G30 bios - memory dividers and voltages.

[Wotcher]

E8600 - Q820A599
Kingston HyperX PC3 11000 DDR3 1375MHz (KHX11000D3LLK2/2G)

I was having trouble with stability issues even at 4.0GHz (500*8) using the 1:2.665 divider to run my memory at 1333MHz so in a fit of absolute frustration I set the FSB to 400MHz, the mutli to 10 and the memory divider to 1:4 CPU400 and low and behold instant stability with a minor memory voltage adjustment.

Even though I am now limited to 4.0GHz cpu (400*10 as the 1:3.2 divider doesn't appear to work either, probably needs more tweaking but I have to go to work :mad: ) I'm quite happy with the result despite the fact my memory is slightly overvolted at 1.725V (there is no 1.70V option :()

Stability testing to continue tonight.



Current settings:

CPU Clock Ratio Unlocked ................ Enabled
CPU Clock Ratio ............................. 10 x
CPU Clock Ratio + 0.5x ................... Disabled
Target CPU Core Speed .................. 4000
CPU Clock (FSB) ............................ 400MHz
CPU Clock vs Memory Speed ............ 1:4 CPU400
System Memory Speed ................... 1600MHz
CPU Clock Skew Control .................. 300 ps
CPU Clock Amplitude Ctrl ................. 800 mV
PCI Express Clock .......................... 100MHz
PCI Clock ..................................... 33.6MHz
Spread Spectrum ........................... Disabled
Over Clock Recovery ...................... Enable

All Voltage Control Sub Menu

********** Affect CPU MHz **********
CPU Voltage Multiplier ..................... x 1.000
CPU Voltage setting ....................... 1.225V
CPU Target Voltage ....................... 1.225V
CPU Current Voltage ...................... 1.2112V
CPU Default Voltage ....................... 1.2250V
CPU Vdroop Compensation ............... Disabled

********** Affect FSB MHz **********
CPU PLL Voltage Setting ................. 1.50886V
CPU VTT Voltage Setting ................ 1.200V
CPU GTL Ref . 0 Voltage .................. Default
CPU GTL Ref . 1 Voltage .................. Default
CPU GTL Ref . 2 Voltage .................. Default
CPU GTL Ref . 3 Voltage .................. Default
NB Voltage Setting ........................ 1.250V
NB GTL Ref . Voltage ...................... Default
SB Voltage Setting ........................ 1.500V

******** Affect Memory MHz ********
DRAM Voltage Setting .................... 1.725V
DRAM Pull-Up Voltage ..................... Default
DRAM Pull-Down Voltage ................. Default
DRAM Ref . Voltage ........................ Default


http://img171.imageshack.us/img171/1205/memsetck8.png http://img171.imageshack.us/img171/8093/cpuztg9.png

[Sailindawg]

Good work Wotcher, but do you really need CPU Clock Skew Control at 300ps? I'm running 8x450, on a 65 nm cpu, with CPU Clock Skew Control set at 100. I have tried 200 ps, but it was unstable. However, when trying to run at 4 G, I had to raise CPU Clock Skew Control to 200ps. I was also runing at 9x434 to get 4G. Is your cpu that unstable at a lower CPU Clock Skew Control?

I have also found what you have regarding RAM dividers. I have found that above 400 fsb, the 1:3 & 1:3.2 dividers work best for me. I'm using the P05 BIOS though. Now let's see you get 8x500!

[Wotcher]

Might be a typo on the skew, I'm pretty sure it was 200, will confirm after work.

I'm most pleasantly surprised that the memory clocked so high tbh, in my experience HyperX has never been much chop above rated spec, it runs tighter timings fine but never really oc'd well.

I did originally have the 4GB of memory in there but read speeds in Everest were ~5K, I was getting 10K+ with 2GB at 1333MHz with the cpu at 4.0GHz so I dropped two modules out and read went up to ~9.5K, as I said, more tweaking required :D

I tried 500*8 with 1:3 and 1:3.2, wouldn't even post, I'd get the "machine gun" beep from the speaker which requires hard power off to stop, even oc recovery doesn't kick in to save the day with a restart.

