Slashdot Mirror
Celeron overclocking mania
This month's C'T is out,
with a
review of the Celeron 366 and 400 which exist in a socket
370 variety. Intel first moved to slots because the L2 of the
Pentium Pro was expensive: it took up Fab capacity for low margin
parts (SRAM). Cartridges could use others' SRAM. But now, the trend with smaller processes, is to put
the L2 on the die, making cartriges expensive. Hopefully this
won't mean the end of the SMP-Celeron days.
Intel is also expected to add a circuit which prevents over-clocking in its
chips... a shame since Danish
company Asetek, which makes CPU cooling equipent,
claims they can overclock Celerons
to 600Mhz, while running at 550Mhz has been
independently verified in Ohio.
do you have any idea how many second-tier and third-tier vendors are out there selling overclocked systems? Intel will be doing everyone a favor by putting a stop to this.
Why should *I* be punished because some criminal dealer wants to sell fraudulent boxen?
I'm considering building a dual celeron system and doing the modifications, and I'm wondering if anyone has done it and has any comments.
I've never really done any serious hardware hacking, but the procedure doesn't seem too difficult. Is it easily reversible though? Can I plug the hole I drilled with a conductor and expect it to work?
Johan Jansson
Care to elaborate as to WHY this is good?
If Intel wants to sell us the same damn chip core at 5 different speeds, well, ok, fine, but I personally think it's foolish to buy a good core that is rated for 300MHz and not try to get it to run at 400MHz or higher. There's no difference between much of this stuff. My 300 runs at 450 without a hitch, and it's not some kind of dangerous trick. It's Intel messin' with your head, selling you things that are way underpowered and under rated.
Check out the Celeron FAQ for more information on how easy it is, and on how stable these systems really are.
That should read:
"This fortnight's c't"
There's a new one every two weeks.
that's the problem, many (lower-tier) vendors are overclocking the chip and not telling the customer. the issue is not simply a case of Intel screwing the customer (although their prices are way too high without overclocking)...
When we designed the first 486 notebook, cooling the cpu was our chief concern. One of our engineers came from an aerospace background. He suggested rubberized aluminum substrate as the contact element with a variety of other layered bonding agents as dispertants.
This became our super-secret "heatpipe." and we used this medium as a transfer the heat to a traditional heatsink.
Bottom line...our startup company got bought out by an industry heavy weight so they could own the technology.
Ehhhh...that was in '91 and possibly a miracle then and my non-disclosure has expired. I don't profess to even pretend to understand the theory or chemistry....but it worked....
We put a 486-66 over clocked in a TI-2000 case.
TLTCA (to lazy to create account)
Z
The procedure is pretty simple with the proper parts. Get a #75 drill (part# 5876-409-75 from Mouser Electronics) and wrap the shaft with tape until it will stay in your drill. Better yet, get the right drill, but those things are expensive!.... RadioShack sells this "High Tech Silver-Bearing Solder" that worked great. I belive it is .015" in diameter. I had some trouble with some other solder the first time. Make sure to grab some desoldering braid if something does need to be taken off.
Actually drilling the pin is pretty quick. Make sure to mark the depth on the drill bit so it doesn't go too deep. It may help to tape down the wires while trying to solder them. By far the most difficult part is soldering the "bridge" on the RP6. I managed to get it without a magnifying glass but wasn't easy. Don't keep the iron there too long or it may take it completely off.
Technically, I belive the chip will still work with the drilled hole. I don't think it serves any purpose. The pair I did works great now but had a problem at first because one of the wires was shorting out on the metal heatsink. They don't seem to be stable at 450 MHz for some reason, but that is probably a cooling problem or something else.
Socket-370 uses the same protocall as the Slot-1 interface. It just looks different. I think there are even plans to make Socket-370 to Slot-1 adapters. It has absolutely nothing to do with Socket-7. Intel did this because it's cheaper to make a socket chip than a slot one.
