Rage3D State of the Union - AMD 790GX Review Rage3D State of the Union - AMD 790GX Review... is just as engaging as reviewing motherboards! This isn't because the chipset guys are less talented, nor is it because the motherboard isn't an important part of any system ... no, it's because motherboards work as intended with annoying frequency. No magic driver updates that bring X extra FPS in popular game Y, no amusing PR battles being fought around them. They simply work and ceased being a primary determinant of performance long ago. The delta in performance between an entry level motherboard and the latest and greatest uber-enthusiast proposition is rather insignificant, all else being equal.
Of course, there are reasons for which expensive boards aren't cheap (although there is none for which the prior phrase is dumb, it just came out that way): better power delivery circuitry, more PCB layers, more BIOS options, more this, more that. Whether or not the above enchilada is relevant to the vast majority of users is debatable. All in all, the take-home note is that motherboard reviews aren't all that sexy ... unless something is actually broken, making one of the competitors stands due to its insufficiencies. Did we get lucky, or was it all just a drag? Read on to find out!
The 790GX
What we have here is AMD latest greatest chipset, oh joy! Wait ... scratch the "greatest" moniker; it's not quite that. The 790GX is AMD's latest value segment offering - this means aggressive pricing, adequate features and an IGP (Integrated Graphics Processor) as the value oriented customer may not want to spend extra for a discrete graphics card that might be underutilized in his HTPC/work system. Oh, and you want it to work without much hassle, because having your HTPC crash right in the middle of that really awesome movie, or be a pain to set up in the first place, is a sure way of alienating a large chunk of your potential customer base. All of the above have been thrown into the mix, with a slight addition of extra spices, as you'll soon see, but let's not be hasty and get ahead of ourselves - let's check out the newcomer:
Starting at the top, we're greeted by:
AM2+ Socket: Home is where the heart is, and this is where your new K10 (Phenom and derivatives), or your older K8 (Athlon etc AM2 CPUs work just fine) processor will reside; Deneb, the upcoming 45nm K10 refresh should also be supported, albeit a BIOS update may be required.
790GX Northbridge: Fabbed on TSMC's 55nm process, it has an interesting bag of tricks attached: 22 PCI-E 2.0 lanes, 16 of them dedicated to the 2 PCI-E X16 (physical, X8 electrical if both are populated) slots, an HD3300 IGP with VGA/DVI/HDMI connectivity included, as well as an optional chunk of dedicated RAM (the performance cache bit AKA sideport memory); it communicates with the CPU via a 5.2 GT/s HT3 link (hypothetical, that assumes full 2.6GHz HT frequency, which isn't supported by any AMD CPU currently available), whilst for southbridge connectivity a 4X PCI-E 1.0 link is employed.
750 Southbridge: Traditionally ATI (which means AMD nowadays) southbridges have ranged from truly awful to pretty bad - the latest attempt, the SB600, was in the "pretty bad" category, with flaky AHCI support, limited SATA/USB connectivity compared to the competition and a penchant for causing the dreaded "A clock interrupt was not received on a secondary processor within an allocated time" crash, which plagued quite a few. The new 750 southbridge attempts to break that mold.
Now, that we have an idea of how the topology of a 790GX motherboard looks, it's time to focus on the two really intriguing things that it brings: the HD3300 IGP and its sideport, and the SB750 with its mysterious and supposedly magical Advanced Clock Calibration.
In order to clear potential misconceptions, we'll introduce the HD3300 by stating that the it is simply an overclocked HD3200 (the IGP in the 780G). The overclock, however, is a rather hefty 200MHz, so the bonus in performance should be consistent. In case you were wondering what's under the hood, take a peek:
The HD3300 is functionally a carbon copy of the RV610 (HD2400), albeit shrunk to 55nm, so you'll get all of the above goodies packed into your northbridge should you opt for a 790GX based motherboard. Differences appear once we start looking at how memory is dealt with, since an IGP has to rely primarily on system RAM:
To access RAM, the IGP has to go through the Integrated Memory Controller (IMC) on the CPU, with which it communicates via the HT-link between the northbridge and the CPU - there's a latency cost associated with doing this.