After several hours of failed attempts I was just happy to be up and running again, roll on 4.00pm so I can get back to it.

:)

EDIT: It was on 300ps, backed it off to 100ps to see if there is any effect, none so far, iirc I had 100ps skew on the Asus P5E3 at 4.0GHz as well.

EDIT 2: IntelBurnTest x20 stable (max 60°C), prime95 blend next, Superpi 1.5 1M - 11.7xx



EDIT 3:

I've got all 4GB of memory in again, it's working fine.

C1E set to Auto

CPU Clock Ratio Unlocked ................ Enabled
CPU Clock Ratio ............................. 8 x
CPU Clock Ratio + 0.5x ................... Disabled
Target CPU Core Speed .................. 4000
CPU Clock (FSB) ............................ 500MHz
CPU Clock vs Memory Speed ............ 1:2.665
System Memory Speed ................... 1333MHz (7-7-7-20-1N-PL9, tRFC 60)
CPU Clock Skew Control .................. 300 ps (as suggested in bios) ;) @ sailindawg
CPU Clock Amplitude Ctrl ................. 800 mV
PCI Express Clock .......................... 100MHz
PCI Clock ..................................... 33.6MHz
Spread Spectrum ........................... Disabled
Over Clock Recovery ...................... Enable

All Voltage Control Sub Menu

********** Affect CPU MHz **********
CPU Voltage Multiplier ..................... x 1.000
CPU Voltage setting ....................... 1.250V
CPU Target Voltage ....................... 1.250V
CPU Current Voltage ...................... 1.223V
CPU Default Voltage ...................... 1.2250V
CPU Vdroop Compensation ............... Disabled

********** Affect FSB MHz **********
CPU PLL Voltage Setting ................. 1.50886V
CPU VTT Voltage Setting ................ 1.200V
CPU GTL Ref . 0 Voltage .................. Default
CPU GTL Ref . 1 Voltage .................. Default
CPU GTL Ref . 2 Voltage .................. Default
CPU GTL Ref . 3 Voltage .................. Default
NB Voltage Setting ........................ 1.281V
NB GTL Ref . Voltage ...................... Default
SB Voltage Setting ........................ 1.500V

******** Affect Memory MHz ********
DRAM Voltage Setting .................... 1.686V
DRAM Pull-Up Voltage ..................... Default
DRAM Pull-Down Voltage ................. Default
DRAM Ref . Voltage ........................ Default

Currently tested to 4 hours prime95 blend test and 20 passes IBT, all stable, reverted to P05 bios too btw, next step will be 4.25GHz, I had a quick play and it appears the only alteration required to the settings will be upping the vcore to 1.30V, all in all a successful weekend for me, roll on next weekend. :)

http://img243.imageshack.us/img243/828/cpuz1sm1.jpg http://img243.imageshack.us/img243/6652/cpuz2bv8.jpg

http://img243.imageshack.us/img243/896/superpi40ghz4x1gbramrx7.jpg

http://img184.imageshack.us/img184/1931/40ghz1333mhz777201npl9ko2.png

http://img184.imageshack.us/img184/2513/40ghz4hoursoq4.jpg

[JohnZS]

Very VERY impressive Wotcher :clap:
Especially seeing as you are running 4x1GB of RAM (many motherboards just do not like this).
Just one quick question.


CPU VTT Voltage Setting ................ 1.200V

Did you set the VTT yourself or is that what the motherboard recommended for you?

AFIAK 45nm processors run best with VTT 1.10-1.15V but 500Mhz FSB would probably require more than this.

Once again great job and good to see you are happy with your new board.
I sadly still have my P5E3.......for now

John

[cadaveca]

1.2v setting in bios = 1.1v for 45nm.

it does not work @ 1333mhz due to improper SPD reading. it will set 7-6-6 instead of 9-9-9-24 JEDEC from SPD.

Your DIMMs should be GTR or GTS, and should scale very well.

I found I could run 7-7-7-18-60-4-5-5 PL6, and mem @ 1.5v, but continued boots @ 1.5v vDIMM was not possible, with chipset volts set to 1.156...but 1.25v works fine. 4 DIMMS.

