Can I still overclock a locked AMD chip?

[Cobra]

Just wondering - I know that AMD have been locking the multiplier on their Athlons, since people were buying, overclocking and locking, then selling on.

I assume that since they have locked the multiplier on the CPU, it is still possible to gain an overclock by upping the FSB. The 2500 has a multiplier of 11, so upping the FSB from 166 to 200MHz should give XP3200+ speeds. Even though it's locked, is this still possible?

I'll be building my system a week tomorrow. Not to say I'm looking forward to starting that baby up and seeing it fly!

If Hardcoreruss is reading, my PC fetched £150!

Cobra

[bugster]

Yes, you can overclock an AMD by increasing the FSB. As with all overclocking ventures, nothing is guaranteed.

[Cobra]

Fingers crossed then!

Thanks,

Cobra

[Piccoro]

Athlon XP-M 2500+ are unlocked.
The moble chip.
Same chip except AMD cherry picks the best CPU yields since they can run at 1.8ghz @ 1.4v rather than needing 1.65v. They use them for moble chips in laptops but they are the exact same thing. Buy one of these and you are getting a very nice chip and should be very OCable. The question is how OCable.

If you stick with the standard XP chip with locked multipliers, best to get a nForce2 board which as PCI/AGP locks vs the VIA boards.
OC the FSB and you are also OCing PCI/AGP clocks. An unstable system is usually what you get without the locks when OCing FSB.

[saurabh_fzr]

This one of mine is slightly off the subject of the post, but if you have an option what would be better playing around with FSB or Multiplier.
Personally I think it should be multiplier, especially if your FSB is already 800 MHz, I could be wrong also comments please .

Saurabh

[Cobra]

I will indeed have an nForce2 motherboard - I will be using an Asus A7N8X-E Deluxe.

saurabh_fzr, AMD chips are unlike Intel chips. Intel quad-pump their FSBs, so a 200MHz board-side becomes 800MHz CPU-side. AMD double-pump FSBs, so 200MHz becomes 400MHz CPU-side. AMD became wise to the fact that leaving their multipliers locked was losing them money, so they locked them up to stop people overclocking the chips and selling them on at a higher price. I was originally going to get an XP2800+ so if the overclock should fail I would still have a speedy chip. However, in light of the multiplier lock, I chose an XP2500+ because the multiplier is 11. 11 x 200 = 2200MHz = XP3200+, which is my target speed.

I will overclock no higher than FSB400. All I want is XP3200+ speeds and I'll be happy. Therefore, I can't see that the rest of the board, such as PCI ports etc. will suffer since they are all designed to run at DDR400 speeds.

If this is wrong, someone please correct me!

As for cooling, I am using a Coolermaster Aero 7+, which is a full-copper heatsink with a big blower on top. There are also a few case fans and a dual-fan PSU extracting directly off the CPU cooler. I will be using Coolermaster premium thermal compund instead of standard TIM (Thermal Interface Material) to give that extra boost. Heat should not be a problem!

Thanks for the input. My order should be arriving on Tuesday, so I will be building up on Friday night. Hopefully the next time I post on this thread will be on my nice new machine, to say the overclock was a roaring success. With all those fans, I expect it will be a roaring something, anyway!

Thanks again,

Cobra

[Piccoro]

Intel already locks their multipliers so why not join in. hehe

I would like to have both Multiplyer and FSB option.
Normally you increase the FSB til you cannot any further because the total clock of the CPU reached its limit at that voltage. Lets say you have a 2200+ (1800mhz = 133*13.5). Lets say you increase the FSB til you hit 170 and hit a wall because the clock speed is 2295mhz and you can't go beyond that because of the board or the voltage. Normally you are stuck there.

