There's been a lot of talk from people over what the absolute maximum voltage for Sandy Bridge cpu's are, and the "safe maximum."
Note: if you don't like reading: please skip to the next thread (tl;dr). But if you want some more information about this "VID" stuff, read on.
The VID range, is a range of voltages, between minimum and maximum VID range, which the CPU's can be calibrated to request as an automatic voltage (when the CPU vcore is set to "automatic", and which can be read from a register inside the CPU, based on its operating parameters).
That's NOT the same as maximum safe voltage.
The tech documents actually explain this.
This is from tech document 1 of 2 (January 2011):
The VID is set for EACH individual CPU, and it's set that TWO IDENTICAL CPU's might have different VID levels!
Now it would take an intel engineer to come in here and explain why one CPU has a default VID at 3.4 ghz, of 1.212v, while another CPU has a default VID at 3.4 ghz at 1.250v. However, at least from pretty standard testing over the last few years, CPU's with lower VIDS tend to overclock farther than CPU's with higher VIDS. However, it was also determined over on xtremesystems, that CPU's with lower vids, that overclocked better, sometimes ran hotter (aka "leaky chip").
Also, it's clear from the way the VID chart progresses from 3.4 ghz to 3.8 ghz, that Intel directly tests the cpu's at these fully supported speeds (Turbo boost is 100% guaranteed by Intel). And the vid skyrockets between 3.4 to 4 ghz. But then the default VID for each multplier seems to bottom out (I suspect that the chips are not tested at higher multis, but the VID is made so that the chip will at least be able to complete POST; someone with knowledge of VRD 12 can hopefully fill in more information).
Intel tests their cpu's to see how they perform at harsh operations (which is one reason why Linpack exists), and I suspect their CPU's are either binned according to several specific torture tests, or have the VID signal implanted into the CPU based on how well it does on these tests.
The 1.52v that you see on the charts is the absolute MAXIMUM VID that is allowable under the VRD 12 specification, and if you look at the binary field charts, you can find that out from the binary on/off bits. It does NOT mean that the particular CPU sample can handle that VID or even has it implanted in its voltage request table at high frequencies. In fact, there is probably NO cpu even made that has a default VID as high as 1.52v at a x58 multiplier, for example.
Now what's interesting is, older CPU's used a VID range from 0.825v-1.3625v or 1.4v, but the binary VID chart goes up to 1.6v on those. But only goes up to 1.52v here.
ANOTHER problem is that on the previous cpu's, intel specified an absolute maximum voltage, beyond which (at least on air cooling) that the cpu's WOULD start taking damage, and this absolute maximum was HIGHER than the "Vid range" shown in the number figure, but always LOWER than the "Vid chart" shown on the binary bit chart.
Needless to say, this isn't exactly user friendly.
What's even worse, the specification update is for ERRATA.
Someone really needs to contact Intel and have them explain "what happened to the absolute minimum and maximum ratings" section.
In summary, NO ONE Knows the "absolute minimum and maximum ratings" for the voltage, because Intel has NOT specified them. Also, NO ONE knows why Intel is using the "Vid range" from the binary vid chart (0.25v-1.52v) for these processors, when the previous cpu's had different results for the binary vid chart (-0.3v (I think; I remember it was negative voltage) to 1.60v), and the vid range (0.825v-1.3625 (or 1.4v). with an "absolute maximum" in yet another table that was always higher than the "VID range". But I have no idea why the binary VID chart had this range, when the VID range was much smaller.
Until someone can explain if the VID range was increased deliberately for VRD 12, and the CPU's made more tolerant to vcore, or if Intel made an error, by just copying the binary VID chart (very, very likely), that will have to wait until someone can find out the truth.
Edited by Falkentyne - 3/30/11 at 11:09pm
Note: if you don't like reading: please skip to the next thread (tl;dr). But if you want some more information about this "VID" stuff, read on.
The VID range, is a range of voltages, between minimum and maximum VID range, which the CPU's can be calibrated to request as an automatic voltage (when the CPU vcore is set to "automatic", and which can be read from a register inside the CPU, based on its operating parameters).
That's NOT the same as maximum safe voltage.
