• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

2003 week last unlocked Barton was manufactured

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.


c(n*199780) Senior Member
Feb 18, 2002
EDIT: Dec 19, 2003, It's finally official: ""...we now have a policy to lock down the frequency of all the products so there’s one product that will continue to be open and it’s the FX product line."
Henri Richard, AMD Senior Vice President of Worldwide Sales and Marketing

For those of you that haven't heard, new Bartons, Thoroughbreds and Durons are factory multiplier locked, (not even wire tricks work with them.)

No motherboard can unlock these CPUs and there is nothing that can be done to bridges or pins that can change the locked multiplier because they are not connected to the circuitry.

We're starting a data base of locked & unlocked chips:

locked: All week 43 (0343) to 46 Bartons and Thoroughbreds and most (but not all) post week 46 Bartons and Durons and:

AQZEA 0338

AQXCA 0340


AQXFA 0339

AQYFA 0342


AQZFA 0342 TPMW and UPMW (multiple confirmations of both locked and unlocked chips)


Thoroughbred JIXIB 0339
Thoroughbred AIXHB 0340 UPMW
Thoroughbred AIXHB 0341 SPMW

unlocked: Most made before week 35 and:

AQXFA 0340 MPMW and TPMW and UPMW and VPMW

AQZFA 0342 TPMW and UPMW (multiple confirmations of both locked and unlocked chips)

You see there is no stepping or week pattern, the only way to increase your chances is to try to get a pre week 35 Barton...

There is no visible physical difference between locked and unlocked chips because it appears changes were made inside the CPU core itself.

c627627 said:
You can't tell the difference between locked and unlocked new chips with a naked eye or through codes.

The best we have is:

• 2003 week 46 and later Bartons and Thoroughbreds are locked.

• Some weeks 35 to 43 are locked, some are not, and that includes Thoroughbreds, Thortons and Durons.

• There are rare cases of post 2003 week 46 unlocked Bartons and Durons.

EDIT: There are no hidden pins, + Hoot took one for the team, if anyone can help him out, read about his sacrifice at the end of this post 1.

EDIT: Multiple reports now of Thoroughbreds weeks 39 and 40 being locked just like new Bartons.

EDIT: L1 and L3 bridges are closed on both locked and unlocked Bartons. However, on the locked Bartons, they do not connect to the core points which lead to the Internal Multiplier Control.

No motherboard can unlock these Bartons and there is nothing that can be done to bridges or pins that can change the locked Barton multiplier because they are not connected to the circuitry.

UPDATE: If it's not a permanent electronic fuse we're dealing with: * a hidden control pin may exist * Details of how to look for one are below. Thank you hitechjb1 for spelling it all out. EDIT: It does not exist.

EDIT: November 9, 2003: Official AMD Forums moderator: "...this is business and AMD can't afford to keep shelling out unlocked 2500+s while losing $300 on each chip."

EDIT: 2500+ Bartons were the first to be reported locked but Singapore is now first to report a 2600+ Barton lock, no word yet on locked 2800+ and up Bartons.

EDIT: AQZFA week 42, UMPW is being reported as unlocked in Australia. There were conflicting reports of other same stepping and week 2500+s being locked and unlocked but now the entire CPU markings AQZFA 0342 UMPW match for both locked and unlocked 2500+ Bartons.

If this is true, game's over as far us trying to figure out steppings & codes for locked Bartons, AMD locked whatever they got their hands on that happened to be in stock.

Majority of the locks were done to post week 39 CPUs. Some appear to have shipped out before they were locked.

Since Thortons are Bartons with 1/2 the cache, they're 2000+, 2200+, 2400+ see table here: http://www.c627627.com/AMD/AthlonXP/), there's a good chance all lower rated Athlon XPs week 43 and later are locked and many of the week 39 through 42 are as well.

Also see the CODES thread for info on how to tell a Thorton from a Thoroughbred:

* * *

Note that you may have to save & exit a second time in BIOS to get Abit NF7 to accept the new multiplier if you have an unlocked CPU.


ol' man said:
AMD is locking them because this summer someone was remarking them after locking them at a higher multi and then reselling them.

Same reason intel started locking them too back in the late 90's. Some one was buying the 233 chips and remarking them as 266 and 300MHz chips. I mean they did it too alot of them. Supposedly this summer resellers where getting whole shipments in with some of the CPUs being remarked.

EDIT: The so called Super unlock:

Basically, the method involves modding of the locked chips.

'PowerNow' feature allows voltage and multiplier changes in order to preserve battery life.

So if you have motherboard/chipset/BIOS support for PowerNow which not all boards do, nForce2 does not, you can use this feature on locked desktop CPUs.

You also need to boot into Windows with your locked multiplier and the FSB with which you can boot into Windows but you cannot change the FSB inside Windows, only the multiplier. Ironically, you can change the FSB too inside Windows with nForce2s but they have no PowerNow feature so it's of no use with super locked chips.