If Sideport memory is present, and this isn't mandatory and is up to the motherboard maker to include it or not, it is accessed via a 32-bit wide physical connection - this means lower latency, as well as avoidance of possible congestion of the CPU-NB link or of the IMC; this is pretty much identical to how discrete graphics cards behave.
When using both main RAM, Unified Memory Architecture (UMA), and the Sideport, there are 2 possible operating modes:
Interleaving enabled: this allows the IGP to request data from the UMA as well as the Sideport in parallel, equating with the possibility of 2 simultaneous reads/writes, as opposed to normal operation when only one 32 bit read or write can occur; for initial Sideport implementations, UMA size equaled Sideport size, but more fine-grained control is now possible, and different Sideport:UMA ratios are allowed.
Interleaving disabled: total memory for the IGP is equal to UMA size + Sideport size, with Sideport memory being used preferentially until filled, after which system RAM is utilized.
All the 790GX boards we've seen allow control over interleaving.
As you'll soon see though, the HD3300 doesn't pack enough processing power to be really sensitive to memory access latencies/bandwidth outside of scenarios where low resolutions/quality settings are employed. But before that, let's dissect the second novelty.
Many things could be said about the SB600, ATI/AMD's last southbridge before SB750 (we'll ignore SB700 as that was not a full step in the right direction, in spite of being an improvement), many of which would be mean and require using 50 Cent worthy expressions. What was wrong with it, you ask?
Limited SATA connectivity - only 4 SATA ports supported
Problematic AHCI - probably also due to its lack of support for 64bit DMA combined with it reporting to the OS that support is present
RAID limitations - RAID levels 1, 0 and 10 only, with the rather popular 5 mode conspicuously absent
Vista+SB600+ some (many) hapless Phenoms:
Why blame the SB600 Southbridge for that? After all, that blue-screen can happen without it being anywhere near - under certain circumstances (insufficient voltage being delivered to the CPU) Intel boards get it as well. Whilst that is correct, seldom does one get it when operating at stock on non-SB600 boards - for example, our 9850 that would only function if underclocked with the MSI K9A2, with even the slightest upwards adjustment of the multiplier causing the above marvel of white and blue to conquer the screen, and stock operation being unstable at best. Getting a 9950, we had better luck ... so dud CPU, right? That might've explained it, until we plugged the 'faulty' 9850 into one of the 790GX boards and found it to be stable.
Another data point supporting the aforementioned BSOD-SB600 link is the ill-fated DFI ICFX2300, which was based on the RD600-SB600 combo, the last Intel board with an AMD chipset. That one was a nightmare as soon as one installed Vista ... it seemed as if clock interrupts were never received! The sole common point between that rather old Intel board and newer 790FX AMD boards is SB600.
To be completely accurate though, the problem is likely caused by a rather complex CPU-clock generator in NB-SB-AMD driver(s) interaction, with the SB600 being the weakest link. Now, after that rather long innuendo, what's up with the new SB750 and how does it patch things up/improve?
Natively supports 6 SATA ports (not necessarily impressive, but an improvement nonetheless), whilst also adding support for RAID 5
AHCI actually works properly this turn (at least it did in our experience, even when using SATA optical drives)
USB 2.0 connection count is bumped up from 10 in SB600 to 12
it adds secret sauce:
Something that's advanced, and handles calibration of clocks must be really good, right? It actually is to a great extent, although it's not entirely clear how it works its magic (and getting an answer from AMD is ... difficult). The biggest benefit it brings, in our opinion, is not the improved overclocking headroom, although that's what's advertised/touted everywhere since it's bound to draw a lot of attention, but its bringing of stability at stock settings to all Phenoms, hapless or not. Remember the troubled 9850 from before? Plug it into a SB750 equipped board, set ACC to auto, and say "au revoir" to the dreaded BSOD ... even a 200 MHz OC becomes a possibility, on stock settings. Fiddle with ACC a bit, and the OC grows to 400 MHz - Noaicee!
