Changing T-bred Clock Rates?

Belgarion · Post by **Belgarion** » Thu Nov 21, 2002 7:45 pm

I personally have never had an interest in overclocking, so I've never followed the overclocking web sites. But now I've been thinking about clock rates and underclocking/undervolting for a new quieter system.

I'm thinking that an nForce 2 motherboard (maybe Asus or Abit brand) plus an Athlon XP 1700+ T-bred with some PC2700 would be nice bang for the buck. From what I've read on other sites, I've gathered the following:

1) Some, if not all, T-breds on the market are not multiplier locked (is this really true?)

2) Athlons perform best on nForce(2) boards with sync memory timings

So my question is this: assuming an unlocked CPU, could you up the fsb from 133 to 166 while lowering the multiplier to keep the actual clock speed (about) the same? For example, instead of running at 1467 = 11x133 change it to 1500 = 9x166 or 1417 = 8.5x166. That would improve memory performance by running sync with the 333 MHz DDR without changing the base clock speed of the chip very much. With an SLK-800 and maybe a little undervolting, I'd think it should run reasonably cool (I'm not looking for absolute silence here). Sound ok?

Being somewhat new to messing with non-stock clock rates, I am still wondering about a few things.

1) Is there any reason why a CPU rated for a certain speed at 133 wouldn't be happy with the same speed at the 166 fsb? Are overclocking limits only clock rate and not fsb dependent?

2) Any ideas how much you could undervolt a T-bred without lowering the clock rate significantly? I know the T-breds are supposed to overclock reasonably well; does this imply anything about underclocking?

Well, I hope this isn't too many questions.

TIA

Post by **MikeC** » Thu Nov 21, 2002 8:22 pm

1) Is there any reason why a CPU rated for a certain speed at 133 wouldn't be happy with the same speed at the 166 fsb? Are overclocking limits only clock rate and not fsb dependent?

No there shouldn't be. It's the multiplier that is locked on CPUs, FSB is controlled by the mobo.

2) Any ideas how much you could undervolt a T-bred without lowering the clock rate significantly? I know the T-breds are supposed to overclock reasonably well; does this imply anything about underclocking?

Usually, if you are not OCing, 0.1V down is no problem; some can handle 0.2V with stability, only way to find out is to try it. Good OC ability implies wider latitude for non-standard settings, again just try it.

Have you read the Undervolting/clocking article?

Post by **Rusty075** » Thu Nov 21, 2002 10:19 pm

Be aware that unlocking the multiplier on an XP is not a task for the easily frightened, or for those with shaky hands either.

A good how-to can be found here:

http://www.ocaddiction.com/articles/howto/xp_unlock/

Other than that your ideas seem like good ones to me. Inventive even. Makes me kinda wish I had thought of it. If you do decide to do it try to do some before and after benchmarking to see what the multiplier changes do to the performance. I'd venture to guess that the FSB increases will balance out alot of the clock decreases.

As for undervolting, my 1900 is undervolted from 1.75 to 1.6. Below that it gets squirrely after running at full load. I don't know how typical that is, but it'll give you an idea of what's possible.

Keep us posted.

Gxcad · Post by **Gxcad** » Thu Nov 21, 2002 11:45 pm

I've also heard rumors that the lower speed thouroughbreds aren't multiplier locked (IE unlocked from the factory), can anyone confirm or deny?

-Ken

Post by **Rusty075** » Fri Nov 22, 2002 12:20 am

I seem to remember that some of the review samples that webisites got were unlocked, but that all consumer chips would be locked as usual. I'll try and find a link. Both AMD and Intel have been locking the multipliers in one way or another for a long time now, I doubt they'll stop now.

As a side note, it looks like unlocking thoroughbreds is easier than palomino's. AMD has done away with the trenches between the bridges. So its more like it was in unlocking the thunderbirds. (although the pencil still won't work, too much resistance or something)

Belgarion · Post by **Belgarion** » Fri Nov 22, 2002 4:23 pm

MikeC wrote:Have you read the Undervolting/clocking article?

