Thanks TONY - drive strength helps coldbug

[mike]

Just finally saw the light and understood better what Tony had been preaching for a while now - different Drive / Data strengths can help greatly to minimize the coldbug.

I am sure most of the extreme cooling already know - but I think a lot of coldbgged CPUs can be helped to certain degree.

The reason for me posting about this - is just to say thank to Tony for posting about this.

THANKS TONY!

PS Hopefully I'll be able to reach 3.6GHZ soon!

[Plywood99]

^^
More Info Please!!!

[Tony]

Just finally saw the light and understood better what Tony had been preaching for a while now - different Drive / Data strengths can help greatly to minimize the coldbug.

I am sure most of the extreme cooling already know - but I think a lot of coldbgged CPUs can be helped to certain degree.

The reason for me posting about this - is just to say thank to Tony for posting about this.

THANKS TONY!

PS Hopefully I'll be able to reach 3.6GHZ soon!

Well thank you m8...im glad it helped you

[Theo404]

^^
More Info Please!!!

ditto

[mike]

Here is what Tony wrote in the RDX thread:

Cold bug

As you chill a cpu the drive strength from the memory controller increases, so you start overdriving the memory.

So, you have to weaken the drive, you can ONLY do this on the motherboard. You have to switch in resistors that alter the drive weaker or stronger depending on the resistance used. This is also one of the ways you tune a motherboard to clock ram well or a specific type of ram..IE...TCCD etc.

Now its not a full on cure to the cpu overdriving the ram, but if tweaked correctly it will allow a higher fsb to be reached.

So...Cold bug is not a serious issue, its just because the cpu's are not designed to run -50 from the onset, but to say nothing can be done on the motherboard is not true....the only way to even half fix the issue is with tweaks on the motherboard or design the cpu to run at -50 or lower.

Testing at ATI of Gruper and halibut proved this theory correct, if it didn't work we would not have designed it into the board. It won't work for everyone as every cpu is different, but it will help.

Any further explanation should have to come from Tony - as far as my knowledge to this goes would be totally insufficient.

All I can share is one of my many experiences of struggling with the coldbug that gives a secnario of where Tony's advise worked

I received from our XXX-IP Steven a X2 4600+ that had no coldbug - no coldbug insofar that it would not crash at a certain temperatures as most AMD will. I could pull the core temperatures down to -100C ( the container bottom temps ~ -160C) and it wouldn't crash.

But my OCs would be almost on the same level as running on cold water.

Trying lots of various things in fighting to get the clocks up I often would set all the ram settings to default - which would often even worsen things I decided to go back to Step 1 and slowly work my way down in temperatures.

I don't really remember the real settings but from my memory it was like that:

1) Air (~ 40C core)
with Drive Data strengths 7/4 - CPU clocks ~2.8GHZ

2) Single Stage Phasechange ( ~ -15C core )
with Drive Data strengths 7/4 CPU clocks ~2.8GHZ
with Drive Data strengths 7/2 CPU clocks ~3.1GHZ

3) Dry Ice (~ -40C core)
with Drive Data strengths 7/4 CPU clocks nothing - lol
with Drive Data strengths 7/3 CPU clocks ~2.8GHZ
with Drive Data strengths 5/2 CPU clocks ~3.3GHZ

4) LN2
with Drive Data strengths 7/4 - CPU clocks ~2.8GHZ

haven't gotten around to run it yet again - but hopefully I'll get to 3.6GHZ on the dual core X2 this time around with better settings
Again I don't reall remember the right settings clocks - but this should help to get an idea...



In the past I had rigged the perfect ram and system settings on air - so that under cold I could only focus on finding the propper vcores and mhz.
Turns out that the ram timings - in particular the drive and data settings can vary greatly under different temps. The settings might only need a slight shift - again in particular the data and drive strength - that can turn a perfect rig into no boot and vice versa. The window is VERY SMALL of good and bad ( crash). The colder it gets then narrower the range gets.
This small range of good and bad - often made me stop early in the quest of combatting the Cbug - espcially when running tube ( dry ice / ln2) your exploration of the ram sttings seems too tedious of a task and I have that feeling more than a few people have run into similar problems like that.