Since AMD was reasonably cagey when we approached them about the "A clock interrupt request ..." error, it's not likely they'll advertise that one of the benefits of ACC is doing away with that. However, this is a biggie, because in order to run you must at least be able to sustain your own weight - which means that OC'ing becomes relevant only after stock stability has been achieved.
As for the OC’ing boost, it is indeed there, to a greater or lesser extent, depending on how bad or good your CPU is. We'll discuss this in more detail once we've properly introduced the boards.
What we know about ACC operation is this:
setting it to Auto means setting it to +2% on all cores
values over +6% are recommended once higher voltages are used (1.5v CPU VID and higher)
negative values should allow for lower operating voltages at stock
different CPUs like different values, so you'll have to fiddle a bit, simply leaving it on Auto may not always be optimal
AMD says that the Southbridge hooks into the CPU via a direct interface, with exact details considered Intellectual Property and not fully disclosed for competitive reasons. Peeking at an nVidia slide that was published by expreview.com a while ago (http://en.expreview.com/2008/08/20/nvidia-mcp7278-will-support-acc-overclocking.html), it seems that this happens via the JTAG interface found on Phenoms.
Something also relatively known, if one knows where to look, is that AMD had intended to add an internal clock generator to SB700, but ran into issues which required motherboard makers to use an external one instead; the SB750 solves this, as did a cancelled prior iteration codenamed SB710, which may be another piece in the SB750-ACC-better overclocking/stability puzzle.
After talking with a very smart Friend, we've come up with two possible ways in which ACC might interact with CPUs to produce the results seen since its introduction (it's a one or the other situation, albeit the second proposal would indirectly involve affecting the first aspects as well):
Change parameters dealing with asynchronous clock domains within the chip/clock crossing, by increasing/reducing the number of sync clock cycles - this would correlate nicely with the increased OC’ing headroom aspect, as well as provide anecdotal evidence that indicates higher ACC values induce a slight reduction in performance in something like Prime 95, the core with the highest ACC modifier finishing last, whereas the opposite holds true for negative ACC values, in which case the core with the lowest ACC value finishes first - however, more testing is required to fully verify this.
Adjust skew/clock slew/ drive strength (adjusting drive strength affects clock slew anyhow) - this possibility correlates with the fact that "weak" CPUs seem to receive the most benefit, whilst good CPUs get almost none, whilst also explaining why extreme ACC values can cause boot failure, and why high positive values are recommended for high voltages (albeit it's hard to know what it is you're adding X% to)
Be aware that the above are at best slightly educated guesses ... there simply isn't enough data out there to deduce exactly what's going on. For all we know, miniature dwarven craftsmen could travel via the SB-CPU link to tweak the latter - at this point in time, the exact mechanism remains a mystery. However, for most people the only thing that will matter is that it pretty much works.
This bit should be short as it includes some closing general considerations, before going straight to presenting the boards being tested, and the testing itself:
Power Circuitry: 790GX boards are engineered to support 140W CPUs, a reference to the older 9950 Phenoms - this responds to the issues encountered with some 780G boards that would die rather spectacular deaths when teamed with high wattage CPUs, due to the inadequacy of their power delivery circuitry.
SB750: The SB750 isn't exclusive to 790GX boards, since it is pin-compatible with the older 600 Southbridge and, as such, is an easy upgrade option for 790FX boards - however, the decision lies in the hands of the motherboard makers; as far as we know, only Asus, Foxconn and DFI have opted for releasing updated 790FX+SB750 motherboards.
Hybrid Crossfire: 790GX/780G motherboards support Hybrid Crossfire, a mode of operation available under Windows Vista in which the IGP teams up with a discrete 3450/3470 graphics card - it is our contention that this mode is rather useless - you're better served by getting a more powerful discrete card; this applies to all rendering schemes that combine an IGP with another card in AFR - whilst 3DMark may love it, the gameplay experience won't be significantly improved, and AFR'ing at truly low framerates is not something you want to do.
Audio: Due to its R6xx lineage, the HD3300 included in the 790GX only supports DD, DTS but only 2 channel LPCM output via HDMI - as one acquaintance of ours put it, had it not been for this limitation AMD might've actually produced the perfect HTPC solution.
