Nehalem-EP......BLOOMFIELD

Printable View

Show 100 post(s) from this thread on one page

10-06-2008, 09:31 PM
AuDioFreaK39

Quote:

Originally Posted by karl_eller

Probably for similar reasons why a lot of us here at XS use way more than stock voltages (or way more than the VID, for that matter) on most of our CPUs: Because we can :p:

Eller

Yeah but Drwho? is an extreme ES tester and he's saying that no one should be running these things over 1.8v.

Quote:

Originally Posted by Drwho?

I keep saying that you will be better to look for modules with 1.5V to start with ... people really concidere this, some hight voltage modules will not start at all, so, you will not have the opportunity to set their voltage in the bios.
I am talking about over 1.8Volts. ( to protect your CPU , the default setting will not allow more than this, and many modules refuse to under volt/clock)
do you understand why?
10-06-2008, 09:35 PM
xVeinx

Coolaler was also adjusting the multis, etc in the process, and I believe that was done to offset the higher voltage to the memory. Not sure how it all works, but there might be some workarounds for the higher voltages...
10-06-2008, 10:23 PM
Drwho?

Quote:

Originally Posted by AuDioFreaK39

Yeah but Drwho? is an extreme ES tester and he's saying that no one should be running these things over 1.8v.

Ok, let s clarify this for ever ... if you run the voltage of the Dimm highter, you will impact the life span of your processor, it is like every overclocking, it is the non obvious impact.
Now, if you choose to reduce the life to few weeks by going crazy on the vDIMM, it is your choice, just do not complain about it :rofl:

Now, I did run up to 2.2Volts, being carefull, just be aware that transistors are like light switch, if you take a light switch of 110Volts, and make it commute thousands of time with 220 volts, the switch will burn out ... that's what you get when you OC ... and now look at the amplitude of regular OC.
Most of the people take a CPU that have based Voltage around 1.1 to 1.3 (depending of brand and model) and they overclock it to 1.4, 1.5, the best OC masters do 1.9 to 2.0 Volts... Higher, you got to mod the vreg.
now, people see 1.5Volts, and many are talking about 2.0 to 2.2Volts for the vDimm, this is a higher amplitude than what the masters of OC do on the CPU Voltage, so, the recommandation we are putting out is simple ... 1.65volts is 10% ... if you go higher, don t get to 70% directly, go there slowly, and if you do , you ll see that with 1.8Volts, you ll get between 2000Mhz to 2200Mhz depending of your luck on your dimms.

So, there is not problem of vDimm, we are just warning because it is new, and I don t want a young kid spending all of his saving for his dream CPU to finish into a :slapass: because the guy did not read the :fact:

The force is strong with Core i7, :lsfight: , and I want every body to enjoy it as it is , without silly mistake to spoil the party.:rocker: :rocker: :rocker: :rocker:
10-06-2008, 10:32 PM
twwen2

I wish i understood this phenomenon better. :shakes:

Is this an issue that has always been present, or is it a result of the move to an integrated memory controller?

I thought that the way around this was separate voltage regulation/phases for the memory, thus making it independent of the CPU voltage and visa-versa. Am i on the right track? :shrug:
10-06-2008, 10:38 PM
T_Flight

I'm gonna put mine on that new Rampage Extreme II. Yep, I'm gonna break down and get it assuming I can get it at or at least close to release. I hope they release that board then.

vDimm? I'll find some low volatge RAM that shows potential. I can always go up, but I can't go down if the thing doesn't POST.
10-06-2008, 10:39 PM
Drwho?

Quote:

Originally Posted by twwen2

I wish i understood this phenomenon better. :shakes:

Is this an issue that has always been present, or is it a result of the move to an integrated memory controller?

I thought that the way around this was separate voltage regulation/phases for the memory, thus making it independent of the CPU voltage and visa-versa. Am i on the right track? :shrug:

That is fairly easy ... statistically, transistors are suppose the transmit only electrons, this is how you design your semi conductor layers. when you increase the voltage, you increase the change to push an atom in its entire self instead of just a electron ... (Covalent collition) if you loose too many atoms ... say bye bye to the semi conductor layer ... then you just lost one of your transistor ...

simple, isn't it?
10-06-2008, 10:39 PM
MomijiTMO

So what I don't understand is why on earth do we currently need 2.1V to get similar performance on a nehalem platform with 1.7V [for example]. Is there some kind of massive inefficiency with the current platform?
10-06-2008, 10:41 PM
Drwho?

Quote:

Originally Posted by T_Flight

I'm gonna put mine on that new Rampage Extreme II. Yep, I'm gonna break down and get it assuming I can get it at or at least close to release. I hope they release that board then.

vDimm? I'll find some low volatge RAM that shows potential. I can always go up, but I can't go down if the thing doesn't POST.

every OC master should have a set of DIMM 1033 DDRIII to face the dim voltage manually ... then, shut down, place your other DIMMs and you are done. :yepp:
10-06-2008, 10:44 PM
Drwho?

Quote:

Originally Posted by MomijiTMO

So what I don't understand is why on earth do we currently need 2.1V to get similar performance on a nehalem platform with 1.7V [for example]. Is there some kind of massive inefficiency with the current platform?

It is part of moore's law ... DDRIII is getting better, and Paul and other Memory experts at Intel went and enable new requirement for memory. We were working on it for the last 3 years. You did not need it for Yorkfield, so, the industry did not do, now, you need it, they are all running to catch up QiMonda :rofl:
It is supplies and demands, people need it, it is for sale
10-06-2008, 10:46 PM
twwen2

Quote:

Originally Posted by Drwho?