If you can adjust the multiplyer then lets say you knocked it down to 11.5 (11.5x170=1955mhz).
The total clock is well below what we were able to hit before. Why would we want to decrease the multiplyer?
Well, you should be free to increase the FSB to 200mhz now.
11.5*200mhz=2300mhz
Not only do we have the same 2.3ghz clock but we now have a lot better FSB.
Some older boards are limited to their FSB so you could increase the clock even more by upping the multiplyer. My old MSI KT266 board couldn't go beyond 145-150mhz fsb.

[Cobra]

I'm in luck then, because all the AMD Bartons run at 1.65V. So I take it the voltages are locked too, then, as well as multipliers? It'd be nice ot take it up to 1.7 or 1.75V for stability!

Cobra

[Piccoro]

The mainboard adjusts the voltage.
Most can do 1.7 and I think the Asus nforce2 DX can do 1.8/1.85 or something.

[Cobra]

I was thinking about that and I relised that it would be fairly impossible for AMD to lock a voltage!

That's good then, I plan to do a DivX encode (three or four minutes of video on slowest) on the chip before overclocking and take speed and temperature measurements, then overclock it and try the same thing again. I can post up with the results. I understand that they'll not hold true for everyone but it'll give a rough idea of how much speed you gain by overclocking the 2500.

Thanks,

Cobra

[rallynavvie]

You can still unlock the processor, but it involves modifying it and thus voiding any sort of warranty. However there are shims that will do this mod for you and not void the warranty on the CPU. It's called the XP-TMC. Here's the link:
http://www.upgradeware.com/english/product/xptmc/xptmc.htm
With this thing you can change the multiplier and not get screwy results by increasing FSB. Read the FAQ and specs on their site for better info.

Someone here may be using them to spoof certain processors into thinking they're MP 2400s

[Cobra]

I have read in a lot of places now that upping the FSB board side on an Athlon XP2500+ from 166 to 200 will make it into a 3200+. It will even POST as a 3200. I don't really need to be changing the multiplier, but I originally asked the question incase the processor forced BIOS to revert to standard settings. You know when you first install a CPU to a mobo, and it knows the settings already? That's what I was worried about.

All sorted now, though, and I'll post up the results of the overclock when it's built.

Cobra

(Edited once due to typo)

[rallynavvie]

Just realized the XP-TMC won't work on "locked" chips. Oy, I need some coffee. Some of the so-called "locked" chips are not even locked.

What are you using for cooling on that? OCing a 2500 to a 3200 is a pretty big deal since the Barton 2500 is the lowest clocked AMD chip when compared to its rating. Make sure to have overly adequate cooling AND a very good power supply.

[mujahid7ia]

I have read in a lot of places now that upping the FSB board side on an Athlon XP2500+ from 166 to 200 will make it into a 3200+. It will even POST as a 3200. I don't really need to be changing the multiplier, but I originally asked the question incase the processor forced BIOS to revert to standard settings. You know when you first install a CPU to a mobo, and it knows the settings already? That's what I was worried about.

All sorted now, though, and I'll post up the results of the overclock when it's built.

Cobra

(Edited once due to typo)

Yeah, mine POSTs as a 3200+ now
How can you tell if it's locked? I tried changing the multiplier from 11.5x to 11x and it wouldn't boot until I cleared the CMOS using a jumper. Does that mean it's multiplier locked?

[Cobra]

I think that if it reverts back to the original multiplier, it could be a locked chip. I can't see why you'd be having trouble taking the multiplier down again - surely AMD wouldn't have a problem with people underclocking their chips.

What chip are you using to have a mulitplier of 11.5? Would that be an XP2600+ Barton?

Cobra

[mujahid7ia]

No, actually I'm using a 2500+ Barton, and I meant from 11x down to [I don't remember].

[Cobra]

If you were running it at a multiplier of 11.5 then it can't be locked:

166 x 11 = 1826MHz, which is XP2500+ standard clock speed
166 x 11.5 = 1909MHz, an 83MHz boost over XP2500+
200 x 11 = 2200MHz, XP3200+ speed
200 x 11.5 = 2300MHz, a 100MHz boost over XP3200+

If you take that chip below the multiplier of 11, then it's going to be underclocked.

