cache speed from 1/2 core to 1/3 core? Why on earth would you do that? Let's say your core frequency is 800 MHz, and your L2 runs at 400. If you overclock it to 900 but your L2 cache is only running at 300, surely you're getting worse performance overall than you were before... http://www.tgwbp.addr.com/cgi-bin/wwwboard.cgi http://www.tgwbp.addr.com/cgi-bin/wwwboard.cgi http://www.tgwbp.addr.com/cgi-bin/wwwboard.cgi
  • Pricewatch:
    Athlon 700 MHz - 496 at tufshop.com
    497 at econopc.com...
    4 listed ~500

    2 listed ~529

    more 583+
  • I had a pair of bitched-up Cel366s I (I think that was the speed) -- anyway -- one I tore apart and the die was welded directly to the heatsink pad, which is opposite of most chips and I believe the same as the 'flip-chips'. I don't beleive that process was unique to the E-series P3s.

    The other I drilled a hole in the corner and is on my keychain. :-)
  • Most of what we did on the departments alphas at Iowa State sucked cycles--when they were running, which was a small fraction (Gad, SAS is a pig). It turned out that (most of the time) an extra 32M or 64M for each extra user was enough to avoid most of the swapping, so that unless both jobs launched at the same time, the person sitting at the unit would never even notice the extra users.
  • I had wrote a lot in response, but it was a lot of stuff no one probably cares about. In short, I'll say this. AMD has said they could produce a 1GHz chip right now, but they're holding back for marketing reasons alone. So they're obviously not using the full potential of their great yields. Because their yields are so good, probably every single CPU they send out marked at 500mhz is good enough to run at 650, but the guys in marketing say they need to keep a 500mhz CPU on the market. This is the beauty of overclocking. The chip companies feel pressed to have a variety of speeds, the ubercheap consumer version and the uberexpensive 'power user' version. The smart ones among us realize the cheap ones usualy come from the same batch as the expensive ones, and the mhz rating isn't worth much. Intel most notably did this with the Celeron - which was originaly just a neutered P2, but now it's a P2 with 128k of full speed cache instead of 512k of half speed. For gaming, a lot of cache is useless - you really want raw power. The Celerons had to be clocked low because they were the 'cheap' version, but they could easily do more than there rating usualy. This was why there was such a big explosion with the 300a, the only reason they put out a 300mhz chip is because they thought the market needed one. There was also quite a bit of fuss over the 266 within the overclocking community, but it was fairly well contained. My point is, AMD and Intel are 'underclocking' as you call it. Most of their new chips can easily hit 650-750mhz, but they produce chips below that because that's what the market needs... Afterall, they couldn't just mark everything at the mhz it can run at, then they wouldn't make as much profit - god forbid.
  • The AC said:
    The optimal performance increase with 2 CPU SMP is about 45%.
    Linux is below 20%, Solaris is close to 40%.

    Huh? Why 45%? Is this Quake performance or some kind of general claim? It sounds awfully low.

    I have made many CPU intensive tasks on a 4 CPU Solaris machine that I have shared with others (also doing intensive computations) and have not noticed anything like this. I have not made any timings, but each CPU is faster than what I have on my desk and it certainly looks like a linear speedup until the machine runs out of processes.

    Lars
    __

"Mr. Spock succumbs to a powerful mating urge and nearly kills Captain Kirk." -- TV Guide, describing the Star Trek episode _Amok_Time_

Working...