Well that's just it, it doesn't really hurt to have the frequency on max from the point where you go into BIOS and set CAS, there's no real affect on the clocking margin from having the frequency up high as you're loading Windows when you're ready to OC.
Raja i used on the fly description in my article so users can distinguish between software and hardware changes not some lame ass gimmick marketing you are trying to make it out to be :p:
its just a simple switch that saves a reboot when you're subzero so you can flick to different frequencies during initial CPU testing over a few sessions. it is just a little easier. Its all about user control. Does it help, well if i tell you one way or the other you wont trust me so its best people do their own testing and see what works best for them with their setups.
Actually my initial post explained why the switch was there - because it's an analogue VRM so needs a GPIO if it is to be set by software. Your response to that was that it's not there to look pretty so I asked why it was deemed beneficial to have it in hardware. You're really only playing with transient response and to an extent overshoot/undershoot duration, I suppose it depends on how important you think that is to do in the OS rather than just "set and forget" once in BIOS.
really nice! :) Maybe a good evolution path for my X58A-UD7(rev1) since i'm not leaving x58 that soon :)
i am not sure if there is a benefit of it being hardware over software,
it does help with stable frequencies though,
990x i tested, 6500 only with the lower PWM frequency settings, when i jack it to 750 it will now bench 6600.. this setting definitely helps and i dont think it matters if its in a hardware switch, controlled by bios or is plugged into your mind via usb... if it works, it works...
if you think it is urban myth, please take some CPU and try different settings, same bios settings, same subzero temperature and try different settings..
lol i definitely will take hardware over bios anyday, for me its just another placebo effect that i am actually doing somthing.
Either way its there, its a great feature, and there is no reason to argue over it being hardware over software, its just not something you need on an analogue PWM which makes it neat. Plus it works.
pro- did you try pushing it to 1mhz?
Would the transient response at say 400KHz be potentially worse than the UD7 or other analogue PWM board due to the lack of electrolytic caps? What about the tantalums ripple handling?Quote:
Originally Posted by raju
sorry in my above post i meant 800 not 750,
1000 is tonight bro ;)
My point is that it's not something that makes a difference from where it's set Pro - so not a requirement. Unlike jacking CPU frequency within the OS - which needs to be done because a processor may not be unconditionally stable at XXXX frequency.Quote:
Originally Posted by youngpro
I was told it wasn't there to look pretty so asked what it was doing that a conventional setup can't do, and you've proven that it does the same thing.
You could have set 800KHz in BIOS and OC'd to the same level. I'm glad it isn't an urban myth that you need this control in the OS to do something special :)
i never tried to say it was :shrug:
Quote:
Originally Posted by CryptiK
Of course it would be worse at 400KHz than having it set up higher. As for how it compares to the other Gigabyte implementations at a given switching frequency, the only way you're going to know that for real is to have someone do transient load testing with the right equipment. There are too many factors involved to call it based on raw levels of capacitance alone.
Your ripple is going to be lower at higher switching frequencies too, which negates the need for larger levels of bulk decoupling. There are optimum levels of decoupling for every PSU, add too much and you risk making the performance of the power supply worse.
Tantalum caps are generally favored over electrolytic types because of their better hf response and durability; microfarad for microfarad, their esr will be lower so you don't need the same level of bulk capacitance for the same "effect". From a raw performance standpoint MLCC caps are better still (lowest esr at hf which is where "most" of the padding is needed on multi phase setups that operate at high switching frequency). After a certain point, socket inductance becomes the dominant factor and no matter how much decoupling or "bandwidth" you have, it'll be will be lost via the interconnect; it's then down to the local decoupling on the rear of the CPU.
On a side-note, unless the controller is having issues with a high switching frequency at higher levels of current draw - there is no need to manipulate it up and down during benching sessions - you could go into BIOS and set 800KHz or 1MHz because you're benching and leave it at that. The only reason to pull it back down is for the sake of energy efficiency. The relative location of the frequency switch on this board also makes it a candidate for condensation, you'll need to seal around it and ensure there is no chance of moisture "bridging" the contacts.
-Raja
from what i have learned tantalum capacitors have many benefits over electrolytic, many of which are very beneficial to subzero overclocking, first they usually are able to run at much lower temperatures, second they have much better volumetric efficiency meaning they can store more than equally rated electrolytic capacitors, third they seem to not degrade over time and actually are rated for longer lifespan. While they are more expensive, they do offer many benefits, lower series resistance(ESR) seems to be huge, because that was one of the biggest problems with just tossing in electrolytic caps, and it seems that is why they are able to raise the bulk capacitance on this board. Correct me if i am wrong but MLCC caps also exhibit very low series resistance? correct?
I think they were added on this board for 2 huge reasons, #1 it is much easier to insulate a board with tantalum caps, and i think that is why every capacitor(other than the MLCC Caps(the really small ones like on the back of the socket) were replaced. This can also help to insulate the PCI-E area.
#2 I think hicookie really pushed for high performance, as in lower ripple and better transient performance.
Raju- while you said there is a direct correlation between switching frequency and transient performance, i think there are many other variables that can impact transient performance as well. i am not disagreeing, there is a huge correlation between transient response and lower overshoot and undershoot and transient. There is also is inverse relationship between efficiency and switching frequency, which is why GB has aimed to increase the phase count, because that way they can still get pretty damn great efficiency and handle loads much better.
i think capacitance and overall inductance has a huge impact on transient response. Of course increase inductance seems to have a negative impact on switching frequency, and from what i have seen, GB is using inductors with half the inductance rating or those previously used(plus the new inductors are rated at 50A!!), as well as a HUGE increase in polymer capacitance. I do agree that there is a balance point for every VR for capacitance. In the past it seems GB has had success using much lower bulk capacitance. From my calculations they are using 8000uF of polymer bulk output capacitance(not counting MLCC capacitive) from the tantalum caps, which is something i have not seen on a GB board.
of course not all of us, if anyone other than the mobo companies have a VRTT, I am sure GB has been putting theirs to good use.
in the end this board is going to be fantastic, and I cannot wait to write about it!!!!!
