FX60 owners with coldbug read this !

After testing my FX60 using a stock vapochill LS, even modest temps of -23c envoked the cold bug and I couldn't get past 260mhz without lockups. See my resluts HERE

But after some testing with the 233 and 250 memory dividers (available with RevE chips and DFI Expert & DFI NF4 SLI) I found that the couldbug has nothing to do with memory clocks but instead is a HTT limitation.

This has been confirmed by Ozan Baseski and his results with a FX60. He too was able to run the memory with the 233 divider keeping his memory speed at a higher level than HTT and encounted no problems with coldbug. His thread can be found HERE

This will work with any RevE chip that has coldbug issues but only the FX60 has the higher multis to make it practical. On the DFI NF4 SLI you can use the 704-2BTA bios, it has the 233 and 250 memory dividers enabled in bios.

I'd like to see others confirm this and post there results here

interesting. will test it out when i get my phase change unit :)

i think i suggested that to someone who was complaining of cold bug and low HTT but it wasn't working out either.....just can't find the post...............it's great it worked out for ya......who knows maybe this will be what many needed....arn't those dividers a bit buggy or is it just CPUz and other programs not picking them up correctly?

Quote:

---

Originally Posted by dinos22

i think i suggested that to someone who was complaining of cold bug and low HTT but it wasn't working out either.....just can't find the post...............it's great it worked out for ya......who knows maybe this will be what many needed....arn't those dividers a bit buggy or is it just CPUz and other programs not picking them up correctly?

---

Cpuz 1.32 reads them correctly so does Clockgen 1.0.4.1

I've tried both the 233 and the 250 dividers and they both worked flawlessly, no more coldbug ;)

Quote:

---

Originally Posted by Major

Cpuz 1.32 reads them correctly so does Clockgen 1.0.4.1

I've tried both the 233 and the 250 dividers and they both worked flawlessly, no more coldbug ;)

---

great find, now i only need a FX or a miracle to get my opteron unlocked :(

another question this htt problem is mobo related or cpu related??

Quote:

---

Originally Posted by metro.cl

great find, now i only need a FX or a miracle to get my opteron unlocked :(

another question this htt problem is mobo related or cpu related??

---

Its a CPU issue

This will work with any RevE chip, if your using an Opteron or X2 you wont be able to increase your OC but atleast you can get the most from your memory again and improve performace for a given OC.

Thanks Major:toast: Nice find:)

Yes Major, I think your premise is correct. Although I can now boot at a higher HTT after some memory tweaks, I can confirm that a high HTT with a lower multi and a mem divider (which should work, as the final CPU and mem frequencies have previously worked fine) fails.

What I fail to grasp is exactly why this should be so. How can the CPU "see" the HTT value? Once the multiplier is applied, one would be forgiven for thinking that it is only the multiplied clock that's presented to the CPU, and its origin shouldn't matter a jot. Any ideas?

Edit: brain in neutral this morning. Plainly the multiplier is part of the CPU, so the raw clock is present.

Great find Major!

I'll have my FX60 under my Mach II today. I'll be sure to post back here for confirmation.

Quote:

---

Originally Posted by mdzcpa

Great find Major!

I'll have my FX60 under my Mach II today. I'll be sure to post back here for confirmation.

---

Looking forward to some results !

Sorry, still struggle along here. Having a heck of time getting a good mount. I'm going to be putting the IHS back on soom if this keeps up.

BTW, just under modest testing so far, Majors theory seems correct. I was getting idle lock up at 260HTT at -20c. Lowering the HTT, upping the multi, and using the 233 divisor DOES work:toast:

As soon as I can get a good mount and decent temps, I'll be on my way. Gamed all day at 3.2ghz. But temps above 10c keep me limited. I'm trying another remount again.

I think I will go ahead and buy a Phase unit now.Was going water because of the cold bug.This is what I have now. Thanks for the info.

FX-60
XP-120 Delta fan
A8N32-SLI Deluxe
Two Gig G.Skill HZ
Two X eVGA 7800GTX KO in SLI
PC Power&Cooling 510 SLI

You guys never thought of doing this before? I was wondering why no one did it with unlocked chips 6 months ago when I was a n00b, I figured there must have been some reason so I never asked

@Mike

IMO its pretty hard getting a solid mount with the MachII head on a naked core. Keep trying I'm sure your FX has 3.4 in it ;)

Quote:

---

Originally Posted by Repoman

You guys never thought of doing this before? I was wondering why no one did it with unlocked chips 6 months ago when I was a n00b, I figured there must have been some reason so I never asked

---

Well the 233 and 250 multis only work with RevE chips to begin with, and it wasn't until the 704-2 bios that they were enabled.