I used an electric 3/8" drill. Big overkill though. The material is pretty soft and goes through pretty fast so a hand powered drill should be fine.
Where does it say hardware engineer in your title?
First the new Celeron fits in Socket-370, not Socket-7. Second it uses the same P6 GTL+ based bus which is completely incompatible with Socket-7.
Unfortunately, considering that all the differently clocked chips come from the same die and are tested for what their ``recommended'' speed is, which is then stamped onto the chip casing, it would be pretty difficult to encode that info into the chip itself.
At least I assume that's how they still do it.
Sounds pretty warm to me. If you are using a OEM Celeron, did you use heatsink grease? If not, do...available for $2 at Radio Shack. Also, my Celeron heatsinks came with thermal pads with were almost 0.5 mm thick. I scraped them off and used a thin layer of grease instead. Then, I superglued a 3-inch fan perpendicular to fan on the heatsink, so that the air stream blows across the heatsink. And of course, you *need* to have a inlet case fan if you intend to run the system with the case on. Total "cost" for extra cooling mods...about $8 for the fan, grease, and superglue. My CPU temp generally stays below 100 F at 450 MHz and 2.1V.
Well, I for one do not believe in everyone's God-given right to exploit other people.
This is simply a case of one group of people using force to steal the resources of the many to use for their own private good (a description that indeed fits all capitalism). I see no reason to tolerate it.
/Vilhelm Sjöberg.
>Intel is also expected to add a circuit
>which prevents over-clocking in its chips..
My guess on how Intel will *actually* do this is by one of three ways: with a fuse, a circuit, or microcode.
Adding a circuit & transistors introduces the problem of reliability & cost, & I know a lot of Intel employees routinely overclock their processors at home, so I'd consider this to be the least likely path. (It was an Intel employee who introduced me to Tom's Hardware Page.) Fuses can be blown. (It just takes a little research & trying.) And microcode -- like any software -- can be reverse-engineered.
Hmmm. It looks like I'm making a case for Open Source CPUs, doesn't it?
(Anonymous because I might have disclosed what I shouldn't have.)
I'm not against anyone overclocking, as long as the product is guaranteed. I don't really *CARE* what the manufacturer of the raw parts SAID they will do, I care about what it WILL do.
I want the ability to fry my chips & boards at will... not some stupid protection mechanism that won't let me.
If we don't demand what we want, then we won't get it.
Well... not every "cam shaft lube" (?) would be a
silicone grease. Also, most heat sink greases
are a grease, plus a filler compound that has good heat transfer capabilities, therby giving the mix a better transfer than the plain grease.
I'd pick a "real" heat sink compound.
Jon Pike
I have done it several times... I use a #74 drill, .020 inches, and chucked it up in a Xacto knife handle. (one of the type that has a double split blade holder, not a single) Works pretty good as a drill chuck..
:-) /. some months back.
:-)
Definatly do it by hand, with a drill motor it would be extremely easy to drill too far, break the bit, etc.. Getting a very sharp and tiny solder tip for your iron is a must, too..
Seriously, a microscope is a big help. Its pretty tiny stuff. Also, I use high temp tape for masking. I have a lot of SMT experience, which helps.
I'm the guy that will do this for $100 a pair, if you aren't the adventurous type.. or drink lots of coffee!
I've posted on
Even though I get many who sound interested, then dissapear, I'm still doing the service for those who want a pair, professionally done. Truly interested parties only...
I have a home system that runs a pair at 450/464, very stable under Linux and NT, for some months now.. been beating it with some pretty arduous tests.
You have to pick a good pair, before you start, is a key thing, I have discovered.
Oh, and to the question of can you "go back" after the conversion.. There's nothing to keep a processor from running single with the mod, THOUGH, if you drill the hole, you probably should leave the pin attached to the pullup resistor, rather than floating, that could be a problem. If you unsolder that resistor pak by accident, that may disable the cpu from running as well...
