Quick and dirty A64 clocking guide **UPDATE** X2 OC information!

Page 16 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Originally posted by: Kai920
OK I have some questions:

1. Max HTT/FSB: I got to ~278 on my mobo, anything higher can POST but booting to Windows will result in BSOD. An earlier post in this thread said going into BIOS is good enough. Should I raise my chipset voltage?

2. Max chip capable: if my max HTT is 278, then theoretically my max CPU is 278x9= 2502MHz (multiplier of 9 on the 3000+). I primed for 9 hours at 2502MHz okay, but load temps were 65C :Q Voltage is 1.52ish. Is it running too hot?

2a. Which option should I use under Prime 95 torture test - "Max FPU stress"?

3. Memory MAX - I have Corsair Value Select so I set it to 2.5-3-3-8 1T. You said "Raise HTT up to boards max..." am I starting from 200HTT and going up in increments? In other words, set HTT and memory both at 200MHz and slowly raise the HTT until I fail memtest86 test 5 & 6?


1. 278 x .95 = ?

2. You can try and raise chipset voltage to 1.7ish for higher HTT -- may not do anything since it could be mem-controller too but it's worth a try. Nah that's not too hot.. is this stock HSF?

2a. Small FFT's is fine -- large FFT's is fine too.

3. If I were you I'd apply 2.8 volts and try 2.5-2-2-8 1T or even 2-2-2-8 1T @ 200 and begin raising HTT from there. Yes you're on the right track.:)
 

suklee

Diamond Member
Oct 9, 1999
4,575
10
81
Originally posted by: Zebo
Originally posted by: Kai920
OK I have some questions:

1. Max HTT/FSB: I got to ~278 on my mobo, anything higher can POST but booting to Windows will result in BSOD. An earlier post in this thread said going into BIOS is good enough. Should I raise my chipset voltage?

2. Max chip capable: if my max HTT is 278, then theoretically my max CPU is 278x9= 2502MHz (multiplier of 9 on the 3000+). I primed for 9 hours at 2502MHz okay, but load temps were 65C :Q Voltage is 1.52ish. Is it running too hot?

2a. Which option should I use under Prime 95 torture test - "Max FPU stress"?

3. Memory MAX - I have Corsair Value Select so I set it to 2.5-3-3-8 1T. You said "Raise HTT up to boards max..." am I starting from 200HTT and going up in increments? In other words, set HTT and memory both at 200MHz and slowly raise the HTT until I fail memtest86 test 5 & 6?


1. 278 x .95 = ?

2. You can try and raise chipset voltage to 1.7ish for higher HTT -- may not do anything since it could be mem-controller too but it's worth a try. Nah that's not too hot.. is this stock HSF?

2a. Small FFT's is fine -- large FFT's is fine too.

3. If I were you I'd apply 2.8 volts and try 2.5-2-2-8 1T or even 2-2-2-8 1T @ 200 and begin raising HTT from there. Yes you're on the right track.:)

1. So 264 is the max HTT? If it's stable in Windows @278, can I leave it at that? I'll try higher chipset voltage and see if it has any effect.

2. I'm using my SN25P's heatpipe. (Shuttle website with good airflow diagrams)

2a. Noted

3. Noted will try that. I was confused last night and set it at 278HTT right from the get-go :eek:
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Try and reapply thermal paste, mount heat sink assembly really tight to heat spreader, if that's an option. Maybe probe is inaccurate.
 

suklee

Diamond Member
Oct 9, 1999
4,575
10
81
CPU temp probe is strange... when priming it'll show ~60C under load. as soon as I stop the test, it'll go down to 47C! what the?

Anyway I've got a pretty stable OC that has primed over 9 hours. critique and comments welcome:

278FSB x 9 = 2502MHz @1.45volts (3900+ rating?)
CPU Stepping: YBBLE 0525FPMW
HTT = 834 (278 * 3)
Memory = 183MHz @2.5-3-3-8/1T, default voltage (2:3 divider)
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Originally posted by: Kai920
CPU temp probe is strange... when priming it'll show ~60C under load. as soon as I stop the test, it'll go down to 47C! what the?

