Author Topic: woohoo, great overclock (Read 4354 times)

radiator · « **on:** May 25, 2003, 08:20:42 am »

just sorted out overclocking my pc, and got a shock as i managed to push it further than i thought.

it was an athlon 1700 (or, 1.47Ghz in proper money) and i managed to get it running at 2.133Ghz...

not bad, not bad at all

i have had it running at 2.2Ghz, but it actually slows down (i run viper racing as a test/benchmark (the only game on my pc) and @ 2.2 the whole game slows down.....no jerkyness or anything, just really slow and smooth?!?!)

skirge66 · « **Reply #1 on:** May 25, 2003, 06:58:59 pm »

I work for a chip making company...those slower chip are from the same batches and product lines as the faster models...all depends on how they sort and test my friend...that's why they "CAN" be overclocked...there is a reason they are sold as a slower chip...not going to tell you what to do, but i would get a bigger fan if you want it to run a while. don't say you weren't warned. good luck.

CthulhuLuke · « **Reply #2 on:** May 25, 2003, 07:18:57 pm »

Hmm.. that's interesting, cause I was thinking about how you could essentially get really good chips and really crappy chips when making processor chips just by chance. So also technically some guys could be getting 1.8ghz chips as 2.0ghz chips, and other guys could be gettin 2.4ghz chips as 2.0ghz chips, am I correct on that? Or is there a pretty tight standard these chips gotta stay on?
btw, what did you use to overclock your processor? I've heard a few methods, but nothing solid enough to get it up another 500mhz.
-Luke

radiator · « **Reply #3 on:** May 25, 2003, 07:29:23 pm »

Quote from: skirge66 on May 25, 2003, 06:58:59 pm

I work for a chip making company...those slower chip are from the same batches and product lines as the faster models...all depends on how they sort and test my friend...that's why they "CAN" be overclocked...there is a reason they are sold as a slower chip...not going to tell you what to do, but i would get a bigger fan if you want it to run a while. don't say you weren't warned. good luck.

yeah, it's a certain model of athlon (certain 'stepping' code) which means it's probably really a 2200 (or something) thats been underclocked because of high demand for the slower/cheaper cpus

and i do have a big fan (a huge 100% copper heatsink with an 80mm fan strapped to it

) - and just the right amount of Arctic Silver thermal compound...

running like a charm at the moment

radiator · « **Reply #4 on:** May 25, 2003, 07:36:00 pm »

Quote from: CthulhuLuke on May 25, 2003, 07:18:57 pm

Hmm.. that's interesting, cause I was thinking about how you could essentially get really good chips and really crappy chips when making processor chips just by chance. So also technically some guys could be getting 1.8ghz chips as 2.0ghz chips, and other guys could be gettin 2.4ghz chips as 2.0ghz chips, am I correct on that? Or is there a pretty tight standard these chips gotta stay on?
btw, what did you use to overclock your processor? I've heard a few methods, but nothing solid enough to get it up another 500mhz.
-Luke

as i said in the above post, i think that on this occasion, AMD weren't making enough 1700 chips, and the demand was high...so they simply slowed down some of their higher end cpus, and badged them as 1700's

normally, when the companies test the cpus, they'll try running them at different speeds, to see how fast they can go before becoming unstable. once they've found that speed, they'll knock it down a few (hundred) Mhz and badge them at that speed.

as for overclocking it, i just changed the multiplier setting in my mobo's bios (there's also an option to change the CPU frequency, which lets you overclock in smaller fractions, but this means connection L1 bridges on the cpu)
the bridges look like this:
. . . . .
. . . . .

you've gotta connect top to bottom with a conductive material, but, these things are like .2mm in size, and spaced at about .7mm......no way am i trying that!!)

skirge66 · « **Reply #5 on:** May 25, 2003, 08:29:04 pm »

actually if a chip is sold as a 2.5ghz...it was tested and met the standards to run as a 2.5 it will not be running at say 2.0 or something slower, however if there are imperfections in the chip that would make it unreliable running at a faster speed they will underclock them to a speed that they will function under normal circumstances reliably...by overclocking them you can exploit a minor defect causing the circuit to fail chip to melt etc...this will normally void any warrenty you have on your cpu.

