Help OCing my first AMD - an Opti 165

RanceJustice

Supreme [H]ardness
Joined
Jun 9, 2003
Messages
6,628
Hello all. I'm preparing to go over my first AMD overclock ever - my Dual Core Opteron 165 on a Biostar Tforce6100 939 mATX mobo. So far my system idles at around 28c stock, with a Zalman 7000B-CU as the cooler. I'm planning to follow Eclipse's OC guide exactly, unless anyone had some better methodolgy to add since it was written. I'd appreciate a little help in the process. Here are some difficulties I've encountered so far, clipped from another thread. At this point it seems some terms don't match up with what my Bios likes to name things.

"Am I to understand that what he (Eclipse) calls "HT Link Multi' = HT Frequency? Is "CPU Multi" = "Hammer Fid control"? Also, when he talks about the "Ram divider 100mhz 1:2" I don't see any ratios in the Bios. Does this mean to set "Memclock frequency" to 100mhz? Thanks!"

Any solutions here? Also, should I do anything different on the grounds that my proc is 1) dual core and 2) an Opteron? Someone told me that Opterons should be treated like FX level consumer procs.
 
i should probably update the guide a bit and add better explainations to the terminology. :D

yes, the ht link multi is also the ht frequency. do you have it in options of like.. 1000mhz, 800mhz, 600mhz etc etc? just divide by 200mhz to get what the multiplier will be

hammer FID control is just a fancy name for the cpu multiplier.

and the memclock frequency is indeed the "memory ratio".

all good guesses on your part, i'm impressed :D

as for 1&2, you shouldn't need to do a whole lot different. you will probably want to run two instances of prime95, one for each core. most likely one of the cores will top out slightly higher than the other one, so.. might want to keep tabs on that.
 
(cf)Eclipse said:
as for 1&2, you shouldn't need to do a whole lot different. you will probably want to run two instances of prime95, one for each core. most likely one of the cores will top out slightly higher than the other one, so.. might want to keep tabs on that.

May I suggest SP2004? Don't have to go through with all the affinity shenanigans :p. Just easier, and more info. Personal preference ;).
 
Alex41290 said:
May I suggest SP2004? Don't have to go through with all the affinity shenanigans :p. Just easier, and more info. Personal preference ;).
good thinking. i've actually yet to use that program, but it does look a lot easier :D
 
(cf)Eclipse said:
good thinking. i've actually yet to use that program, but it does look a lot easier :D

I use it and its good, although I can get my computer to [email protected] volts its only about stable for 30minites in prime before it errors, if I give it @1.5volts then its okay but the temp gets to about 61C :eek: Of course I can't have a load temp of 61C so I always seem to fall back to my [email protected] setting. (don't need more anyways, and I can always go higher later when I feel like killing my cpu)
 
holy crap! you still using stock cooling? 61c is rediculous for 1.5v.. :p
 
(cf)Eclipse said:
holy crap! you still using stock cooling? 61c is rediculous for 1.5v.. :p

no I changed to a Artic Cooling Freezer 64, although I have tried running 1.585 volts and it only got up to 63C (10 minitue duel prime), at [email protected] I'm getting 48-50C load currently.

Edit: with stock cooling I tried 1.5volts and got up to 68C :eek:
 
sculelos said:
no I changed to a Artic Cooling Freezer 64, although I have tried running 1.585 volts and it only got up to 63C (10 minitue duel prime), at [email protected] I'm getting 48-50C load currently.

Edit: with stock cooling I tried 1.5volts and got up to 68C :eek:

I strongly advise you to not go that high with the voltage :p

What are ambient temps where you are? That seems a little high, especially with some aftermarket cooling.
 
Alex41290 said:
I strongly advise you to not go that high with the voltage :p

What are ambient temps where you are? That seems a little high, especially with some aftermarket cooling.

about 75F (23C) where I am right now, although it's night time so it's cooling off now, since I have my window open.

on other thought I have succesfully primed at [email protected] for an hour now.

Both cores doing small ffts and large-in place ffts at the same time, also I only did the 1.585 volt thing once for an experement. (Tried 2900Mhz but both cores failed within minites)
 
Okay. I followed the guide as instructed. HTT crashed out at 305, so I guess its maxed at 300. My CPU went from 1800 > 2501 at stock voltage, and 44c temp during the test. Hope I'm off to a good start. Since my temp was lower than 55c, I was thinking of upping the voltage. Unfortunately, my mobo makes it difficult to see what stock equates to!

