Nehalem overclocks to 4GHz+

D

DaddyDC650

Limp Gawd
Joined
Dec 31, 2005
Messages
311
Awesome if true!

It's not all rosy with Nehalem. We've learned that in spite of its 4GHz+ overclocking potential on air, that there are some limitations. You simply cannot get more than 1.65V to the chip which might be a limiting factor for high end overclocking.

Its gets better as the CPU and memory voltage have to be synchronous which implies that not all DDR3 memory will run with Nehalem. At this time you can forget all DDR3 2000 modules with voltages over 1.65V as they simply won't work on current Nehalem motherboards.

We know that Intel is aware of the issue but we don't know can it actually fix it. Memory people most certainly won't like it if it stays that way.

http://www.fudzilla.com/index.php?option=com_content&task=view&id=9074&Itemid=35

http://www.fudzilla.com/index.php?option=com_content&task=view&id=9072&Itemid=1
 
lol I don't think most people consider Yorkfields very "overclockable".
 
I wonder if that's through unlocked multipliers on an Extreme Edition (or equivalent engineering sample) chip, or though the raising of the reference clock speed?

Definitely good to see though.
 
I will actually just take a 3.2 Nehalem at stock speed and be really happy lol :)
 
If they are talking about the 3.2 they mean the EE, unlocked multi. To hit 4GHz on the base 2.66Ghz version, you have to bump the bus by quite a bit. But everything else based on the ref clock should have their own ratios on it...200Mhz reference gets you 4Ghz CPU.

Thats a big jump. Im debating air or water right now, if water I might try to push for that 200 number :D Air...maybe just 166.
 
I don't buy that the CPU and memory voltages have to be synchronous. Doesn't really make any sense and it would put serious limitations on your memory usage.
 
Considering that DDR3 reference voltage is 1.5v and nehalem will be running something like .9-1.3v, locking those together is NOT something that intel would have done.
And who doesnt consider ~1.5x overclocking that people get from yorkfields to be "good" overclocking? Just because they arent doing 4.5Ghz like some wolfdales doesnt mean they dont overclock better than almost anything in the past.
 
Peteman100 said:
I don't buy that the CPU and memory voltages have to be synchronous. Doesn't really make any sense and it would put serious limitations on your memory usage.
Click to expand...


well intel may have not designed the chip that way. i know AMD with the IMC that if u put too much voltage to the memory it will fry the IMC.

 
Unknown-One said:
I have never heard of this happening on ANY chip with an IMC (AMD or otherwise).

Got a source to back that up?
Click to expand...
In the beginning of 90nm 939 days when some users were still running Winbond memory chips which craved voltage, folks who used stock voltages or CnQ with their memory at 3.3v would experience degrading OC maximums until stock speed would not run without added voltage. IIRC it was a diode that isolated the IMC from the termination of the memory and if there was too much of a difference, this diode would be damaged. I believe the safe minimum VCore was DDR / 2 - .325 = V. When people were pumping 3.3-3.5v through the memory and a VCore of only 1.3-1.35, it became a problem. I've not read of any similar issues with AM2 systems.
 
PliotronX said:
In the beginning of 90nm 939 days when some users were still running Winbond memory chips which craved voltage, folks who used stock voltages or CnQ with their memory at 3.3v would experience degrading OC maximums until stock speed would not run without added voltage. IIRC it was a diode that isolated the IMC from the termination of the memory and if there was too much of a difference, this diode would be damaged. I believe the safe minimum VCore was DDR / 2 - .325 = V. When people were pumping 3.3-3.5v through the memory and a VCore of only 1.3-1.35, it became a problem. I've not read of any similar issues with AM2 systems.
Click to expand...

Yup. That.
 
GLSauron said:
Considering that DDR3 reference voltage is 1.5v and nehalem will be running something like .9-1.3v, locking those together is NOT something that intel would have done.
And who doesnt consider ~1.5x overclocking that people get from yorkfields to be "good" overclocking? Just because they arent doing 4.5Ghz like some wolfdales doesnt mean they dont overclock better than almost anything in the past.
Click to expand...
Well, I was mostly referring to the Q9450's (as there aren't as many Q9550 and Q9650 results floating around)..and the majority of the results I've seen do not point to Yorkfields being all that great, especially considering the Q6600's and the die shrink. Anyway, I'm sure some people got 4 GHz out of their Q9450s, but I wouldn't say the chips did well overall. *shrug*
 
gwai lo said:
Well, I was mostly referring to the Q9450's (as there aren't as many Q9550 and Q9650 results floating around)..and the majority of the results I've seen do not point to Yorkfields being all that great, especially considering the Q6600's and the die shrink. Anyway, I'm sure some people got 4 GHz out of their Q9450s, but I wouldn't say the chips did well overall. *shrug*
Click to expand...

I think a lot of that has to do with the lower multiplier on the Q9450s. With the 8x multiplier you need to be able to run 450FSB to get to 3.6Ghz which is where many of us have Q6600s running. To hit 4Ghz, you need to be able to run 500FSB. Even with motherboards getting better, 500FSB still isn't the easiest thing to hit on a regular basis.

 
Yep, I got the Q9550 for the 8.5 multiplier for easier med-range overclocking without overtaxing FSB. Of course, I'm looking for long term overclock, not 1-2 year, therefore, the max I'll go is 3.6 but would be satisfied at 8.5 x 400 = 3400.
 
Well you can't expect all chips to overclock like E6300. That's just setting the bar way too high.
 
PliotronX said:
In the beginning of 90nm 939 days when some users were still running Winbond memory chips which craved voltage, folks who used stock voltages or CnQ with their memory at 3.3v would experience degrading OC maximums until stock speed would not run without added voltage. IIRC it was a diode that isolated the IMC from the termination of the memory and if there was too much of a difference, this diode would be damaged. I believe the safe minimum VCore was DDR / 2 - .325 = V. When people were pumping 3.3-3.5v through the memory and a VCore of only 1.3-1.35, it became a problem. I've not read of any similar issues with AM2 systems.
Click to expand...

It is a known issue with the Phenoms. the most ddr2 vdimm they can take is 2.1-2.2v

I found out the hard way thanks to newegg selling me ram that was rated for 2.1 but said 2.35 on the label once i got them so i set the ram to 2.35vdimm not thinking and the cpu only lasted ~ 5 minutes.

 
DaddyDC650 said:
Awesome if true!

It's not all rosy with Nehalem. We've learned that in spite of its 4GHz+ overclocking potential on air, that there are some limitations. You simply cannot get more than 1.65V to the chip which might be a limiting factor for high end overclocking.

Its gets better as the CPU and memory voltage have to be synchronous which implies that not all DDR3 memory will run with Nehalem. At this time you can forget all DDR3 2000 modules with voltages over 1.65V as they simply won't work on current Nehalem motherboards.

We know that Intel is aware of the issue but we don't know can it actually fix it. Memory people most certainly won't like it if it stays that way.

http://www.fudzilla.com/index.php?option=com_content&task=view&id=9074&Itemid=35

http://www.fudzilla.com/index.php?option=com_content&task=view&id=9072&Itemid=1
Click to expand...
I don't think the voltage limitations will be an issue cause the stock volts on a Nehalem are somewhere in the .80-.90 range. 1.65V would probably kill a Nehalem without extreme cooling like a high end cascade or LN2. I would say that you can probably expect to hit overclocks well into the 4Ghz realm on a mere 1.1-1.2V on air.

Edit: Also there is pretty much no way that they would link the vcore and the vdimm. That would just be dumb.
 
You must log in or register to reply here.
Back
Top