Stream Computing: the IGP can be used for Folding@Home, albeit its performance for that particular task is less than impressive, as well as doing CAL/Brook+ programming, for those who are interested in doing GPGPU work.
With all of the above in mind, it is now time to meet the boards themselves!
Yeah, we know, martinis and motherboards don't mix ... but there's a Bond in everyone, so be lenient! Ehm, back to the main topic now: we'll be playing with 2 790GX based motherboards from opposite ends of the pricing spectrum:
Gigabyte MA790GP-DS4H: this is a high-end incarnation of the 790GX, with an 128 MB DDR3 Sideport clocked at 1333MHz and a few other goodies which we'll detail in its respective section- this board was 128.99$ on Newegg as this article was going into editing, and was kindly provided by AMD
Biostar TA790GX A2+: here we have the cheapest 790GX available on Newegg, bringing only 64MB of 800MHz DDR2 Sideport goodness, as well as being slightly more Spartan in general - Newegg has this one available for $99.99, and this board we got from the store, to keep and cherish forever and ever ... or until we blow it up.
As a measuring stick, we'll be using the Asus P5E-V HDMI:
Asus P5E-V HDMI: based on the Intel G35 chipset, this has been in one of our work PCs for quite a while, proving to be a very good, reliable and ultimately likeable board - by using it we'll be able to look at how AMD and Intel CPUs match up as well as see how their IGPs handle each other
There may be some cringing because we're using a G35 based board for comparison instead of the recently released G45. However, given the small differences between the two, and the lack of availability that characterized the latter (at least in certain parts of the world), the decision was justified. Basically there are 2 important things added byG45: 2 Extra EUs(Execution Units) and a 133MHz clock bump for the IGP, as well as HD-Video decode assist, with the base architecture being basically the same otherwise (it's also on a smaller process compared to the G35). As you'll soon see, the G35 needs far more graphical oomph than the clock boost and the EU increase can bring in order to become competitive and, as for HD-Video decoding, it's not hard to extrapolate its effect. We would've liked to have the G45 in for the sake of completeness, but it wasn't possible - the same applies to nVidia's recently released 9300 chipset.
Test Configuration
* You'll note that we've also included the MSI K9A2 Platinum V2, which is based on the 790FX chipset. We'll be using it for one test only, in order to gauge the general level of optimization the newcomer boards have achieved, since the MSI is fairly mature and has had numerous BIOS releases, which means it's fairly well tuned by now. Since it lacks an IGP we used a HD 3870 video card.
There's nothing out of the ordinary about the rigs we built, aside from the rather lavish amount of RAM, and even that one's debatable given the low memory prices we're seeing these days. Using this somewhat more exotic configuration allowed us to test how the motherboards deal with such a demanding scenario - it's known that not all motherboards could handle 4 stick arrangements, though this continues to improve.
Pricing
It was also to ensure that the competing platform from Intel was similarly priced, as well as located in a similar performance envelope - as such, we opted for the Q6600 CPU, which is the cheapest Intel Quad-core that Newegg sells, as well as being an excellent processor - some call it the best in history and, when factoring everything in, we're not sure we'd disagree with that claim. Once the math is done, we end up with these prices for each respective combo (we chose retail pricing for the CPUs):
Gigabyte MA790GP ($138.99) + Phenom 9950 ($184.99)=$323.98
Biostar TA790GX A2+ ($99.99) + Phenom 9950 ($184.99)=$284.98
Asus P5E-VM HDMI* ($124.99) + Core 2 Q6600 ($189.99)=$314.98
* the P5E-V HDMI that we used is not listed on Newegg so we used its little P35 brother as reference; the G45 variant of it is ~ $10 more
It seems we've managed to meet the price parity criteria. Once you get to the tests themselves you'll see that we also met the CPU performance parity one, as the 9950 and the Q6600 are fairly evenly matched most of the time.
let's take a peek at what each board has to offer:
Now, time to delve into the details.
