ABIT KG7 RAID
Introduction
It seems that Abit have entered this class of motherboard market later than the others. Epox have had their 8K7A and 8K7A+ boards out for a while and Asus with the A7M266 which has been out for quite a few months, all based on a hybrid AMD/VIA chipset, utilising AMD761 northbridge and the ubiquitous VIA 686B southbridge.
Since VIA don’t seem to have done very well with their KT266 chipset with their own northbridge and motherboard makers have had trouble making the northbridge work, they’ve turned to AMD and the 761 northbridge so they can release a Socket A DDR range of boards.
The 761 is a high performing solution, providing the path between memory and CPU, the PCI interface (including AGP, which to all intents and purposes is a special PCI bus, indeed, it shows up as PCI bus 1, with PCI0 being the main bus). It supports DDR memory running at 100 MHz or 133 MHz so it’s compatible with all the PC1600 (1.6 Gb/sec bandwidth, 100 MHz bus) and PC2100 (2.1 Gb/sec bandwidth, 133 MHz bus).
So in pairing the 761 with the 686B Southbridge, the motherboard makers are able to deliver a performance and stable DDR solution to the market.
The Alpha EV6 bus the seventh generation AMD CPU’s use seems to have scaled very well. It runs at 200 or 266 MHz depending on the front side bus frequency (100 or 133). Andy Gibson, an Overclockers UK staff member has seen 190 MHz front side bus, giving a 380 MHz EV6 bus on an Epox 8K7A+.
I wonder if the original EV6 engineers envisioned that their little DDR bus would scale so well. Looking at the EV6 bus like this, let us see just how well the AMD seventh generation platform has scaled.
With all this in mind, let’s take a look at the KG7-RAID’s features.
The KG7-RAID is well featured and is on a par with the Epox 8K7A+ in terms of features. Here’s a quick rundown.
· Supports AMD Athlon 700 MHz – 1.33 GHz or future Socket A processors.
· Supports AMD Duron 600 MHz – 950 MHz or future Socket A processors.
· AMD 761 Northbridge
· VIA VT82C686B Southbridge (686B)
· AGP 2x and 4x support in both 3.3V and 1.5V modes (1.5V at 4x only)
· Highpoint HPT370A IDE Hardware RAID controller
· Softmenu III BIOS Technology using an Award 6.00 BIOS
· 2 x IDE ports + 2 for the HPT370A
· 2 USB ports on the backplane + 2 on supplied fly-off cable
· ATX form factor
· 1 x AGP, 6 x PCI
· Wake on LAN, IrDA header, Wake on Ring, SM-Bus header
· Hardware monitoring including voltages, temperatures and fan speeds
Examining the list, it’s good to see old favourites from past Abit boards make an appearance. This is my first Abit board since the KT7A-RAID and it’s good to see Softmenu III back for jumper less tweaking. Having everything at your fingertips in the BIOS and not having to rely on dip switches or jumpers to adjust things like DDR voltage, CPU multiplier and Vcore, unlike on the Epox and similar boards, is perfect.
The removal of ISA, a common feature on new boards is a bonus and 6 PCI slots are very welcome. Features like Wake on LAN and the IrDA header, while not used by everyone, are cheap to implement and keep those who need the features happy. Using the IrDA to provide a link to a PDA is just one example.
On thing to note is that Abit don’t explicitly state that the KG7 supports Athlon MP/4/Palomino CPU’s. I’m happy to note that the new CPU works great. In fact, this review of the KG7 was conducted using a 1.2MP CPU.
So, with all those features and support for the new CPU’s out of the box, how does the KG7-RAID stack up?
Taking a closer look at the KG7
Let’s take a closer look at the layout of the KG7-RAID.
From the picture and looking at the actual motherboard, we can see the layout is fairly standard. I have a few minor niggles with component placement on the board however. The placement of 5 huge capacitors on the south edge of the socket may put some people with large heatsinks off from purchasing the board. You really do have a restricted area around that south socket edge. I wish I still had my Swiftech MC462 for testing with. My Overclockers Hideout Reactor waterblock fits fine but I hear the Vapochill evaporator needs some modification to persuade it to fit the KG7-RAID. I’m taking delivery of a Vapochill next week so I’ll soon find out.
The placement of IDE connectors at 90 degrees to the vertical edge of the motherboard (when fitting in an upright case) still bothers me since I think the parallel layout would be possible on the board with a small redesign. Placement of the IDE connectors so low on the board is also a poor feature. In this case, the RAID connectors for the HPT370 are the culprits with the floppy connector the worst, being the lowest down and also at 90 degrees. The old KT7A-RAID didn’t suffer from any of these flaws and neither does the IWill KK266-R, my previous motherboard.
Lastly on my list of layout niggles is the ATX connector being at 90 degrees to the top horizontal edge. For me at least, on this board, it means an extra twist of the cable and obscures the 2 fan header to the right of the connector. A minor issue, but one not present on the IWill. It’s a far cry from the insane layout on the Epox 8KTA3+ so it’s not all bad. The actively cooled northbridge is becoming routine and expected so no surprises there.