Anyway I've got a pretty stable OC that has primed over 9 hours. critique and comments welcome:

278FSB x 9 = 2502MHz @1.45volts (3900+ rating?)
CPU Stepping: YBBLE 0525FPMW
HTT = 834 (278 * 3)
Memory = 183MHz @2.5-3-3-8/1T, default voltage (2:3 divider)

Nice :) see if you can crank down timings a bit on ram? OTOH hand try for high latency OC as directed.

 

imported_rod

Golden Member
Apr 13, 2005
1,788
0
0
Originally posted by: Kai920
CPU temp probe is strange... when priming it'll show ~60C under load. as soon as I stop the test, it'll go down to 47C! what the?

Putting the CPU under load (i.e. running Prime95) will increase the temperature. Then when you stop priming, it falls back to idle, and the temps drop.

RoD

 

suklee

Diamond Member
Oct 9, 1999
4,575
10
81
Originally posted by: rod
Originally posted by: Kai920
CPU temp probe is strange... when priming it'll show ~60C under load. as soon as I stop the test, it'll go down to 47C! what the?

Putting the CPU under load (i.e. running Prime95) will increase the temperature. Then when you stop priming, it falls back to idle, and the temps drop.

RoD

But I'm talking about 1 second after I hit the stop button. surely the CPU temp won't go from 60C to 47C within split seconds?
 

aatf510

Golden Member
Nov 13, 2004
1,811
0
0
Originally posted by: Zebo

UPDATE: Dual core CPU stability testing
X2 is not any different than a single processor overclocking but for volts and when prime testing CPU(s) for stability:

Volts should not exceed 1.45 for safe operation!

Prime95
1. Make two folders with Prime95
http://mersenne.org/gimps/p95v2413.zip

2. Starting two instances: Launch both instances from each folder and choose: torture test > small FFT's > click OK.

3. Next, launch taskmanager and set affinity of the first instance of prime to CPU 0 only. The other instance of prime to CPU 1 only.

1.45v? What would that give out? 2.4GHz Max?
 

suklee

Diamond Member
Oct 9, 1999
4,575
10
81
Originally posted by: Zebo
Nice :) see if you can crank down timings a bit on ram? OTOH hand try for high latency OC as directed.

Tried a 166MHz divider: 278x.83 = 230MHz @2.8V, boots into Windows stable but froze up on BF2. backed down to 133MHz divider = 183MHz @default voltage.

How much real world performance difference am I looking at with 230 vs. 183 memory?
 

Smooshieus

Junior Member
Aug 1, 2005
3
0
0
I've been reading this thread even before I ordered my first ever AMD rig about 6 weeks ago and must say I'm very impressed at the amount of help going around. Zebo's beginner's OC tutorial is really one of the best I've found and I've been using it to try to see how far my system can go without risking too much (I envy your graphics card b.t.w.).

My last OC was a P4 2.66 and this stuff is completely different from what I'm used to; especially in the fact that now almost every timing and clock speed seems to be editable. It took me quite a while to figure out what everything means since I decided to go for an AMD64.

Because this is my first post here I feel a bit guilty asking for help with what's been explained so many times already but I just seem to be stuck. I've been testing maximum timings on my system today (not yet using memtest86 and only short tests with Prime95) and came up with the following numbers:

Max FSB approx. 1050 MHz
Max Memory HTT 225 MHz @ 2-2-2-5, 235 MHz @ 2-3-3-6, 250 MHz @ 2.5-3-3-6 (all SEEM stable enough to boot windows, I still have to run memtest86 overnight)
Max Core Speed 2605 (11x236.8) @ 1.5625V already fails Prime95 after 7 to 10 minutes. According to CPU-Z the voltage peeks at 1.60V often and idles at around 1.58V.

