CPUs with low VID

[Cookie Monster]

When I got myself an E0 stepping Q9550, the vid of it was 1.1375V (checked via coretemp).
If I understand correctly, this chip is required less voltage to operate at stock frequencies. However does this also relate to overclocking? i.e require less voltage to run at say 3.4GHz compared to a Q9550 with a vid of 1.25V?

 

[cusideabelincoln]

Yes, with a lower stock Vid there is a better possibility the chip will either overclock higher or overclock to the same speed as another chip but with a lower voltage.

Although 1.13 seems pretty low for a stock speed Q9550. Are you sure Speedstep was disabled? Power saving features will downclock and undervolt the chip when idle, although I'm not sure about the specifics of the Q9550.

 

[Idontcare]

Cookie, cusideabelincoln's post is correct (of course), I'll just add that the technical information underpinning his post and your topic post is called a shmoo plot if you wanted to track down any deeper fundamental info info on it.

 

[Cookie Monster]

Speedstep/C1E disabled.

Thats what I thought to since most Q9550s run at stock of 1.2~1.25V, yet atm Im running my chip at a vCore of 1.168 according to CPUZ (Set as 1.2V in BIOs). Although it drops down to 1.128V (CPUZ) during prime 95 which is abit alarming :Q. Whats more surprising is that there hasn't been a single crash.

 

[BTRY B 529th FA BN]

Can the mobo have something to do with the vcore or is it strictly a chip thing (god my grammar has gone to shit)

 

[Cookie Monster]

Originally posted by: Idontcare
Cookie, cusideabelincoln's post is correct (of course), I'll just add that the technical information underpinning his post and your topic post is called a shmoo plot if you wanted to track down any deeper fundamental info info on it.

Ah thanks. Theres always great value from your post IDC :thumbsup:

 

[Cookie Monster]

Im not sure atm. Just how reliable is CPUZ when it comes to reading voltages?

edit - Wow that was whack. I posted a reply to BTRY B 529th FA BN and somehow my reply ended up above his post.

 

[Idontcare]

Originally posted by: Cookie Monster
Im not sure atm. Just how reliable is CPUZ when it comes to reading voltages?

edit - Wow that was whack. I posted a reply to BTRY B 529th FA BN and somehow my reply ended up above his post.

CPU-z, as with all software-based voltage programs, is reliant on the accuracy of the readings as reported by the hardware on your motherboard.

To my knowledge CPU-z itself does not introduce any further error to the readings. It just tells you what it is being told by the hardware that is responsible for tracking the voltages.

That hardware will of course vary from mobo to mobo, but no one has ever documented or reported any extraneous examples of really piss-poor CPU-z Vcc readings as being solely attributable to a bad motherboard. Meaning regardless whether your cpu is really living with 1.13V, your readings relative to those of others is to be expected as "accurate enough" for our kind of enthusiast hobby work.

 

[Cookie Monster]

Here is the reading from CPUZ for the default core voltage.

Q9550@stock
1.104V ~ 1.048V (when loaded)

Assuming Everest is pretty close with the power consumption reading of the CPU, Ive never broke ~60W barrier even with linpack (overclocked). The temperature never broke above 56C either. Would it be possible to assume that this chip is equivalent to say a Q9550S? I think apoppin owns one of these puppies. (Would be nice to know what you VID was for the Q9550S, and the default voltage reading by CPUZ)

 

[Idontcare]

Originally posted by: Cookie Monster
Here is the reading from CPUZ for the default core voltage.

Q9550@stock
1.104V ~ 1.048V (when loaded)

Assuming Everest is pretty close with the power consumption reading of the CPU, Ive never broke ~60W barrier even with linpack (overclocked). The temperature never broke above 56C either. Would it be possible to assume that this chip is equivalent to say a Q9550S? I think apoppin owns one of these puppies. (Would be nice to know what you VID was for the Q9550S, and the default voltage reading by CPUZ)

Yes absolutely...understand that Q9550S is not some special stepping or design, it is your same garden variety Q9550 that demonstrated exceptionally good Vcc vs. GHz during parametric testing (needs low VID at stock clocks) and so was binned out as an "S" chip. Same thing they do when binning for mobile chips out of the "general population" - I am speaking to the LV's and the ULV's.

Prior to the creation of the "S" SKUs those chips still existed and were binned/sold as simply low VID standard Q9550's.

That said, the power-consumption estimates most folks throw around for the cpu's are based on the published TDP specs and not actual measurements of the CPU's power-consumption. (or sometimes they use system-level power consumption, better than nothing but still not cpu power consumption)

I like lostcircuits reviews because they go to great lengths to isolate and analyze the CPU's contribution to the overal power-consumption. For example, their QX9650 cpu was found to use a mere 64.8W at full load. So your <60W observation is certainly plausible.

 

[jackstar7]

Just to add to this a little of my own experience, I have both a Q9550 and an X3360 (yes, same thing) both with VIDs of about 1.25 or greater, and both undervolted at stock speed down to 1.08 or so via bios, and completely stable.

Anyone who doesn't need more than the speed this chip comes with might want to consider undervolting to bring down temps and power consumption, if those are concerns.

I'm running each in a Shuttle, so keeping things cool is a priority.

 

[imported_Shaq]

Yet there is one max voltage for a CPU. It is a lot different for say a 1.1 VID chip than one at 1.3 if the max is 1.5. If you raise both to 1.5 the 1.1 chip would be damaged much more quickly. I always thought there should be a VID+x, which would indicate how much higher than VID would be safe for a given CPU. Say up to .15v higher than VID would be the max voltage for a given CPU. I don't think this is spelled out anywhere but if the i7 max is 1.55v than that would be the max for the max VID chips-1.375v and not the D0's that are 1.1v VID. If any of that makes sense. lol But practically speaking heat will limit it to those parameters pretty closely anyway unless a person is using exotic cooling.

 

[Idontcare]

Originally posted by: Shaq
Yet there is one max voltage for a CPU. It is a lot different for say a 1.1 VID chip than one at 1.3 if the max is 1.5. If you raise both to 1.5 the 1.1 chip would be damaged much more quickly. I always thought there should be a VID+x, which would indicate how much higher than VID would be safe for a given CPU. Say up to .15v higher than VID would be the max voltage for a given CPU. I don't think this is spelled out anywhere but if the i7 max is 1.55v than that would be the max for the max VID chips-1.375v and not the D0's that are 1.1v VID. If any of that makes sense. lol But practically speaking heat will limit it to those parameters pretty closely anyway unless a person is using exotic cooling.

I don't think you quite understand the correlation between why a particular chip can function with a particular VID whereas another chip cannot, versus how and when and where and why voltage degrades and damages a chip.

As I interpret your post, assuming I am doing that correctly, the premise you state is incorrect.

A 1.1V VID chip is to be expected to have the same lifetime statistics as a 1.3V VID chip if both chips are overvolted to 1.5V. (assuming all else is held constant, chips are the same design/stepping and made with the same process tech, etc...i.e. we are comparing two otherwise seemingly identical i7 920's for instance)

Vcc needed for stable device operation is determined by the transistor threshold voltage whereas the voltage-induced deterioration (in particular see the section on Black's Law) is determined by the dimensionality (geometry) of imperfections and process-induced non-uniformity's present in the metal wires and insulating dielectrics in the transistors as well as the interconnect.

Those imperfections do not typically effect the sub-threshold voltage, as such the voltage needed for device operation versus the voltage needed to destroy the device are two independent attributes of an IC.



(I hope my post doesn't come across as rude or harsh, its late here and I am finding myself at a loss for words as to how to kindly imply/state your post is in error to the best of my knowledge.)