P67A-UD7 Most In-Depth Preview/Review!!

[pr@$r1g]

Awesome man

Just ripping apart heatsink ,in depth info of every microchip ..... I really like

I don't know im going to buy this or not ,i personally like to keep looking on every mobo graphics boards ,there's so much to learn

Never saw something like this before.

[Leeghoofd]

nice work man and a a good looking Giga lineup.

But really wondering how all these boards/features will be worthwhile as the real OC potential is in the CPU (K model) and not in the board...

[Sn0wm@n]

this smiley explains it all ... best thourough review ive seen in a while

[LardArse]

I dont see what switching frequency or response time has to do with 'digital' or 'analogue'. IMO the main difference is just companies use the word 'digital' when you can use software to adjust loadline and stuff without additional circuitry while analogue relies on GPIOs to switch between different settings. A very vague way to judge a pwm for an end user unless in comparison of flexibilities which boil down to what he can truly control via bios/softwares.

[darckhart]

hey great review! thanks especially for the power lesson and the heat tests of the cooling system. a few questions:

1. to clarify, these are still pcie2.0 x16 slots (not 2.1, not 3.0)? the NF200 provides the SLI ability by doing "something fancy" (i don't quite understand how it works) so you can get SLI x16/x16 off the 8 lanes, so is the same "something fancy" done so that crossfire x16/x16 is possible? how are the "extra" lanes gotten if the cpu doesn't actually got them and there's no "extra" chip like NF200 on the ati side of things?

2. can you try putting in some graphics cards to check for clearance? i am interested in a) extremely long cards (4870x2) running into/covering the sata/usb/etc connectors at the edge of the board, and b) particularly pceix16 slot 4 (possibly #3 too) where the heatsink and caps may be too high and prevent fully shrouded cards (gtx260) from actually locking in tightly.

3. from your photo of the back panel, the yellow ports, i guess 2xesata, 2xusb2, a 1394, and what the heck is that tiny little yellow one?

thanks!

[Clint]

Awesome review! Very tempted even though I decided to wait until next shrink.

[sin0822]

I dont see what switching frequency or response time has to do with 'digital' or 'analogue'. IMO the main difference is just companies use the word 'digital' when you can use software to adjust loadline and stuff without additional circuitry while analogue relies on GPIOs to switch between different settings. A very vague way to judge a pwm for an end user unless in comparison of flexibilities which boil down to what he can truly control via bios/softwares.

Switching frequency is important as it has to do with the duty cycle and how much time the phases are on/off higher frequency more on time= more power consumption =higher output. They are all directly correlated, and I have found that Digital PWMs are good as long as you use Chil and not volterra. I did not say higher switching frequency was better, I stated that Digital VRM design does NOT REQUIRE high switching frequency then i went and listed the pros of Digital PWMs such as ability to control switching frequency (which isn't a feature of all digital PWM's only the highend, but that is what we are comparing) . Digital PWMs ability to to work at lower switching frequency is a plus, duty cycle is decreased and directly correlated to it is temperature and efficiency. I personally don't see why the user needs to control any aspect of the PWM (many don't even know what it stands for let alone give them the ability to change vcore switching frequency isn't smart). Maybe release a special BIOS that lets pro users like you to change it, just like Gigabyte released GOOC bios for rev 2.0 x58 boards that allowed higher voltage range, OCP disable, and other features normal users should not be able to control.

Something more I want to add is that Digital PWMs do take time(ADC and DAC) to convert analogue to digital and back, and this is their biggest downfall, as when comparing vout to vref the difference, aka the difference from what is outputted to what requested by user take a while, with analogue PWMs you don't need to deal with the conversion of analogue to digital. When you want an OC like 4.5-5ghz the board probably knows what is better. These boards are smarter than us, we some of them at least.