To me it seems my system can do much better considering the results I've seen with 3000+ and 3200+ CPU's in this thread. Also the temp during Prime95 doesn't exceed 52 degrees, idling at 41, but the voltage is already above what's recommended here. What can I best do to try to achieve higher results? Would an even higher voltage give me a worthwhile gain in stability or perhaps enough to raise my HTT further?

Any help would be greatly appreciated!
 

aatf510

Golden Member
Nov 13, 2004
1,811
0
0
By FSB, do you mean HyperTransport?
You should definately keep it under 1000MHz AT ALL TIMES.
Higher than 1000MHz would give you zero performance gain, and would probably coz stability problems.
Also, trying using 2T command rate while o/cing memory.
 

imported_wyrmrider

Senior member
Dec 6, 2004
204
0
0
Check the basics
Ht to 3T if over 250 bus (HTT) disired
memory divider looks like 333 or 5/6 or 2/3 should work depending on your board
get your memory below the max you listed- prefer down to around ddr 400 (200 bus ) with divider i.e. 250 cpu needs 4/5 divider or "166" or 333
300 cpu would need 2/3 (or 133 or "266"
but every motherboard has different terminology for the bios settings to do this

agp-pci locked?
sata in correct slots for your motherboard (if sata)
report back
wyrmrider
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Originally posted by: Kai920
Originally posted by: Zebo
Nice :) see if you can crank down timings a bit on ram? OTOH hand try for high latency OC as directed.

Tried a 166MHz divider: 278x.83 = 230MHz @2.8V, boots into Windows stable but froze up on BF2. backed down to 133MHz divider = 183MHz @default voltage.

How much real world performance difference am I looking at with 230 vs. 183 memory?

Not much:) 2-3% in games..nothing you'd notice.

Smooshieus -- you need to isolate CPU first, oh and Asus does'nt like 1T (command rate) at high HTT -- try 2T.
 

suklee

Diamond Member
Oct 9, 1999
4,575
10
81
Originally posted by: Zebo
Originally posted by: Kai920
Originally posted by: Zebo
Nice :) see if you can crank down timings a bit on ram? OTOH hand try for high latency OC as directed.

Tried a 166MHz divider: 278x.83 = 230MHz @2.8V, boots into Windows stable but froze up on BF2. backed down to 133MHz divider = 183MHz @default voltage.

How much real world performance difference am I looking at with 230 vs. 183 memory?

Not much:) 2-3% in games..nothing you'd notice.

Cool :cool: Maybe I can up the RAM to CAS2 since it's running under 200. But anyways I'm already very happy with my setup and OC, it knives thru BF2 like it's buttah!

Thanks Zebo :beer::beer::beer:
 

Smooshieus

Junior Member
Aug 1, 2005
3
0
0
Wow, thanks! Those were some pretty fast reactions.

Originally posted by: wyrmrider
Check the basics
Ht to 3T if over 250 bus (HTT) disired
memory divider looks like 333 or 5/6 or 2/3 should work depending on your board
get your memory below the max you listed- prefer down to around ddr 400 (200 bus ) with divider i.e. 250 cpu needs 4/5 divider or "166" or 333
300 cpu would need 2/3 (or 133 or "266"
but every motherboard has different terminology for the bios settings to do this

agp-pci locked?
sata in correct slots for your motherboard (if sata)
report back
wyrmrider

Just for the record, I'm using an A8N-SLI (basic). PCI-E and PCI are locked at the standards of 100 and 33.33 and my only SATA HD is in slot 3 (which I believe to have read is locked along with 4 on the A8N series).

For the sake of CPU testing I've set my HT multiplier to 3x even though I haven't gotten my CPU to run at HTT's over 250 (yet?:p). I also set a 4/5 divider on memory and left the timings higher than normal (3-3-3-6) although with the current bus under 200 I can run it fine at 2-2-2-5. I've also set the command rate to 2T as toattett and Zebo suggested.

Unfortunately I'm not seeing an improvement in the core speed I can reach as Prime95 still spits out errors at anything over 2600 MHz (11x236) within a number of seconds. This still doesn't seem like my max due to a stressed temp of 52 degrees.