For the CPU voltage it is by default at "StartUp" giving no value. Above it in a field I cannot change it says *Cpu spec Voltage* 1.350 Now I'm not sure if that is what it actually is set at, or simply voltage the MB reccommends. Also, the Hammer Fid Control/CPU Multi also has "StartUp" as its choice by default, instead of an actual number.

What's up with this? Thanks for the help so far.
 
Xaeos said:
Okay. I followed the guide as instructed. HTT crashed out at 305, so I guess its maxed at 300. My CPU went from 1800 > 2501 at stock voltage, and 44c temp during the test. Hope I'm off to a good start. Since my temp was lower than 55c, I was thinking of upping the voltage. Unfortunately, my mobo makes it difficult to see what stock equates to!

For the CPU voltage it is by default at "StartUp" giving no value. Above it in a field I cannot change it says *Cpu spec Voltage* 1.350 Now I'm not sure if that is what it actually is set at, or simply voltage the MB reccommends. Also, the Hammer Fid Control/CPU Multi also has "StartUp" as its choice by default, instead of an actual number.

What's up with this? Thanks for the help so far.

For the voltage and multiplier, just google "CPU-Z" which gives you plenty of information about your processor and motherboard. Stock volts for the opty should be about 1.35v and the multiplier for the 165 is 9x. Nice OC to start though :)
 
Thanks! Using CPU-Z revealed exactly what you stated, except there seemed to be some relatively frequent voltage fluxuations. Nothing major, but like the difference betwee 1.328 to 1.344/1.366, and back down again. I'm assuming this is normal?

So back to the OCing. I increased to voltage in small increments and ran the test again, the latest being 1.450, which crashed as soon as I got to 2.6, with a temp of around 50c. Since I'm using a Zalman 7000B-CU, does that qualify as "good air"? Should I add more voltage and try again, or have I basically reached my limit. If so, that's unfortunate considering I've heard about 165s getting to 2.7 or 2.8 on Air :( Thanks!
 
Xaeos said:
Thanks! Using CPU-Z revealed exactly what you stated, except there seemed to be some relatively frequent voltage fluxuations. Nothing major, but like the difference betwee 1.328 to 1.344/1.366, and back down again. I'm assuming this is normal?

So back to the OCing. I increased to voltage in small increments and ran the test again, the latest being 1.450, which crashed as soon as I got to 2.6, with a temp of around 50c. Since I'm using a Zalman 7000B-CU, does that qualify as "good air"? Should I add more voltage and try again, or have I basically reached my limit. If so, that's unfortunate considering I've heard about 165s getting to 2.7 or 2.8 on Air :( Thanks!

Voltage fluctuations are normal within reason, of course.

What speed are you using the Zalman on? I know it has a "silent" setting, make sure it's not at that. If it is, you might have just hit a wall with it...what stepping is it, if you know? If not, that's cool, but it would help to know because we know what certain steppings generally hit.
 
My Zalman is connected to an external fan controller, so I don't think its on the silent setting (I assume the silent one is a function of the FanMate 2). I tried increasing the voltage to 1.475, and got it to 2605 for about 2 minutes before one of the cores takes a dive.It hit about 50-52c during that time. It may be important to note that only one of the cores seems to fail SP2004, not both of them.

According to CPU-Z Stepping = 2. I'm not sure if that's what you're looking for. The guy who sold it to me said it was a "0550" that he got to 2.7ghz at 1.475.

Edit - my temps are measured via a trial version of Everest, in case anyone was wondering.
 
Repeated the test, with the same results. Any ideas? Do I have a good stepping? I hope I'm not overvolting too much.
 
I am tyring my first OC with an Opteron 165, ASUS A8N-SLI MB Corsair Memory 1 GM, OCZ 520 ADJ-SLI 520 W Psu.

I set the CPU Freq to 290 X 9 = 2.6Ghz. However, I had to set the Ram Freq to 266 Mhz
and could only get the memory to 380 Hhz (dual ch).

I have the HTT set to 3 X. How can I get my memory up to 400+ ?

If I set the memory Freq to 333 Mhz, I can only raise the cpu freq to about 2.2 Ghz which puts my memory up to about 420 Mhz which is ok for memory, but too low for cpu speed.

What is the safe temp range for the Op 165?

Terry
 
Anyone?