I have never heard of underclocking to sell a cheaper model...that's not good business smarts. what you get is a crossover of product lines...what i mean as an older process becomes more mature, it becomes more robust, you tend to produce more chips that meet the higher end requirements for that line, and can hold strong at a higher speed, where as the newer process yields lower but the older model chip and newer chip may very well be athelon umptysquat and can be sold on the market as the same cpu...of course this isn't true when it comes to significat advances that say have a substantial technology improvement like front end bus speed or such that cannot be simulated if you know what i mean.

radiator · « **Reply #6 on:** May 25, 2003, 09:18:06 pm »

i'm pretty certain that's what AMD did for this last batch of 1700 CPUs

for a small company, it would be easier to underclock a higher spec cpu, rather that running off another batch of 1700's (which would take time) - there was a stupidly high demand for lower end chips at the time, so they just underclocked a few batches, which gave them enough time to produce 'proper' 1700 chips

yes, it's bad business, but AMD are such a small company (compared to Intel - the Micro$oft of CPU manufacture IMO) that the time it would take to run off some 1700 chips would result in loss of sale

u_rebelscum · « **Reply #7 on:** May 26, 2003, 04:12:47 am »

Quote from: skirge66 on May 25, 2003, 08:29:04 pm

I have never heard of underclocking to sell a cheaper model...that's not good business smarts.

It is a bad idea for high demand, "clocked-as-rated" chips to be underclocked, for sure. But if no one will buy them, better to cut the profit than get nothing.

Always happens now a days with both AMD and Intel. When the demand for $300 (x) ghz is lower than the supply of them, while the demand for $200 (x*80%) ghz is higher than the supply, the chip makers "underclock" some x ghz chips so all of chips are sold. Actually semi-good business smarts.

Don't you remember some of Intel's celerons being underclocked P3's with one cache bank disabled, even when they tested fine?

shmokes · « **Reply #8 on:** May 26, 2003, 06:59:27 pm »

Remember the 486 DX v. 486 SX? The DX had a built in math coprocessor and the SX did not. They both came out of the same batch, though, and the coprocessor was burned out for the SX chips after the fact. The SX chips were, thus, more expensive to produce than the DX chips due to this extra step in the manufacturing process, but they were considerably cheaper to buy since they did not have the built in Math Coprocessor that their DX siblings had.

skirge66 · « **Reply #9 on:** May 26, 2003, 07:35:13 pm »

celeron / p3 ? umm...sorry but these are 2 different product lines...no they didn't come from the same batch.

generally the higher demand is based on price, so you lower the price on the higher end model to what is acceptable to the average consumer which generally wants the fastest thing they can afford. chip companies always try to maintain the greatest interest in their newest products...look how fast the price on high end P4's came down...it's very common to lower prices like this to stimulate market interest and growth...companies don't want the public to decide to "settle" for their lower end stuff, they want people to keep wanting to upgrade, also they don't want to be stuck filling high volume factories with old technology that doesn't build profit.

Mike · « **Reply #10 on:** May 27, 2003, 02:32:58 pm »

I'm not a fan of AMD's because i think they run too hot. But if you want to see another chip that can really be over clocked look at the new celerons. I run my 2.4ghz celeron at 3.2GHZ and it runs ice cold you don't even really need any special cooling. I've had the thing running all the way up at 3.8 ghz with some added cooling power. I didn't feel comfortable running it there so i lowered it back down to 3.2GHZ.

radiator · « **Reply #11 on:** May 27, 2003, 06:27:52 pm »

each to his own

i'm not a fan of Intel, never have been, and would never consider buying a 'budget' chip like the celeron (or the amd duron for that matter)

but considering this cpu only cost me

Mike · « **Reply #12 on:** May 28, 2003, 10:53:42 am »

