Micron D9 GMH\GKX\JKH etc club

[Mafio]

In the last week I got a couple more CPUs to play with and also repaired the motherboard.

The first CPU is an E6850, decent clocker too bad for the low FSB wall (~ 550 MHz single stage cooled, ~ 520 MHz when air cooled).
Managed to run 32M at 4770 MHz with a very good efficiency.

E6850 530x9
Cellshock PC8000 4-4-4-4 636 MHz trfc 26 PL 7 - VDIMM 2.9 volt just to be safe, 2.7 or 2.8 volt should be enough



http://hwbot.org/submission/2585225_

The second CPU is a Xeon E3110, still have to try it on cold (not the best days to fireup a single stage, 30?+ C in Milan today) but looks good at least FSB wall wise.
Managed to boot with both motherboard and CPU air cooled at 650 MHz of FSB.

E3110 650x6
Cellshock PC8000 650MHz 4-4-4-4 trfc 26 PL 10 - VDIMM 2.9 volt just to be safe



http://valid.canardpc.com/1cxjxy

The trick to run this particular Cellshock kit at such high frequency/low latency combination and FSB : DRAM ratio 1:1 is play with drive strengths, in this case I used +3 on all settings in the drive strength's page.
Drive strengths on auto -> no more than 585 MHz in 1:1
1333 MHz OC profile -> boot up to 610 MHz in 1:1
+3 on everything -> boot up to 650 MHz 1:1
PL9 could be possible, I'll work on it in the next days.

[Don_Dan]

Excellent results!

[Calathea]

Insane Bit worried you might wreck the sticks though, that much volt and so much drive strength. I read somewhere it's actually the latter that is really dangerous for mem chips. On the other hand, it's probably worth the risk, right Mafio?

[Mafio]

We only live once.
Today I found out one stick of my 1500+ MHz capable Ballistix is dead.

[Don_Dan]

Today I found out one stick of my 1500+ MHz capable Ballistix is dead.

Sorry to hear about your loss, but that's the reason why you can never have enough D9s...

[Mafio]

Sorry to hear about your loss, but that's the reason why you can never have enough D9s...

You are right, I should have bought more back in the days.

E3110 610x7
Cellshock PC8000 4-4-4-4 610 MHz trfc 26 PL 9 - VDIMM 2.8 volt
Drive Strength +3
- efficiency is off because of no waza and task manager open -



I hope to reach at least 620/625 MHz with PL 9 and CAS4.

5fth place on the bot with the Xeon E3110, efficiency is quite good considering I'm probably the only one using P45 and DDR2
E3110 580x0
Cellshock PC8000 4-4-4-4 580 MHz trfc 26 PL 9 - VDIMM 2.6 volt



http://hwbot.org/submission/2588569_...min_6sec_649ms

[Don_Dan]

You are right, I should have bought more back in the days.

chew* was right!

E3110 610x7
Cellshock PC8000 4-4-4-4 610 MHz trfc 26 PL 9 - VDIMM 2.8 volt
Drive Strength +3
- efficiency is off because of no waza and task manager open -

I hope to reach at least 620/625 MHz with PL 9 and CAS4.

5fth place on the bot with the Xeon E3110, efficiency is quite good considering I'm probably the only one using P45 and DDR2
E3110 580x0
Cellshock PC8000 4-4-4-4 580 MHz trfc 26 PL 9 - VDIMM 2.6 volt

http://hwbot.org/submission/2588569_...min_6sec_649ms

Do you think you can reach 6x666MHz CL4?

[Mafio]

Do you think you can reach 6x666MHz CL4?

I don't seem to be able to reach anything higher than 657 MHz, tried single channel, dual channel, CAS4, CAS5, Cellshock PC8000 and Crucial PC8500 (the only stick left) without any luck.
650 MHz is ok for boot and short benchmarks, 655 MHz is a no go, windows crash, 660 MHz always results in post error core 52; from my experience usually it is something memory related (incorrect GTL, incorrect CPU or PCI-E clock skew, not enough VDIMM/Drive Strength) but this time I can't find the right settings for the life of me.
I am inclined to think the RAM themselves are not the problem anyway, using memory divider 5:6 they can boot easily at 665 MHz 4-4-4-4 super tight subs and run 1M (it's 3 am, there is no time for more exaustive testing )



In windows using setfsb I can run up to 657 MHz of FSB, no one MHz more (I already did it 3 years ago with another CPU): http://hwbot.org/submission/2177382_...eme_657.46_mhz
I clearly am missing something but I don't know what it is.

[Don_Dan]

I don't seem to be able to reach anything higher than 657 MHz, tried single channel, dual channel, CAS4, CAS5, Cellshock PC8000 and Crucial PC8500 (the only stick left) without any luck.
650 MHz is ok for boot and short benchmarks, 655 MHz is a no go, windows crash, 660 MHz always results in post error core 52; from my experience usually it is something memory related (incorrect GTL, incorrect CPU or PCI-E clock skew, not enough VDIMM/Drive Strength) but this time I can't find the right settings for the life of me.
I am inclined to think the RAM themselves are not the problem anyway, using memory divider 5:6 they can boot easily at 665 MHz 4-4-4-4 super tight subs and run 1M (it's 3 am, there is no time for more exaustive testing )

In windows using setfsb I can run up to 657 MHz of FSB, no one MHz more (I already did it 3 years ago with another CPU): http://hwbot.org/submission/2177382_...eme_657.46_mhz
I clearly am missing something but I don't know what it is.