Part of the issue with 4 DIMMs is lack of 1T...but yes, adding the second 2 DIMMs seems to introduce more loss of bandwadth than should be nessecary, and no settings I have tried has aleviated this.

[Wotcher]

Hi John, thanks for that. (Did you have a look at the link I pm'd you?)

1.2v is the bios default but as cadaveca points out in reality it is 1.1v, saaya has pointed this out in one of the other posts here.

@ cadaveca, my modules are Elpida IC's, I can boot to 6-7-7-6-18-1N-PL9 but it fails prime95 blend instantly and I lose nearly two seconds in SuperPi. (more tweaking required)

I was getting 10,4xx memory read with 2GB, then ~5K with 4GB and the G30 bios so I reverted back to the P05 and it all came good as you see above, Mem read is 9,7xx, sure I took a ~600 hit but given time I think I can claw back some of that ;)

[cadaveca]

have you tried 7-7-7-7-18? PL7? PL will have 300-400MB/sec, per level, in the least...

on G30, 4DIMMS, 1333 7-7-7-18 = 8k for me.

[Wotcher]

PL7 is instant freeze :(

7-7-7-7-18-1N-PL8, 7-7-7-6-18-1N-PL8, 7-7-7-5-15-1N-PL8 all resulted in around 10,1xx

The best I could manage tonight is the screenies below, been a hard day for me today and I'm too tired to go further tonight, at present the below is 30 minutes Prime95 blend stable and counting. This gives me a start point for the weekend session if nothing else. :up:


http://img208.imageshack.us/img208/594/85234682fx6.jpg


http://img208.imageshack.us/img208/2724/777201npl8rg3.png

[cadaveca]

Lol you got 500FSB. is very different than 1333/333fsb. 500FSB requires PL8 here too. As such, what you have posted is very good, I suppose, as voltage is low all around...

[Sailindawg]

Nice work Wotcher :up: Very nice clocks and with the 4 dimms! Who says cheap RAM can't be used! Nice low volts too. I ought to get one of those 86's too, but I'm liking my quad.

[Wotcher]

Thanks mate, I'm quite happy with the results so far.

:yepp:

[cadaveca]

OK, so I was playing a bit last night...my cpu is not as good as yours, but I did have some luck...

CPU Clock Ratio Unlocked ................ Enabled
CPU Clock Ratio ............................. 8 x
CPU Clock Ratio + 0.5x ................... Disabled
Target CPU Core Speed .................. 4000
CPU Clock (FSB) ............................ 500MHz
CPU Clock vs Memory Speed ............ 1:2.665
System Memory Speed ................... 1333MHz
CPU Clock Skew Control .................. 300 ps
CPU Clock Amplitude Ctrl ................. 900 mV
PCI Express Clock .......................... 100MHz
PCI Clock ..................................... 33.6MHz
Spread Spectrum ........................... Disabled
Over Clock Recovery ...................... Enable

All Voltage Control Sub Menu

********** Affect CPU MHz **********
CPU Voltage Multiplier ..................... x 1.000
CPU Voltage setting ....................... 1.475V
CPU Target Voltage ....................... 1.225V
CPU Current Voltage ...................... 1.462V
CPU Default Voltage ....................... 1.2250V
CPU Vdroop Compensation ............... Disabled

********** Affect FSB MHz **********
CPU PLL Voltage Setting ................. 1.50886V
CPU VTT Voltage Setting ................ 1.350V
CPU GTL Ref . 0 Voltage .................. Default
CPU GTL Ref . 1 Voltage .................. Default
CPU GTL Ref . 2 Voltage .................. Default
CPU GTL Ref . 3 Voltage .................. Default
NB Voltage Setting ........................ 1.650V
NB GTL Ref . Voltage ...................... Default
SB Voltage Setting ........................ 1.500V

******** Affect Memory MHz ********
DRAM Voltage Setting .................... 1.725V
DRAM Pull-Up Voltage ..................... Default
DRAM Pull-Down Voltage ................. Default
DRAM Ref . Voltage ........................ Default


Mem @ 6-6-6-15-48-4-5-5.PL7

[quake6]

NB voltage is quite high... Try 1:2.4...I achieved 570MHz stable with 1.594v NB and 1:4 (333) 500MHz (mem 8-7-6 PL7) with 1.594,didn't try lower...