Lets start by discussion the Biostar TA790GX A2+:
 |
 |
Board Design: in spite of being a full-ATX board, there are only two rows of screw-holes, with the edge near the DIMMs/SATA connectors being hole free - this is unusual
SATA: the use of angled SATA connectors ensures that long video cards, like the 4870X2 or the GTX280 can be used without fear of losing SATA connectivity
PCI-E: the PCI-E x16 slots are reversed, so the bottom one (orange) is the master slot that can run at full x16 rate, the top one (greenish) can run only at x8 maximum and is intended for use in Crossfire configurations
Cooling: the cooling solutions employed for the NB and the SB aren't spectacular but do their job done adequately
 Northbridge Cooling |
 Southbridge Cooling |
Back Panel: on the back panel we find 4 USB ports, HDMI, D-SUB and DVI video connectors (HDMI and DVI-D cannot be used simultaneously, since the chipset uses the same channel to control HDMI and DVI) PS/2 Mouse and Keyboard ports, 1 Gigabit LAN port and the 8.1 audio panel
Debug Tools: Biostar decided to add two very useful features, namely a pair of debug LEDs that aid one in figuring out what's the cause of certain no boot situations and power and reset buttons, for those using the motherboard outside of a case who aren't all that keen on shorting jumpers
Bios Chip: the BIOS chip isn't soldered, which is something most overclockers will appreciate, since it makes it somewhat harder to completely kill a board
CMOS Clear: the CMOS Clear jumper is placed in a rather disadvantageous position: once the video card is in place, access becomes impossible (unless the card itself is shorter than usual)
Capacitors: Biostar opted for using solid-state capacitors only for the power delivery portion of the board, electrolytic ones being in place everywhere else
Power Delivery: speaking of power delivery, as you can see we're dealing with a 4-phase setup, with 3 MOSFETs per phase - not quite impressive when considering that higher end Phenoms are quite demanding in this area, but Biostar says that their solution is adequate, and were confident enough to rate the board for 140W CPUs
Overclocking: if you plan on overclocking with this board, you'll HAVE to add cooling to the MosFETs ... or be prepared for an explosive experience - they get quite hot under normal operation with the 9850 or 9950, and once we started upping the volts the situation became quite toasty to say the least; we used a high CFM 120mm fan blowing directly on them for cooling, but for typical, closed case conditions, a heatsink should suffice - factor in that distance between mounting holes equals ~80mm when selecting a cooler
And now, the Gigabyte MA790GP-DS4H:
 |
 |
Board Design: the 790GP has the complete 9 hole complement, unlike its colleague
SATA: if Biostar made an excellent choice, going with angled SATA connectors, Gigabyte opted for what is one of the worst possible solutions in this area: not only are the connectors straight, they're also placed in such a fashion that if one were to opt for a Crossfire configuration with dual 4870s, for example, 3 of them would become unusable ... try it with 4870X2s and you'll be left without SATA ports
PCI-E: if the message wasn't clear enough, 4 USB headers are splurged right on top of the slave PCI-E slot (slots aren't reversed on the 790GP)- this, coupled with the SATA placement create a rather clear “Crossfire NO!” feeling, although the 790GX is quite capable on that front, with only dual 4870X2s possibly suffering from less than 16 PCI-E 2.0 lanes available to them
Back Panel: back panel connectivity is quite similar between the two boards, with the 790GP adding a S/PDIF Out optical connector and an IEEE 1394a port
Debug Tools: although there are no helpful debug LEDs, nor are there useful buttons for powering up the system/doing a reset, there are 2 advantages that the 790GP has: the CMOS clear jumper can be accessed with 2 video cards installed (although it requires having small hands/good tools), and, more importantly, it uses a Dual-BIOS approach which should, in theory, ensure that the board is never lost due to BIOS problem, since there should always be a safe copy to fall back upon
Capacitors: another difference from the TA790GX is the absence of electrolytic capacitors - you'll find only solid state aluminum caps on the 790GP
Cooling: if Southbridge cooling is quite similar to what we saw before, for the Northbridge and the PWM Gigabyte opted for a beefier, heat-pipe using solution - this approach has the advantage of cooling the hot MOSFETs that were naked on the Biostar, as well as creating more surface area for dissipation by linking the NB and VRM heatsinks ... however, the FETs get hot, and in the process heat the NB to higher levels compared to those it would normally reach; we ran into overheating issues when overclocking/overvolting, issues which caused graphical corruption, and which forced us to add active cooling to the PWM