My FX60 is the first chip I've had that the coldbug became an issue with my stock LS, so I've not had to deal with it b4.

Quote:

---

Originally Posted by stone_cold_Jimi

What I fail to grasp is exactly why this should be so. How can the CPU "see" the HTT value? Once the multiplier is applied, one would be forgiven for thinking that it is only the multiplied clock that's presented to the CPU, and its origin shouldn't matter a jot. Any ideas?

Edit: brain in neutral this morning. Plainly the multiplier is part of the CPU, so the raw clock is present.

---

mentioned this a few days ago.. ;)
http://www.xtremesystems.org/forums/...1&postcount=19

Quote:

---

Originally Posted by ozzimark

mentioned this a few days ago.. ;)
http://www.xtremesystems.org/forums/...1&postcount=19

---

I think your right on the mark ozzimark ;)

Great find!

Kind of a side note, but what is everyone's FX-60 doing on air/water?

So if HTT is infact the limitation, then the FX is infact really special due to the higher multiplier. :)

Glad to hear, maybe I'll have to invest/ build a single stage phase.

thanks major :D

i'd like to think i'm that lucky ;)

the real question is of course.. why? what is it about the clockgen part of the cpu that doesn't like running at cold temps.. and only for 90nm? why isn't intel effected from what we can see? and what are the changes between the "memory controller revisions" that make such a big difference in how it acts in the cold?

maybe a difference in the transistor types, fast vs slow (thin gates vs thick gates iirc, but i'm probably getting way off base here :p: )

lastly.. something that could reallllly help the fx users.. options for higher than 250mhz ratios :p:

Quote:

---

Originally Posted by ozzimark

thanks major :D

i'd like to think i'm that lucky ;)

the real question is of course.. why? what is it about the clockgen part of the cpu that doesn't like running at cold temps.. and only for 90nm? why isn't intel effected from what we can see? and what are the changes between the "memory controller revisions" that make such a big difference in how it acts in the cold?

maybe a difference in the transistor types, fast vs slow (thin gates vs thick gates iirc, but i'm probably getting way off base here :p: )

lastly.. something that could reallllly help the fx users.. options for higher than 250mhz ratios :p:

---

heheheh 150 HTT .....300/350 ratio hehehehhe

Quote:

---

Originally Posted by dinos22

heheheh 150 HTT .....300/350 ratio hehehehhe

---

LOL

Quote:

---

Originally Posted by dinos22

heheheh 150 HTT .....300/350 ratio hehehehhe

---

well, not to get toooo far off topic, but one thing i've always wished for was a directly selectable memory divider, none of this "ratio" sillyness, but it doesn't seem that this is really all that possible from how amd set things up :(

AHHHHHHHHHH!

Just when things were getting cooking when my PCP&C 510 SLI ASL took a dump :( The rig just shut down. Thank god it didn't take anything else out with it. All I have is a 350 Antec Smartpower I use for testing. The rig fires up fine with that. Just not enough juice to continue tonight.

Sorry for the off topic Major. But your trick works nice!

great find here definantly something ill keep in mind in the coming weeks.

Damn awesome find! Someone remind me of this thread when my cascade comes in... I'll definately be interested in trying this technique!

Quote:

---

Originally Posted by ozzimark

mentioned this a few days ago.. ;)
http://www.xtremesystems.org/forums/...1&postcount=19

---

:eek:

Fantastic! I'd never realised that the clockgen was internal to the CPU. Nice1 oz.

I know this i silly and offtopic, but since we know where the memory controler is located in the chip, and that this is what is causing the coldbug, why not keep the memorycontoler at a temperature it likes..
You could maybe install a very small heating element right on top of the memerycontroler connected to a variable powersuppley, so you could control the heatoutput. When you then go subzero you might encounter the coldbug, but then you turn up the power for the heating element, and thereby making the memorycontroler warmer.
Then it is all about trying to find the thin line between heat from the heating element, and the cold from whatever cooling you are using. And at the same time making sure that it is only the memorycontoler you are heating up, and not the rest of the chip.