Yeah i'm waiting for the 64-bit processors too. I'm not opposed to AMD processors I just think the athlons run way too hot. I had an athlon XP1800 and the thing was always overheating even though it was never overclocked and I had like 7 fans in the case. The reason i like the budget processors is that when i build a new system I can spend my money on the motherboard rather than the processor so I can upgrade longer. And this is the way I see it. My Celeron 2.4Ghz costed me $100 and over clocked runs about like a 2.5ghz P4. A 2.5Ghz P4 would have ran me around $200. In a year or so I can spend another $100 and get a 3ghz hyperthreading celeron and overclock it. And it will probably run like a 3.2Ghz P4 or so. So for the same money long term my computer stays faster.

HeadRusch · « **Reply #13 on:** May 28, 2003, 12:38:42 pm »

I fried an Athlon just by booting up my PC and not having the heatsink properly attached. I saw smoke whisp up from the chip, I shut it down..and I was out $60 bucks for the chip

Having said that, unless another 1.6Ghz P4 comes out (my pc I'm using now is a 1.6 at 2.4 Ghz) I'll probably only buy AMD. Their performance per clockcycle is superior to overclocked P4's. In fact, my MAMECabs 1.4Ghz athlon and radeon 8500 benchmarks as fast or faster than my overclocked P4.

Whats the current overclocking king in the P4 family now.......like I said, my 1.6 at 2.4 doesn't really benchmark as fast as a true 2.4 would......are the 2.8's getting up to 3.8 or.....

Do any of these allow Cruisin USA/World to run at full speed?

radiator · « **Reply #14 on:** May 28, 2003, 12:59:49 pm »

hehe, the only thing that scares me about building a computer is attatching the heatsink ontop of the cpu...

a few years back, i splashed out on an athlon 1Ghz (which was pretty beefy at the time) - but saved a bit of money by getting a 2nd hand heatsink off eBay

the guy that sold me the heatsink said "I've lapped it, so it's nice and smooth"...yeah, he lapped it, but it wasn't perfectly flat anymore, it had a slight (and i mean really slight) angle...

...but i didn't notice it and proceeded to install the thing...

turned on the computer, nothing happened...

...i'd heard lots of nasty storys about people cracking the core of the cpu by using dodgy heatsinks, and this is exactly what i'd done!

there was a tiny chip on the corner of the cpu core...

shmokes · « **Reply #15 on:** May 28, 2003, 03:29:20 pm »

I've slipped while installing or removing heatsinks two or three times and stabbed the motherboard with the tip of a flathead screw driver pretty hard. Pisses me off to no end. Why not enclose those tabs so that doesn't happen instead of leaving the sides open for a slip path? Some heatsinks tabs do wrap all the way around the head of the screwdriver, but I can't understand why they all don't.

Luckily the boards have seemed to continue working in each case but the poor design is mind boggling.

atomaka · « **Reply #16 on:** June 06, 2003, 04:07:40 pm »

AMD a small company? I don't know about that. They aren't as large as Intel, but you can't call them small. I highly doubt they produced higher level chips, and then decided to underclock them. That makes zero sense.

radiator · « **Reply #17 on:** June 06, 2003, 08:09:00 pm »

yeah, i meant a small company compared to intel...

taken from hexus.net:
"Demand is a strange animal. Sometimes both Intel and AMD have to fulfill demand for various grade of CPU by using certain wafers, and like any cost-conscious business, AMD try to minimise production costs by employing the minimum number of fabrication lines as is possible. Therefore it's quite possible to earmark a particularly pure wafer, initially destined for XP2400+ speeds, to lower speeds. Market demand dictates all."

so there you go

atomaka · « **Reply #18 on:** June 06, 2003, 11:00:57 pm »

Quote from: radiator on June 06, 2003, 08:09:00 pm

yeah, i meant a small company compared to intel...

taken from hexus.net:
"Demand is a strange animal. Sometimes both Intel and AMD have to fulfill demand for various grade of CPU by using certain wafers, and like any cost-conscious business, AMD try to minimise production costs by employing the minimum number of fabrication lines as is possible. Therefore it's quite possible to earmark a particularly pure wafer, initially destined for XP2400+ speeds, to lower speeds. Market demand dictates all."

so there you go

Indeed. Good quote, but I'd still need confirmation before I'd actually believe it

They just suggest the possibility.

