Delidded my i7-3770K, loaded temperatures drop by 20°C at 4.7GHz

[BonzaiDuck]

VID is never a value that you set.

VID is the value of voltage that the processor requests at a given speed. You cannot alter this.

. . . But determined by your BIOS choices, no? In other words, if I select something like 1.38725 for VCORE, the processor might request 1.42V at the OC'd speed setting. If I choose 1.35, the VID could be 1.385V . . . etc.

 

[Ferzerp]

. . . But determined by your BIOS choices, no? In other words, if I select something like 1.38725 for VCORE, the processor might request 1.42V at the OC'd speed setting. If I choose 1.35, the VID could be 1.385V . . . etc.

Nope.

This is for mobile, but desktop is the same.

See PDF page 92

http://www.intel.com/content/dam/ww...rd-gen-core-family-mobile-vol-1-datasheet.pdf

 

[BonzaiDuck]

Nope.

This is for mobile, but desktop is the same.

See PDF page 92

http://www.intel.com/content/dam/ww...rd-gen-core-family-mobile-vol-1-datasheet.pdf

Thanks. I was aware of a distinction between vcore and VID -- that vid is a "signal" or "what the processor requests." Of course, those distinctions are clear with the QX9650 article . . . I'll take a closer look at the spec sheet . . .

[BACK AGAIN] I come away with this with the same conceptions I had before, that there is a set of VID values for various speeds, that the processor can transition between VIDs. Am I missing something here? Or are we thrashing over semantics?

 

[Kenmitch]

Thanks. I was aware of a distinction between vcore and VID -- that vid is a "signal" or "what the processor requests." Of course, those distinctions are clear with the QX9650 article . . . I'll take a closer look at the spec sheet . . .

[BACK AGAIN] I come away with this with the same conceptions I had before, that there is a set of VID values for various speeds, that the processor can transition between VIDs. Am I missing something here? Or are we thrashing over semantics?

Pretty much the way it works....It's the 1.52v part that most can't live with. I figure if it would fry the chip then Intel would have never allowed such high voltage. The VID will also fluctuate by type of load such as IBT vs Prime or Cinabench etc. Guess it depends on features being hit along with how many cores, etc.

My 2550k has taken 1.7v's and lives to tell about it.

I think Idontcare once pumped 2v's plus into one of his chips(2600k?) and it lived. Of course it was one of those oops moments for him.

 

[BonzaiDuck]

Pretty much the way it works....It's the 1.52v part that most can't live with. I figure if it would fry the chip then Intel would have never allowed such high voltage. The VID will also fluctuate by type of load such as IBT vs Prime or Cinabench etc. Guess it depends on features being hit along with how many cores, etc.

My 2550k has taken 1.7v's and lives to tell about it.

I think Idontcare once pumped 2v's plus into one of his chips(2600k?) and it lived. Of course it was one of those oops moments for him.

Thanks. I think we're talking about the same thing here. My approach and limitations: I want everything to be lower than 1.4 (on my sandy Bridge). If I can get the sustained voltage for any load or idle to hang below 1.35+, it's better.

And I can believe IDC had an "oops," because he does a lot in phenomenally minimal time. I take plenty of time to do very little. It's either old age or willfulness.

 

[Idontcare]

Pretty much the way it works....It's the 1.52v part that most can't live with. I figure if it would fry the chip then Intel would have never allowed such high voltage. The VID will also fluctuate by type of load such as IBT vs Prime or Cinabench etc. Guess it depends on features being hit along with how many cores, etc.

My 2550k has taken 1.7v's and lives to tell about it.

I think Idontcare once pumped 2v's plus into one of his chips(2600k?) and it lived. Of course it was one of those oops moments for him.

Yeah, I goofed when first using the offset-mode feature for overvolting with Asus' AI Suite and accidentally pumped ~2.4V through my 2600k for about 30s before I happened to glance at the voltmeter and about nearly had a heart-attack while I scrambled for the power switch on the PSU

It was none for the worse though.

 

[dqniel]

You mean "At 100% LLC . . . . vcore rises under load," don't you? I was very cautious about this with the SB when I started, preferring "No LLC," and then "25%" LLC. As long as it allows me to do what I did -- at same time reducing idle ("turbo") VCORE from 1.38 to 1.35, I feel pretty good about "50%."

Anyway, we diverge and digress. "TIM and de-lidding . . . "

Nope. I meant 0%. Seems backwards to me, too.

 

[WhoBeDaPlaya]

Nope. I meant 0%. Seems backwards to me, too.

You know, I wonder if the poor shmuck BIOS engineers got the 0% and 100% settings bass ackwards
Which BIOS version are you using? My board originally came with the retail version but I neglected to test LLC on that (immediately updated to 1.40).