I don't know if I'm missing the point here - I'm a bit nackered right now!

Anyway, I'll be building this time tomorrow night! I hope it works, or else I'll need to wait ages to get it (since I have to travel home from Uni to build it). In all the machines I build, it seems that the hard drive is the most common thing to be dead on arrival.

I'm tired of having to use the library PCs at uni - the one I'm using just now is an ancient Celeron machine running XP Pro - it takes a full minute to log on!

Anyway, fingers crossed - the next post I make here should be on my new machine!

Cobra

[Piccoro]

nvm

[mujahid7ia]

If you were running it at a multiplier of 11.5 then it can't be locked

I wasn't running on 11.5; I attempted to run on a multiplier lower than 11x and it wouldn't boot.... Does that mean I'm multiplier locked, or should I try raising the multiplier to check?

[Cobra]

I really don't know - I would have thought that if you can't change the multiplier at all, then it's a locked chip.

I put my PC together tonight. The chip I got (XP2500+ Barton) was GREEN! It was definitely a Barton though - the CPU die was longer than normal and it said XP2500 on the chip itself.

I'm running an Asus A7N8X-E Deluxe motherboard.

When I try to up the FSB to 200MHz, it fails the POST and remains on a black screen, and it says (talking through the speakers): "Failed power on self test due to CPU overclocking". It then refuses point blank to boot until I drop the CPU back to 166MHz. I know this is wrong.

Does anyone know how I can bypass the Asus test for the CPU to allow my system to boot up without the board stopping me?

Thanks very much!

Cobra

[Cobra]

Sorted the problem this morning: I missed one setting in BIOS that allows overclocking good-style.

I have managed to push the chip to FSB 191, so with the multiplier of 11 it's running at 2101MHz (over 1833MHz). The VCore is up from standard 1.650V to 1.825V.

Funny thing I noticed, the machine passed all burn-in and error checking tests - the CPU was flawless at this speed but when performing a full scan with Norton AntiVirus, it would be like someone pressed reset. Weird! Upping VCore from 1.800V to 1.825V sorted that out though and it's performing perfectly.

I am running a max heat burn-in just now but it looks like it's not going to get over 60 deg C. I believe this is acceptable, and therefore I don't want to push for FSB200 because I'd need to increase the VCore even more, meaning more heat. I'm not sure if it's even safe to do so, as in I don't want the CPU dying after six months!

Anyway, it's sorted now. I have a machine that flies, and I'm dead chuffed about it. Once I sort all my stuff out on it and make sure it's working fine, I'll post up a comparison in DivX and DVD encoding speeds to try to illustrate how much of a boost overclocking a lower processor gives in real terms.

Thanks for the suggestions and advice!

(A damned fast) Cobra

[mujahid7ia]

Wow, everyone I know (including myself) with a 2500+ has been able to go to 200MHz FSB easily, no VCORE changes. Are you using default HSF? I'm using a ThermalTake Volcano 9+ and it cools OK, but the Max speed was SO DAMN LOUD it was like a F/A-18!!!

[Piccoro]

I'm not sure if you are tying to run the FSB and Memory in a 1:1 ratio or not.
If you leave the memory at 166mhz then you may beable to clock the FSB higher.
Or you can increase the timings on the ram and try both the ram and fsb at 200.

I really would'nt run it at 1.8volts. Thats my opinion though. I just think that would reduce the life of the cpu. Not like 6months but I don't think you would be hitting anything around 2 years @ 1.85volts either (or 1.8 for that matter).

[mujahid7ia]

I'm not sure if you are tying to run the FSB and Memory in a 1:1 ratio or not.
If you leave the memory at 166mhz then you may beable to clock the FSB higher.
Or you can increase the timings on the ram and try both the ram and fsb at 200.

