My x3350 overclocking experience

[Leonardo]

You have to give it more voltage!

Probably, but keep in mind, he's overclocking a 45nm Yorkfield! The rules may be a bit different now.

 

[torin3]

Probably, but keep in mind, he's overclocking a 45nm Yorkfield! The rules may be a bit different now.

Right! I'm going to be going up in FSB until it has a problem, then adjusting the motherboard voltages first, before I up the Vcore. At 6x multiplier, I maxed out at 435 FSB with 1.39Vcore actual.

And people, please note that I'm still working on this. I'm updating it literally seconds after I capture the screen shots.

Oh, and here is the 420x8 run.

 

[Hazaro]

Probably, but keep in mind, he's overclocking a 45nm Yorkfield! The rules may be a bit different now.

They are, but 1.35V load is still fine.

Also this:

I don't think you need more vCore at all, I think you need more FSB (vTT) voltage instead. The I/O pins seem to be the inhibitor of FSB on these chips; my E8400 isn't stable at more than ~466 FSB unless I start upping the vTT by about 0.010v -- 500FSB needs something like 0.020v more, and the max ~535FSB requires almost 0.03v more. I haven't tried more than that, because 535 x 8 = way more speed than my processor can really attain

So go looking for vTT or FSB voltage options in your bios and give those a little bump... You might be pleasantly suprised. Also, don't forget that quad cores are more stressful on your northbridge, so give it a bit more love too while you're there.

Give a little love to some of your other parts. 435 seems abysmally low.

 

[torin3]

They are, but 1.35V load is still fine.

Also this:


Give a little love to some of your other parts. 435 seems abysmally low.

Yep, I'll take it up to 1.35V load if it needs it.

I'm just working my way up until the motherboard actually needs the extra voltage. No point pouring it on before it needs it.

430x8 SuperPi 1M (and this is with the ram at 1:1)

 

[MikeyLikesItSI]

Results are looking better by the post. Keep up the good work!

 

[torin3]

Results are looking better by the post. Keep up the good work!

Will do. About to try 440x8. I think I'll need to bump some voltage for this to work.

Also, under load the Vcore dropped to 1.16x volts. May have to bump it just a bit too.

 

[Gautam]

Intel has raised the FSB for no reason other than to reduce the multi and make OC less rewarding. The FSB isn't a bottle neck for the q6600 and it didn't need to be increased. Intel is obviously listening to the OC community and responding appropriately to word of free FSB room.

No, Intel has raised the FSB because they don't care about overclockers. They care about the other 99.999999% of the world. (you know, the part that doesn't frequent these forums) For this part of the world, that leaves things at spec, they get to advertise higher front side bus and easily pull performance gains without having to do any higher scaling.

Yes, everyone's bitter about the high FSB parts, and I'm not thrilled about it either, but its not like Intel is doing this just to tick off a fraction of a percent of their user base.

 

[MikeyLikesItSI]

Are you guys saying you miss the 18-23 range multipliers??? LOL Back then we all bitched about our prescotts and barely making 300 fsb!! now 425+ is the easy norm for us OCers! LOL

 

[torin3]

Ok, after a bit of a diversion, I was finally able to boot into windows at 440x8.

Too much FSB termination voltage is a bad thing However I was able to see why it stopped dead after 435 FSB. That is the point at which you need above the purple voltages and go into the red for the Northbridge. 1.57v on the NB and I can boot into Windows, but it isn't stable. Also, at this level, I have switch to the 333 strap instead of leaving it on auto.

I wasn't able to get it more than five minutes priming, even if I took it up to 1.63v on the NB, and 1.3Vcore. I tried the NB GTL settings, but 67x, 63x, and 60x didn't work, and 60x was even more unstable.

Any suggestions on where to go? It is annoying that I can't really tell where the auto settings are putting the voltages. The BIOS only lists the Vcore in the hardware monitor.

I'm pretty sure this should be able to get more FSB. I'm just a little concerned because 1.61V and 1.63v on the NB was getting to be more unstable. I've got active cooling on the NB (40mm fan).