Yes. Good article, BTW.

That table labelled "Various Undervolted & Underclocked Settings Achieved" definitely has some interesting results. I notice from the last two lines that the heat produced at 7.5X133 is less than 10X100 (due to much lower Vcore in the former case). And yet, the memory bandwidth is still higher (due to the higher fsb). This kind of win-win situation was basically what I had in mind. Take a relatively low-clocked chip and up the fsb while lowering the multiplier. This way you don't have to buy an expensive and hot 333 MHz fsb T-bred to be able to get the 166/333 memory bus.

As for the multiplier lock, the comments I've seen on review sites (don't have any handy but I can try to look them up) and Newegg's XP 1700+ T-bred page make it seem like they are factory unlocked:

http://secure.newegg.com/app/Custrating ... 19-103-351

I suppose even if I got one that was locked, I could evaluate whether to try to unlock it or just run it at stock settings. According to http://users.erols.com/chare/elec.htm, the 1700+ T-bred tops out just under 50 W, so I figure something like a SLK-800 with some quiet fan cooling should be acceptable. Especially under ordinary circumstances, since I'm not too likely to push it hard all the time with the software I run.

Asmordean · Post by **Asmordean** » Sat Nov 23, 2002 6:33 pm

My TBred 2400+ is unlocked (all L1 bridges connected). So far as I can tell, everyone that has a TBred has reported it having intact L1 brigdes.

There is one L5 that is severed but NForce2 and KT400 chipsets can work around it.

To date I have not been able to get my CPU to work at 166MHz. I have gotten it up to 2266MHz with the default voltage but can't do much else.

The lowest I can get it to run is 133x12 (1600MHz or a 1800+). For some reason I can't get it to work at 6.5x

ez2remember · Post by **ez2remember** » Fri Nov 29, 2002 2:55 am

There is one L5 that is severed but NForce2 and KT400 chipsets can work around it.

There are quite a few motherboards on the market that does not require you to connect the L1 to the L3 bridges. These motherboards are usually based on the KT400 chipset. I have such a lucky M/B, the Asus A7V8X does not require any unlocking of the CPU.

I have a athlon XP 2200+, I have ran this at 1300Mhz @ 100FSB, same voltage with no problems. I also want to run this at 166FSB, but until I upgrade to PC2700 its not possible according to the manual. i.e. 166FSB x 2 = 333Mhz DDR memory required.

Old Dude · Post by **Old Dude** » Wed Dec 04, 2002 4:06 am

I bought an XP 1700+ Thoroughbred yesterday. I haven't had much time to test it yet but that much I've found that at least it does work at 11.5 and 10.5 clock ratios (default 11.0) without any modding. So it seems it can be both over- and under-clocked. My mobo is Soltek SL-75DRV5 which is a KT333 mobo.

peaceful_moi · Post by **peaceful_moi** » Sun Jan 12, 2003 7:28 pm

The AMD web site shows that the Typical and Maximum Thermal Power (dissipation) of the lower speed T-Bred B's increases substantially. The 1700+ T-Bred B's are at shown at 55.7W and 59.8W, compared to 44.9W and 49.4W for the 1700+ T-Bred A's. The Vcore is up from 1.5V to 1.6V, which can't account for all the 11W increase, or can it?

I don't understand this. The thermal dissipation for the 2200+ T-Bred B's went down, so I was waiting for the 1700+ T-Bred B to come out. But for quieter computing should I run out and buy a 1700+ T-Bred A now, before they're all gone?

ez2remember · Post by **ez2remember** » Sun Jan 12, 2003 11:50 pm

peaceful_moi wrote: I don't understand this. The thermal dissipation for the 2200+ T-Bred B's went down, so I was waiting for the 1700+ T-Bred B to come out. But for quieter computing should I run out and buy a 1700+ T-Bred A now, before they're all gone?