 Northbridge Cooling |
 Southbridge Cooling |
Power Delivery: closing the VRM topic, observe that Gigabyte went with a 5-phase solution (4+1 actually) ... we didn't remove the heatsink to show you the FETs since the board was loaned to us by AMD and it's not nice to rip apart other people's toys ;)
The Overclocking part of this article won't be extremely long, because it's more of a first-contact. We didn't resort to unorthodox cooling, we didn't push the volts all that far, we didn't seek to eek out the last little megahertz ... and our chips weren't that great to begin with. Remembering the target market, think of it more as what a casual user, rather than an enthusiast, would try to do. Our goals were multiple:
See what ACC did for the “Chimp” Phenom (the hapless 9850 that tortured us as we were working on the 4870X2 review)
Reach the mighty 3GHz mark with full stability, and possibly explore beyond - this is trickier than it may seem at first sight
Now, as any enthusiast worth a hill of beans will tell you, overclocking AMD CPUs is a rather masochistic way of having fun, and this is exacerbated with the Phenoms. Forget the usual bump FSB, bump/reduce multiplier, bump volts, rinse-clean-repeat approach that worked quite well and brought good results with Intel chips. The Phenom brings heaps of fun and variables to tinker with, but its sensitivity is higher. It's also no Q6600 G0 that does free 3.2GHz overclocks with ease. Time to see what each contender tempted us with, overclocking wise:
Chimp to Champ? 9850 Overclocking
Most of our overclocking adventures happened on the Gigabyte board, mostly due to its cooling solution (we've yet to attach a heatsink to the TA790GX and we weren't keen on killing it). Of course, our first action was to plug in the 9850, enable ACC on Auto and see how it fared ... and fare well it did: whereas on the K9A2 Platinum even a modest 2.6GHz OC was impossible, here we could get 2.8 with only a moderate voltage bump; going above this mark, however, seemed impossible, even when bumping up CPU voltage to 1.45 (our self imposed limit for CPU voltage, on this occasion). Was more possible? Indeed it was ... as we proceeded to tinker with ACC, we discovered that one of the cores (the weakest one?) favored a setting of +6, instead of the +2 enabled by Auto mode, and after making the necessary adjustments, 2.9GHz was achieved:
Above we see it graduating 2 Hours of OCCT 2.1.0 Beta 5 testing, which is nothing to sneeze at (in the K9A2 it couldn't achieve a similar feat even at stock clocks). Of course, it had to suffer through some more OCCT (at 6 hours we deemed it safe for use under load) - and this is only half of the Phenom stability equation, and the easy half at that! You see, Phenoms have a nasty habit of locking up/crashing, when idle. As far as we can tell, this was mostly a problem with earlier chips, since the 9950s we tried appear issue free in this regard, even on the SB600 powered board. The “Chimp”, however, was a troubled individual, so we went on to use it/leave it on for a day, with the overclocked settings. Luckily for us, it did not falter (otherwise we'd have wasted quite a few hours for nada) ... so, in this case, ACC worked in a stellar fashion adding 400MHz (500 if you consider that before we needed to underclock the chip to 2.4 for stability).
Back in Black - 9950 Overclocking
It was now time to bring out the big guns: the 9950s. The one we bought and used in the 4870X2 review was a goodish chip: it could do 3GHz on the K9A2, albeit not quite stable; an ability that ultimately brought about their deaths, as one of the MOSFETs blew whilst pumping current to the overclocked heart of our dear Phenom - another story for another time. And this is where things changed: whilst ACC did bring full stability to our 3GHz overclock, it did not push the boundaries. We could boot at 3.3GHz, but no amount of voltage could stabilize it; 3.2GHz was partially OCCT stable, but it would eventually reboot after 1-2 hours (we're not excluding VRM weakness/inadequacy); at 3.1GHz we had nearly nailed it, but it locked up at idle, so we were left at the nice round 3GHz mark, which was both stress and idle stable (in the picture below note that we had moved to a newer OCCT beta):
The AMD provided 9950 didn't fare all that differently, the sole discernible difference being that it could achieve 3.0GHz with a slightly lower 1.375 (in the BIOS) voltage - however, higher OCs were still not fully reliable. The Gigabyte had a penchant for over-volting the CPU by about 0.02 volts (! - this was worse with the initial F1/F1b bioses) so when looking at the OCCT pictures, factor in that the 9850 was actually set at 1.425 Vcore in the BIOS, with the 9950 at 1.400.
