2100+ owners: min vcore for 200fsb?

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Originally posted by: shady06
it depends on the stepping. is a tbred A or B?
BUT, even if it's a T-bred B, you're going to have to raise your ram's timings to at least 2-3-3-6. Plus, I would raise the vcore to at least 1.70v, and probably to 1.75v.

edit: Since it posted at 1.65v, you probably won't need to raise the vcore higher for 10x200, but it's possible that you may. The reason I would raise the vcore to 1.75v is because quite a few people's T-bred B's run at up to 12x200 on that vcore.:D
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: myocardia
Originally posted by: shady06
it depends on the stepping. is a tbred A or B?
BUT, even if it's a T-bred B, you're going to have to raise your ram's timings to at least 2-3-3-6. Plus, I would raise the vcore to at least 1.70v, and probably to 1.75v.

edit: Since it posted at 1.65v, you probably won't need to raise the vcore higher for 10x200, but it's possible that you may. The reason I would raise the vcore to 1.75v is because quite a few people's T-bred B's run at up to 12x200 on that vcore.:D

it was a fairly conservative run so i actually had ram at 8-3-3-2.5 [it has passed memtest overnight at these settings].
initially i just want to match my clock speed [12x166 = 2gigish] but on the faster fsb...if i get a more powerful fan for the hsf then i might shoot higher.

i guess if it crashed at 1.65 then i should try for 1.675? i know 1.7 would be more fool-proof but heat is an issue too; the temps [when i ran the sandra cpu burn-in at 1.65, 10x200 were 55C load so that was acceptible to me..]
 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Wigwam, you shouldn't have hardly any rise in your temp, as long as you keep your vcore under 1.70v actual. If it skyrockets, it's because you don't have enough air flow through your case.
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: myocardia
Wigwam, you shouldn't have hardly any rise in your temp, as long as you keep your vcore under 1.70v actual. If it skyrockets, it's because you don't have enough air flow through your case.

whilst airflow is not perfect, it isnt poor - i have a side intake, a rear [panaflo] vent as well as a rear pci-slot-exhaust under the gfx card, the antec psu venting [dual fan config taking heat from the psu and out back], and a top [panaflo] vent. [i know, input not matched to vents!]

running the machine 12x166 at 1.6v [really 1.62 according to mbm5] i get 42-44 idle and 48-51 load
running it 10x200 at 1.65v as i did yesterday i got 44-46 idle and 54 load with the sandra cpu burn test
admittedly the proof of the pudding is in the trying but certainly even a 0.25v increase does up the temp a little and upping by 0.5 will surely bump it up more.....:eek:
i [arbitrarily] decided that 55c was going to be my ceiling for load temps as that was what it was running with at 12x166 with the stock hsf
maybe i am being overly cautious with that upper temp?
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: TerryMathews
Just run it at 1.85V. It won't hurt the processor, it lies within AMD's technical specifications.

the voltage per se might not but it would run damned hot!
 

TerryMathews

Lifer
Oct 9, 1999
11,473
2
0
Originally posted by: Wigwam
Originally posted by: TerryMathews
Just run it at 1.85V. It won't hurt the processor, it lies within AMD's technical specifications.

the voltage per se might not but it would run damned hot!

V=IR says temps shouldn't increase by more than 12% over 1.65V.
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: TerryMathews
Originally posted by: Wigwam
Originally posted by: TerryMathews
Just run it at 1.85V. It won't hurt the processor, it lies within AMD's technical specifications.

the voltage per se might not but it would run damned hot!

V=IR says temps shouldn't increase by more than 12% over 1.65V.

yeah true but that still take me into >60C country....
 

TerryMathews

Lifer
Oct 9, 1999
11,473
2
0
Originally posted by: Wigwam
Originally posted by: TerryMathews
Originally posted by: Wigwam
Originally posted by: TerryMathews
Just run it at 1.85V. It won't hurt the processor, it lies within AMD's technical specifications.

the voltage per se might not but it would run damned hot!

V=IR says temps shouldn't increase by more than 12% over 1.65V.

yeah true but that still take me into >60C country....

As long as that 60oC is measured at the core's diode, you're fine. .13u Athlons and Durons are rated to a diode temp of 90oC.
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
hhhmmm

ran it at 9.5x200 at 1.675v and it seemed stable enough [and yes, the temps werent much diff from 1.65, with 56C load] :eek:
so, i then ran it at 10x200 at 1.70v:
ran the sisoft cpu burn in okay with max temp 58-59C , but locked when i ran 3dMark01 and prime95.....:frown:

having pretty extensively tested the ram at 200MHz at 8-3-3-2.5 [its rated speed] and infact also at 6-2-2-2 at this speed and no memtest errors.
the fact that it it can live with running at 9.5x200 i guess means that i am going to have to squeeze up the juice even more to 1.75v to run it at 10x200?
oh man...
rolleye.gif


those guys who claim to run 2Ghz @200Mhz fsb on stock vcore must surely have been speaking out of their bottoms!:confused:
 

TerryMathews

Lifer
Oct 9, 1999
11,473
2
0
No they wern't lying. Each chip OCs differently.