Recently, by pure laziness - or axiousness to find out what the max is before I even know the cpu - I had been prepping my rigs under cold for the first boot and realized that my OCs started to get better and better than other peeps - especially with weeks of CPUs that weren't supposed to be doing too well under cold...

All this and many other experiences helped me to see the value in what Tony had said above. Hopefully Tony can elaborate some more on his findings.

[eva2000]

and the board was ?

[HydroX]

DFI RDX200 ?

Here is what Tony wrote in the RDX thread:

[MaxxxRacer]

sheesh.. ppl still dont know... Atleast some ppl know though..

[kimandsally]

OK can someone explain what the drive strength numbers relate to on the DFI Expert?

Does it go from 1 being the weakest to 15 being the strongest? Or do they work odds and evens?

[ManagHead]

Interesting. To change this, you have to fysically change things on the mobo?

[don_vercetti]

could drive/data strength help using water/cold air at all?

[5days]

I read this in the RDX200 thread a while back too... I thought you extreme Oc'ers all knew this... :O

[freecableguy]

This was all discussed months ago, but I like that it is coming back into light. I believe the 1-15 settings on the Expert are broken into two groups: 1-7 and 8-15, with the higher numbers being stronger.

[GoThr3k]

so why not use 1 ???

[stone_cold_Jimi]

Erm... the flaw in this plan is that you can't boot into the BIOS to change the drive stength! I've been plotting to boot on air, change the BIOS setting, rebuild the mobo for phase and booting again for fun and happiness. Great, until you have to do a CMOS reset to defaults; then you have to rebuild to air etc etc.

So still a PITA then.

[don_vercetti]

you can normally modulate the coldness though surely, e.g. with DI alter the acetone levels to make it warm enough to get into the BIOS.

[Super Nade]

Hey guys, what resistors and stuff do you have to mod on the Mbo? Tony details and pics man!

PS*NF4 Ultra-D here.

[un0kai]

Hey guys, what resistors and stuff do you have to mod on the Mbo? Tony details and pics man!

PS*NF4 Ultra-D here.

You can change this from the bios...

[s7e9h3n]

This was all discussed months ago, but I like that it is coming back into light. I believe the 1-15 settings on the Expert are broken into two groups: 1-7 and 8-15, with the higher numbers being stronger.

Nope....the settings on the expert are 1-8 for 1gb or less of memory and 9-15 for 2gb. Odd numbers signify "weak" settings while the even numbers are "strong" settings.

@ Mike: I thought I had given you some settings to try already with this cpu??? And for those who haven't seen these, here are a couple of pics of the 4600+ which mike is using right now (note the core temps)

[xtrempinoy]

Nope....the settings on the expert are 1-8 for 1gb or less of memory and 9-15 for 2gb. Odd numbers signify "weak" settings while the even numbers are "strong" settings.

what im confused? i have 2gigs of memory, and a DFI expert. so even numbers are strong? and odd numbers are weak? i don't get that? then what should i put for drive strengths? i have a Vapochill LS for cooling. im guessing 15/4 for drive strengths???

[Tony]

Here is how expert differs from say the LP Ultra D etc.

On Ultra D we had 8 levels of drive, actually its 4 levels in normal and 4 in weak. This is why on the 704-2bta i changed the wording to help you all understand. on eXpert you have 15 levels, this is corresponding losely with the 15 levels available on the CPU memory controller, the issue is you have to switch resistors in and out with various values to set these levels...no resistor change no drive change..simple as that.

Expert is NOT 9 to 15 for 4 dimms thats not correct, to get 4 dimms at 1T you have to apply massive amounts of drive so this is why 9 to 15 come into play, the fact is you can still use 9 to 15 with 2 dimms and on TCCD this may help you get cas1.5...which btw is a true cas1.5 and not what you see on the ultra D.