The XP 1700+ is at really low price right now, I wonder if AMD will continue to make any more of these chips. Anyway I think the XP 1700+ ver A or B, whichever one is better will be fairly easy to keep quiet and cool.

So I would say go for it, the best AMD heatsinks cost about the same price in the UK as the chip itself, which is a bit ridiculous.

peaceful_moi · Post by **peaceful_moi** » Wed Jan 29, 2003 8:44 pm

I bought the ASUS A7N8X and a Tbred A 1700+. Here’s what I’ve learned so far.

I had to reload Windows XP. Otherwise, it wouldn't boot. (Funny, because my old HDD withWin Me booted just fine.)

Underclocking works well. The Tbred A has a factory unlocked frequency multiplier, so all multiples from 6.5 to 13 are available in the BIOS of the A7N8X. In fact, the Overclockers.com A7N8X FAC states the following about the AMD Tbred processors: “All XP2000+ and lower rated tbred cpus will be unlocked from the factory, allowing multipliers 5-12.5 to be adjusted in the bios of ANY MOTHERBOARD that supports the processor and has multiplier selection.”

Pretty easy. There’s nothing to do to the CPU bridges to unlock the multiplier.

Undervolting does not work on the A7N8X. I have NOT been able to lower the Vcore voltage to the CPU below the 1.5v default – though the A7N8X Manual states that it is possible to set Vcore as low as 1.1v. The menu in the BIOS displays the 1.5v default and voltages selections up to 1.85v, but no voltages lower than 1.5v. This may not be too big an issue, since the Tbred A 1700+ runs pretty cool at default frequency and voltage. (I bought the Tbred A, rather than the Tbred B because the Tbred B has a higher Thermal Power requirement, as I noted in a previous post.) If I find an answer to this, I’ll edit this message. (Maybe we can get a BIOS hacker to create an undervolting BIOS for us.)

EDIT 2/14/03 Still no word from ASUS on undervolting.

Another issue is that the variable speed Q-Fan feature does not work on my board. This is not a real problem for me, since I have a PCMods “Bay Bus” control for my fan speeds.

The board ships with BIOS version 1001G, which I did not try, because the messages on the enthusiast sites indicate it has several problems. I first flashed the latest 1002.003 Beta BIOS and had lots of trouble with the BIOS “losing” changes I made and booting to 100MHz Front Side Bus. I had similar problems with the new 1002 BIOS. I am currently using the original shipping BIOS 1001C, which seems to work well, with the exception of the Vcore and Q-Fan problems.

You might be interested to know that there are other users who are having issues with the A7N8X BIOS. You might want to check the message boards yourself, to understand the various problems people are having. A long thread on this is at amdmb.com.

EDIT 1/31/03
I tested increasing the Front Side Bus (FSB) and lowering the CPU multiplier, and it works fine. My Tbred A 1700+ is running just fine at 6.5x166=1086MHz. The memory is running in sync at 166 (333). It seems to be much cooler than the Duron 950 that I’m replacing. CPU temperature reported is 98F, according to MBM. A CompuNurse sensor stuck into the big AlphaPAL 8045 heatsink with a slow fan reads 95F.

EDIT 2/14/03
The reason I'm running so much cooler may have to do with a MAJOR undocumented feature called S2K Bus Disconnect. Here's a quote from X-Bit Labs: "The second move on AMD's way had to do not with the CPU protection against burning, but with the temperature reduction during work. Now AMD will require all mainboards applying for certification to support S2K Bus Disconnect function, which will allow reducing the average consumed power and the heat dissipation of the CPU in most Windows applications without any performance losses." This seems very important. There's more at: http://www.xbitlabs.com/cpu/athlonxp-3000/