Technically, if you must know, the better clocking CPUs will be the ones at the center of the wafer, with overclockability falling off the further you are from center. Of course, the location of the CPU cut from the wafer isn't etched into its die anywhere, so you can't really tell where it was from.

Typically, mobile chips or low voltage chips are cut from the center.
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: TerryMathews
No they wern't lying. Each chip OCs differently.

Technically, if you must know, the better clocking CPUs will be the ones at the center of the wafer, with overclockability falling off the further you are from center. Of course, the location of the CPU cut from the wafer isn't etched into its die anywhere, so you can't really tell where it was from.q]

fair enough. my piece of crap must have been cut from the end of the wafer!
tried running 10x200 [and even 10x198] at 1.75v and no joy: crash city.....
what is frustrating is that i get this speed and above on a lower fsb [12/166] and can run fine 9.5x200 but it doesnt like 10x200 even with lots of juice.

hhhmmmmmmmmm
:disgust:
 

Pilsnerpete

Platinum Member
Apr 4, 2002
2,060
0
0
For 200 fsb, eh? I have the exact same mobo and ram as you, and I've found that mine won't run stable @ 200 w/default timings either. I'm thinking it's something wrong with the board, but once I build my sister's IC-7 based pc, I'll have something else to test this ram by. To eliminate another variable.
 

edmundoab

Diamond Member
Apr 21, 2003
3,223
0
0
www.facebook.com
I'm running a Tbred B 2100+ at
FSB 205 X 11,

Vcore = 1.725 V
Vdimm = 2.7V
Memory timings at 2-2-2-11, Corsair XMS TWINX PC3200LLPT

Very Stable.

I tried 1.700 V on the Vcore at that settings and it failed prime95 after about 2 hours.
Bumping to 1.700 and using 2.6V on Vdimm didnt help much

so I decided to stick to 1.725 on Vcore and 2.7 on Vdimm
 

drewdogg808

Golden Member
Oct 1, 2000
1,513
0
71
running 200 fsb also depends alot on your motherboard and ram. my 2100+tbred b does 11.5x200 @1.65v, 12x200 @ 1.8v on air. it also does 12.5x200 @ 1.85v on water.
i remember testing it on default voltage and it would do 10x200 fairly easily.

helps that i'm running on nf7-s and mushkin pc3200 lv2 black.
 

TerryMathews

Lifer
Oct 9, 1999
11,473
2
0
In my opinion, the way to overclock if you're going for maximum (and you have an unlocked processor you can set the multiplier on), is to adjust the multiplier until you find the highest speed your processor will run stable at. Jot that down. Then set the multiplier low and find the highest FSB and RAM speed you can run at. Jot that down. Divide processor speed by FSB speed and you've got your ideal multiplier. That should give you an excellent starting point.

I ran every Athlon I owned at 1.85V. Never had a problem and I've owned more than a dozen including one heck of a furnace (Dual 2100+ @ 2.26GHz each)
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: TerryMathews
In my opinion, the way to overclock if you're going for maximum (and you have an unlocked processor you can set the multiplier on), is to adjust the multiplier until you find the highest speed your processor will run stable at. Jot that down. Then set the multiplier low and find the highest FSB and RAM speed you can run at. Jot that down. Divide processor speed by FSB speed and you've got your ideal multiplier. That should give you an excellent starting point.

I ran every Athlon I owned at 1.85V. Never had a problem and I've owned more than a dozen including one heck of a furnace (Dual 2100+ @ 2.26GHz each)

yes, i accept as gospel the principle you outline here, but it seems to fall down abit:
i have gotten the thing to run [ie load into winxp] at 13x166 = 2158 even at default voltage
it runs and is prime95 stable at 12x166 = 2004 at default vcore
and since it runs at 9.5x200 at 1.675vcore, by rights it "should" run at 10x200....especially at 1.75v.

this is what i am having trouble understanding - why it wont. i am 99.9% sure it isnt my ram as i have tested it extensively at 200mhz upto 6-2-2-2 :evil:
 

TerryMathews

Lifer
Oct 9, 1999
11,473
2
0
Well, is your motherboard certified for 200MHz FSB? Some of the NF2s aren't. If it's not an NF2 Ultra 400 or something like that, you may have trouble hitting that high of a FSB.
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
yes. that is partly the reason i am so peeved by all this: i cherry picked my components just to be able to overclock...and run 200fsb:
a7n8x deluxe revision 2 [certified to 200mhz fsb]
2100+ t/b B AUIHB
twinmos/winbond bh5 ram
everything is doable on paper - i seem to be missing a trick and cant for the life of me discover what it is
 

drewdogg808

Golden Member
Oct 1, 2000
1,513
0
71
Originally posted by: Wigwam
yes. that is partly the reason i am so peeved by all this: i cherry picked my components just to be able to overclock...and run 200fsb:
a7n8x deluxe revision 2 [certified to 200mhz fsb]
2100+ t/b B AUIHB
twinmos/winbond bh5 ram
everything is doable on paper - i seem to be missing a trick and cant for the life of me discover what it is

can you run it at 10x200 with one stick of ram or both sticks in single channel? my previous 8rda had trouble running 2x512 sticks in dual channel even though each stick runs well above 200 with memtest. when i ran 1 stick or put them in single channel, 200 was not a problem again. it was sorta wierd...i read somewhere that the memory controller was stressed when running 1 gig in dual channel so it was completely stable. but i don't know how true that theory may have been.
 