I get to about 2.6 (when it crashes) at 1.475volts and 52c - is this too high a temp or voltage?
 
Rand the test again. At 2600 it crashes, FSB is around 289 or so. Running at 2574, it seems to be okay. Temps up to 51-52 again at load.
 
Xaeos said:
Repeated the test, with the same results. Any ideas? Do I have a good stepping? I hope I'm not overvolting too much.
What you found in CPU-Z is not your stepping.

The stepping to your CPU is on the physical CPU itself. It's something like "CCBE1" or something. If I can recall correctly, the 165's with "CCBE1" stepping can usually get ~2.8-3Ghz on average, but these stepping 165's are rare.
 
Xephian said:
What you found in CPU-Z is not your stepping.

The stepping to your CPU is on the physical CPU itself. It's something like "CCBE1" or something. If I can recall correctly, the 165's with "CCBE1" stepping can usually get ~2.8-3Ghz on average, but these stepping 165's are rare.

CCB1E is the stepping. They are shipping out fairly reliably from Monarch now, as I know a few people that ordered some recently including myself that got one.
 
Is there any way I can tell what that stepping is, without having to remove my heatsink and look at the proc (I'm assuming thats where the number is found)?

Also, should I just accept the 2.5ghz for now and move on?
 
Xaeos said:
Is there any way I can tell what that stepping is, without having to remove my heatsink and look at the proc (I'm assuming thats where the number is found)?
Sadly, nope :(

Xaeos said:
Also, should I just accept the 2.5ghz for now and move on?

Of course not!
 
Ran the test a few more times and without increasing voltage above 1.475 I can't get it do any higher than 2.5 stable. Temps are in the early 50s, and right now I don't have my case closed, so I'm worried about hitting 55c at load if I go any higher. FSB will go up to 300 stable, so now I'm working on the ram.

Its a 2gb Gskill kit DDR500 3-4-4-8. I went into the bios to change it to this, which means putting memory frequency at 250mhz, right? However, ClockGen reads it as being 225mhz instead. Likewise with CPU-Z. What's going on?
 
which memory ratio did you set?

remember that ram speed is cpu/x where x = ceiling( cpu multi / memory ratio) ;)
 
Xaeos said:
Ran the test a few more times and without increasing voltage above 1.475 I can't get it do any higher than 2.5 stable. Temps are in the early 50s, and right now I don't have my case closed, so I'm worried about hitting 55c at load if I go any higher. FSB will go up to 300 stable, so now I'm working on the ram.

Its a 2gb Gskill kit DDR500 3-4-4-8. I went into the bios to change it to this, which means putting memory frequency at 250mhz, right? However, ClockGen reads it as being 225mhz instead. Likewise with CPU-Z. What's going on?

if your FSB is at 300mhz, you need to set the ram ratio to 5:6 for it to be 250mhz.
 
While waiting for you guys to answer I bumped the voltage to 1.5v and managed to get the CPU stable at 2605! No surges in temperature either, load still about 51c. Score! Maybe I should go back a step or two - right now I've not actually "set" any overclocks, just tested for the best I can get.

CPU stable = 2605mhz , HTT stable = 300. My ram is rated as DDR500 and I've run it stable at 3-4-4-8 in memtest at both 200mhz and 250mhz (set in bios) - of course, getting to windows it kicked it back to 225mhz. Reading your memory post, Eclipse, I don't see any mentions of dividers greater than 200mhz, so maybe I did something incorrectly? I guess I should ask where I should go from here?

Reading the OC guide, it talks about bumping the HT 3mhz in bios - maybe my bios gives funny names to things yet again - the only thing that can be increased by multiples of 3 is "CPU Frequency" which starts at 200 and goes up in 1mhz increments. I'm guessing this is it?

Thanks for the help again guys. Feeling like a noob once again ^^;;;
 
Anyone? Should I increase "CPU Frequency" to change the HTT during ram OCing? It does seem to be the only thing in bios that can be edited in 1-3mhz steps.

Also, what should I be "shooting" for as far as mhz and timings for optimal performance? Ie, I want to somehow run my CPU at 2605mhz, and keep the HTT at or under 300mhz. Any more experienced overclockers have an idea as to what's the optimal way to handle this?
 