I also compared the Sisoft Sandra memory benchmark, placing the 2x Crucial 2700 memory sticks in Sockets 1&3 (different memory controllers) and also in Sockets 1&2 (same memory controller). The difference is minor, a 0.9% improvement. So the “dual channel” memory controller feature does not appear to be that important. (The reports I’ve read say that it comes into play with integrated on-board graphics.)[/b]

blinky · Post by **blinky** » Thu Jan 30, 2003 8:07 am

if you've got an unlocked cpu, for best performance jack up the FSB as high as you can (there's mobos and memory out there that can let you run 200MHz+ synchronous, don't do it async), then just lower the multiplier to run the cpu at near-stock speeds for a big performance boost, with minimal extra heat to cope with. The cpu doesn't care what the FSB is, all it means is that a 333MHz FSB cpu has a different multiplier to a 266MHz FSB; given that FSB speed the chip will run at the rated speed.

Asmordean · Post by **Asmordean** » Thu Jan 30, 2003 3:48 pm

I have an XP2400+ and an A7N8X. The L1 bridges of the CPU are intact. No modification made to the board or CPU. Here is what I have tried:

BIOS 1001C - Runs only at 133MHz FSB. Won't run anything else. Could only go as low as 13x (1730MHz) and as high as 17x (2261MHz)

BIOS 1002.003 - Runs at 100/133/166. Can go below 13x, seems I need to do some sort of a wire trick. Not willing since I overclock 90% of the time. Stable between 13x100 1300MHz all the way up to 14.5x166 2400MHz (if I give it a bit more power on the core.) It likes to forget settings all the time and defaults back to a 1800+ setting. It freaks out often when booting and I have been getting lockups for no reason that I never got before.

BIOS 1002.006 - Runs the same as .003. Not sure about stability. It only came out yesterday. It is supposedly a "final" BIOS even though AsusUpdate still says beta but beta isn't desplayed a POST anymore.

I run it at 2200MHz because my water setup is able to almost passivly handle the thermal output of this. At 1800MHz I can go totally fanless but I don't mind a small amount of fan noise.

At 2400MHz I have to keep the fan cranked at nearly full speed just to keep the CPU below 40°C.

peaceful_moi · Post by **peaceful_moi** » Thu Jan 30, 2003 4:39 pm

BIOS 1002.006 - Runs the same as .003. Not sure about stability. It only came out yesterday. It is supposedly a "final" BIOS even though AsusUpdate still says beta but beta isn't desplayed a POST anymore.

With 1002.006 do you have access to Vcore lower than stock/default Vcore?

Gxcad · Post by **Gxcad** » Sun Feb 02, 2003 5:01 pm

Hey I only scanned the thread so I may be repeating information but I've recently ordered 2 1700+ t-breds from newegg.com for very cheap and both versions were Tbred "B" cores (the supposedly same 9 layer core used in higher end tbreds). I've gotten one to 2000+ on 1.45vcore and the other to 2000+ on 1.4vcore. Obviously I'm keeping the 2000+ on 1.4vcore

so thats a 200mhz oc with .2vcore below default, I've heard at higher vcores the JIUHB's are hitting ~2200mhz with good air. The 2100+ that newegg is selling for $95 also seems to be a killer overclocker (though bang for the buck the 1700+ is the obvious choice at nearly half that price) often hitting 2400mhz+ on good air, especially week 48 chips seem highly sought and report of up to 2600mhz on good air. I can cool my 1700+ oc to 2000+ @1.4vcore easily with my swiftech mcx462+ and overhead panaflo 5v on zalman bracket with no case fans EASILY. Something to think about. I could oc it higher if I weren't limited to 133fsb (Abit KG7) so I'm thinking of upgrading to KT333 platform like a KX7

. Anyone want a cool and fast chip better get these chips at newegg, they seem to all be Tbred B's from my experience.

-Ken

Changing T-bred Clock Rates?

Changing T-bred Clock Rates?

Experience with A7N8X and Tbred 1700+

BIOS 1002.006