I think the demand grew for the 1700+ because it was such an "overclocking beast" (as I've seen it called on the Anandtech Forums) and not that it became the overclocking beast because of the demand.

I'd also be curious to see pricing on the actual manufacturing level for a CPU. I mean, now that I think about it, other than R&D, wouldn't all CPUs cost approximately the same?

skirge66 · « **Reply #19 on:** June 07, 2003, 02:23:47 am »

I guess being there, working with these various lines, and knowing what does and does not run through our lines, and what they are used for counts for nothing...damn...I should just quit my job now

u_rebelscum · « **Reply #20 on:** June 07, 2003, 04:19:23 am »

Quote from: atomaka on June 06, 2003, 11:00:57 pm

Quote from: radiator on June 06, 2003, 08:09:00 pm
yeah, i meant a small company compared to intel...

taken from hexus.net:
"Demand is a strange animal. Sometimes both Intel and AMD have to fulfill demand for various grade of CPU by using certain wafers, and like any cost-conscious business, AMD try to minimise production costs by employing the minimum number of fabrication lines as is possible. Therefore it's quite possible to earmark a particularly pure wafer, initially destined for XP2400+ speeds, to lower speeds. Market demand dictates all."

so there you go

Indeed. Good quote, but I'd still need confirmation before I'd actually believe it They just suggest the possibility.

A few people have popped off the heatsink in pre-made (dell, hp and compac, IIRC) systems, and the printed speed on the CPU was higher than the rate they thought they bought! And sure enough, when they overclocked it, the number printed on the chip was a little under how high they could overclock it to. This is the extreme end, propably where the CPUs were already put together, but a huge order from one of the huge customers of a lot of the slower cpus was not able to be met at the last moment because of a delay or something; so the only chips at hand, the faster chips, had to take the place.

I read this off http://www.hardocp.com/ and http://aceshardware.com/ a long time ago IIRC. Probably not enough proof for you, but thought I'd mention it.

Quote

I think the demand grew for the 1700+ because it was such an "overclocking beast" (as I've seen it called on the Anandtech Forums) and not that it became the overclocking beast because of the demand.

Of course: overclockers make only a small bit of the demand; most is from OEM customers rather than BYO people. Sort of like us BYOAC people are to game controller manufacturers. $:-\$

Quote

I'd also be curious to see pricing on the actual manufacturing level for a CPU. I mean, now that I think about it, other than R&D, wouldn't all CPUs cost approximately the same?

Manufacturing price for different speed chips in the same batch are the same. There are differences between, for example, the Athlon Thoroughbred & the Athlon Barton. Also, it is possible to "target" a silicon disk for higher speed CPUs at a cost of a lower total yield. But a single silicon disk makes a bunch of CPUs, and each CPU is tested to see what speed it can run.

A single disk can yield all high speed, or all low speed, but usually it's a mix of high, medium, low, and broken CPUs. The companies try to predict the number of each, but random and chance play into the actual out come. The "total yield" is the number of working CPUs. The "yield split" is the number of fast compared to number of slow CPUs. Problems occur when the yield is worse, in yield split or total yield, than expected, or better than expected. The first is obvious; the second means too many chips. Too many means a glut in the market, which can drop the prices too quickly.