Xaeos said:
Reading the OC guide, it talks about bumping the HT 3mhz in bios - maybe my bios gives funny names to things yet again - the only thing that can be increased by multiples of 3 is "CPU Frequency" which starts at 200 and goes up in 1mhz increments. I'm guessing this is it?
yup, that is indeed what most would call the htt or if their still stuck in the intel mindset, fsb :p

just consider it the base clock. sets the ht link speed with the ht link multi (ldt sometimes), sets the cpu speed with the cpu multi, and from the cpu speed the ram speed is set with an integer divider that is calculated with the memory ratio and cpu multi.

sounds complicated, but if you think of it in a linear fasion, it makes a lot of sense :D
 
Thanks again Eclipse. Here are what I think are my final stat limits...

Cpu - 2605mhz @ 1.5v
HTT- 300mhz
RAM - 232mhz at 200Mhz 1:1 divider 3-4-4-8 T1, (Gskill HZ 2gb kit)


I'm still a little confused on exactly how the "formula" applies now, to setting the OC. Does everything above look feasable for obtaining the 2.6ghz? Does my ram seem reasonable? The reason I ask is that its DDR500, which I thought would mean it at least would get to 250? Or is that another method of measurement?

I'm a little sketchy on the formula -

Is CPU final speed = Base clock (HTT?) x CPU Multi (Like 9x Max)?
HTT final speed = Base Clock x HT Link Multi (1x-5x)?
RAM speed = CPU final speed x.... ???

Thanks!
 
Xaeos said:
Thanks again Eclipse. Here are what I think are my final stat limits...

Cpu - 2605mhz @ 1.5v
HTT- 300mhz
RAM - 232mhz at 200Mhz 1:1 divider 3-4-4-8 T1, (Gskill HZ 2gb kit)
1:1 divider would put the RAM at 300MHz. If you are using 300MHz HTT, the divider is probably something like 166MHz or 4:5. If you are 1:1 at 300MHz, that is good :). If not, try setting the divider at 4:5 as that would put your RAM at 240MHz, so you could probably tighten the timings. If you are running 1:1 at 300MHz, keep it :p.

Xaeos said:
I'm still a little confused on exactly how the "formula" applies now, to setting the OC. Does everything above look feasable for obtaining the 2.6ghz? Does my ram seem reasonable? The reason I ask is that its DDR500, which I thought would mean it at least would get to 250? Or is that another method of measurement?
You are correct. DDR500 is 250MHz. It's PC4000 too:
RAM FSB * 2 = DDRXXX
DDR * 8 = PCXXXX

Xaeos said:
I'm a little sketchy on the formula -

Is CPU final speed = Base clock (HTT?) x CPU Multi (Like 9x Max)?
HTT final speed = Base Clock x HT Link Multi (1x-5x)?
RAM speed = CPU final speed x.... ???

Thanks!

CPU speed = HTT * CPU Multiplier (9 max in your case, yes)
HTT Bus speed = HTT * HT Multiplier (yes, between 1x-5x...this is shown to make very little difference, so you pretty much should go with 3x any higher than 250 for stability)
RAM speed = HTT * divider (if it's 4:5 it's HTT * (4/5) )

Hope that helps. One last thing...if you are at 300MHz HTT with the 9x multiplier, that's 2.7GHz not 2.6 :D Hopefully it is the first and you simply made a typo.
 
Alex41290 said:
RAM speed = HTT * divider (if it's 4:5 it's HTT * (4/5) )
NO!
(sorry)

i've been on a very long vendetta now. since the memory controller on K8 operates at the same frequency as the cpu (intel's memory controller is on the chipset which runs at fsb speed, so the fsb*ratio rule works)

so, there is a divider to get the ram speed from the cpu speed, so the real memory divider is normally described as cpu/x where x is:
ceiling(cpu multi / memory ratio)

so if i was running at 300mhz with a 9x cpu multi and 166mhz memory ratio (5:6)

the cpu speed is 300x9 = 2700mhz
memory divider is 9 / (5/6) = 10.8 -> 11
which makes the ram speed 2700mhz / 11 = 245.45

try not to get the ht link speed and the htt bus mixed up. i really hate the confusing names. if it was me i'd call it: "chipset link clock" and "base clock", because it's more.. descriptive, but whatever :p
 
Thanks again for your help, Eclipse, Alex.

I'm slightly more confused :D :D :confused:

Using the three steps for testing detailed in the OC guide...

I got CPU stable to 2605
I got the HTT stable to 300 (borked out at 305, so -5 = 300)
RAM was stable at 232mhz Base Clock, during the test set up as in the OC guide.