I really would'nt run it at 1.8volts. Thats my opinion though. I just think that would reduce the life of the cpu. Not like 6months but I don't think you would be hitting anything around 2 years @ 1.85volts either (or 1.8 for that matter).

If the memory is PC2700, then it might be wise to keep it @ 166. But, if you have PC3200 RAM, such as myself, definitely take it to 200 and it will be stable.

[Cobra]

My RAM is PC3200 and is running in sync with the FSB.

It just isn't stable past this point without more and more voltage.

I'm a little disappointed, but at least it's given me a good speed boost for free.

Cobra

[mujahid7ia]

Is it name brand? What company? Mine are Corsair ValueRAM. If the RAM is PC3200 you should definitely reach 200MHz, its designed to handle that? Bad RAM??

[Cobra]

That's a good point - I bought the cheapest RAM I could lay hands on. It's branded MATRIX. It is rated at PC3200, but you're right - it could be dodgy.

Asus reckon it's the chip itself, they say I've just been unlucky and got one that doesn't take well to overclocking. The A7N8X-E supports (unofficially) an overclock up to FSB250 and multipliers up to 20x. Reckon I could squeeze 5000MHz out of it, perhaps with a VCore of 12V?

I'm only 100MHz away from XP3200 speeds, so I'm not too bothered just now. I set my memory timings to agressive mode, but it makes very little difference. My machine is perfectly stable at FSB191 (=2101MHz). That's still +14% on both clock speed and FSB, for free.

Maybe next time, eh?

I have ripped one of my DVDs (Pirates of the Carribean) and I'll test how much of a boost this overclock gives me in real terms as soon as I can.

Cobra

[mujahid7ia]

That's a good point - I bought the cheapest RAM I could lay hands on. It's branded MATRIX. It is rated at PC3200, but you're right - it could be dodgy.

Asus reckon it's the chip itself, they say I've just been unlucky and got one that doesn't take well to overclocking. The A7N8X-E supports (unofficially) an overclock up to FSB250 and multipliers up to 20x. Reckon I could squeeze 5000MHz out of it, perhaps with a VCore of 12V?

MATRIX? lol

Yeah, could be the chip or the RAM, but I think its that cheap RAM, PC3200 incapable...

[Ducatti20]

What kind of temps are you guys getting overclocked?
It probably is the ram holding you up... Corsair is the overclockers champion...
If you are hovering over 45C idle and over 50C under what you consider load, you might want to consider liquid cooling... my system was hanging out around 50C load with my fans drowning out all other noise... then I built my own water cooling system and hear my hard drive spin up..... it's that quiet and temps are under 40C load now..

[Cobra]

It's a shame that I can't test the RAM itself at DDR400 to see if it is what's embuggering my system.

I had to clock back to FSB189, so 2.07GHz now because it was encountering errors when encoding DivX. The 30MHz drop seems to have done the trick - it's deadly stable now.

My system runs 42 deg C idle and under a burn test maxed out at 59 deg C. My cooling is as follows:

- Steel case
- One fan at front, bottom of case IN
- One fan on side window, IN
- One fan on rear of case, between graphics card and CPU cooler to intercept and extract hot air, OUT
- Dual fan PSU extracting directly off CPU cooler, TWO OUT
- Coolermaster Aero 7+ full copper HSF, blower fan

It's one helluva lot of fans, but it's surprisingly quiet. There's not really a fan sound as a rush of air, so it's easy to work around the machine.

I would love to have water cooling, but I can't afford it. This machine cost me £250 (including sale of my old PC for £150 and salvage of graphics and hard drive). That literally broke the bank - I need to be really careful for the rest of the semester.

I am finding this machine is more than fast enough, and I'll try to do a deal with the RAM at some point. At least when I need to upgrade my machine, I can drop in new RAM and a new CPU and off I go - unlike my old Dell!

Now that I have finished tweaking for stability, I can finally start to test encoding times. I have an encode running right now, and I'll post up with the results ASAP, assuming it works!

Cobra