So with this so called super unlock method, remembering that MHz Speed = [FSB] x multiplier

FSB can be changed only before you boot into Windows but
multiplier can only be changed after you boot into Windows.

...with limited success so far.

It's not a real unlock at all.

stamasd said:
1. closed L5, 2nd bridge from the right with pencil

2. cut L8 for 1.5V according to the mobile chart (same as L11 for 1.6V), i.e. C:C:C (see notation on Fab51's page); actually I may have cut the first (from the top) bridge on L8 as well, because CPUMSR shows a default volytage of 1.45V instead.

3. cut L6 for a max multiplier of 19x. Why 19x? It's high enough that you actually never get to use it, and it's very easy to do, the highest multiplier which can be set with just 2 cuts :CCC: - in other words, cut the top and the bottom bridges on L6.

That's it. It just works, on a A7V600 mobo. It's simple enough that I will probably mod the same my other CPU, an unlocked XP2500+. I've began to like changing multis from within the OS, and I don't see any downside to it. For use on the A7V600 I'll probably even leave out the L8 cuts altogether, since I set the Vcore manually in the bios anyway, and the "default" Vcore means nothing. *NOTE* the L8 cuts may be required on other motherboards, if they read it - you don't want to accidentally push 2V in your chip. Also note that cutting bridges on L8 will not change the voltages needed to run the CPU - if before the mod it needed 1.8V to reach 2200 MHz, it will still need the same 1.8V after the mod for the same speed.

For a locked Barton? I dunno, probably the same. The only thing that may be different is the location of the L8 cuts, if you want to set a different default Vcore. IMHO there's no need to do all 5 cuts for L6, I can't imagine a setup where you'd want to use a multiplier higher than 19X on these chips. Unless you push the FSB down to 100MHz of course. :)


hitechjb1: Are there any hidden multiplier pins from the NC said:
The main idea is to see

1. Whether there are some VCC or VSS pins, measured from VSS, have large resistance, say, > 100 ohm. Normal VCC to VSS should be much smaller, and of course normal VSS to VSS should be zero.

2. Whether there are some NC pins, measured from VSS, have relatively small resistance, say, < 10 M ohm.

3. All resistance measured from VSS is sufficient, at least for the first trial.

4. In principle, ONLY measuring VCC, VSS and NC pins from VSS is sufficient. But if one can find someone to help, may well just measuring every pins and put the resistance number from VSS in the pin layout map and analyse them after the measurement.

AMD pin layout map, on Page 55

If suspecting there are some hidden pins (not specified in data sheet) among the
- NC pin group, or
- VCC pin group, or
- VSS pin group,
one can make a resistance map or table of each pin to VSS, and look for difference.

With a probe of a DMM connected to VSS, measure the resistance of each pin to VSS, and make a table. Then compare the resistance table to the pin table. One may want to measure each pin twice with the polarity reversed.

It should not be too difficult to do, especially if one can find a helper to write down the numbers while doing the measurement.

If a pin can be measured in 4 sec, a single pass of 478 pins would take about 32 minutes. Two pass would be about an hour.

If there is/are multiplier controlling pin(s) hidden among these NC, VCC, VSS pins, the resistance of that pin(s) would stand out differently from the other regular NC, VCC, VSS pins.
- If a pin is not a VCC pin or a VSS pin, and if it goes to the input of logic gates, or FET terminals, even if there is pullup or pulldown impedance, would be >> 10 ohm.
- If a NC pin turned out to be not NC, then the resistance would be much smaller than 10M ohm.

Have to set the DMM resistance to x1 or the lowest setting, and keep it that way, so that the setting and biasing point from the DMM to the pins remains as constant as possible.

I checked a few NC, VSS, VCC pins using a 1800+.
- The resistance from VCC to VSS should be very low, of the order of 10 ohm or less (exact number depending on the DMM) and polarity. It should be similar for locked and unlocked Barton.
- Also the NC pins, if indeed is NC, should have very high impedance (>> 10M ohm).

Hoot said:
Okay, here's the scoop on the possible existence of a second set of pins. I have spent the past three evenings ohming the 5 known BP_FID pins on my unlocked Barton to all the other pins on the bottom and sadly, though not unexpectedly, I must report that none of the other pins, regardless of claimed function, be that Vss, Vcc, N/C, etc, are connected to them.

Now, here's where it gets "creepy". I eventually snapped and did something I never dreamed I would do. On the Locked Barton, I heated the core with a micro hydrogen flame torch, to keep the heat locallized to the core until the sealant around the edge softened, as well as all the solder dots, be they bismuth, indium, or whatever low temperature metal AMD uses, melted. I then lifted the core off the carrier.

Yes, we're talking CPUicide here. Following the traces from the L3 bridges to their destination solder dots in the array on top of the carrier, beneath where the core sits and the dots leading inside it touch, I discovered they were actually connected to the core's corresponding dots just like all the other traces.