 

[BonzaiDuck]

You know, I wonder if the poor shmuck BIOS engineers got the 0% and 100% settings bass ackwards
Which BIOS version are you using? My board originally came with the retail version but I neglected to test LLC on that (immediately updated to 1.40).

Maybe some idiot at the "BIOS factory" was thinking "0% vDroop" when those choices were made. But with my P8Z68 board, reviews judged that the LLC feature had been done "nicely" so that one of the intermediate settings provided perfect cancellation of the droop, as opposed to a setting on the extreme end of the setting range . . .

Sad, really, that a consumer has to discover by trial and error what the idiot intended . . .

 

[WhoBeDaPlaya]

Managed to get 4.4GHz stable with an average Vcore of 1.236V under IBT load.
Unfortunately, changing the RAM from 4x 4GB DDR3-1333 CL9 (what I had on hand) to 4x 8GB DDR3-1600 CL10 (just got it from Amazon) busted down my max stable OC to 4.3GHz

 

[BonzaiDuck]

Managed to get 4.4GHz stable with an average Vcore of 1.236V under IBT load.
Unfortunately, changing the RAM from 4x 4GB DDR3-1333 CL9 (what I had on hand) to 4x 8GB DDR3-1600 CL10 (just got it from Amazon) busted down my max stable OC to 4.3GHz

What sort of RAM? Take a closer look at the RAM voltage spec, and see if you have some safe headroom to bump up the VCCIO voltage a tad. I don't know what's "safe" or reasonable for the Ivy chips -- right now my experience is based on the Sandy. But there should be a comfortable range of voltage settings that won't endanger your IB, and you might only need to bump up VCCIO by 0.0025 to 0.01+V.

I'm assuming you are running these RAM sticks at their stock, spec settings? Your main b**** is getting through adequate RAM testing on the new sticks for 32GB. It takes forever with even half that amount. I waited patiently for almost three days, "Ah tell ya!"

 

[WhoBeDaPlaya]

They're Komputerbay sticks, rated at 1.5V, 1600, 10-10-10-27.
I kept the same OC settings (except dropping the multi to 40x from 44x) while testing the sticks in memtest86+ (left it running for a full day with no errors). Just wasn't happy at 4.4GHz.

 

[BonzaiDuck]

They're Komputerbay sticks, rated at 1.5V, 1600, 10-10-10-27.
I kept the same OC settings (except dropping the multi to 40x from 44x) while testing the sticks in memtest86+ (left it running for a full day with no errors). Just wasn't happy at 4.4GHz.

It could be something else, but the VCCIO might be the focus of attention here. Of course, in my own experience, I was testing 1600 RAM to run at 1866. And you would think that your CPU over-clock would remain separate and stable from the RAM. But it's something you might look into.

I"m also assuming that your 4.4 OC did not involve fiddling with the bCLK?

 

[WhoBeDaPlaya]

It could be something else, but the VCCIO might be the focus of attention here. Of course, in my own experience, I was testing 1600 RAM to run at 1866. And you would think that your CPU over-clock would remain separate and stable from the RAM. But it's something you might look into.

I"m also assuming that your 4.4 OC did not involve fiddling with the bCLK?

Will poke around with that and nope, definitely did not mess with the BCLK.

Heck, also forced TurboBoost to be disabled (Asrock boards have an annoying habbit of maitaining Turboboost when applying a single multi to "All Cores" - have to specify them separately) and turned off Speedstep, C1E, C3 and C6.

Will be running heavy calculations on this pretty 24/7/365, so the power saving features don't really matter

 

[Dufus]

VID is never a value that you set.

VID is the value of voltage that the processor requests at a given speed. You cannot alter this.

With the older processors such as core2 you could change VID but with the newer processors it seems only turbo VID's can be changed.

Loadline at 100% should conform to Intel specs where vcore should drop respective to current draw (load) from VID, i.e a VID of 1.2V with little load should produce a vcore close to 1.2V whereas with a high load vcore might drop to 1.1V for instance, if at that load the loadline was set to 50% then that should give 1.15V vcore.

Anyway just got a 3770k to play with, some strange results...



Forgot to add, if your using fixed voltage then VID is ignored altogether for how much voltage the core gets.

 

[Ferzerp]

But the point is VID is the voltage the proc is requesting and set during manufacturing. You don't change it. Ever. You may not provide the VID requested voltage, but it is not something you change. It is based on the individual proc, and the speed it is currently running at.

Claiming you are changing it is a little like claiming a 2600k that you run 100Mhz over stock is suddenly a 2700k.

 

[dqniel]

You know, I wonder if the poor shmuck BIOS engineers got the 0% and 100% settings bass ackwards
Which BIOS version are you using? My board originally came with the retail version but I neglected to test LLC on that (immediately updated to 1.40).

I'm on 1.30 for the Pro4-M.