But the need for an analogue system still is there, and the fact that Gigabyte has been able to implement such an Analogue VRM design is phenomenal, the ability to control voltage within 2mv of a digital system is impressive to say the least(as in ASUS's =1.205,1.210,1.215,1.220)vs (GIGABYTE's = 1.2065, 1.2130,1.2195,1.226). I have no idea how they implemented 3 levels of LLC support like Asus's Rampage 3 series. What does all this prove?? Well that analogue can in fact perform up to par with Digital. But you still have the huge benefits of an all analogue system such as that huge 24 phase array, to no need to conversion of signals. But Digital has its own gains just as precise software control, much lower temps, and better control over board real estate.
Response time and vcore switching frequency are important when it comes to VRM design. While the exact number really is unimportant, i wanted users to see the difference between analogue and digital VRM design because they are VERY different, the loadline and an equation and is easier controlled by the digital VRM as is voltage precision. As you saw on the R3E they had 3 Levels of LLC controller, Gigabyte had the same thing but on an Analogue system so it is possible just tougher to implement because there isn't any programmable firmware, as much as having it pre-programed(or separate IC). But here is where switching frequency and response time come into play, the loadline equation is fed to the algorithm processor and then to the digital to analogue converter, it takes time to act on the loadline spec because of all the ADC, just when LLC is enabled loadline spec would take more time to compensate for the negative feedback of an overshoot than an analogue controller would even though overshoot is very minimal with highend digital and analog PWMs. Digital has to go through 3 complex steps A to D Algorithm then D to A, analogue is more just A to A response time is much quicker such as when going idle to load.

Switching frequency is directly related to power output and I wanted to specifically tell people why and what ranges, so when they are benching their digital ASUS or/EVGA board are at 200Hz and are trying to hit 5ghz and have no idea why, it is because they haven't set switching frequency to 800hz+. For analogue VRM design is thought it was also important, as with lower switching frequency comes more efficiency, and lower amperage from 300KHz-500KHz, so the system can rest just like with digital VRM, then at 1MHz the driver MOSFETs can output current like no tomorrow. i think what is most important was to teach people about digital vs analogue VRMs, and you hear a lot about Digital VRMs, and there just isn't too much info people think just because it is digital and new that it is better, I felt people should here the pros and cons of both.
The main part of both PWM are the same, both have DVID input SVID input, channel current sensor, delta sigma encoder, and some type of monitoring systems(OVP, OCP, temperature). In an analogue PWM controller delta sigma encoder handles (OVP,OCP, temp)monitoring and current monitoring and balancing and then relays signal to the soft-start and fault logic processor to monitor vref and vout and to check the error, then that signal goes to a ramp generator/analog pulse control module, then the signal goes to an analog pulse module per channel, then out to MOSfet Driver that can be fed less voltage than 12v making it much more efficient because the driver MOSFETs DC-DC conversion is made easier.

on the other hand in a Digital PWM you do not have the soft start and fault logic, instead you have the delta signal encoder directly to the voltage error analog to digital converter, all the things I listed that the two PWMs have in common goto the ADC(analog to digital converter) or their own ADC or the voltage error analog to digital converter where vref is compared and corrected to vout, then the signal proceeds to the digital subsystem which includes current balancer, transient controller, and PID controller which goes to the digital PWM generator so all the channels use the same PWM.

Both are closed loops with internal negative feedback.

in the end both systems come to the same final vout, digital is more precise and user has the ability to control the PWM, while an analogue system has higher efficiency and strong current output and because each channel has its own separate pulse module you are able to use many different configurations of driver MOSFETs to raise efficiency and lower temperature. Digital PMW tech is in its early stages about 4-5 years old in motherboards, while analog systems achieve the same results just by pure hardware design.
The reason I decided to do this now is because sandybridge s going to rely most heavily on the board's ability to deliver power, and I just wanted peopel to know power output, efficiency, precision, ability to control output and quickness of both VRM designs. (PM me if you have more questions I don't want this thread to be an argument over analogue vs. digital. ) serial Vid is supposed to be a 3 wire ultra fast system and I am unsure about the new processor design is going to have the CPU's PCU control a lot of what PWMs used to control or not, but it is a way for them to communicate and speed would mean something.