-Steve
Raju i would like your input?
MLCC caps have far lower ESR than electrolytic and tantalum types over most of the frequency range but are susceptible to "drift" in capacitance according to temperature (losing upto 50% of their capacitance at -50 Celsius). The upshot is that you can get away with almost 50% of the capacitance with MLCC versus tantalum across most of the workable range and they still have lower ESR. Just depends on what your design goals are and the capabilities of the FETs and controller.Quote:
Originally Posted by sin0822
One would assume it's a better solution for overclocking than what is on the other GBT boards, otherwise there would be no real reason to change anything.Quote:
Originally Posted by sin0822
Given current demands of modern CPUs and the buck controllers/FETs that are available, 8~10 phases is plenty. That way you don't waste excessive space, complicate the topology to the point where ripple and other ringing artifacts manifest or create issues with balancing across a larger power plane. Of course, if the design brief was to chase high efficiency and maximum raw current handling at the expense of the other scenarios, then I suppose the brief has been met.Quote:
Originally Posted by sin0822
The comment I made above about switching frequency was relative to a single design - ie increasing the switching frequency of a particular VRM versus running it slower (provided the system can handle it). Hence why I said each VRM would need to be evaluated in transient tests to find out which is superior at a given operating frequency.
I've got nothing against the board at all, I just found the choice of using hardware mode on switching frequency a curious one given where the switch is placed and the board's target audience. Like you I hope this motherboard is the start of something new. We all want products to improve and as with all things, there is a balance between what is really beneficial and the overall execution. Asking questions and evaluating the purpose and need for various features is all part of the process - an interjection of objectivity instead of raw bally-hoo isn't a bad thing.
-Raja
Didn't Gigabyte make a video showing a UD9 supporting like 1500 watts of power to just the CPU socket? Isn't that like 3x the max power required on even like a 7+Ghz 990x?
I am glad that we agree on what i said :)Quote:
Originally Posted by raju
Callsign vega this one: http://www.youtube.com/watch?v=YsFG0Pv-7bM
Most of it was already in this http://www.xtremesystems.org/forums/...0&postcount=61 post above, but you missed it somehow ;)Quote:
Originally Posted by sin0822
WRT inductance and high switching freqeuncies, the inductors are storage, the larger the inductance the more time it will take to charge them and the more bulk capacitance you will need.
hahah you know what this looks like to me, Raja and sin0822 fighting over a toy in kindergarden,
no it's mine, no it's mine :D
lol kind of, i just glanced over it, i was just trying to go over some things that you missed lol, but yea we are saying the same thing. Anyways you want to see some sick voltage regulation take a look at the DMM: http://www.youtube.com/watch?v=iTHqmgePj2I&hd=1 6.6ghz real world performance, from 1.831-1.8350 and no more.Quote:
Originally Posted by raju
Are we? lol I thought I was saying something else Sin. :D
nah i am trying to agree the best way i can lol. :)
That's cool, over the past few weeks we've covered small tidbits on things like where VDDQ comes from, switching frequency and current thresholds, and now a small one on higher inductance needing more bulk capacitance rather than the other way round. It'll take time for that to sink in to the point where it really clicks, so take your time. It took me a while just to get here, and I'm not anywhere near what's needed at EE level to design this stuff from the ground up (factoring loop gain and phase margin).Quote:
Originally Posted by sin0822
-Raja
well its not just that, i do a lot of my own research, i have a subscription to IEEE and that has shed a lot of light on a lot of things. I actually got the info on inductance and capacitance from the OCmag article and some forums for EE, but yea you said the same thing, but i tried not to copy what you said..
I hope you understand it before you post because some of the stuff that I and others have had to correct over the past few weeks suggests otherwise. The info in datasheets and whitepapers is helpful if you understand what it means otherwise what you present can be wide of the mark.
-Raja
Did is the Gigabyte has confirmed that it intends to introduce full UEFI implementations (friendly graphical interface with 1024x768 resolution,support mouse,more languages - Asus efi bios currently support 6 languages ,Msi efi bios support 15 languages and other efi features )) to their bios the socket 1155, AM3 + and future sockets ??? because for now uses it hybrid efi bios.
http://gigabytedaily.blogspot.com/20...37625918667235
quotation from page http://www.gigabyte.com/products/pro...px?pid=3761#ov
DualBIOS™ 3TB+ HDD Support (Hybrid EFI Technology)
Hybrid EFI Technology combines the benefits of GIGABYTE's mature BIOS platform including stability and compatibility with 3rd party products with 3TB+ HDD support from EFI technology, allowing GIGABYTE to offer the best of both worlds through a quick and easy BIOS update using GIGABYTE's @BIOS utility that is freely available from the GIGABYTE website.
one could surely do with the Bios would be chip version of the classic elements of efi bios and full efi bios with support for mouse-friendly graphical user interface, support for multiple languages.
Did Gigabyte hybrid efi bios, running in 32-bit or 64 bit mode and is immediately available to boot the system, all memory. If your motherboard bios P67 B3 operates in 16 bit mode and is available for the latter only 1 MB of memory??