Learn something new every day: FSB actually does depend on the CPU

[c627627]

... I thought it was the RAM and mobo.

Palominos: even when unlocked may not do higher than 150 FSB.
Thoroughbred A's: Can't do 200 FSB even with PC3200 RAM and new nForce2 mobo.

Graphic67 wrote on 12-24-03:
Just look at the track record of AMD cpu's (and roadmaps) over the last two years. The chip has, until the release of the Thoroughbred B core, been the limiting factor. The Palomino cores were clocked by AMD at 133MHz fsb and most overclockers were able to reach close to 150

(AMD was denying that 166MHz cpus would be out any time soon). The Thoroughbred A which was a 130nm shrink of the Palomino did better, reaching 170 to 180MHz. (AMD was saying nothing about 166MHz parts).

Then the modified design of the Thoroughbred B (and extended cache Barton) hit the streets... And over the space of a few months AMD was able to produce 166MHz and even 200MHz fsb parts.

I have not done this, but I suspect that if you have an unlocked Palomino and drop it into a newer nForce2 board, you will still find a fsb limit around 150 even at a lower multiplier.



Gautam wrote on 12-24-03:

It's a phenomena that no one can really explain. Some sort of intrinsic properties allow some processors to "handle" higher fsb's. But this is only an issue for hardcore overclockers. Some DLT3C 1700+'s seem to hit a wall at 230 regardless of multiplier in the same systems that, with a different processor, could go farther. But for most people, the fsb wall of a processor is irrelevant.

c627627 wrote on 12-24-03
How come?
If the mobo and RAM can hit high FSBs, why do people say a CPU can limit how high FSB can OC, I mean if you can lower the multiplier, how's the CPU a limiting factor?

 

[ZachM]

I know that one very well because I used to have a Polomino. Couldn't FSB overclock worth a darn. Usually the RAM and Mobo are the main limiting factors in FSB though.

 

[ColtIce]

i can add something to this as well.i have a 1700 tbred a that will run 1914@174 fsb all day long, prime stable but won't even post at 9.5x200.it wont post at 190x10 either.i know its not the mobo or ram as i have ran both at 210 fsb with my 2500 no problems.

 

[Drisler]

..I've seen pallies..do 200fsb...with very low multipliers of course.

 

[c627627]

Drisler, post links to those Palominos running above 150 FSB please if you can.

 

[rogerdugans]

I have messed with a few Pally cores...(newest I have in fact- 4 of them) and all have no trouble at 170mhz. Two will hit 180 easily, and have been run at 190, although at high voltages and not for long.

But it is very true that a cpu can be the FSB limitation: back in the old days before pci locks , it was common to spend time figuring out how how each component COULD run and then finding the best compromise-
{cpu multi/fsb}/ ram fsb/ {pci/agp bus speed}.

It is not uncommon to find an "old" system set at a lower raw speed than it CAN run at simply because the overall system is faster and more stable with different settings.

BTW- the lowest I have seen any XP Palomino chip stop at for FSB speed was about 166.

 

[t_biggs]

My TBred A does a 220 FSB just fine, I just have a low multiplier. I think the only thing you learned was they can't do high FSB at high multipliers.

 

[The_Jizzler]

gautams comment on the B's hitting a wall at 230 rings very true for me. ive had 4 abit nf-7, three s's and one plain. and i never could get one over 230 no matter what. even with voltmods and all the trix. the only common factor in thos rigs were xp1700 and xp2100. i tried a couple of different ram types, all made it to 230, different PS, different everything. never made it past 230. but i always thought that the cpu in theory has no fsb limitation, only final clock speed is limited.

 

[CrashOveride]

I have seen a great number of pallys running at over 150fSB... I can think of 3 or 4 right now that I have seen with my own eyes...
Mine and 2 of my friends'

 

[Drisler]

Drisler, post links to those Palominos running above 150 FSB please if you can.

This good enough?
http://www.vr-zone.com/guides/AMD/AthlonXP/

There you see several hitting OVER 200fsb.

not forgetting Tbred As (there some there)
http://www.vr-zone.com/ocdb/AMD/Tbred/





So you see, both theories aren't quite true....or probably limited to those two users' experiences ONLY.

 

[asj]

at stock voltage, stock multi my 1600+ will do 182fsb! that was it maxing my pc2700 ram out. but now I have pc3200 ram so I'll see how far it'll go!