That is fairly easy ... statistically, transistors are suppose the transmit only electrons, this is how you design your semi conductor layers. when you increase the voltage, you increase the change to push an atom in its entire self instead of just a electron ... (Covalent collition) if you loose too many atoms ... say bye bye to the semi conductor layer ... then you just lost one of your transistor ...

simple, isn't it?

Ah yes, i get the fundamentals of electron transfer and such.

What i don't understand is this:
We have been running our CPUs at 1.4-1.5v and our RAM at 2.0-2.2v without problems.
Why does Nehalem have a problem running these RAM voltages? Is it because the CPU and RAM share the same voltage?
10-06-2008, 10:51 PM
Drwho?

Quote:

Originally Posted by twwen2

Ah yes, i get the fundamentals of electron transfer and such.

What i don't understand is this:
We have been running our CPUs at 1.4-1.5v and our RAM at 2.0-2.2v without problems.
Why does Nehalem have a problem running these RAM voltages? Is it because the CPU and RAM share the same voltage?

The memory controler of Nehalem is in 45nm, the mem controler of Yorkfield is in 65nm in the X48 ... When you increase integration, you get a change to decrease the voltage and become more power efficent. Overclockers are the only one looking for voltage increase ... :clap:

The memory guys are ready with better memory. It is all logical, there is not problem.
10-06-2008, 10:52 PM
MomijiTMO

Ahh ok so we cant use the current DDRIII sticks but the new ones out will be at similar speeds but operating at a lower voltage.
10-06-2008, 10:55 PM
Drwho?

Quote:

Originally Posted by MomijiTMO

Ahh ok so we cant use the current DDRIII sticks but the new ones out will be at similar speeds but operating at a lower voltage.

you can use the previous one if you are tricky enough, but I recommand to move to new DIMM sticks, you platform will be all new gen.
10-06-2008, 10:55 PM
CZS

Drwho?

Do you believe that memory manufacturers will have low voltage "performance" ram ready by i7 launch?
10-06-2008, 10:57 PM
Drwho?

Quote:

Originally Posted by CZS

Drwho?

Do you believe that memory manufacturers will have low voltage "performance" ram ready by i7 launch?

I know for sure that the big players do.
10-06-2008, 10:59 PM
CZS

Quote:

Originally Posted by Drwho?

I know for sure that the big players do.

Excellent and thanks again for taking time to answer questions :up:
10-06-2008, 10:59 PM
T_Flight

I'm reading you DrWho. :yepp: I'm pretty sure I understand it, but will get the particulars later. I'm definitely not gonna OC that thing until I have a few hours on it, and after I've had a chance to study it...alot. I will need to do testing anyway to get a baseline on what it does at stock settings. No disasters will happen here. :)

I'm definitely gonna run new JEDEC standard RAM to start with. :yepp:
10-06-2008, 11:01 PM
gallag

DrWho, Thanks for taking the time to clear this up.
10-06-2008, 11:03 PM
T_Flight

Indeed! Thanks a bunch. DRWho has answered alot of my questions here, and has been a big help. :up:
10-06-2008, 11:08 PM
savantu

Quote:

Originally Posted by twwen2

Ah yes, i get the fundamentals of electron transfer and such.

What i don't understand is this:
We have been running our CPUs at 1.4-1.5v and our RAM at 2.0-2.2v without problems.
Why does Nehalem have a problem running these RAM voltages? Is it because the CPU and RAM share the same voltage?

In the past chipsets were build on a n-1 process generation.As a result , the IMC was older tech than the CPU.It also ran at far lower frequency in a higher voltage environment.

Moving it on die , means using the latest generation process which is more sensitive to voltage ( the goodies Intel uses at 45nm seems to have exacerbated that ).In other words , for traditional chipsets they could use bulkier , more resistant transistors that run at low frequency.Not the case anymore.

A solution seems to be playing with the independent voltages.Increase NB/IMC voltage so the difference to Vdimm stays the same or is lower.
10-06-2008, 11:10 PM
Kondik

Thanks Doc, much appreciate your help and the info. Ready to buy the nehalem *waves with his AMEX*
10-07-2008, 12:35 AM
twwen2

Right, that makes sense. Thanks all for taking the time to explain it all to a noob :rolleyes:.
10-07-2008, 12:17 PM
Blauhung

Quote:

Originally Posted by Drwho?

The memory controler of Nehalem is in 45nm, the mem controler of Yorkfield is in 65nm in the X48 ... When you increase integration, you get a change to decrease the voltage and become more power efficent. Overclockers are the only one looking for voltage increase ... :clap:

The memory guys are ready with better memory. It is all logical, there is not problem.

hehe, X48 is Bearlake on 90nm:p:
10-07-2008, 05:35 PM
dinos22

Quote:

Originally Posted by Drwho?

That is fairly easy ... statistically, transistors are suppose the transmit only electrons, this is how you design your semi conductor layers. when you increase the voltage, you increase the change to push an atom in its entire self instead of just a electron ... (Covalent collition) if you loose too many atoms ... say bye bye to the semi conductor layer ... then you just lost one of your transistor ...

simple, isn't it?

hi mate

thanks for the explanation. Is there any way to separate the vdimm voltage from CPU IMC and have there been any discussions in designing it that way to possibly reduce RMA hassles and allow for greater performance

Dino
10-07-2008, 06:00 PM
Drwho?

Quote:

Originally Posted by Blauhung

hehe, X48 is Bearlake on 90nm:p:

oupssss , correct, it is 90!

Show 100 post(s) from this thread on one page