Well, I'm up about an hour too late, so I need to call it a night today, but I can get back on this tomorrow.

Thanks for the comments everybody.

 

[GTengineer]

13.88 sec at only 3.4GHz! that is pretty bad arse! I do 14.0 sec at 3.8GHz with my Q6600! Still I must avoid the temptation till Nehalen

 

[prime81]

Ok, after a bit of a diversion, I was finally able to boot into windows at 440x8.

Too much FSB termination voltage is a bad thing However I was able to see why it stopped dead after 435 FSB. That is the point at which you need above the purple voltages and go into the red for the Northbridge. 1.57v on the NB and I can boot into Windows, but it isn't stable. Also, at this level, I have switch to the 333 strap instead of leaving it on auto.

I wasn't able to get it more than five minutes priming, even if I took it up to 1.63v on the NB, and 1.3Vcore. I tried the NB GTL settings, but 67x, 63x, and 60x didn't work, and 60x was even more unstable.

Any suggestions on where to go? It is annoying that I can't really tell where the auto settings are putting the voltages. The BIOS only lists the Vcore in the hardware monitor.

I'm pretty sure this should be able to get more FSB. I'm just a little concerned because 1.61V and 1.63v on the NB was getting to be more unstable. I've got active cooling on the NB (40mm fan).

Well, I'm up about an hour too late, so I need to call it a night today, but I can get back on this tomorrow.

Thanks for the comments everybody.

If your running an X38 board such as the Maximus the NB being at 1.65-1.7 is not uncommon. Mine with 500 fsb is running 1.59bios and 1.63 in windows. Although I'm on a E3110 chip right now, I plan to play around with my Q6600 this weekend to see if I can get 500 fsb out of her, even if its only 6 x 500 itll be blazing.

Here's my settings if it's any help...

Ai Overclock Tuner : Manual
CPU Ratio Control : Manual
- Ratio CMOS Setting : 8
FSB Frequency : 500
FSB Strap to North Bridge : 400
PCI-E Frequency: 100
DRAM Frequency: DDR2- 1000
DRAM Command Rate : 2
DRAM Timing Control: AUTO
CAS# Latency : AUTO
RAS# to CAS# Delay : AUTO
RAS# Precharge : AUTO
RAS# ActivateTime : AUTO
RAS# to RAS# Delay : Auto
Row Refresh Cycle Time : Auto
Write Recovery Time : Auto
Read to Precharge Time : Auto

Read to Write Delay (S/D) : Auto
Write to Read Delay (S) : Auto
Write to Read Delay (D) : Auto
Read to Read Delay (S) : Auto
Read to Read Delay (D) : Auto
Write to Write Delay (S) : Auto
Write to Write Delay (D) : Auto
DRAM Static Read Control: Dis
Ai Clock Twister : Man
Transaction Booster : Dis setting 1

CPU Voltage : 1.3750
CPU PLL Voltage : Auto
North Bridge Voltage : 1.59
DRAM Voltage : AUTO
FSB Termination Voltage : 1.36
South Bridge Voltage : 1.050
Loadline Calibration : Enabled
CPU GTL Reference : 063
North Bridge GTL Reference : 067
DDR2 Channel A REF Voltage : Auto
DDR2 Channel B REF Voltage : Auto
DDR2 Controller REF Voltage : DDR2 REF
SB 1.5V Voltage : Auto

 

[HDCHOPPER]

GTL REF. volts seems to be the answer on these new chips ( just from reading around the forums)

good luck !

 

[ghost_recon88]

Yep, I'll take it up to 1.35V load if it needs it.

I'm just working my way up until the motherboard actually needs the extra voltage. No point pouring it on before it needs it.

430x8 SuperPi 1M (and this is with the ram at 1:1)

Stop teasing and give it some gas already. 1.5 volts now! Max out the FSB now!

 

[orion456]

they get to advertise higher front side bus and easily pull performance gains without having to do any higher scaling.