So all in all, a decent layout with tiny niggles that may not affect everyone, depending on case and device placement.
Performance
Performance is probably the main reason for a motherboard upgrade so let’s take a look at the performance of the KG7-RAID using the common benchmarks.
First off, a rundown of the test machine:
· Abit KG7-RAID
· AMD Athlon 1.2MP Processor (9 x 133)
· 256Mb Corsair XMS2400 DDR (single stick)
· Asus V8200 GeForce3 Deluxe (21.81 Detonator XP drivers)
· SoundBlaster Live! Player 1024
· Adaptec 39160 64-bit Dual Channel U160 SCSI controller
· 2 x 73Gb Seagate Cheetah U160 10,000rpm Hard drives
· 2 x 20.6Gb IBM Deskstar UDMA100 5,400rpm Hard drives
· Creative 12x IDE DVD, Plextor 12/10/32s SCSI CDRW
Being a DDR motherboard, the memory performance of the board comes under some scrutiny. Sandra is the preferred method of analysing memory performance and bandwidth amongst the hardware websites and for parity, that’s what I’ll use here.
As we can see, the memory bandwidth of the board is particularly impressive. The Athlon MP helps with the memory benchmarks in Sandra, especially in the ALU test (the first number), with a significantly higher score than the equivalent Thunderbird core Athlon at identical speeds (with a test 1.2 Thunderbird scoring 626 on the ALU test vs. 724). Overall it is an impressive start from the KG7-RAID.
Next up we have the Sandra CPU benchmark at stock speed.
Again, as well as showing off the KG7, the CPU really shines here, offering very good raw CPU horsepower when compared to CPU’s like the Pentium IV. The MP with its data prefetch unit and other core enhancements when compared to Thunderbird is able to show a good turn of speed, even at stock speed. It really is a decent improvement over Thunderbird and while the CPU carries a price premium at the moment, they are a good investment.
In the multimedia benchmark in Sandra at stock speed, we can see again that the MP runs the SSE test by default, not the 3DNow test, due to it’s implementation of Intel’s SSE instructions. It posts a good score, edging up on the 1.33MP reference CPU in Sandra and biting at the heels of the 1.6 GHz P4.
Next up we have 3DMark 2001 from MadOnion. This is primary a test of graphics performance, but it also stresses the CPU and memory bandwidth of the system and is still a useful tool for comparing performance of systems on the same motherboard and the same clock speed and same graphics chipset.
This score compares very favourably against a similarly clocked CPU on similar motherboards like the Epox 8K7A+. Using the MadOnion online result browser to gather scores for a 1.2MP CPU at the benchmark settings, we see that KG7-RAID manages a score of 6368, placing it right where we expect it, compared to other systems in the ORB which again score around 6400, going up to 6900. Yes, we’d expect higher scores from the systems in the ORB, but remember, the 1.2MP isn’t a common CPU and most of the systems in the ORB using the 1.2MP are using the Tyan Thunder K7 dual Socket A motherboard. Given time, we will see a lot more 1.2MP scores in the ORB.
The MP on the KG7-RAID keeps up with the same CPU running on the Tyan. While 3DMark doesn’t use the 2nd processor at all, it’s nice to see the KG7-RAID can keep pace with a motherboard costing near triple the price and aimed at a much more demanding market.
So, to conclude the testing at 1.2 GHz, the KG7-RAID is a favourable performer. I wish I could have directly pitted it against an Epox 8K7A+, the current 761/686B leader but that’s possibly for another review. Using Sandra and the ORB, we can see that the system performs very well. Using the MP in the KG7-RAID helps as you would expect. They are a good match.
Overclocking performance
Abit boards are generally considered motherboards for the enthusiast, for he or she who likes to extract some free performance via overclocking and tweaking. This board is no exception. SoftMenu III makes a very welcome appearance on the KG7-RAID and in this implementation, paired with the 761 northbridge; Abit has given us an insane amount of adjustable options with regards to tweaking and overclocking. You get complete control of the DDR memory timings and comprehensive control of all the tweakable elements of the PCI configuration registers and the 761 chipset including AGP options.
Full voltage adjustment on all 3 voltages (DDR, CPU, I/O) and multiplier and front side bus adjustment on the CPU, all from within the BIOS are class leading. The Epox with it’s messy combination of dip switches and jumpers, spread all over the motherboard are clumsy. Having it all in the BIOS is a major bonus with the power to tweak almost all aspects of the system at your fingertips.
So how does it perform when overclocked? Is it overclockable at all? I’m glad to say answers are good ones. It performs very well when overclocked and tweaked and yes, it’s very overclockable.
Previously I had a pre week 31 Athlon 1.2MP CPU which did around 1480 @ 1.85v in the KG7-RAID. Just recently however I stumbled across a new 1.2MP owned by friend of mine who happened to sell me the original 1.2 I had. A quick bit of wheeler dealing and I found myself in the possession of a new week 31 chip. It’s still a 1.2, but this time able to clock a lot higher.