[cadaveca]

I can do PL8 @ 1.25v, PL7 REQUIRES 1.625vbios/1.65v real.


Chipset load temp is 45c, under stock cooler iwth included fan blowing AWAY from heatsink.

I'm just haing alot of fun seeing how these HCF0 play...they do really well under low volts..so i think most current tri-channel kits are HCF0...7-7-7-18 @ 1333 requires 1.495v bios only.

[Wotcher]

Nice cadaveca, are you able to run at 1333MHz 7-7-7-20 at 1N? That 2N is holding you back sustantially imo, with such good numbers at present I think 1N at PL8 or 9 would probably result in higher Read bandwidth.

:up:

[cadaveca]

4x1gb sticks.:shrug:

[quake6]

With 4 sticks I can't run at 1T... It just don't want to change at 1T,even if it is 1T in bios...

[Wotcher]

That's intriguing, I wonder why mine runs at 1N, both Everest and Memset report it as 1N and the performance increase between 1N and 2N when set in the bios would suggest that it is indeed 1N.

Have you tried leaving everything else on auto and just setting the Command Rate to 1?

Sure the CAS latency will default to 9 (depending on your memory) but it would be interesting to see if 1N is achievable.

I have some serious doubts about my bios chip, I recently loaded up settings for a previously stable 4.25GHz overclock only to find it is no longer stable, absolutely nothing has been changed in the pc except the profiles.

The last time this happened I reset the cmos, it's almost like the bios chip is retaining some information from the previous settings and/or not implementing the new ones completely or reaches a maximum number of changes and fails to implement more, I know it sounds weird but it is the best way I can think to describe it. I'll reset the cmos and manually load the 4.25GHz settings this weekend to be sure.

[cadaveca]

A few more....:D

[cadaveca]

sry for double post, but here's 550FSB...I have a crappy cpu, and I can keep going...

[Wotcher]

what's your VTT, NB voltage and temp at +500MHz FSB?

[cadaveca]

vtt is 1.35 in bios(1.25v) 45c load temp, 1.65v. had to up ram to 1.801v bios for 550fsb, that's it.

[Wotcher]

I've had a weird experience today, I reset my cmos via the m/board switch and manually entered the same settings as I was previously running only to have my pc running like a pig in mud. Superpi 1m was around the 21sec mark.

I reloaded the previous OC via OC Gear (exactly the same settings in the bios) and things returned to normal, Superpi 1M is 11.779.

To those having no luck running 1N with their memory could you please humour me and try the following.

set memory timings to By SPD - reboot - run superpi 1m - post result

leave memory timings at default except command rate - set to 1T - run superpi 1M - post result

set memory timings to module default - e.g. 7-7-7-20 @1333MHz (cadaveca)

run superpi 1M - post result

Did you end up running 1T????

It appears (on my system at least) that the bios is not releasing prior settings completely which creates an amalgam of settings.

This has become apparent through several instances, my cpu used to run 4.0GHz fine at 1.225V but when I manually reset the bios values it would not be stable without 1.25 vcore

Having reloaded the settings I achieved via the above explained method I am now back to 100% stability at 4.0GHz and 1.225V

The same applies for my 4.25GHz oc which was perfectly stable at 1.30V once then proceeded to fail IBT within minutes if the settings were entered manually.

I'm not sure if the culprit is Memset or Aegis Panel or the bios chip itself but something is screwing around with the bios settings.

Any input appreciated

EDIT: Please make all changes in the bios, not using Aegis or Memset, also I have nfi which voltage cpuz is reporting as vcore. :shrug: The closest would be NB at 1.281V
:toast:

http://img407.imageshack.us/img407/7089/test1jw2.png

[cadaveca]

i think the problem may be I use bios only, and it seems you use memset?I jsut pulled my butt out of bed, so I'll have a go at this in a couple of hours.

[cadaveca]

OK, so 1t= no boot. Maybe time for you to reflash bios...

[Wotcher]

Thanks for trying it mate, mmmmmm, I'm really quite perplexed now.

I'll study it a bit further tonight after work.