Also, if the yield split is so good the company knows it can yeild anything faster (heat, power, internal timing limits for example), if the company releases all chips at the highest rating, in the short term, price for all high end chips will drop, driving down all chips at lower speeds. In the longer term, since consumers wouldn't see any more speed increases until design changes to fix the limit were done, there would be no high end chips to fill those "bleeding edge" buyers. The "earlier than planned" switch to the high speeds as the "normal" speed would mean the company would have "lost" the money it could have made if it decreased some of chips speed in this (and the next few) batch, and slowly ramped up, and slowly dropped the prices.

High yield splits have happened to both AMD and Intel. When they thought they could ramp up the speeds even further in the future, they didn't "under-clock" many chips; they tried to increase market share instead; when they knew they were reaching the limits of the current design, they "under-clocked" as many chips as needed to keep the profits up.

Huh, went a little too much there. $:-\$

Review:

a) Costs are basically the same for high, medium, and low speeds. Unless tricks are done to increase the average speed, at the cost of total working CPUs.

b) Sometimes it makes market sense to release all chips at their max rated speeds.

c) Sometimes it makes long term financial sense to release some chips at rates under there possible max speeds.

d) Companies are in trouble if the yields are low.

radiator · « **Reply #21 on:** June 07, 2003, 06:31:57 am »

just remebered something to backup my claim...

remember when the first Athlons wree released (the Slot A models...looked like the old PIII'3) - well, I was the first in my area to get one (opted for the 500Mhz version)...

well, a few months after buying, I read somewhere that in around week 37 of production, they actually used 650Mhz cores, which didn't pass the AMD quality control at that speed, and was underclocked and sold as a 500Mhz...

so, I took out the screwdriver, and very carefully popped the case off the thing (and I mean carefully...this CPU cost me around

skirge66 · « **Reply #22 on:** June 08, 2003, 07:45:59 am »

for what it's worth just this week, we "re-introduced a deceased product line into the fab...why? supply and demand...newer chips just cost too much to underclock them on purpose...I can't speek for amd...but at intel if we have a higher demand on certain products we simply add more wafer starts to the "mega fabs", and if need be we tag them as priority lots to move them quicker through the line.

u_rebelscum · « **Reply #23 on:** June 08, 2003, 11:57:43 pm »

Quote from: skirge66 on June 08, 2003, 07:45:59 am

for what it's worth just this week, we "re-introduced a deceased product line into the fab...why? supply and demand...newer chips just cost too much to underclock them on purpose...I can't speek for amd...but at intel if we have a higher demand on certain products we simply add more wafer starts to the "mega fabs", and if need be we tag them as priority lots to move them quicker through the line.

It takes weeks before the silicon disc will make it all the way through the process to the finished computer on the consumer's desk. Intel is probably faster than AMD at silicon-to-computer time, but "time is money" too.

You are making chips to match the projected demand. Example: if a farmer's demand for hamburger is higher than expected (ie: not enough fresh ground meat), and the demand for steak is low, it will ground the the extra steak and make it into the less profitable but "at-least-it-sells-and-makes-money" hamburger, rather than losing the customer's order (ie: buys from the other farmer named AMD), plus the possibly the meat getting too old before being sold as steak (ie: the newer P4 revision is released at the same speed, making the old P4 at that speed like beef jerky).

You don't know exactly how fast each of the chips will actually run, especially at the start of a new chip revision, and even more so at the start of a new chip line.

"Newer chips just cost too much to underclock them on purpose" is only true if the high speed silicon doctoring results in a lower total yield, and there is a demand for the high speed. Otherwise the cost of a fast chip is almost the same as the cost of a slower chip. Your only loss is the "opportunity cost" against selling at the higher speed. Are you saying Intel's total yield is very low for the doctored silicon compared to the less doctored slower speeds? (Don't answer if it's a trade secret.)

Maybe Intel doesn't under-clock any more, but other's have found cases, in the past, of Intel under-clocking (to lower Mhz) and down chipping (PII down to celeron).