Currently running at between 1.52 and 1.55 on the KG7-RAID, our theory is that the new 1.2’s, at least week 31 or later are the new 1.4 cores. It certainly does 1.4 at default volts and hits 1.5 easily. While the current limits of the MP core are yet to be hit, Macci, currently top of the 3DMark 2001 charts and the first person to break 10000 has had a 1.2MP at 1585 and holds the top MP score on the ORB.
Onto the benchmarks.
Running at 160 x 9.5 for a total of 1520 MHz, the KG7-RAID and Athlon MP really start to stretch their legs. We see an increase of 700 3DMarks for our increase in front side bus speed and CPU clock which is quite an increase from a simple CPU adjustment. The extra 27 MHz on the memory speed and front side bus will give a handy increase in memory bandwidth, approaching the 2.4 Gb/sec of PC2400 DDR memory running at 166 MHz. Analysing the memory benchmark from Sandra we see around a 26% increase in the scores, primary due to the extra 27 MHz memory speed. Finally, looking at the CPU and CPU multimedia scores we again see a healthy increase across the board, which is to be expected given the increases in clocks.
Overall, the 1520 MHz speed of the 1.2MP is a comfortable overclock and should be doable by all Week 31+ CPU’s, unless I got lucky. The 160 MHz front side bus speed is a comfortable speed for the KG7-RAID and during test I’ve seen it as high as 170 MHz which is a superb speed for this class of board. Excellent stuff from the motherboard and equal to its class peers.
Conclusion
So, what about the KG7-RAID as a whole, taking into consideration features, layout, initial performance out of the box and subsequent performance when overclocked and tweaked? This is the first DDR board I have had and I come to the platform from the excellent KT133A solution, the IWill KK266-R. I only switched from the KK266-R when I found out that revision 1.1 IWill’s don’t support the Athlon MP. I’ve seen a screenshot of an MP booting at 1.6 in a Kryotech SuperG system on the KK266-R but presume it was a later 1.2 revision board.
I used the MP not working in the IWill as an excuse to purchase the KG7 and make the move to DDR. With the demise of my super Thunderbird CPU and the inability for the IWill to boot its replacement, I finally gave in to temptation and made the switch to DDR. Am I happy with the new performance? I have to say I am. As someone who didn’t adopt the new platform as soon as it was launched, preferring to sit back and let things mature a bit, I like the increase in performance borne out the benchmarks. I have to say real world performance has increased too. I don’t think I’d have been so convinced if I was running the KG7 with a similarly clocked Thunderbird CPU, but the Athlon MP makes all the difference to me in this new system.
I’ve been archiving some old VHS to DivX and making copies of some DVD’s in DivX format so that I can take them with me on my travels with my laptop. The Flask performance over my old Thunderbird/KT133A setup is more than I hoped to expect and very welcome and I’ll revisit the board and CPU with some extreme cooling at a future date and I’ll use Flask to illustrate the speed. I was hoping to give you some Flask numbers in this review, but I’ve overrun the deadline for the review enough as it is. Even with the CPU clocked slower than my old Thunderbird chip, the MP’s new core tweaks and the DDR memory bandwidth help Flask a lot. Not so much in quicker encoding rates since they weren’t slow to begin with, but being able to encode at higher quality in the same time period.
SETI, whilst I don’t have number is significantly faster on the MP than the Thunderbird at the same clocks.
I do have one sore point with the board however, despite the features and performance which are first rate. This board and many other KG7’s and KG7-RAID’s are suffering from severe boot problems. This system can take up to 10 minutes and lots of work with the reset switch to bring it to life. Powering on spins up the drives, but the VGA card isn’t initialised and there is no screen activity. Holding down the reset button for a few seconds and letting will bring the machine to life eventually, but it can take many tries and quite a few minutes to coax it into life. The system will not soft reset from software or the BIOS either. Initially fingers were pointed at an incompatibility with Corsair PC2400 memory modules, but I have since tried a Crucial PC2100 stick and the system exhibits the same symptoms. I’m hopeful a BIOS update will cure the problem and I’m waiting on Abit replying to an email I sent outlining the problems I have.
Overall, the upgrade and switch to DDR was cost effective and relatively painless and the performance increase is most welcome. The boot difficulties are extremely annoying but when the system does boot, it’s very quick in use and I enjoy the power that the DDR and Athlon MP bring. The KG7-RAID seems to be a very capable platform for the new Athlon CPU’s. With KT266A boards about to hit, I don’t think the 761/686B combo will hold the AMD performance crown for much longer, but at the moment, the KG7 is a fine board and in RAID spec, even better. Overclocking is great and there are plenty of tweaks to keep the enthusiast happy until KT266A and NForce hit.
Provisional thumbs up to Abit on this one, as long as they solve the boot issues! Well done.