That shoots down my theory about the paths leading to the multiplier control circuitry having been etched back, cut, or whatever, before the core was mounted onto the carrier.

So where does that lead us, other than recommending me for committment to an appropriate institution? Either, as some of you have proposed, they gave those circuit inputs a lethal shot of current or there may be some microcode embedded inside the core which can be programmed to ignore the input logic levels to the multiplier control circuitry. I seriously doubt the cores themselves are different, though I'm certainly not going to trash my remaining unlocked Barton to compare them. I'm already having enough difficulty dealing with my self-loathing, not to mention the $90 price tag to satisfy my curiousity.

Score: AMD 1, OC'ers 0

I'm bushed after three nights of late hours and am off to the refuge from my deed, known as sleep.

Just in case anyone who both appreciates the sacrifice and has a real job want to help out, I'm not too proud to accept contributions. ;)

Hoot the Horrible
Last edited:
Bartons work with PC2700 RAM whose default is 166.

You can increase the FSB. However, with 2500+ multiplier locked at 11, you're looking at 11 x
(for example)
[175] = 1925
[180] = 1980
[190] = 2090

That's still below the above 2.3 GHz 2500+ can do.

If you want a higher overclock with PC2700, they want you to buy the 2800+ with its 12.5 multiplier.

newegg today:
2500+ $85
2800+ $136.00

That's 60% more.

Yes, PC3200 will get you a higher FSB and closer to Barton chip's limits but most people use PC2700 (and PC2100) RAM.

...although there's still an option of running async, overall performance is best insync...
Last edited:
..."when everything seems fine"?
They haven't made a profit since the middle ages.

When quarterly report comes out in a few months, it very well may be the first one which will not involve posting a loss since.... when was the last time they posted profits, I can't even remember.

They give us CPUs that cost two or three or more times less then Intel chips and what do we do, we give them like 16% of the market (to Intel's 83%).
aww man i was planning too build an AMD system based upon this chip. were do u think i can get a barton thats not locked? :rolleyes:
Excellent posting!

Wish this had been here a couple of days ago when i started looking for a 2500...found a week 39...it was locked.

now i've found a week 32 (AQZEA) & a 34(AQXEA).
Which one would you choose?

Thanx again!
Last edited:
bleh. i already took appart the amd heatsink to note its crappiness. not sure if i can exchange my chip as that retail package is like... well i think everyone knows about them... bugger.
so even when people set a specific multiplier, and not try that 12.5x and below multiplier unlock, the chip still boots with the default multiplier?

Sonny pointed this out: could it be that below the new packaging the L1 bridges are simply not joined? Could this problem be solved if the L1 bridges were connected if that was the case?
I don't really understand why they would do this when you can still OC by pushing up the FSB. How does this help them make more money?
We have yet to see visible physical difference which not one person has noticed thus far despite posting pics of L1 bridges, etc.

bk94si, post 3 in this thread describes some of the resulting overclocks with PC2700, the most widely used RAM with 2500+ Barton out there: well below the chip's 2.3 or even 2.4 GHz limits.

With locked 2800+ you get the 12.5 x multiplier instead of 11 x, do the math by multiplying the FSB your RAM can do with 12.5 x instead of 11 x.
bk94si said:
I don't really understand why they would do this when you can still OC by pushing up the FSB. How does this help them make more money?

People have to buy faster more expensive chips.

I got mine tho.
i have a week 31 aqxea that i just got from newegg, so if you want an unlocked i would try there.
I just ordered 2 Barton 2500 from Newegg this week for a pair of computers I'm building for my bro-in-law. They should be delivered tomorrow. I'll let you guys know what week they are.

Sure, we're all outraged, but understand that this is the logical thing for AMD to do. However, note that AMD is still leaving die-hard overclockers an escape path; the FX-5x. You now have to pay dearly to enjoy the luxury of an unlocked multiplier, but it is an open option. With people buying their cheapest processor en masse, they need to take some kind of action. Its good business, and in the long run, we'll be glad that they did, as they'll still be around, hopefully for a while.
No, because they've been waiting for two years.

AMD Net losses for the past two years, check out how Q4 2002 ended for them:

Q3 2001: -$98m (Athlon XP launched)
Q4 2001: -$16m
Q1 2002: -$16m
Q2 2002: -$185m
Q3 2002: -$254m
Q4 2002: -$855m
Q1 2003: -$146m
Q2 2003: -$140m
Q3 2003: -$31m

^wargod said:
well couldnt they have done it after i ordered in a week? :(
Last edited:
looking at those numbers it's pretty amazing that AMD did not die off at the beginning of 2003.

now it looks like they might bounce back. I sure hope they do!

I kind of hope Intel really ****s up on Prescott, and can't actually bring a product out. Just so that AMD gets some profits and marketshare back.