I know it i silly but if it was possible you could get round the coldbug... or frie your chip :D

Quote:

---

Originally Posted by Leverpospade

I know this i silly and offtopic, but since we know where the memory controler is located in the chip, and that this is what is causing the coldbug, why not keep the memorycontoler at a temperature it likes..
You could maybe install a very small heating element right on top of the memerycontroler connected to a variable powersuppley, so you could control the heatoutput. When you then go subzero you might encounter the coldbug, but then you turn up the power for the heating element, and thereby making the memorycontroler warmer.
Then it is all about trying to find the thin line between heat from the heating element, and the cold from whatever cooling you are using. And at the same time making sure that it is only the memorycontoler you are heating up, and not the rest of the chip.

I know it i silly but if it was possible you could get round the coldbug... or frie your chip :D

---

That sounds like a LOT of testing to get this to work. Probably very hard to find even the right materials to do such a thing. But in priciple its the most logic workaround I guess. Only if it is indeed only 1 area on the chip that causes this problem, which most of us have no idea of.

Anyway it's great that there is now the optionto get your mem upto better clocks, and some more understanding of how this coldbug works.

What is AMD's official stance on this coldbug issue, do they give any information or try to find solutions? I think that they should at least be a little bit open minded about this because extreme OC'ers are making their chips he fastst in the world which makes for good advertisement.

Fantastic, now I want one.

Quote:

---

Originally Posted by v0dka

That sounds like a LOT of testing to get this to work. Probably very hard to find even the right materials to do such a thing. But in priciple its the most logic workaround I guess. Only if it is indeed only 1 area on the chip that causes this problem, which most of us have no idea of.

Anyway it's great that there is now the optionto get your mem upto better clocks, and some more understanding of how this coldbug works.

What is AMD's official stance on this coldbug issue, do they give any information or try to find solutions? I think that they should at least be a little bit open minded about this because extreme OC'ers are making their chips he fastst in the world which makes for good advertisement.

---

As far as I have understood it is the memorycontoler rev. which influence if the chip is coldbuged, and how bad. Therefore it is the memorycontroler which is "the bad guy".
But as I said it is silly, but if it could work it would help a lot of benchers.
It sure is a lot of work, and it require lots of testing, but then again that is a part of the overclocking and benching "game".

It migth be worth a try.. I have a variable powersuppley (0,3 v - 24 v), but it need some repairs and some vapochill heating elements. If I could find an old coldbuged Winnie it could be fun to try. :)

Quote:

---

Originally Posted by Leverpospade

As far as I have understood it is the memorycontoler rev. which influence if the chip is coldbuged, and how bad. Therefore it is the memorycontroler which is "the bad guy".

---

but if you read what major says, he's able to clock the ram higher, while the htt is limited, making me strongly suspect that it's not the memory controller, but the clockgen, which inputs the htt and turns it into the cpu clock.

Quote:

---

Originally Posted by ozzimark

but if you read what major says, he's able to clock the ram higher, while the htt is limited, making me strongly suspect that it's not the memory controller, but the clockgen, which inputs the htt and turns it into the cpu clock.

---

So no one knows for sure what is causing the coldbug? And why it only appears on 90 nm AMD cpus??

Clockgen or memorycontoller, whatever it is that is causing the coldbug, you should be able to get around this if it is only a small part of the chip which doesn't like cold...

BTW where did you find the map of the opteron architecture?? Could you link to architecture maps of the other cores??

if i was an engineer at AMD, i'd probably know, but also probably be under an NDA of some sort that prevents me from revealing it?

as for the die plot, got it from amd's website ;)
http://www.amd.com/us-en/Corporate/V...72_573,00.html

http://www.amd.com/us-en/Corporate/V...~69715,00.html

or if you wanna see the dual cores.. cause that's what's important here... :D
http://www.amd.com/us-en/Corporate/V...E11210,00.html

but sadly, there's no marked plot like the first one i linked, but i'm gonna guess that the clockgen is somewhere in the middle :toast:



Major.. something probably obvious, but i've never seen a direct answer one way or another. does LDT voltage effect anything?

It is in the same position

And now I'll stop highjacking this thread before i get kicked from the forum :D :D

http://www.amd.com/us-en/assets/cont...e_marked_E.jpg

great news you have:banana:
now we like to see the fx60 owners testing and posting the results:fact:
i,m waiting my prommy ricky moded:D and need to test too

major you know what
i tried that b4 but it actually solve the problem from locking the post screen but when i try load into windows it lock up
but i use DI at 200/250.... maybe too cold

Quote:

---

Originally Posted by mdzcpa

AHHHHHHHHHH!