Also yea that is great have software control power delivery on a board and let users control the software, when is the last time you had software that never crashed? What happens if an error in software computing occurs? All hardware or hardware + software for reliability? Something so new should really have time to mature before it is implemented into the top tier products for micro processors, you don't see too many digital PWMs controlling server parts.

but hey i re-read my power delivery section and found a big typo 800KHz not HZ, and I fixed a few things to make it more understandable and to clear it up.

hey great review! thanks especially for the power lesson and the heat tests of the cooling system. a few questions:

1. to clarify, these are still pcie2.0 x16 slots (not 2.1, not 3.0)? the NF200 provides the SLI ability by doing "something fancy" (i don't quite understand how it works) so you can get SLI x16/x16 off the 8 lanes, so is the same "something fancy" done so that crossfire x16/x16 is possible? how are the "extra" lanes gotten if the cpu doesn't actually got them and there's no "extra" chip like NF200 on the ati side of things?

2. can you try putting in some graphics cards to check for clearance? i am interested in a) extremely long cards (4870x2) running into/covering the sata/usb/etc connectors at the edge of the board, and b) particularly pceix16 slot 4 (possibly #3 too) where the heatsink and caps may be too high and prevent fully shrouded cards (gtx260) from actually locking in tightly.

3. from your photo of the back panel, the yellow ports, i guess 2xesata, 2xusb2, a 1394, and what the heck is that tiny little yellow one?

thanks!

1. the NF200 has the ability to multiply and split one lane of PCI-E frequency two ways and add bandwidth, since SLI is Nvidia's native territory you don't need another IC to Double the bandwidth and divide the lanes in two which in theory is what NF200 does. Lets say you have a 1 lane one way road, you add extra cement to the edges of each side of the lane, and then divide in the middle when you reach two lane street width requirement. You can think about it like Intel's Hyper threading technology, but that definitely isn't how it works, I just wanted you to think about how Intel divides one core into two threads, that actually works and provides windows 8 logical cores from 4 physical cores.

2. There is no problem with clearnance for extra long graphics cards, the 90 degree angle of the SATA ports allows for extra long card as does the low profile P67 PCH block. you are set on this aspect

3. That is a mini IEEE 1394a port I believe.


Thanks everyone once again!

[LardArse]

I have found that Digital PWMs are good as long as you use Chil and not volterra.

How did you find that out? How do you judge a PWM?

[sin0822]

That was more of an opinion. Not a fact. You really like EVGA boards huh? I have seen the designs of both, it comes to company preference. Why did Abit, DFI, ASUS all use Chil? ASUS said they tested and came to CHil because of the options it gave them and it made a lot of sense. I liked how ASUS designed the R3E with Chil Digital PWM and still carried over half Analogue VRM with driver MOSFET, Digital PWMs aren't really better than one another it comes down to what is more attractive. Volterra has a problem in that #1 they don't wanna post spec sheets, #2 they have proprietary products. That is just so weird for an electronics maker, Interstill doesn't make you use Interstill driver MOSFETs. Volterra makes you use Volterra MOSFETs as well as Volterra specific inductors. Not only does that drive the price of the board up it also limits your options as an engineer, what if you don't want to use such a high switching frequency for the MOSFETs? What if you don't like to create a situation where hey this thing is burning up at stock, what if we OC? then it just goes right up with it.