So, this is what my system is running at now:
HTT @ 236.13 (x3)
CPU (x11) @ 2597, 1.5625V
Mem @ 185.5, 3-3-3-6 2T, 1.85V... also stable @ 185.5, 2-2-2-5, 1.80V
 

imported_wyrmrider

Senior member
Dec 6, 2004
204
0
0
Smooshieus
drop down and post this in the motherboard forum
could be bios or??? voltage??? heatsink compound???, case cooling excellent??
you are going to need some specific help on this one
you might see what happens with x10 and x9 sometimes that helps with diagnosis
when you are through testing 1t should help (a little) you can tune memory last
the 1t problem may be a bios problem which could be fixed later right now go with Zebo's advice
now we want it (memory) out of the equation

anyone else???
 

Smooshieus

Junior Member
Aug 1, 2005
3
0
0
Thanks wyrmrider, I'll head over to the MB forum for further help.

In case someone may find it usefull I'll post my results since yesterday here.

I've tested the effect of 1T and 2T command rates at Mem Bus speeds of 187, 218 and 235 MHz (on a 5:6 divider) and in every case I get a 1GB/s bandwidth boost or more running at 1T and no apparant effect on my max core clock. It just seems to stick under 2.6 GHz, even with memory running under 200 MHz at 2T.

To be able to work I'm now temporarily settling at 3500+ @ 2574 (9 x 285) with memory running a 5:6 divider for 234 MHz with 2-3-3-6 latency at 1T for a boost in bandwidth which averages around 6.5 GB/s. My HT is at 3x giving me an 855 MHz FSB.

To answer your questions:
Bios: 1008 (latest official. there's a 1010 beta available, though i'm not much of a beta-bios fan)
Voltage: 1.5750V
Heatsink compound: I'm using the MX-1 paste which comes packaged with my Arctic Freezer 64
Case cooling: Dual 12" fans in front and back. (Silentmax 580 PSU has dual fans exiting the case). I think the most important thing here is that my CPU temp never exceeds 52 degrees and my system temp is rarely seen over 40. With temps like that I think a 3500+ should be breaking 2700 or 2750 under 60 degrees without much trouble; but maybe that's inexperience talking.


Out of interest:
Could raising the voltage above Zebo's recommended settings damage the processor without causing high temps (<60 degrees)? In other words, can temperature and voltage cause damage independantly?
 

imported_wyrmrider

Senior member
Dec 6, 2004
204
0
0
yes voltage can do damage without high measured temps
high voltages can cause high temps at the individual transistor level which would not be seen by the average- outside the die- temp probe
BVVC
Did you see any change with the VC changes you made? Is your VC reading accurate?
I mentioned termal compound as too much or spotty application can cause parts of the die to be hot eventhough the temps read ok
also some compounds take several days to "settle in"
you might try reapply or artic silver and see if you can document any difference

from where you're at the devil will be in the details

you have a popular motherboard- there should be (read that as there are) several forums
chipset cool?
Sorry I do not have that board so just guessing at this point
good work
keep us posted

wyrmrider
 

OahnMacleod

Junior Member
Aug 8, 2005
5
0
0
Here's what I just built...

Asus A8N-SLI
AMD 64 3200+ (O.C.'d to 2700 270x10)
Thermalright XP-90 w/ Zalman ZM-F2 92mm fan blowing on CPU
PQI Turbo Ram 2 Gigs (2x1024) 8-4-4-3-2T set to 8-4-4-2.5-1T
HTT = 4x

All power settings are set to Auto, haven't raised voltage on anything yet.

Did a few speed tests with Nvidia nTune and everything seems pretty high now.

CPU is running at 34 degrees Celcius. 80mm fan intake, 120mm fan rear exhaust as well.

Can anyone explain what all the numbers mean in regards to ram settings? I think
I'm pretty good right now, and haven't tried pushing the FSB of the CPU past 270 yet.

I have no clue what the dividers and other things mean and have tried reading through
several explanations here and on other sites.

Thank you in advance...