There is no substantial performance boost from stock CPU speeds and higher FSB. In fact there is just as often a performance loss. The FSB isn't a bottleneck until a quad hits around 13x multi.

http://www.nordichardware.com/Guides/?page=1&skrivelse=517

I'm guessing that popular articles showing how to get a 2.4 ghz CPU up to 3.2 for free, instead of paying $1k to Intel for a 3.0 ghz CPU, was a definite embarassment and is the real reason for raising the FSB from 1333 and soon to be 1600. At 1600 there won't be much OC room left. How else to explain locked multipliers? It certainly wasn't because regular consumers want to play with upping the multiplier, its most definitely enthusiasts that do that.

Intel wants you to get what you paid for, and to pay for what you get! You want 3.2....a grand please.

 

[Leonardo]

I'm guessing that popular articles showing how to get a 2.4 ghz CPU up to 3.2 for free, instead of paying $1k to Intel for a 3.0 ghz CPU, was a definite embarassment and is the real reason for raising the FSB from 1333 and soon to be 1600.

I just don't see that. Out of every CPU sold by Intel, what percentage are overclocked? .001%, if that? We overclockers are but a miniscule component of the CPU market. Maybe I'm wrong, I just find it hard to believe that the relative small number of us would really influence a design change very much. For sure, holding title to the fastest, most powerful consumer processor is worth a lot to Intel (or AMD...again, someday...maybe) for prestige and for dazzling the stockholders, but we overclockers could all stop buying CPUs tomorrow and it wouldn't affect Intel's bottom line.

 

[Craxbax]

Yeah, 13.88 Super PI @3.44ghz is nice evidence of the Q9450 performance potential. He equaled my Q6600 @ 3.46 when he ran it at 3.08ghz.

 

[Albuquerque]

Too much FSB termination voltage is a bad thing However I was able to see why it stopped dead after 435 FSB. That is the point at which you need above the purple voltages and go into the red for the Northbridge.

I wasn't referring to FSB termination voltage in my post above; that's something else -- CPU vTT is where I was suggesting, but different boards call it different things. It isn't the same as termination voltage though...

And do you really need that much northbridge voltage? Yikes... I've tested my board to 575FSB with zero additional MCH voltage, so I have to wonder if it's a board issue rather than a voltage issue...

 

[torin3]

I wasn't referring to FSB termination voltage in my post above; that's something else -- CPU vTT is where I was suggesting, but different boards call it different things. It isn't the same as termination voltage though...

And do you really need that much northbridge voltage? Yikes... I've tested my board to 575FSB with zero additional MCH voltage, so I have to wonder if it's a board issue rather than a voltage issue...

The problem is there is no VTT voltage. Here is pretty much what I have (pulled from another post about the same motherboard with the settings I can remember from here at work):

A.I. -MANUAL
CPU RATIO - 8.0
FSB - 430
FSB STRAP - 333
PCIE- 100
DRAM RATIO - 1:1
DRAM COMMAND RATE - Auto


STATIC READ - (Don't remember)
AI CLOCK TWISTER - (Don't remember)
TRANS ACTION BOOSTER - AUTO

V-CORE - 1.3000
CPU PLL - AUTO
FSB TERMINATION VOLTAGE - AUTO
DRAM VOLTS - 2.0
NORTH BRIDGE VOLTS - 1.57
SOUTH BRIDGE VOLTS - AUTO
LOADLINE CALIBRATION - ENABLED
CPU GTL VOLTAGE - AUTO
NORTH BRIDGE GTL VOLTAGE - AUTO
SOUTHBRIDGE 1.5 VOLTAGE - AUTO

CPU SPREAD SPECTRUM - DISABLED
PCIE SPREAD SPECTRUM - DISABLED

 

[Albuquerque]

Ahh, GTL is likely what they're calling it. Give that a go...

 

[torin3]

Ahh, GTL is likely what they're calling it. Give that a go...

I've tried it at 63x and 59x. I'm not really sure which way is better. When I get home, I'll post the available options.