A large problem when engineering motherboards that are made for overclocking (maybe 5-10% woudl use DICE,LN2, Phase change and I am part of that 5-10% we aren't many) is that the majority of people are going to use air and water cooling and thus not really need or want to add more heat to things they do not think about like VRM. When you are dealing with power delivery you are dealing with a system that generates heat in an linear fashion in relation to switching frequency. Even bigger is the drop in efficiency as temperature goes up and it isn't something to ignore it is actually pretty huge. The truth is that OCed CPU is going to draw a hell of a lot of current, MUCH more than at stock. As an engineer you want options for what you can do, one method is to add heatsinks, but they only do so much. When you have MOSFETs that are just too much for the board, you want the ability to switch them out, or add more of them to even out workload and have them to work at a lower switching frequency that offers more efficiency and lower temperatures. These new high-tech Driver MOSFETs work in load varying conditions they will jump from 300KHz to 800KHz to 1.3Mhz as more current is needed, so if load is light they might run at 300KHz, if you are going for 5.5ghz you want to them to work at high as they go, but when you do that heat is going to jump because it is related to switching frequency too which is related to current output. But if you had the option to add double the phases then in theory you could work at half the switching frequency(if they were linearly related) which is more than half less temperature because the jump in switching frequency is not linear. Volterra doesn't make that option available. With them its this is how many MOSFETs you need for this controller and here is where you can buy the inductors, because you have to use this design.

In school I had a class were the professor would mention Chil as a ground breaking company, the professor(he was a researcher because I want to a research institution) would always talk about how Chil has an example of this Chil implemented the delta sigma encoder to reduce the high resolution to a lower resolution signal blah blah. you begin to relate things in your mind. Plus i don't care too much for EVGA motherboards as a whole because of quality concerns I have with them, but the classified is an excellent benching board with Liquid N2 or DICE if I could get some Liquid He that would be nice. Just freeze the whole damn board and it will be straight.

in all reality the differences are very small, and Volterra does make good components.

[LardArse]

I just form my preference based on real world testing , load tests, transient wave forms, and most importantly overclocking on the various implementations rather than theology.

[Confuzzled]

I really like the look of the board.

However I have become wary of things with "Ultra" on them over the last couple of years.

I would like to be proven wrong.

[LardArse]

I got you, so which do you prefer? I have only taken Gigabyte boards subzero. I have never had the chance with R3E or classified but seen Overclocks and seen them go under in person.

I wasn't trying to bash Digital PWMs, they are great for many things. My personal preference has been towards real world testing as well, but all my overclocking experience from R3E and Classified boards I didn't have the chance to take them below 0. Cpu yes VRM no.
I have heard now that Chil PWMs do have problems going cold, which is a very big downfall.
i was just pointed out the reason why Gigabyte picked analogue VRM for this board.

from my perspective as a sub-zero over-clocker i have a preference towards Volterra, but that is prolly too sweeping since judging individual implemenations of individual regulators is more accurate than going by brand name.

[sin0822]

I understand your perspective and i welcome it. I always am willing to learn. you are a better person than I am as you don't judge the book by its cover, some of us aren't as practical.

you obviously have more knowledge when it comes to the subzero aspect of it because of your personal experience and background. I only have perspective from what I have seen and done and learned. You taught me something new today thanks man.

I really like the look of the board.

However I have become wary of things with "Ultra" on them over the last couple of years.

I would like to be proven wrong.

yea ultra and true and other adjectives are sometimes/most of the time misleading. Like True 2oz copper PCB true USb 3.0 and so on. usually it is a small improvement and not a whole new product. Gigabyte just uses it because they have been for a while Ultra Durable 3 has been around for some time, But they don't put ultra on anything else I don't think.

[sin0822]

Added PDF by request:
File name: P67AUD7Preview.pdf File size: 1.98 MB

[Johnny87au]

thanks sin gonna check it out now

[sin0822]

You are welcome I put it in PDF format so that you can print it, a lot of people have requested it this way, it makes sense if you are on the go or want to read it to your son!!!(get him started early!)

[PatRaceTin]

I would expect Top of the line 1155 mobos from many manufacturers review round-up.