Good Luck, (or mail me!!)
Jon Pike
jonpike@deltanet.com
NOOOO! The thermal conductivity of plain silicone is pretty low unless you add an appropriate flake/powder filler such as nickel.
Dave K
daking at infinet dot com
Sounds high, but then again Intel rates it up to 80C. My setup is:
CPU: 300A OEM, don't remember the stepping.
Core voltage: Set to 2.2V, actually runs 2.1V. At the next lower setting of 2.0V (actual 1.9V), it would crash on occasion in Win 95 but seemed fine in Linux and NT4.
Speed: Runs 100% stable at 464 and below. Crashes frequently at 504 MHz. I run it at 450.
MB: Asus P2B
Heatsink: Single fan shrouded PII sink drilled to match holes in CPU card and attached with plastic screws & nuts. HS pad removed and a thin film of RS heat sink grease applied.
Case: Mid tower with case fan in front and extra fan at top rear. PS fan blowing out of case.
The P2B uses a Winbond chip for HW monitoring and has connectors for 2 thermistors. I have one thermistor mounted right onto the center of the heat sink, covered in the thermal grease. I read anywhere from 37-40C in Win 95 and 28-35C in Linux depending on ambient air temperature and what I'm running. Others have reported that the Winbond chip + P2B reads approximately 4C high across the board due to a bug, so that puts my max temp at 36C. Of course, this is on the other side of the heat sink from the CPU so there is probably a temp gradient of a few degrees across the 2-3mm of heat sink metal + grease.
Dave K
overclocking is stupid intel crap. you don't see overclocked ppc, sparc, alpha, etc. overclocked.
Pardon my Ignorance, but why is all the hardware hacking needed? I've been considering a dual Celeron option for a while, but had no idea that I'd need to tinker with the hardware like that.
:)
--=>*=- Matt (waiting for passord in the mail
Pardon me but alpha cpus HAVE been sucessfully over clocked. Go over to www.kryotech.com and look at the 600mhz alpha over clocked to 767mhz. Any cpu that uses an external clocking source could be over clocked. The problem is that the motherboard for that cpu may not provide any functionality to adjust said external source or that a massive amount of cooling may be needed (kryotech is basically building a bloody refrigerator around the cpu). The celerons are just all of the rage in over clocking because it doesn't require a major hardware investment to do it. With the .25 micron technology in my celeron 300a, it runs quite a bit cooler at 464mhz than the p2 300mhz cpus that I have in my smp machine do at their rated speed. I'd love to push an alpha or ppc or ultra sparc just to see how far they would go. But, I don't want to spend the money it would require to refrigerate the things. Personally I'm a Motorola fan. Motorola cpus have been better than comparable Intel products all of the way back to the pre 68000 days (I think that was around the 8502 cpu era?) Don't be so critical of people trying to get the most of their machines. If you want to run your hardware slower than it's potential... I suggest you visit www.microsoft.com. They'll hook you up.
The box im writing this on right now is a Celeron 300a clocked up to 450. And thats just with a regular fan, nothing extra-special cooling wise. Im sure with a three fan "polar bear" I could get get it higher but im not going to press the issue, a 450 for $100 is fine by me.
course, i'd love to be running a G3 but its all about finances, sigh.
2^5
Intel isn't locking down overclocking to prevent counterfeits; That's just a convenient side-effect. They just want to stop over-clockers.
Hmmm...If I decide to soup up the motor in my car, Dodge doesn't seem to object (besides screwing my warranty).
VENI! VIDI! VICI!
Uh, Socket 370 isn't socket 7.
Ok, that would make sence.
Actually you can use PROM type technology: just burn the relevant values into the register.
Is Intel preventing overclocking due to complaints and returns due to reliability and consumer warranty returns? If so, why not have a section of the CPU dedicated to performance management that has a PROM that records runtime conditions and averages? Things like runtime maximum and mean temperatures and clocking can be recorded. Adding the traces and the read only instructions to access these parameters that could prove any consumer abuse of a returned chip would take less than a thousand transistors.