It seems that AMD's assertion that ACC will primarily benefit weak Phenoms was quite correct: whilst it could turn a chimp into a respectable member of society (the 9850), it did little for gentlemen that already had proper manners (the 9950s).
The Biostar could emulate these results, but as we already mentioned, due to the nakedness of its MOSFETs we didn't really stress it. This will change once we “dress” it up nicely.
Cooling the Beasts
For cooling we used either the boxed Phenom Black Edition cooler (not that great), a Xigmatek Red Scorpion (excellent cooler for the price) or a Thermaltake Silent Water with changed fluid and fans (yes, cringe as it is cringe-worthy: we only used it because it made it easier to maneuver around the CPU area/play with the DIMM-slots, but in terms of cooling it was about on par with the Xigmatek, with slightly lower temps, which is most unimpressive for a liquid cooling solution, and for its price) - as already stated, nothing too exotic. The Xigmatek would be our choice for a low-cost, efficient and silent solution, but there is a caveat: due to the fact that it can only be mounted in a single fashion (either blowing air into the PSU or on the video card, it cannot be turned to blow air out of the exhaust due to how AM2 mounting works), you'll have trouble accessing the first DIMM slot near the CPU, since that'll end up covered by the side of the HSF. This applies to all large Heatsinks, depending on how flexible their mounting method is, since the DIMM slots are quite close to the CPU socket.
Once again, before moving on, we'd like to point out that this was mostly an appetizer ... a proper Phenom OC'ing experience deserves a separated article, and it will probably be tied to the unraveling of the ACC mystery (we're trying to work out what it actually does, with help from the aforementioned Friend).
Now, there are a few interesting things we picked up during our stint with these motherboards, aspects which may be helpful to you, some common to both, some specific to each:
Running the RAM at 1066 MHz was far more of a hassle than we had anticipated and, in fact, neither board could do it properly in a 4x2GB configuration prior to the latest BIOS releases - of course, according to AMD, this should not be possible at all since they don't support 4 dual-sided 1066 MHZ DIMMs on the Phenom, only 2 - it seems like motherboard makers didn't quite agree with this and took it upon themselves to provide support, feat which was managed with the aforementioned BIOS updates.
AHCI worked properly on both boards, even when using a SATA DVD-Writer, which is a change for the better compared to the SB600 using K9A2, which would not work properly with AHCI.
The Gigabyte was plagued by a very annoying, and impossible to solve, cold-boot problem: at times it would simply refuse to boot after it had worked flawlessly the night before, with the only solution being removing the powercord from the PSU and letting it sit for a bit, then starting it up - sometimes this also caused the Dual-BIOS aspect to kick in, with the main BIOS being programmed from the reserve one, and messing up settings in the process; this wasn't fixed with the F2a BIOS, and we can only hope that Gigabyte is working on solving it.
Whilst free from such mysterious ailments, the Biostar didn't always play nicely with RAM (the default RAM voltage is too low for some sticks).
Continuing the prior idea, the Biostar has a somewhat droopy behavior, the exact opposite of the 790GP (not necessarily surprising given its VRM).
Overclock protection never kicked in during our play-through with the TA790GX, so if we made a booboo the only solution was to perform a CMOS clear; however, pleasingly enough, we were able to recover from a bad BIOS flash by applying the AMI BIOS restore routine (detailed http://www.biosman.com/biosrecovery.html, using an USB stick instead of a floppy - this is quite convenient and kudos to Biostar for enabling it (other motherboard makers don't do so and floppies aren't all that ubiquitous these days).