So Oskar looks to have given us 1 3 5 7 all weak and 2 4 6 8 all normal drive on eXpert, then he gives us 9 to 15 in normal drive I think with 15 being full on drive with the memory controller working as hard as it can. The issue is Oskar is not telling the whole story to me so i am guessing a little. His weakest drive resistance is around 300ohms and his strongest is around 20 ohms or so, so 1 is 300 and 15 is 20 with various levels in between. Crosstalk with a massive amount of drive will be huge so he must have the traces to the dimms over 4 layers with the power rails on another layer...pretty slick designing

Now I helped ATI do the same thing on Grouper, Halibut and the new Manta reference boards, the fact was ATI test with 3 stage cascades which Oskar has not got so ATI were able to fine tune the weak drive under phase to help get high FSB's. Now if you have issues booting you need to boot warmer then chill the cpu further and clock in windows, Fugger did the same with the 955 he was testing recently as below -68 the board would not boot, clocking in XP was just fine though.

There are ways to mod boards to alter memory controller drive strength and beleive me it can make a HUGE difference to ram overclocking.i will enlighten you all on compensation capacitors another day.

[s7e9h3n]

Here is how expert differs from say the LP Ultra D etc.

On Ultra D we had 8 levels of drive, actually its 4 levels in normal and 4 in weak. This is why on the 704-2bta i changed the wording to help you all understand. on eXpert you have 15 levels, this is correspong losely with the 15 levels available on the CPU memory controller, the issue is you have to switch resistors in and out with various values to set these levels...no resistor change no drive change..simple as that.

Expert is NOT 9 to 15 for 4 dimms thats not correct, to get 4 dimms at 1T you have to apply massive amounts of drive so this is why 9 to 15 come into play, the fact is you can still use 9 to 15 with 2 dimms and on TCCD this may help you get cas1.5...which btw is a true cas1.5 and not what you see on the ultra D.

So Oskar looks to have given us 1 3 5 7 all weak and 2 4 6 8 all normal drive on eXpert, then he gives us 9 to 15 in normal drive I think with 15 being full on drive with the memory controller working as hard as it can. The issue is Oskar is not telling the whole story to me so i am guessing a little. His weakest drive resistance is around 300ohms and his strongest is around 20 ohms or so, so 1 is 300 amd 15 is 20 with various levels in between. Crosstalk with a massive amount of drive will be huge so he must have the traces to the dimms over 4 layers with the power rails on another layer...pretty slick designing

Now I helped ATI do the same thing on Grouper, Halibut and the new Manta reference boards, the fact was ATI test with 3 stage cascades which Oskar has not got so ATI were able to fine tune the weak drive under phase to help get high FSB's. Now if you have issues booting you need to boot warmer then chill the cpu further and clock in windows, Fugger did the same with the 955 he was testing recently as below -68 the board would not boot, clocking in XP was just fine though.

There are ways to mod boards to alter memory controller drive strength and beleive me it can make a HUGE difference to ram overclocking.i will enlighten you all on compensation capacitors another day.

I never said 9-15 was for 4 Dimms only TBH though, nothing above ~DS9 seemed to work for me. Using 2x512 BH5, I couldn't even get the Expert to boot @ any of those drive strengths. I also can confirm that the Expert does in fact run a true CL1.5 as I can't boot at any memory speeds above 220 using that setting...
Oh, BTW, Oskar will have a cascade....very soon

[Tony]

Im just trying to enlghten the newer members are steer the older members down the right path, ram overclocking has gone mainstream all of a sudden with everyone wanting in but not many knowing how it all actually works.

BH5 can use level 9 and up for real high clocks, its the same on the new RD580 ati board I have here, issue is TCCD may not lie it as much and i know Infineon hates it so you really do have to play with it.

What would be nice is an infinite level of drive from 0 to 300ohms or so, not 15 steps but hunderds and we could then really fine tune for the dimms we use, even 1 ohm steps would be cool but doing it would probably be a nighmare.

Now i know im going to be asked how to apply a mod to enable POTs to set the drive, maxxracer has already asked and i will work with him to see if he can see any success on his board, after that i will let him post the mod if he chooses.

[ozzimark]

0 ohms is a lot of drive

though this brings up an interesting point. i saw someone say somewhere that having a board where everythng voltage related is adjustable with pots.. it was probably maxx who said it, but i agree, it's a damn good idea.. maybe have a jumper to switch from bios controlled to manually controlled to increase the market a tad?

[Plywood99]

Could someone list out the drive strengths for the Ultra D?
AKA, is 1 weak and 8 strong, 2 less weak and 7 less strong???