That way, those of us who want to void the warranty and experiment with full performance can. I am using a Celeron 300A at 464MHz (504 was tried for a few minutes once...) It runs cool and I doubt ion migration will hapen at cool temperatures.
Actually the technically best solution (for end users) would be to encode the chip with a unique identification code that would tell you what speed the CPU is supposed to run at, but wouldn't prevent fools from doing what they will with it.
Need to get some high quality RAM, a case, and a HD. Oh yeah, and a "Celery Sandwich".
About the ONLY thing I'm not happy about is Intel's decision to penalize people who want to experiment with the chips THEY BUY.
I can understand that they want to stop people from remarking chips and selling them OC'ed. But locking the chip down is NOT the way to go about it. You annihilate a market of people who like to tweak the hell out of their systems. Now I can also understand that Intel isn't happy about selling 450 Mhz chips for $80 when they're trying to bleed $400 more out of people. But these people are the legitimate owners of the chips they buy. Why keep them from taking their own risks?
I'll admit. I am NOT a chip designer, but I would think it cannot be that hard to alter the standards for a chip. Have an extra register in the chip that is read at boot time. Something that would read like:
Even if it came up as:
With such a setup, you would PHYSICALLY have to alter a chip (as opposed to just the cosmetic angle of scrubbing off the case and silk screening on another chip ID.
Intel is attacking the wrong group of people with this. They're not hurting the remarkers any. There's TONS of 300a's out on the market and 450 Mhz will be a pretty good standard for another year or so yet. They'll still make a killing ripping people off regardless.
Ah well. I'll take what I can get while it lasts.
Chas DN Team: 4688
Chas - The one, the only.
THANK GOD!!!
Chas - The one, the only.
THANK GOD!!!
Where do you get them and how much do they cost?
Well, google is my new found friend, and I was actually looking for something that would allow me to put a slot 1 cpu into a socket 8 mobo.
...for coming back to the Socket-7 format, yet making their new mobos incompatible for no good technical reason! Grr.
I am against manufacturers overclocking, but I don't want it to be stopped at my level. My Celeron 300 cost me $100 and outpeforms a PII 350 at $300 or so, by overclocking it. If Intel wants to stop overclocking, they need to come up with a solution that reports the origional safe running speed and current speed at boot, so a consumer would be aware of the overclocking. But don't stop the chips from doing this.
On another note, there are Slot 1 Celeron 400's, and the Socket 370 varity. Intel also released Socket 370 Celeron 300 and 333's when the new Celerons came out. So all those looking to upgrade, don't toss that Slot 1 board. Also a company is working on a Socket 370 -> Slot 1, similar to the Socket 8 -> Slot 1 converters for PPros.
I have to say that that is 100% wrong! What intel should do instead is just modify the cpus so that they have an easily machine-readable "recommended" cpu speed on the cpu. The bios could then probe this value and if the speed that the system is running at does not match a warning could be printed. Intel is NOT conserned about the end user but about their own profit margin. They are doing no one a favor but themselves. They have realised that people finally figured out that they market the exact same cpus at different clock speeds and prices. They've been doing this atleast since the 486. It really pisses me off that they even think that they have the right to tell me how I use MY cpu.
-matt
Must every industry company force people to pay for overpriced products? Intel obviously doesn't like the fact that you can get a 450 mhz chip for less than the ~$500 they charge for the P2... you can get a Celeron 300a for $90. Now they're releasing their equally over-priced "Pentium 3", a terribly misleading name given that it's not a 7th generation chipset (like the K7 is). Thank God for AMD....
http://www.frostyplace.com/NewsHTML/1997/G3_Overcl ock/OverClock.html seems to show different information than what you've given. I've seen over clocked iMac's (scary!!!). Remember, it's not what your running on, or what your running that matters. What your doing is the important part.