 

[felinusz]

Maybe this *is* each user's individual experiences. My Thoroughbred "B" 1800+ will do 2600+ MHz for me totally stable - but won't clear 211 FSB stably in 3DMark (only in Prime and memtest86) at 2000 MHz and less (low multipliers).

I guess I need to borrow a Barton and see what my board (NF7-S) is capable of - my RAM can do 233 stably in memtest just fine though.

It isn't too big an issue what with an unlocked multiplier though, and performance gain caused by a higher FSB seems fairly minimal bench-mark wise, for me at least.

 

[dippy_skoodlez]

... I thought it was the RAM and mobo.

Palominos: even when unlocked may not do higher than 150 FSB.
Thoroughbred A's: Can't do 200 FSB even with PC3200 RAM and new nForce2 mobo.

Um.... I had my T-bred 1700+ doing 200fsb fine....

 

[DeathONator]

I've always heard some people say that CPUs can limit your FSB, but never really believe them. Mostly because my tbird could do 229 into windows (w/ my old NF7-S). However, after getting my 2100+ (tbred B) and replacing my old fried NF7-S, I could only get 209mhz into windows (bairly). I'm really hoping my new 1800+ will show that it was the CPU causing the problem and not the mobo. Maybe now I'll hit 250mhz w/ L12 mod, and vddmod... lol * crosses fingures.... and toes... and eyes... legs, arms* ok we'll see.

 

[CrashOveride]

What multiplier for your T bred DathOnator? Cuz a few have problems for some reason, I think 10.5 and 10 or somthing like that.

 

[Baowuf]

I'm currently running a XP1600 Pali (AGOIA) at 11x170, with room for improvement. I've had the FSB as high as 185, though unstable. It needs at least 1.9v for FSB >150mhz . I feel that my limiting factor currently is my A7V8X(KT400) with it's lack of 1/6 divider(or a pci lock). Just finished a WC setup and am working on trying to get 2ghz out of it. Running PC2700 ram currently.

 

[Avg]

I got my palomino ran fine at 8x200, but the fsb was still 266mhz.

 

[Avg]

My bad I think it was actualy 5x200.

 

[Graphic67]

C627627 started this thread off with a quote from a PM that I sent with regard to fsb limitations on overclocking. Since there was some confusion over the issue, I thought I should clarify my previously quoted statement.

Front side bus speeds have been a problem for AMD from a production standpoint for the entire K7 line until the release of the Thoroughbred "B" core and its cache enhanced cousin the Barton. For nearly two years, the K7 cores only had enough headroom with regard to fsb for AMD to market 100MHz fsb models. At the end of the Thunderbird line, AMD was finally able to manufacture cores with enough headroom to manage 133MHz fsb. We were stuck at this speed for almost two years as well. Then came the first of two recent milestones in Athlon production: the die shrink to 130nm. Suddenly, the cores were able to just reach 166MHz fsb, but still did not have enough headroom for AMD to market a 166MHz fsb part. A substantial rearangement of the core into the Thorouhgbred B / Barton configuration finally gave AMD enough headroom in its cores to MARKET 166MHz and even 200MHz fsb parts. If AMD had seen significant numbers of its cpus reaching higher sustainable fsbs, we would have seen more product stratification like we saw in the last days of the Thoroughbred (with 100MHz and 133MHz high end models) or with the top Bartons (166Mz and 200MHz variations at the high end). AMD would have taken the opportunity to increase performance to win market share over intel.

In my time browsing these forums, I have noticed that some members discuss cpu purchasing tactics whigh I would call "cherry picking"... Ordering a couple of cpus, testing them out to find the best overclocker and either returning or selling off the ones that did not measure up to the individual's expectations. I do not agree with this practice, but the result is that a select few are able to achieve spectacular overclocks while neglecting to mention that a significant percentage of the cpus did not reach that level.

 

[snvpa]

I've hit 200 fsb with an nforce 7 mobo with every chip I've had in the past 2 years. Including a duron spitfire core 750 a morgan core duron 1200 a tbird 1100 tbird 1333 4 1700 tbred's 2 1800's a 2000 and 2 2500's. I've not found any evidence that fsb speeds are limited by the cpu itself. Until someone gives me decent evidence to support that I'll stick with my opinions.