I see, impressive nonetheless. It might just be the board that can't do more, especially considering you hit about the same limit with two different CPUs. I remember seeing 6x666MHz only two times anyway, one was done with DDR2 @ CL4 on Biostar P45, by TaPaKaH.

Keep pushing, maybe you'll figure it out!

[darco_2]

I don't seem to be able to reach anything higher than 657 MHz, tried single channel, dual channel, CAS4, CAS5, Cellshock PC8000 and Crucial PC8500 (the only stick left) without any luck.
650 MHz is ok for boot and short benchmarks, 655 MHz is a no go, windows crash, 660 MHz always results in post error core 52; from my experience usually it is something memory related (incorrect GTL, incorrect CPU or PCI-E clock skew, not enough VDIMM/Drive Strength) but this time I can't find the right settings for the life of me.
I am inclined to think the RAM themselves are not the problem anyway, using memory divider 5:6 they can boot easily at 665 MHz 4-4-4-4 super tight subs and run 1M (it's 3 am, there is no time for more exaustive testing )

Hello Mafio. Was reading your posts and saw that you got exactly the same hardware setup that i also have. Few days ago got my hands on some ballistix and started to play with them but nothing hadcore yet, regarding the voltage. Saw you can't pass 657 mark no mather the cl on those rams. I got this results here in dual ch, after 10-20 mins of play with the platform, on 2,2V and this here in single chanel with no more then 2,3V on air cooling. Diden't even give the NB more then 1.3V, diden't play arround with the subs for those results. Seemed pretty easy to get there. Maybe you should try those loose timmings and see if you can get bit further with them.

[Calathea]

Hello Mafio. Was reading your posts and saw that you got exactly the same hardware setup that i also have. Few days ago got my hands on some ballistix and started to play with them but nothing hadcore yet, regarding the voltage. Saw you can't pass 657 mark no mather the cl on those rams. I got this results here in dual ch, after 10-20 mins of play with the platform, on 2,2V and this here in single chanel with no more then 2,3V on air cooling. Diden't even give the NB more then 1.3V, diden't play arround with the subs for those results. Seemed pretty easy to get there. Maybe you should try those loose timmings and see if you can get bit further with them.

I should let Mafio answer but just have to tell you, there's a very good reason people are not using 3:5 divider - not just because it's very slow

[Don_Dan]

Hello Mafio. Was reading your posts and saw that you got exactly the same hardware setup that i also have. Few days ago got my hands on some ballistix and started to play with them but nothing hadcore yet, regarding the voltage. Saw you can't pass 657 mark no mather the cl on those rams. I got this results here in dual ch, after 10-20 mins of play with the platform, on 2,2V and this here in single chanel with no more then 2,3V on air cooling. Diden't even give the NB more then 1.3V, diden't play arround with the subs for those results. Seemed pretty easy to get there. Maybe you should try those loose timmings and see if you can get bit further with them.

Mafio is talking about 657 MHz FSB, not memory clock. He has a 763 MHz memory clock validation in his HWBot profile.
The divider you're using is bugged, please read this thread. I'll also edit your submission on HWBot, sorry.

[darco_2]

yes, saw it too late

Thanks guys.

Will get back later with more results, once i get play bit with those D9.

[Mafio]

Hello Mafio. Was reading your posts and saw that you got exactly the same hardware setup that i also have. Few days ago got my hands on some ballistix and started to play with them but nothing hadcore yet, regarding the voltage. Saw you can't pass 657 mark no mather the cl on those rams. I got this results here in dual ch, after 10-20 mins of play with the platform, on 2,2V and this here in single chanel with no more then 2,3V on air cooling. Diden't even give the NB more then 1.3V, diden't play arround with the subs for those results. Seemed pretty easy to get there. Maybe you should try those loose timmings and see if you can get bit further with them.

Like Don_Dan and Calathea said I was referring to FSB, I don't seem to be able to reach anything higher than 657 MHz no matter what.
RAM clocking using "low FSB" and for example 5:6 divider is not a problem, did ~760 MHz back in the days with this very same board and the now dead crucial ballistix PC8500, I also should have a validate with the cellshock at around 1375 MHz 4-4-4-4.

Now that I am back home (I still don't know how I managed to survive from my trip in spain ) I will try to find out what I am missing to reach more than 657 MHz FSB frequency.

[darco_2]

Topping at around same numbers...something around 653-655. On the latest beta bios. Did you try the older ones, F7-F8?

[Mafio]

Topping at around same numbers...something around 653-655. On the latest beta bios. Did you try the older ones, F7-F8?

I use F8, the latest beta (F9_something) is messed up, super-pi 32M is slow as hell (around 20 seconds slower than F8 if I remember correctly).