The Gigabyte GA-X48T-DQ6 - Redefining the High End?
by Rajinder Gill on January 2, 2008 3:15 AM EST- Posted in
- Motherboards
More BIOS Details
Command Rate (CMD)
The options are Auto, 1N and 2N. A setting of 1N can be used almost exclusively while overclocking, bringing a performance advantage of around 2ns to memory access latency. If memory speeds over DDR3-1800 are desired then a setting of 2N may be required for stability.
CPU/PCI-E Clock Driving Control
The default setting is 800mv, with a range of voltage control offered between 700mv-1000mv. As this is a differential amplifier circuit, increasing voltage may actually decrease the clock signal accuracy due to increased power supply noise. Differential circuits are used in preference to single ended circuits because of their noise rejection and low voltage operating capabilities. Increasing voltage to these circuits in turn increases "nasties" such as overshoot and output clock signal jitter. This in turn counteracts the benefits of using a differential amplifier in the first place. We did experiment with various levels of overvoltage and found no gains in stability whatsoever, further cementing our beliefs that more is not always better.
CPU Clock Skew Control and (G)MCH Clock Skew Control
These two clock skew settings are directly related to the voltage control circuit above. They control the PLL output to both the CPU and Northbridge. Again, as a differential amplifier is used, the level of offset required should never exceed 150ps (Pico seconds) of skew to either the CPU or Northbridge reference clocks. At most, PLL circuits such as these should be "good enough" to retain a jitter level of around 150-200ps (lower is better). For those wishing to experiment, adjustments in the range of 0-200ps are of interest for both of the clock skew functions. As a rule, start with the lowest voltage possible, tune either skew setting, and then monitor for effects if any.
DDR3 Overvoltage Control
Base VDimm is 1.5V. The scale offers between 1.5V-3.05V, more than enough to fry any DDR3 modules. 3.05V is certainly more than enough voltage for extreme benchmarking.
PCI-E Overvoltage Control
This sets
a level of overvoltage to the PCI-E bus. The available voltage range is between
1.5V-2.25V. We recommend staying close to stock voltage; higher voltages cause the
board to shut down during the boot cycle. Most modern graphics cards do not benefit
from higher PCI-E voltages. Increasing this voltage will increase noise and magnetic interference into nearby sensitive circuits. Remember, more voltage always has side effects, it is best to let graphics cards draw their power from the PSU PCI/E connectors rather than through the motherboard.
FSB Overvoltage Control
This voltage setting is more commonly known as VTT. Default is 1.10V, with a maximum of 1.41V available. This voltage is critical for quad-core overclocking. 400FSB will require 1.41V right off the bat if stability is desired. (Ed: When isn't that the goal?) Unfortunately, we have no direct control over GTL (Gunning Transceiver Logic) reference values, which are locked at 67% of VTT for CPU die 1 and 63% of VTT for CPU die 2. On a top-end board like this, we had expected to find GTL adjustments, which can be crucial for finding quad-core CPU stability at high FSB speeds.
(G)MCH Overvoltage Control
Default voltage is 1.45V, and the available voltage scale provides an overvoltage of 0.025-0.775V. There an actual undervoltage of 0.2V from the BIOS set overvoltage. For example, as the stock voltage is set to 1.45V a BIOS setting of +0.50V would imply a voltage of 1.95V. However setting +0.50V gives an actual voltage of 1.75V, suggesting an undervoltage of 0.2V.
Loadline Calibration
Setting this to Enabled reduces Intel's specification voltage droop to the CPU when significant power is drawn. Setting enabled places more strain on the PWM circuit and increases voltage overshoot (within the corrective feedback loop) and settling time from load oscillation, which in turn can affect CPU overclock stability. The trade off is that disabling this setting will require a higher Vcore increasing idle CPU temperatures. Vdroop under load is in the region of 0.06V. Therefore, if a full load voltage of 1.35V is required for stability, a setting of 1.41V will be required for Vcore with Loadline Calibration set to Disabled. The Enabled setting gives a Vdroop of around ~0.02V under load. Users are advised to pick a method that suits their cooling, and use common sense to work out how much voltage is safe at idle if Loadline Calibration is set to Disabled.
CPU Voltage
A range from 0.50V to 2.35V is available, making this board suitable for every type of perceivable over and underclocking. Please note we have not experimented with voltages higher than 1.90V with our cascades cooling, so we cannot guarantee voltages over 1.90V actually work.