[koc]

Great review

Thank you

[Boogerlad]

Are the ite chips used on the cheaper gigabyte p67 boards too? I like how software voltage readings are accurate =)

[sin0822]

I would expect Top of the line 1155 mobos from many manufacturers review round-up.

That would be nice, i just don't have other boards.

Great review

Thank you

No thank you!, and you are welcome, I feel as though you should see the board in depth before buying

Are the ite chips used on the cheaper gigabyte p67 boards too? I like how software voltage readings are accurate =)

Yes the ite for phase switching? yes PCI-E bridge? yes
Super i/o monitoring? yes

The reason I predict voltage monitoring to be very good is because of my experience with it on past GB boards including all the X58A series. From what I hear and will know soon, is that one beta BIOS is throwing off weird readings, but that the official is fine. I will find voltage read points and do a comparison soon enough.

[flashnc]

excellent reveiw...certianly one of the most in depth reviews Ive ever seen.. thank you sin.

One of the questions that has plagued me throughout this whole change has been.

what are going to be the real world advantages of SB over the x58 platfrom in terms of not just overclocking potential, but shear performance?

with so many people going triple screen setups for gaming be it eyefinity, or surround and video cards on board memory usage in high end games reaching into the +1100-1500mbs region ( depending on resolution, the higher obviously, the more usage), how is SB going to impact those types of scenarios?

how will memory bandwith be effected for the better or worse??

I think Im still not informed enough to make a fair comparison to be able to risk telling people in a simulation where it is being said to "wait for SB" before you buy anything new.

This is a forum im very much a part of in helping with issues and also with performance concerns, due to it still being one of those that uses d3d9dlls and 32 bit codeing ( and the limitations that places on higher end systems that arent being utilized ) ( its very cpu dependent, and doesnt take advanatge of higher end vid cards and onboard memory).

any insight would be greatly appreicated..

[sin0822]

i will have a full comparison in depth like this out on the day NDA is up, answering just those questions. Triple Channel Ram as far as I have seen is not much of a gain maybe 10% at most over dual channel, from the improvements I have sent through leaked sources, this platform should have no problem with clocking Ram very high, so that might be a benefit. I don't knwo how this will play out against X58 as X58 is still around and this is the LGA 1156 replacement. Just like everything new the improvements are there, just how much is the question.

here is what i am releasing when SB NDA is up:
C300 SSD review 64gb and I want to say that I will be focusing on SATA6GB natively from the PCH, which is VERY new.
P67A-UD7 performance review, against the X58A-UD5 both at 4ghz clock CPU should be interesting because blck isn't going to be upped on the SB side, I will also go for max clock on sandybridge with this board.
Sandy Bridge Guide on OC as well as Review.

[Johnny87au]

@ sin check out the storage section, c300 256gb running on a p67

[flashnc]

thanx for the reply sin..

While everything im seeing so far in your reveiw points to a very well thought out, and as much as possible optimized board, and certianly a very nice looking board as well, excellent cooling solution..Im curious mostly in how the pcie bandwith will help since it is no longer sharing as with previous platforms.

seems that with the new 580 offering from nvidia and the speeds it will run...in sli, tri sli ..this may well be where this board outshines others...

looking forward to your out of the NDA review

[sin0822]

@ sin check out the storage section, c300 256gb running on a p67

[S] i didn't know we could talk about that, I am guess I can't but that guy can? I can't show OCes either, its ok I guess i wanted to do a full review.[/S]

iTs ok, idk who did that review but they missed the biggest past. If you know anything about SSDs is is that Marvell will allow you to obtain the max 350mb/s per drive, wand with highpoint RR 640 you can get the raid figures from P67, what we(the community of SSD lovers, my audience) wanted to know was how 4k(KB) small file random times improved, as Marvell was destroying those times, and those are the only important figures in OS performance. Rarely will anyone hit sequential large file 350/220mb/s speeds, as the file size required is a lot.

Thanks flash NC i was looking forward to the review as well