Finally, we've stumbled upon a rather annoying aspect: for the time being, a dual 4870X2 CrossfireX configuration is not possible on the TA790GX - the slave card isn't detected by Windows, in spite of being properly initialized by the board. Going through the manual we were surprised to discover that it mentions the fact that CrossfireX is not supported on the 3870X2 either - we've notified Biostar, and it remains to be seen whether or not this is intended behavior or something that will be rectified via a BIOS update.
All this talk about overclocking would be meaningless without BIOS pictures:
 T-series tab |
 CPU FID/VID |
 Voltage configuration |
The Biostar TA790GX opted for an AMI BIOS, in which all relevant overclocking items are grouped under the T-Series tab. Unless you opt for one of the automated overclock presets, you can tinker with most of everything that matters on a Phenom, from HT reference clock to memory timings to NB frequency. Under the CPU tab, significant fine tuning can be performed, as via DID one can use quarter/less than quarter multipliers for increased granularity in terms of core frequency. One slight disadvantage is that NB VID isn't separated from Core VID here, the exposed adjustment adjusts both - in our experience it's possible for NB VID to top out before Core VID, so having separate adjustments would've been useful. Voltage settings are rather coarse, especially when it comes to RAM (0.1 V increments). Finally, the voltage adjustments are self explanatory: do note that Core VID and CPU Over Voltage are additive, so adding 0.05 volts to the former and +0.037 to the latter results in a total 0.087 increase in Core voltage - be careful! ACC is controlled via the EC sub-menu here, or via the ACC sub-menu under the Advanced/CPU tab. And what about the Gigabyte? Easier to present:
A more common Award BIOS is used on the Gigabyte 790GP, with all overclocking related settings grouped under the M.I.T. (Motherboard Intelligent Tweaker) menu. You have to press CTRL+F1 in order to expose all settings, including ACC control. Gigabyte decided to expose NB VID controls, but not Core VID ones, in a somewhat interesting move. Neither Core nor NB divider can be tinkered with, so only integer and half multipliers are available. Voltage settings are far more granular here, using 0.05 increments for the RAM and 0.025 increments for CPU voltage. Summing up, the Gigabyte is somewhat more user-friendly and more refined, whereas with the Biostar employs a more old-schoolish approach. Both are good enough with regards to the knobs they expose, but they're no DFIs (which can be either a blessing or a curse).
In closing, a few words about AMD Overdrive. This was meant to be the great unifying software for overclocking on AMD platforms. Historically, motherboard makers have added their own software solutions for this, but in general they weren't all that great (the two boards we're looking at aren't an exception on this front, Gigabyte has EasyTune whilst Biostar has a T-series overclocking utility). AMD took it upon themselves to solve this difficult problem and to provide a solid, useful and powerful overclocking utility via Overdrive ... to a great extent, they succeeded, but take a peek below:
 Biostar Overdrive |
 MSI Overdrive |
Nope, it's not the color difference that's bothersome (ACC supporting boards get the Black color scheme whereas the others get the red one, by default) - it's the fact that there is such a large delta with regards to the settings that are exposed - the Biostar is very very spartan, whereas the MSI is quite forthcoming. Biostar added insult to injury, as CPU and NB VID couldn't be adjusted (the slight bump was done via the BIOS). Is it AMD's fault? Not really. AOD is dependent on manufacturers implementing the BIOS hooks for the SW. If those hooks (handles, if you prefer) aren't there, the application itself can't do much. So AMD is gambling on motherboard makers being nice, following guidelines and putting everything in but, as we see above, that's not always the case as there's quite a bit of variation between motherboards when it comes to which exact handles are implemented - seemingly no two boards are alike! Other than that, the latest 2.1.4 version of AOD was quite neat - since we're old crones we're still partial to BIOS based overclocking, but we liked it well enough. If AMD actually manages to get motherboard makers to respect a set of ground rules when it comes to implementing the BIOS handles for AOD, as well as fix the few issues still present, AOD will be truly excellent, but the first part of this equation seems hard to solve.