 

[Dufus]

But the point is VID is the voltage the proc is requesting and set during manufacturing. You don't change it. Ever.

Example of running 40x multi and manually changing VID, note that CPU-Z calculation is a little off.

40x VID 1.280V


40x VID 1.325V


40x VID 1.410V


40x VID 1.520V

 

[Ferzerp]

That box that you're calling "VID" is where cpu-z displays core voltage, not VID.

Not sure why you've changed CPU-Z to say "VID".

Again, VID is something the processor is requesting. You don't magically tell it to request a different voltage than it is requesting. You might change parameters which might cause it to request a different voltage ((while idle, it requests a lower voltage, etc)), but you aren't controlling VID.

 

[Dufus]

If you change sensor=1 to sensor=0 in the cpuz.ini file it will display VID instead of vcore. The reason to change it to show VID is that it would not be any use to show vcore to demonstrate that VID values can be changed by software. That 40x multi was run with EIST, C1E, C3 and C6 all disabled so no idle other than C1 which just stops the clock, it does not lower voltages.

As I Stated before

With the older processors such as core2 you could change VID but with the newer processors it seems only turbo VID's can be changed.

 

[Ferzerp]

But you are, by no means, setting a specific VID. That's what the proc is asking for (and is a table that is configured per processor during manufacturing, you don't edit this table).

You're claiming to set it. That just doesn't happen.

edit: maybe the confusion is that you think I am saying that there is a specific VID for each proc and only one VID. That's not the case. It is a full table of values based on current multiplier, load, power state, etc. The settings and current workload may cause it to look up a different value, but that table is set during manufacturing, and you cannot alter it.

edit2: What I am saying is that the VID you see is the result of a table lookup that the proc does based on all the input parameters. Think of it like a read-only spreadsheet on a pressed CD. If you pass it the inputs that cause it to read the voltage from cell A5 (let's say 1.200V), it will always ask for 1.200V. A different processor might have a different value in cell A5, but your processor will always ask for 1.200V when A5 is looked up. When I say "setting VID", I mean suddenly making cell A5 not contain 1.2000V. This never happens. Intel has set this cell during manufacturing, and you can't touch it. You can change something (say the multiplier) that makes it look up cell D3 instead, which might be 1.340V for example, but again, the state that makes it look up cell D3 will always result in a VID of 1.340V. These tables of values are calibrated on a per-processor basis by Intel, and once written, you don't touch them.

 

[BonzaiDuck]

But you are, by no means, setting a specific VID. That's what the proc is asking for (and is a table that is configured per processor during manufacturing, you don't edit this table).

You're claiming to set it. That just doesn't happen.

edit: maybe the confusion is that you think I am saying that there is a specific VID for each proc and only one VID. That's not the case. It is a full table of values based on current multiplier, load, power state, etc. The settings and current workload may cause it to look up a different value, but that table is set during manufacturing, and you cannot alter it.

edit2: What I am saying is that the VID you see is the result of a table lookup that the proc does based on all the input parameters. Think of it like a read-only spreadsheet on a pressed CD. If you pass it the inputs that cause it to read the voltage from cell A5 (let's say 1.200V), it will always ask for 1.200V. A different processor might have a different value in cell A5, but your processor will always ask for 1.200V when A5 is looked up. When I say "setting VID", I mean suddenly making cell A5 not contain 1.2000V. This never happens. Intel has set this cell during manufacturing, and you can't touch it. You can change something (say the multiplier) that makes it look up cell D3 instead, which might be 1.340V for example, but again, the state that makes it look up cell D3 will always result in a VID of 1.340V. These tables of values are calibrated on a per-processor basis by Intel, and once written, you don't touch them.

Amen, bros and sisters, Amen. Stay in Peace.

It's going to take a while until we find out more about the TIMs, temps and performance-reliability currently of interest.

 

[Dufus]

Ferzerp I am well aware of the VID LUT and of course you don't change the values in the LUT, you change the VID request value from the CPU to the VRM via MSR. As shown in the example above, a request is made to the CPU to run at 40x multi, the CPU looks up the VID value for 40x, say the binary value that needs to be transmitted to the VRM to request 1.280V. Now maybe I want more than 1.280V so that VID request value can be changed to a higher value for 40x multi as already demonstrated (40x multi running with a VID request from the CPU to ask the VRM for 1.52V).

 

[-Slacker-]

I'm guessing this question's been asked before in this huge thread, but...

Why doesn't intel resolve this issue already when they could shave off temps in the double digits just by changing the TIM with something better?

 

[dma0991]

I'm guessing this question's been asked before in this huge thread, but...

Why doesn't intel resolve this issue already when they could shave off temps in the double digits just by changing the TIM with something better?

Would shaving off temps in the double digit with solder = more money in Intel's coffers?