Facebook
Twitter
RSS
37 Comments
View All Comments
Ozlaw - Monday, June 23, 2008 - link
I recently switched the inner workings of my computer from a Sriker Extreme, X8000 CPU and an 8800 GTX, to the Gigabyte GA-EX38T-DQ6, with an Asus HD 3870 X2 1gb (and I was about to add another in crossfireX, until I found out AMD was rolling out the 4000's series X2 early, actually possibly in Q3 of this year). If I had known, which is the story of my life and computers, that a newer chipset was coming out so soon, I would have waited for the 48T. The funny thing is that ever since I studied coomputer progaramming in college in the 70's, and began building my own computers when costs became parctical to do so, I suscribe to all the industry news but always miss something big.Oh well I can say that the 38T, since it was designed aroung quad core, including the extremes, instead of relying on BIOS updates to recoginize and work with critical hardware, such as the 680i in my old Striker Extreme was doing (I had the QX6850 in int for a time)in and of itself made an ennourmous difference in my computer's speed and functionality.
THanks for a good article and now I am going to have to decide whether to buy a new motherboard when I buy the new graphics cards as I admit to being on of those weak early consumers of tech products, althhough I am getting better when it comes to Intel's cpu's now that they have become a CPU of the month club.
papatsonis - Wednesday, April 16, 2008 - link
"(G)MCH Overvoltage ControlDefault voltage is 1.45V, and the available voltage scale provides an overvoltage of 0.025-0.775V. There an actual undervoltage of 0.2V from the BIOS set overvoltage. For example, as the stock voltage is set to 1.45V a BIOS setting of +0.50V would imply a voltage of 1.95V. However setting +0.50V gives an actual voltage of 1.75V, suggesting an undervoltage of 0.2V. "
The default voltage of X38/X48 is 1.25v (and in the article explaining tRD values , states that also) , the gigabyte in fact OVERvoltages MCH , when set to normal , to 1.45volt , but when adjusting values, it sets correct values (multimetered) e.g. +0.125 ->~ 1.38v
Galvin - Monday, April 28, 2008 - link
I dont have a volt meter. But I did verify thats what happening.I have the x48 ddr2 version. Normally the NB heatsink is too hot to keep my fingers on it. So if you run less voltage the heatsink should be cooler to the touch. Which is what happened. I set the overvolt for the MCH to 0.025 then rebooted. Waited a few min and I was able to hold my fingers to the heatsink without burning. I did this 2 more times to be sure.
Surprised no one has found this before. That explains why the NB heatsink was crazy hot.
Thanks
neat1 - Sunday, January 6, 2008 - link
http://www.firingsquad.com/hardware/intel_skulltra...">http://www.firingsquad.com/hardware/intel_skulltra...neat1 - Sunday, January 6, 2008 - link
Does this board supp sli or is there still only crossfire support on the x48 chipset? (which would be kind of odd)Looking fwd 2 see the ASUS X48 tested
Regard/neat1
Rajinder Gill - Sunday, January 6, 2008 - link
Hi,Right now Nvidia has no plans whatsoever to allow SLI on the desktop Intel chipssets. Skulltrail may be the only exception to this. Of course, that's assuming Skulltrail ever gets released.
regards
Raja
Rob94hawk - Sunday, January 6, 2008 - link
Benchmarks with skulltrail or just hype?Rajinder Gill - Sunday, January 6, 2008 - link
Hi Rob,Personally I have not seen anything of Skulltrail myself, Intel are notably tight-lipped about publishing anything. Of course we'd all love to have a play...lol
regards
Raja
kilkennycat - Friday, January 4, 2008 - link
Is Gigabyte paying Anandtech for front-cover "pseudo-advertising"? I though Anandtech exclusively reviewed quality products. Seems as if the pre-fitering within Anandtech of products for review needs to be strengthened.This review is so full of wishful thinking about future BIOS updates (whhich may never materialize) that I want to scream. Surely there are computer-related products in a far more mature state begging for serious technical reviews?
Rajinder Gill - Friday, January 4, 2008 - link
At least the preview/review let's people know what to expect at present. I guess I am wishful, because I hate to see good products go to waste via an inadequate BIOS. Immediate maturity with top-end performance boards (the reviews I concentrate on) is hard to find, almost impossible in fact. Occasionally we find a gem, when we do, we write about it. Not every board that comes down the channel is going to make the masses happy, especially in this segment. It's one of the toughest segments of all, the budget minded will never be happy with it. But then these products are not aimed at the budget segment at all. In the PC world there's no such thing as linear performance scaling per dollar. This board needed a 2 stage review, because fo the length of time the cascade cooled results can take. Where this may not be the typical approach, the performance boards are used by a percentage of this crowd. The rest is the more stable down to earth stuff which we endeavour to cover with the BIOS guides. It just so happens this all fell in with the Christmas season, when most companies go into shutdown.We review these boards in a way that the people who have the dollar or inclination to spend this kind of money will use them, no more, no less.
regards
Raja