I've not plugged these into "equation" as of yet, these are just my upper limits. How can I best get my CPU to 2.6ghz, without borking the rest of it up somehow? ^^;;;
 
Xaeos said:
I'm slightly more confused :D :D :confused:
glad i can help <3 :D

seriously though, let's see.. so 2.6ghz cpu and 232mhz ram?

you wouldn't be able to get that ram to 237mhz stable by chance, would you?

if you could, 289mhz htt, 9x cpu multi and 166mhz (5:6) memory ratio

cpu = 289*9 = 2601mhz
ram divider = 9 / (5/6) = 10.8 -> 11
ram speed = 236.5mhz
 
So far I've upped the voltage to 2.8 and set timings to 3-6-6-10 1T and I still get errors in test 5! All other settings are as specified

CPU voltage = 1.5
Memory Voltage= 2.8
CPU Frequency/BaseClock = 237
Hammer Fid Control/ CPU Multi = x5
HT Frequency/HTT Multi = 2x
PCIE Clock = 100mhz
Memclock Freq/Memory Divider = 200mhz
1T/2T = 1T

Does it sound like I got some bad ram? This stuff is...http://www.newegg.com/Product/Product.asp?Item=N82E16820231021 , which seems to recieve accolades for being good RAM. According to the sticker, this stuff should run DDR500 at 3-4-4-8, right? So that means I should be able to raise the BaseClock to 250 in the above scenario, and it should work flawlessly, without having to increase voltage or loosen timings? Or am I incorrect? If this stuff won't even do 237 without throwing errors and it should do 250, then it sounds bad to me?

Did I miss something somewhere?
 
i HIGHLY advise having a dedicated fan to cool the ram with those sticks ;)

also.. four things i guess:

1. you shouldn't see any benefit from looser than 3-4-4 timings
2. try 2T
3. generally 2x ht link multi is bad. besides, at 237mhz you should be able to use 4x without a problem, as the final ht link speed is unde 1000mhz
4. why on earth is the CPU multi at 5x? :p
 
Wish my little mATX case could have room for one. Silverstone SG01-E, seems like there's a fan everyplace but over the ram! H :)

1. Okay, setting timings back to 3-4-4-8 at 237mhz
2. Set 2T- still getting errors in test 5
3 + 4 - I was using the 2x and 5x because you told me to in the OC guide :D

So I set Ht Link Multi to 3x instead and CPU Multi back to 9x - lo and behold it works at 2T, went back to 1T and it still works in test 5! Running one complete pass of memtest now...

So provided this works as spec'd, should I stay at 3x, or move up to 4x? Is there a performance boost here? Crossing fingers, this might actually work :D :D

Also, since the Ht Link Multi (I assume) was causing the problem with my ram, does this mean that either my HTT(FSB) or CPU may actually have higher thresholds then tested with the 2x multi? Or does it not work that way?

Edit: Complete memtest 1 pass works just fine. Time to set clocks as Eclipse instructed and boot it up for some prime. Set HtLinkMulti to 4x for the heck of it.

Edit2: For some reason, even though 289*9 = 2601, both memtest and CPU-Z say my core speed is 2592 instead. Multiplier is X9.0, HTT from CPUZ is 288 Mhz. Is that a rounding error or something?

Edit3: Bah. SP2004 failed its blend test within a minute. Clocks are CPU 2592, FSB/HTT 288, RAM 235

Wonder where the problem is. Boot temp is a little higher than normal, at 39c. Suggestions?

Edit4: Argh. Small FFT SP2004 test (CPU stress) seems fine. Large, In Place FFT test fails almost immediately on core0. Must be a ram issue. Any idea on how to solve this? The damn thing passed memtest!
 
Also, for reference in SP2004, the area that should say "Rounds: 0" sometimes turns into a red box with "paging" inside it, shortly before the test fails. The way it seems to go, Core 1 gets the "paging" box, and about 10 seconds later Core 0 stops.

Never seen it do that before.
 
Changed Ht Link Multi to 3x from 4x. No change, still messes up on SP2004 with a Blend or Large FFT test. Stats currently

Cpu V 1.5
Memory V 2.8
Base Clock 289
CPU Multi x9
HT Link Multi 3x
PCIE Clock 100
Memclock Freq 166
Timing 1T
3-4-4-8
 
Back
Top