Computer geeks sites I used to visit before I was introduced to mame and found BYOAC, with more info on this topic:
http://aceshardware.com/
http://hardocp.com/
http://slashdot.org/
http://www.arstechnica.com/
http://anandtech.com/
http://www.tomshardware.com/

Other links:
http://theinquirer.net/
http://www.theregister.co.uk/
http://www.shacknews.com/

skirge66 · « **Reply #24 on:** June 09, 2003, 07:11:38 am »

ok so without going into excessive details or adding "classified stuff" to it basicly the way things are done

so we start up a new process ( XXX )

(AAA, BBB, CCC)...we will call these our basic products
we have a mobile, a desktop, and server

now currently in the line we still have product (aaa,bbb, and ccc) which are the previous process (xxx) products that are still being used by the public and are still being produced until EOL of the process.

now as the new process matures, products AAA, BBB, and CCC receive upgraded steppings (run faster) you'll end up with multiple products with multiple steppings running through the line, and as we mature further you eventually are able to "shrink" the current process...smaller lines=faster processor again now we have our original 3 products, if only 2 stepping changes that's another 6 products, add in one shrink and that gives you 12 different products in a single process...now you have products that are feeding all of your various consumer groups. the older stuff in turn takes the place of the older process which traverses to EOL. Now another thing you need to keep in mind a large sum of what we produce is 1 stuff not on the market yet, preliminary testing material for the future uprades, and surplus product based on projected market growth. the factories produce a phenominal quantity chips every week. "hot lots" travel from beginning to end of line in a very short time...5-6 lots will fill a very large demand, if you have say 3 factories each with 6 lots accelerated through the line surfice to say...it's not hard to stay up on demand...and it's not hard to adjust to actual demand. intel isn't known to miss many commits...neither is amd for that matter, sometimes it is unavoidable...

neither company is out trying to build market share on 486's...and intel has no need to go back to the producing p2's...though some people would love to pick one up for nothing...the cost to produce them is just too high...but you can bet that they can and do still make piii's and probably will do so until it become impracticle to do so.

my basic point is this...why would you need to underclock a fast cpu? we have more than 1 p4 product going through the line...now there is crossover between those depending on how they sort so a cpu out of a faster product line may be underclocked to run at a reliable speed and sold as lower end but this is nothing close to the norm.

Spawn859 · « **Reply #25 on:** June 14, 2003, 09:58:27 pm »

hey guys, have you ever heard of a copper shim?....

htey cost about 3.95 and will save your athlon processors from damage from a heat sink......

Also if you want to overclock an amd processor use the pencil trick. ie use a .2 mm mechanical pencil and a drivers license or any kind of card to cross the l1 bridges together with the pencil, graphite will conduct electricity... here is the article

http://www.motherboards.org/articlesd/how-to-guides/41_1.html

radiator · « **Reply #26 on:** June 15, 2003, 07:24:27 am »

hehe, I used the 'pencil trick' on my first athlon (1Ghz, running @ 1.4) but it won't work on my athon XP, as AMD has laser cut inbetween the bridges, which means i'd need to fill them with a nonconductive material, then connect the bridges (but not with a pencil, it wont work)


Main	Restorations	Software	Audio/Jukebox/MP3	Everything Else	Buy/Sell/Trade
Project Announcements	Monitor/Video	GroovyMAME	Merit/JVL Touchscreen	Meet Up	Retail Vendors
Driving & Racing	Woodworking	Software Support Forums	Consoles	Project Arcade	Reviews
Automated Projects	Artwork	Frontend Support Forums	Pinball	Forum Discussion	Old Boards
Raspberry Pi & Dev Board	controls.dat	Linux	Miscellaneous Arcade	Wiki Discussion	Old Archives
Lightguns	Arcade1Up	Try the site in https mode		Site News


Unread posts \| New Replies \| Recent posts \| Rules \| Chatroom \| Wiki \| File Repository \| RSS \| Submit news