Wigwam

Senior member
Dec 26, 2002
943
0
0
Originally posted by: drewdogg808
Originally posted by: Wigwam
yes. that is partly the reason i am so peeved by all this: i cherry picked my components just to be able to overclock...and run 200fsb:
a7n8x deluxe revision 2 [certified to 200mhz fsb]
2100+ t/b B AUIHB
twinmos/winbond bh5 ram
everything is doable on paper - i seem to be missing a trick and cant for the life of me discover what it is

can you run it at 10x200 with one stick of ram or both sticks in single channel? my previous 8rda had trouble running 2x512 sticks in dual channel even though each stick runs well above 200 with memtest. when i ran 1 stick or put them in single channel, 200 was not a problem again. it was sorta wierd...i read somewhere that the memory controller was stressed when running 1 gig in dual channel so it was completely stable. but i don't know how true that theory may have been.

interesting [i think so anyway
rolleye.gif
] answer:
it cant run dual channel with slots 1&3 at 200mhz but can at 166 [loads of memtest errors]
it runs dual channel perfectly happily at 200Mhz 6-2-2-2 in slots 2&3 [no errors in 12 hours memtest] and stable at system speed = 200MHz with 9.5 cpu multi....

admittedly i havent run 1 stick in this way but i dont see why it should make any difference; put it another way, if i have to do that then i wouldnt run the machine on this setting anyway
 

drewdogg808

Golden Member
Oct 1, 2000
1,513
0
71
Originally posted by: Wigwam
Originally posted by: drewdogg808
Originally posted by: Wigwam
yes. that is partly the reason i am so peeved by all this: i cherry picked my components just to be able to overclock...and run 200fsb:
a7n8x deluxe revision 2 [certified to 200mhz fsb]
2100+ t/b B AUIHB
twinmos/winbond bh5 ram
everything is doable on paper - i seem to be missing a trick and cant for the life of me discover what it is

can you run it at 10x200 with one stick of ram or both sticks in single channel? my previous 8rda had trouble running 2x512 sticks in dual channel even though each stick runs well above 200 with memtest. when i ran 1 stick or put them in single channel, 200 was not a problem again. it was sorta wierd...i read somewhere that the memory controller was stressed when running 1 gig in dual channel so it was completely stable. but i don't know how true that theory may have been.

interesting [i think so anyway
rolleye.gif
] answer:
it cant run dual channel with slots 1&3 at 200mhz but can at 166 [loads of memtest errors]
it runs dual channel perfectly happily at 200Mhz 6-2-2-2 in slots 2&3 [no errors in 12 hours memtest] and stable at system speed = 200MHz with 9.5 cpu multi....

admittedly i havent run 1 stick in this way but i dont see why it should make any difference; put it another way, if i have to do that then i wouldnt run the machine on this setting anyway

here is a quote from a guy (hitechjb1) on overclockers that knows a ton:
The key to run NF2 above 200 MHz is the motherboard and the north bridge chip and its memory controller in particular. The natural out of the box FSB ranges from 180 - 210 MHz (without vmod). Many have success in getting FSB to 220 MHz by doing vdd mod on the NB.

I have done some tests, even you have good working memory (PC3500) and good MB that the FSB can run up to 230 MHz (in ASYNC) per se, once you put the dual channel memory in, up the FSB in SYNC to somewhere between 195-210 MHz (depend on the MB), and pump memory intensive data such as 3DMark through the NB, the system will crash. The system is totally normal if there is NO memory intensive application running.

This means that the NB will fail at certain memory bandwidth (not just the magnitude of FSB MHz). Repeat, the problem is NOT the memory and FSB itself. It will fail even you have $400 PC5000 modules !!!


i had similar results with 2x512 dual channel at 200fsb, i would be prime stable for 24hrs but 3dmark would crash in an hour. so you can see if it might be a memory controller limitation on your board if you test with 1 stick or in single channel at the speed you want.



 

Wigwam

Senior member
Dec 26, 2002
943
0
0
now that IS interesting - thanks.

mind you, 1 thing i forgot to mention throughout this episode is that i have noticed that mbm5 also reads the cpu speed wrong when i set it to 10x200.
it will report all the other combos i use [any on 166fsb and 8.5/9/9.5x200] but for 10x200 it always reads *1538*

maybe that is hinting at some other problem tantalisingly out of reach of my brain?