Just when things were getting cooking when my PCP&C 510 SLI ASL took a dump :( The rig just shut down. Thank god it didn't take anything else out with it. All I have is a 350 Antec Smartpower I use for testing. The rig fires up fine with that. Just not enough juice to continue tonight.

Sorry for the off topic Major. But your trick works nice!

---

Maybe you just had the OVP kick in on the PCP&C, did you try it again after it set for a while without power applied ?

I'll get out my LS later tonight and do some runs. I'll report back...

Works for me although I couldn't find stability @ 263 using the 233 divider for my 2GB OCZ plat which is weird as it does 280 on my X2 4800 :confused:

I am back with 1GB of VX tight timings @ 245 1/1 for 3.43G which helped me break 17K 05 !! :toast:

Ordered 2GB of Mushkin today so hopefully that will run at 233 / 263.

Nice find, i hope it works out wel. I never read anything about it in other treads regarding coldbug but i do think it would be a kinda obvious solution. No other owners of coldbugged and unlocked procs tried this before?

Anyone else confirm this yet ?

Mikeguava posted some interesting stuff on the FX60, it can be found HERE

Can anyone post results after trying this? Like before and after...thx

:( no one??

Waiting on Mushlkin 2GB to see if I can replicate Major's results. My VX won't go past 253 FSB

very nice find!!

I am having a bit of a delimma here. My Fx-60 works fine under water 3gig to 3150, but will not even boot under cold anymore. It booted when I first installed it, locked up in windows when stock. Commenced overclocking 200X16. Worked great. 17X200 worked great too. So i decided to put my pdp back in 2-3-2-5DDR400 because I'm not worried about memory bandwidth as long as I can get MHZ(reason I baught an FX-60). Needless to say. Have not had a succesful boot since then on Vapo. Any help would be appreciated.

well come on peeps. Where did you all go?

1st tests

http://img322.imageshack.us/img322/8...ndracpu0is.jpg

3dmark06 without tweaks

http://img442.imageshack.us/img442/5...mtweaks0bo.jpg

3dmark01 without tweaks

http://img442.imageshack.us/img442/5...marks016bm.jpg
my prommy need a litle bit more gaz:(
next weekend ricky go see that:toast:

i'm the process purchasing a processor..the fx-60!
along w/

a8n32 sli deluxe
4 gigs corsair xms pc4000
ocz powerstream 520


i'm being told to not purchase the fx-60 and purchase the opt 165 instead.
the opt 165 has the ability of being oc'd to the performance of a fx60 and some stories of it being up to 3.0 on air.

performance is the reason for this purchase.
not so much for gaming but for general rendering.
i'm looking to oc this chip as much as possible since i have the ability to phase cool this chip.

1. questions that i have is that would it make sense to purchase the opt 165 over the fx-60?
2. do all opt 165 chips have the ability of being oc'd to the performance of a fx-60 or is it a specific type of opt 165.
3. 165 i should be looking for (stepping)?
4. is it only the 165? why not the 170 or 175?
5. if i could use the 170 or 175 couldn't i oc the fx-60 beyond the bounds of a opt 165, 170, 175?

then i saw this fx-60 thread and these are more questions that i have.

i do too own a phase cooling system.
vapochill xe.

1. can i purchase just "any" fx-60 chip; let's say from newegg, ebay, etc?
2. what is this stepping i see people talking about and how does it effect the fx-60?
3. do all fx-60 has this cold bug issue?
4. is this cold bug issue something that amd has done purposely?
5. i noticed some have 1.7 for vcore voltages; i am being told that anything over 1.5 will damage the cpu; is this correct?
6. how high have you seen the opt 165 been oc'd to?
7. is there something i should know / purchase before getting involved w/ phase cooling w/ the fx-60 other than the cold bug?

i know it's alot of question but please please help.
it looks like you people know what you are doing and i wanted to ask before dumping my money into a possible problem.
thanks

please....anybody?

Hi nubian,

I think we can sum up your question like this. No, the 165 will not match the performance of FX-60. Some might have a chance to come close but on the whole any FX-60 will overclock alot further than any given 165. If you had a very bad FX-60 and a very good 165 it might be interesting.

Steppings effect overclocks & the ability of a chip to run properly under phase or similar cooling, you can read alot about that in this forum. The same goes for voltages & other considerations running phase cooling.

Good luck.