The tech documents actually explain this.
This is from tech document 1 of 2 (January 2011):
Quote:
|
The processor uses three signals for the serial voltage identification interface to support automatic selection of voltages. Table 7-1 specifies the voltage level corresponding to the eight bit VID value transmitted over serial VID. A ‘1’ in this table refers to a high voltage level and a ‘0’ refers to a low voltage level. If the voltage regulation circuit cannot supply the voltage that is requested, the voltage regulator must disable itself. VID signals are CMOS push/pull drivers. Refer to Table 7-9 for the DC specifications for these signals. The VID codes will change due to temperature and/or current load changes in order to minimize the power of the part. A voltage range is provided in Table 7-5. The specifications are set so that one voltage regulator can operate with all supported frequencies. Individual processor VID values may be set during manufacturing so that two devices at the same core frequency may have different default VID settings. This is shown in the VID range values in Table 7-5. The processor provides the ability to operate while transitioning to an adjacent VID and its associated voltage. This will represent a DC shift in the loadline. |
Now it would take an intel engineer to come in here and explain why one CPU has a default VID at 3.4 ghz, of 1.212v, while another CPU has a default VID at 3.4 ghz at 1.250v. However, at least from pretty standard testing over the last few years, CPU's with lower VIDS tend to overclock farther than CPU's with higher VIDS. However, it was also determined over on xtremesystems, that CPU's with lower vids, that overclocked better, sometimes ran hotter (aka "leaky chip").
Also, it's clear from the way the VID chart progresses from 3.4 ghz to 3.8 ghz, that Intel directly tests the cpu's at these fully supported speeds (Turbo boost is 100% guaranteed by Intel). And the vid skyrockets between 3.4 to 4 ghz. But then the default VID for each multplier seems to bottom out (I suspect that the chips are not tested at higher multis, but the VID is made so that the chip will at least be able to complete POST; someone with knowledge of VRD 12 can hopefully fill in more information).
Intel tests their cpu's to see how they perform at harsh operations (which is one reason why Linpack exists), and I suspect their CPU's are either binned according to several specific torture tests, or have the VID signal implanted into the CPU based on how well it does on these tests.
The 1.52v that you see on the charts is the absolute MAXIMUM VID that is allowable under the VRD 12 specification, and if you look at the binary field charts, you can find that out from the binary on/off bits. It does NOT mean that the particular CPU sample can handle that VID or even has it implanted in its voltage request table at high frequencies. In fact, there is probably NO cpu even made that has a default VID as high as 1.52v at a x58 multiplier, for example.
Now what's interesting is, older CPU's used a VID range from 0.825v-1.3625v or 1.4v, but the binary VID chart goes up to 1.6v on those. But only goes up to 1.52v here.
ANOTHER problem is that on the previous cpu's, intel specified an absolute maximum voltage, beyond which (at least on air cooling) that the cpu's WOULD start taking damage, and this absolute maximum was HIGHER than the "Vid range" shown in the number figure, but always LOWER than the "Vid chart" shown on the binary bit chart.
Needless to say, this isn't exactly user friendly.
What's even worse, the specification update is for ERRATA.
Someone really needs to contact Intel and have them explain "what happened to the absolute minimum and maximum ratings" section.
In summary, NO ONE Knows the "absolute minimum and maximum ratings" for the voltage, because Intel has NOT specified them. Also, NO ONE knows why Intel is using the "Vid range" from the binary vid chart (0.25v-1.52v) for these processors, when the previous cpu's had different results for the binary vid chart (-0.3v (I think; I remember it was negative voltage) to 1.60v), and the vid range (0.825v-1.3625 (or 1.4v). with an "absolute maximum" in yet another table that was always higher than the "VID range". But I have no idea why the binary VID chart had this range, when the VID range was much smaller.
Until someone can explain if the VID range was increased deliberately for VRD 12, and the CPU's made more tolerant to vcore, or if Intel made an error, by just copying the binary VID chart (very, very likely), that will have to wait until someone can find out the truth.
Edited by Falkentyne - 3/30/11 at 11:09pm
SB Rig
(13 items) |
|
SB Rig
(13 items) |
